A Role for the Peroxin Pex8p in Pex20p-dependent Thiolase Import into Peroxisomes of the Yeast Yarrowia lipolytica *

Peroxins are proteins required for peroxisome assem-bly. The cytosolic peroxin Pex20p binds directly to the b -oxidation enzyme thiolase and is necessary for its dimerization and peroxisomal targeting. The intraperoxisomal peroxin Pex8p has a role in the import of peroxisomal matrix proteins, including thiolase. We report the results of yeast two-hybrid analyses with various peroxins of the yeast Yarrowia lipolytica and characterize more fully the interaction between Pex8p and Pex20p. Coimmunoprecipitation showed that Pex8p and Pex20p form a complex, while in vitro binding studies demonstrated that the interaction between Pex8p and Pex20p is specific, direct, and autonomous. Pex8p frac-tionates with peroxisomes in cells of a PEX20 disruption strain, indicating that Pex20p is not necessary for the targeting of Pex8p to peroxisomes. In cells of a PEX8 disruption strain, thiolase is mostly cytosolic, while Pex20p and a small amount of thiolase associate with peroxisomes, suggesting the involvement of Pex8p in the import of thiolase after docking of the Pex20p-thio-lase complex to the membrane. In the absence of Pex8p, peroxisomal thiolase and Pex20p are protected from the action of externally added protease. This finding, together with the fact that Pex8p is intraperoxisomal, suggests (Ti8S) (Ti8P) the Ti8S, Ti8P subjected immunoblotting. Oligonucleotides synthesized on an Oligo 1000M DNA Synthesizer performed with an ABI Prism 310 Genetic Analyzer (PE Applied Biosystems, Foster City, CA). Ready-To-Go PCR beads (Amer- sham Pharmacia Biotech). of lysates by disruption with beads, of E. coli , manipulations of DNA as (28). using a commercially available kit (Bio-Rad). SDS-PAGE per- formed essentially as Proteins transferred to ni-trocellulose for immunoblotting using a wet transfer system, anti- gen-antibody complexes were detected by enhanced chemiluminescence (Amersham Pharmacia Biotech).

Peroxisomes are members of the microbody family of organelles and are found in organisms from yeasts to humans and in most cell types. They are the site of many important biochemical processes that vary depending on the organism or cell type. Among their many metabolic activities, peroxisomes perform the ␤-oxidation of fatty acids, bile acid synthesis, plasmalogen synthesis, cholesterol metabolism, and methanol oxidation (1). Soluble enzymes and other proteins found in the peroxisomal matrix are synthesized on polysomes free in the cytosol (1) and are targeted to peroxisomes by cis-acting peroxisomal targeting signals (PTS). 1 Matrix proteins are targeted most commonly by a PTS1, a tripeptide of the sequence SKL or a conserved variant thereof, located at the extreme carboxyl terminus of proteins. Less commonly, matrix proteins are targeted by a PTS2, a nonapeptide located near or at the amino terminus of proteins and having the consensus sequence (R/ K)(L/V/I)X 5 (H/Q)(L/A) (reviewed in Refs. [2][3][4]. Interestingly, unlike other organelles in which proteins traverse the membrane in an unfolded conformation, peroxisomes are capable of importing oligomeric protein complexes (5,6).
Peroxins are proteins required for peroxisome assembly. A subset of peroxins is required for the targeting and import of peroxisomal matrix proteins. The PTS receptors Pex5p and Pex7p bind PTS1-and PTS2-containing proteins, respectively, and are necessary for targeting these proteins to peroxisomes (reviewed in Refs. [2][3][4]. There are conflicting findings regarding the localization of these receptors in the cell. Both receptors have been found in the cytosol, associated with the peroxisomal membrane or in the peroxisomal matrix (reviewed in Refs. 3 and 4). Pex5p and Pex7p have therefore been proposed to be mobile receptors that bind cargo proteins in the cytosol, dock at the peroxisomal membrane, enter the peroxisome, release their cargo in the matrix, and recycle to the cytosol (2-4, 7, 8). Pex13p and Pex14p are integral and peripheral peroxisomal membrane peroxins, respectively, necessary for PTS receptor docking (reviewed in Refs. [2][3][4]. Since Pex13p and Pex14p interact with each other and with both PTS receptors, it has been suggested that although PTS1-and PTS2-dependent targeting pathways are divergent, the import pathways are convergent (9,10). Pex8p is a peroxin associated with the inside of the peroxisomal membrane, and, in its absence, PTS1-and PTS2-containing proteins are mislocalized to the cytosol (11)(12)(13)(14). Recently, a physical interaction was detected between Pex8p and the PTS1 receptor Pex5p in the yeast Saccharomyces cerevisiae, and Pex8p was suggested to function in protein translocation across the peroxisomal membrane following the docking of Pex5p-cargo complexes (14).
Peroxisomal targeting signals and the peroxisomal import apparatus have been strongly conserved during evolution, but interestingly some divergence has been recognized. In certain mammalian systems, two isoforms of Pex5p have been identified, a short isoform necessary for PTS1 import and a long isoform necessary for the import of both PTS1-and PTS2containing proteins (15,16). In Yarrowia lipolytica, the PTS2 receptor Pex7p has not been identified. In this yeast, Pex20p, a peroxin with sequence similarity to Pex5p, is necessary for the dimerization and peroxisomal targeting of thiolase, a PTS2-containing protein (17). Since Pex20p has been shown to be found in both the cytosol and associated with the peroxisomal membrane, it has been proposed to act as a cycling protein that picks up thiolase in the cytosol, docks at the peroxisomal membrane, and then repeats the circuit (17).
Here we report the identification and characterization of a direct interaction between Pex8p and Pex20p. Pex20p is apparently not required for the targeting of Pex8p to peroxisomes, because Pex8p is still peroxisomal in cells of a PEX20 disruption strain. Instead, we provide data that are consistent with Pex8p being directly involved in the import of thiolase into peroxisomes at a stage following docking of the Pex20p-thiolase complex at the peroxisomal membrane. Our data also suggest that, like Pex7p and Pex5p, Pex20p may accompany its cargo into peroxisomes during import.
Cells of the S. cerevisiae strain SFY526 were transformed simultaneously with a pGBT9-derived plasmid and a pGAD424-derived plasmid. Transformants were grown on synthetic media lacking tryptophan and leucine and tested for activation of the integrated lacZ construct using ␤-galactosidase filter and liquid assays. Both assays were performed according to the manufacturer's (CLONTECH) instructions except that for liquid assays, yeast from 1 ml of culture was used, and cells were broken by three freeze-thaw cycles at Ϫ70°C, and for filter assays, agar plates contained 100 g/ml adenine to reduce the red color of the yeast.
In Vitro Binding Assay with Recombinant Proteins-An in vitro binding assay was performed using recombinant GST, GST-Pex20p, MBP, and MBP-Pex8p purified as described above. Glutathione-Sepharose 4B (Amersham Pharmacia Biotech) was washed with RW buffer (20 mM HEPES-KOH, pH 6.8, 150 mM potassium acetate, 5 mM magnesium acetate, 0.1% (w/v) Tween 20, 0.1% casamino acids, 1 g/ml leupeptin, 1 g/ml pepstatin A, 1 g/ml aprotinin, 2.5 g/ml antipain, 0.21 mg/ml NaF, 0.1 mg/ml Pefabloc SC (Roche Molecular Biochemicals)). GST and GST-Pex20p were tumbled end-over-end in batch with glutathione-Sepharose 4B (2.5 g of GST or GST-Pex20p and 5 l of packed beads per reaction) in 0.5 ml of RW buffer for 20 min at room temperature. Beads were collected by centrifugation and washed three times with 0.5 ml of RW buffer. Purified MBP or MBP-Pex8p (5 g/ reaction) was added to 100 l of RW buffer containing GST or GST-Pex20p linked to beads and tumbled end-over-end for 30 min at room temperature. Beads were collected by centrifugation, and supernatants were retained as the unbound fractions. The beads were resuspended in 0.5 ml of RW buffer and applied to spin filters (CytoSignal, Irvine, CA), washed three times with 0.5 ml of RW buffer, and eluted with 50 l of SDS-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer at room temperature. 50% of bound fractions and 25% of unbound fractions were run on 9% polyacrylamide gels, and proteins were detected by staining with Coomassie Brilliant Blue R250.
In Vitro Binding Assay with Yeast Lysate-Aliquots of glutathione-Sepharose 4B (10 l of packed beads/reaction) were washed four times with 1 ml of RX buffer (20 mM HEPES-KOH, pH 6.8, 150 mM potassium acetate, 2 mM magnesium acetate, 0.5% (w/v) Triton X-100, 0.1% (w/v) casamino acids, 1 g/ml leupeptin, 1 g/ml pepstatin A, 1 g/ml aprotinin, 2.5 g/ml antipain, 0.21 mg/ml NaF, 0.1 mg/ml Pefabloc SC) and resuspended in 0.5 ml of RX buffer. 3 g of GST or GST-Pex20p was added to the beads, and reactions were tumbled end-over-end at room temperature for 80 min. Beads were collected by centrifugation and washed three times with 0.5 ml of RX buffer. YPBO-grown cells of the PEX8-HA c strain (Table I) were collected by centrifugation, washed three times with water, and washed once with RX buffer. Cells were disrupted in RX buffer by agitation with glass beads. The lysate was clarified by centrifugation at 20,000 ϫ g in a microcentrifuge for 45 min at 4°C. 120 l of the cleared supernatant (10 mg of protein/ml) or of RX buffer alone was tumbled end-over-end in batch with glutathione-Sepharose-bound GST or GST-Pex20p for 2 h at 4°C. Beads were washed twice with 0.8 ml of RX buffer, applied to spin filters, and washed four times with 0.5 ml of RX buffer. Bound proteins were eluted in 40 l of SDS-PAGE sample buffer according to the manufacturer's

Function of Pex8p and Pex20p in Thiolase Import
instructions. Proteins in equal fractions of each sample were separated on 10% polyacrylamide gels, transferred to nitrocellulose, and subjected to immunoblotting. Epitope Tagging of Pex8p-A modified PEX8 gene encoding Pex8p tagged at its carboxyl terminus was made by inserting a DNA fragment encoding two copies of the influenza virus hemagglutinin (HA) epitope in frame and downstream of the PEX8 ORF. A fragment containing the PEX8 ORF flanked by 2.0 and 1.2 kilobase pairs of genomic DNA at its 5Ј-and 3Ј-ends, respectively, was excised from p50LD (13) with HindIII and ligated into the HindIII site of pGEM7Zf(ϩ) to make the plasmid pG7PEX8. Next, a BglII site was introduced immediately upstream of the stop codon of PEX8 using sequential PCR. Essentially, two regions of the PEX8 gene flanking the proposed BglII site were amplified from pG7PEX8 using oligonucleotides RP and 761 for the 5Ј region and oligonucleotides UP and 760 for the 3Ј region (Table III). The two products were used as templates for a second round of PCR with oligonucleotides UP and RP. Since oligonucleotides 760 and 761 have BglII recognition sequences and complementary 5Ј-ends, the second round of PCR generated a single fragment containing PEX8 with a BglII site immediately upstream of the stop codon. This product was digested with HindIII and ligated into pGEM7Zf(ϩ) to generate the plasmid pG7-PEX8B. Next, a fragment with BamHI/BglII termini encoding the peptide DPLAMYPYDVPDYAAMYPYDVPDYAAMGKGE, which contains two repeats of the HA epitope (underlined residues) (18), was ligated in-frame into the BglII site of pG7PEX8B to generate plasmid pG7PEX8-HA c . The region of pG7PEX8-HA c amplified by PCR was confirmed to be correct by sequencing.
A fragment containing the modified PEX8 gene was excised from pG7PEX8-HA c with HindIII and integrated into the genome of the pex8 -1 mutant strain (Table I) by homologous recombination to replace the pex8 -1 allele of PEX8. The pex8 -1 allele has a substitution mutation in the PEX8 gene (A1924T), creating a premature stop codon 2 and cannot grow on oleic acid medium (13). Transformants were selected for reestablished growth on oleic acid medium (YNO agar) and characterized by Southern blotting and electron microscopy. One strain, PEX8-HA c , having the correct genotype and wild-type morphology, was chosen for further study.
To make a PEX8-HA c expression plasmid, PEX8-HA c was excised from pG7PEX8-HA c with HindIII and ligated into the HindIII site of the E. coli/Y. lipolytica shuttle vector pINA445 to make the plasmid pPEX8-HA c encoding Pex8p-HA c . pINA445 contains the Y. lipolytica LEU2 gene for positive selection of yeast transformants and the Y. lipolytica ARS68 gene for autonomous plasmid replication in Y. lipolytica cells (19).
Coimmunoprecipitation-Coimmunoprecipitation was performed by immunoaffinity chromatography using protein A-Sepharose CL-4B (Sigma). All steps were performed at 4°C unless otherwise specified. YPBO-grown PEX8-HA c cells were collected by centrifugation, washed three times with water, and washed once with RX buffer. Cells were resuspended in RX buffer and lysed by disruption with glass beads. The total cell lysate was clarified by centrifugation at 100,000 ϫ g in a TLA120.2 rotor (Beckman, Palo Alto, CA) for 15 min, diluted to a protein concentration of 10 mg/ml with RX buffer, and divided into 500-l aliquots. 4.5 l of immune serum or 18 l of preimmune serum (volumes of serum containing equal numbers of IgG molecules) was added to each of two aliquots, and reactions were tumbled end-over-end for 30 min. Protein A-Sepharose, preblocked in RX buffer containing 8% bovine serum albumin for 1 h, was pelleted by centrifugation and resuspended in RX buffer to a final concentration of 20% (v/v). 50 l of the protein A-Sepharose suspension was added to each reaction, and reactions were tumbled end-over-end for 40 min. Antibody complexes were pelleted by centrifugation and resuspended in 0.5 ml of RX buffer. Samples were then applied to spin filters and washed three times with RX buffer, and antibody complexes were eluted in SDS-PAGE sample buffer at room temperature. Proteins in 20% of each eluate, along with 10 g of protein of the cell lysate, were separated by SDS-PAGE, transferred to nitrocellulose, and immunoblotted.

CAGGAAACAGCATGAC
Function of Pex8p and Pex20p in Thiolase Import as homogenization buffer. Essentially, a postnuclear supernatant isolated from YPBO-grown cells was fractionated by differential centrifugation at 20,000 ϫ g for 20 min into a pellet (20KgP) enriched for peroxisomes and mitochondria and a supernatant (20KgS) enriched for cytosol and high-speed pelletable organelles. The 20KgS fraction was further subfractionated by centrifugation at 200,000 ϫ g for 1 h to yield a pellet (200KgP) enriched for high-speed pelletable organelles and a supernatant (200KgS) highly enriched for cytosol. Peroxisomes were purified from the 20KgP fraction by isopycnic centrifugation on a discontinuous sucrose gradient (21). Subfractionation of Peroxisomes and Protease Protection Analysis-To subfractionate peroxisomes, 1.6 ml of ice-cold Ti8 buffer (10 mM Tris-HCl, pH 8.0, 5 mM EDTA containing 1 g/ml leupeptin, 1 g/ml pepstatin A, 1 g/ml aprotinin, 2.5 g/ml antipain, 0.21 mg/ml NaF, 0.1 mg/ml Pefabloc SC, 0.2 mM benzamidine hydrochloride) was added to 160 l of 20KgP fraction containing 240 g of protein. Samples were incubated on ice for 30 min with occasional gentle agitation to disrupt peroxisomes. Half of each disrupted 20KgP (20KgP-D) was separated into a supernatant (Ti8S) enriched for soluble proteins and a pellet (Ti8P) enriched for membranes by centrifugation at 200,000 ϫ g at 4°C for 1 h. Proteins from equal fractions of the 20KgP-D, Ti8S, and Ti8P were separated by SDS-PAGE, transferred to nitrocellulose, and subjected to immunoblotting.
Protease protection analysis was performed on 20KgP fractions isolated in the absence of protease inhibitors as described (22). 0 -32 g of trypsin was combined with 60 g of protein of the 20KgP fraction in the absence or presence of 0.1% (w/v) Triton X-100 in a final volume of 160 l. Reactions were incubated on ice for 20 min and terminated by precipitation with trichloroacetic acid. Proteins from equal fractions of each reaction were separated by SDS-PAGE, transferred to nitrocellulose, and subjected to immunoblotting.
Immunofluorescence Microscopy-Strains were grown at 30°C in YEPD overnight, transferred to YPBO, and grown for 9 h. Cells were fixed for 30 min at room temperature and processed for immunofluorescence microscopy as described (23), except that spheroplasts were prepared in 100 mM potassium phosphate, pH 7.5, containing 1.2 M sorbitol, 40 g/ml Zymolyase-100T (ICN, Aurora, OH), 0.4% (v/v) 2-mercaptoethanol for 30 min at 30°C with gentle agitation. Images were captured using a Spot Cam digital fluorescence camera (Spot Diagnostic Instruments, Sterling Heights, MI).
Antibodies-Guinea pig and rabbit antibodies to Y. lipolytica Pex20p (17), rabbit antibodies to Y. lipolytica peroxisomal isocitrate lyase and to thiolase, guinea pig antibodies to Y. lipolytica Pex2p (24), and rabbit antibodies to Y. lipolytica Kar2p (25) have been described. Mouse monoclonal antibody 12CA5, which recognizes 9 amino acid residues of the influenza virus HA protein, was purchased from the Berkeley Antibody Company (Richmond, CA). To raise antibodies to Y. lipolytica Pex6p, a DNA fragment with EcoRI/HindIII termini encoding the first 452 amino acid residues of Pex6p was amplified from Y. lipolytica genomic DNA by PCR with oligonucleotides 424 and 425 (Table III). This fragment was ligated into the corresponding sites of the vector pMAL-c2 in frame and downstream of the malE ORF encoding MBP. Antibodies to the fusion protein were raised in rabbit and guinea pig as described previously (26,27). Rabbit antibodies to Pex20p and guinea pig antibodies to Pex2p were preabsorbed with permeabilized spheroplasts of the respective disruption strains prior to use in immunofluorescence microscopy.
Miscellaneous-Oligonucleotides were synthesized on an Oligo 1000M DNA Synthesizer (Beckman). Sequencing was performed with an ABI Prism 310 Genetic Analyzer (PE Applied Biosystems, Foster City, CA). DNA was amplified using Ready-To-Go PCR beads (Amersham Pharmacia Biotech). Preparation of yeast lysates by disruption with glass beads, growth of E. coli, and manipulations of DNA were performed as described (28). Protein concentrations were determined using a commercially available kit (Bio-Rad). SDS-PAGE was performed essentially as described (29). Proteins were transferred to nitrocellulose for immunoblotting using a wet transfer system, and antigen-antibody complexes were detected by enhanced chemiluminescence (Amersham Pharmacia Biotech).

Detection of Y. lipolytica Peroxin Interactions with the Yeast
Two-hybrid System-A limited yeast two-hybrid screen (30) was performed to identify physical interactions between Y. lipolytica peroxins Pex1p, Pex2p, Pex5p, Pex6p, Pex8p, Pex9p, Pex16p, and Pex20p. Others have used this methodology successfully to detect interactions between peroxins (for examples, see Refs. 10 and 31-33). Chimeric genes were generated by ligating Y. lipolytica PEX genes in frame and downstream of sequences encoding one of the two functional domains (AD or DB) of the GAL4 transcriptional activator. All possible combinations of plasmid pairs encoding AD and DB fusion proteins were transformed into the S. cerevisiae strain SFY526, and ␤-galactosidase filter detection assays were performed initially. Physical interaction between peroxins was expected to bring the two domains of the GAL4 transcriptional activator together and result in activation of transcription of the integrated lacZ construct and detectable ␤-galactosidase activity in the form of blue coloration of the yeast colony in the presence of the substrate 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (Xgal). Potential interactions were detected between Pex8p and Pex5p and between Pex1p and Pex6p using the filter detection assay (data not shown).
To confirm the results of the filter assay, liquid ␤-galactosidase assays were performed. Cell lysates of strains synthesizing both Pex1p and Pex6p fusion proteins showed much greater ␤-galactosidase activity than lysates of control strains synthesizing either one or the other of the fusion proteins (Fig. 1A), demonstrating that Pex1p and Pex6p interact physically. Ly- sates of strains synthesizing both Pex8p and Pex5p fusion proteins also showed very high ␤-galactosidase activity, but so did the lysates of control strains synthesizing one or the other of the fusion proteins (Fig. 1B). Using the Smith-Satterthwaite test for statistical significance (34), it was determined that the levels of ␤-galactosidase activity from strains synthesizing both DB-Pex8p and AD-Pex5p or both DB-Pex5p and AD-Pex8p were significantly higher than those of the corresponding control strains at confidence levels of 95 and 90%, respectively, indicating that Pex8p and Pex5p interact physically.
Lysates of strains transformed with constructs encoding DB-Pex8p or DB-Pex20p showed high levels of ␤-galactosidase activity, making it difficult to identify an interaction between Pex20p and Pex8p. Therefore, a plasmid encoding DB-Pex20p-(1-76), a fusion between the DB domain and the amino-terminal 76 amino acid residues of Pex20p, was generated. DB-Pex20p does not have intrinsic transcription activation activity. Lysates of strains synthesizing both AD-Pex8p and DB-Pex20p-(1-76) showed significantly higher ␤-galactosidase activity than did lysates of control strains synthesizing one or the other of the fusion proteins (Fig. 2, compare row 1 with rows 5 and 6), demonstrating that Pex8p and Pex20p interact physically. To characterize this interaction further, fusions to smaller domains of Pex8p were assayed. DB-Pex20p-(1-76) still interacted with AD-Pex8p-(12-644), containing Pex8p truncated at both its amino and carboxyl termini, but did not interact with AD domain fusions to more extensive amino-or COOH-terminal truncations of Pex8p (Fig. 2, compare row 2 to  rows 3 and 4). Therefore, Pex8p physically interacts with the amino terminus of Pex20p, and the 11 amino-terminal and 27 COOH-terminal amino acid residues of Pex8p are dispensable for this interaction.
The Interaction between Pex8p and Pex20p Is Direct and Autonomous-Some interactions between peroxins that have been detected using the yeast two-hybrid system have been shown to be indirect (10,33). Therefore, it is possible that an endogenous S. cerevisiae protein may have bridged the observed interaction between Pex8p and Pex20p. To determine whether the interaction between Pex8p and Pex20p is direct and autonomous, an in vitro binding assay was performed. The ORFs encoding Pex8p and Pex20p were fused to the 3Ј-ends of DNA sequences encoding MBP and GST, respectively. The chimeric genes were expressed in E. coli, and the fusion proteins were purified. GST and GST-Pex20p were immobilized on glutathione-Sepharose beads and tested for their ability to bind MBP or MBP-Pex8p. Proteins in bound and unbound fractions were separated by SDS-PAGE and visualized by staining with Coomassie Brilliant Blue (Fig. 3). MBP-Pex8p interacted with GST-Pex20p, while GST or MBP alone did not interact with MBP-Pex8p or GST-Pex20p, respectively. We conclude that the interaction between Pex8p and Pex20p is direct and autonomous.
The Pex8p-Pex20p Interaction Is Specific-An in vitro binding experiment was performed to determine whether GST-Pex20p could interact specifically with Pex8p tagged with the hemagglutinin epitope, Pex8p-HA c , in a yeast lysate. GST and GST-Pex20p were immobilized on glutathione-Sepharose beads. A total cell lysate was prepared from PEX8-HA c , a strain synthesizing Pex8p-HA c , and incubated with immobilized GST or GST-Pex20p. The proteins bound to each column and the proteins of the total cell lysate were separated by SDS-PAGE, transferred to nitrocellulose, and analyzed by immunoblotting with various antibodies (Fig. 4). Pex8p-HA c interacted specifically with GST-Pex20p, while the peroxins Pex1p and Pex6p did not. Interestingly, peroxisomal thiolase, which has been shown to interact with Pex20p (17), did not bind GST-Pex20p (Fig. 4). Since an interaction between thiolase and Pex20p also could not be detected in the yeast two-hybrid system, 2 it may be that peptides fused to the amino terminus of Pex20p inhibit its interaction with thiolase.
Pex8p-HA c and Pex20p Coimmunoprecipitate-To determine whether Pex20p was capable of interacting with Pex8p-HA c in vivo, Pex20p was immunoprecipitated under native conditions from a total lysate of cells grown in YPBO medium, and coimmunoprecipitating proteins were analyzed by SDS-PAGE and immunoblotting with various antibodies (Fig. 5). Although all proteins tested were present in the total cell lysate, Pex8p-HA c coimmunoprecipitated specifically with Pex20p but not with peroxisomal isocitrate lyase or the peroxin Pex6p. Therefore, Pex8p-HA c interacts with Pex20p in vivo. As previously reported (17), the precursor form of thiolase also specifically coimmunoprecipitated with Pex20p.
Pex20p Is Not Required for Targeting Pex8p-HA c to Peroxisomes-Since Pex20p binds thiolase in the cytosol and is nec- essary for its targeting to peroxisomes (17) and because Pex20p interacts with Pex8p-HA c , we investigated whether Pex20p might also bind Pex8p in the cytosol and act in targeting it to peroxisomes. To test this hypothesis, Pex8p-HA c was localized by subcellular fractionation of cells of the wild-type strain E122, the PEX20 disruption strain pex20KO, and the original PEX20 mutant strain pex20 -1 (Table I), each transformed with the expression plasmid pPEX8-HA c . For each strain, a postnuclear supernatant fraction was separated into 20KgS enriched for cytosol and high-speed pelletable organelles and 20KgP enriched for peroxisomes and mitochondria (17). The 20KgS fraction was further divided into 200KgS highly enriched for cytosol and 200KgP enriched for high-speed pelletable organelles, including high-speed pelletable peroxisomes (17,20). Proteins in equal portions of each fraction were separated by SDS-PAGE and analyzed by immunoblotting (Fig. 6). Pex8p-HA c was present in the 20KgP fraction enriched for peroxisomes and in the 20KgS and 200KgP fractions enriched for high-speed pelletable peroxisomes but not in the highly enriched cytosol fraction (200KgS) of cells of all strains. For all strains, Pex8p-HA c cofractionated with peroxisomal marker proteins in a sucrose density gradient (data not presented). Together, these results suggest that Pex20p is not required for the targeting of Pex8p to peroxisomes. As expected, thiolase was preferentially localized to peroxisome-enriched fractions of cells of the wild-type strain and to cytosol-enriched fractions of cells of the PEX20 mutant strains.
Pex20p and a Small Amount of Thiolase Associate with Peroxisomes in Cells of a PEX8 Disruption Strain-Interaction between Pex20p and Pex8p may occur at the level of the peroxisomal membrane during the targeting or import of thiolase. Consistent with this scenario, only a small fraction of thiolase is peroxisomal in a PEX8 disruption strain (13). To further elucidate the role of Pex8p in the import of thiolase, the localization of Pex20p and the subperoxisomal localization of peroxisome-associated thiolase were determined in cells of the PEX8 disruption strain, pex8-KA (13) ( Table I). Double labeling immunofluorescence microscopy of YPBO-grown cells demonstrated that as expected, Pex20p showed a diffuse localization characteristic of a cytosolic protein (17), while the peroxisomal membrane protein Pex2p and the peroxisomal matrix protein thiolase showed a punctate pattern of localization characteristic of peroxisomes in cells of the wild-type strain E122 (Fig. 7). Interestingly, in cells of the pex8-KA strain, Pex20p colocalized with Pex2p to punctate structures characteristic of peroxisomes, while as expected, thiolase had a diffuse localization characteristic of a cytosolic location. These results suggest that while Pex20p is predominantly cytosolic in wild-type cells, as has been reported previously (17), a significant fraction of Pex20p associates with peroxisomes in cells lacking Pex8p.
Pex20p and thiolase were also localized by subcellular fractionation of cells of the wild-type E122 and pex8-KA strains. In E122 cells, Pex20p was found largely in the 20KgS fraction enriched for cytosol and high-speed pelletable organelles, while thiolase was primarily in the 20KgP fraction enriched for peroxisomes, as expected (Fig. 8A). In pex8-KA cells, by contrast, Pex20p was found primarily in the 20KgP fraction and thiolase was found primarily in the 20KgS fraction, although a small amount of thiolase fractionated to the 20KgP, since some thiolase is peroxisomal in pex8-KA cells (13). The results of subcel- lular fractionation are consistent with those of immunofluorescence analysis.
The suborganellar locations of Pex20p and thiolase in wildtype cells and in pex8-KA cells were now compared. The 20KgP fraction of wild-type and pex8-KA cells was disrupted by incubation in dilute Tris buffer. The 20KgP-D was then separated by centrifugation into a pellet fraction (Ti8P) enriched for membranes and a supernatant fraction (Ti8S) enriched for soluble proteins. Proteins in equal portions of each fraction were sep-arated by SDS-PAGE and analyzed by immunoblotting (Fig.  8B). Pex20p localized primarily to the Ti8P fraction from both wild-type and pex8-KA cells along with peroxisomal membrane protein Pex2p, whereas thiolase was primarily found in the Ti8S fraction from both strains. These data indicate that in both wild-type and pex8-KA cells, Pex20p associates with the peroxisomal membrane. In addition, since thiolase was present in the soluble Ti8S fraction derived from pex8-KA cells, either thiolase is in the matrix of peroxisomes in the absence of Pex8p or membrane-associated thiolase is released during the extraction incubation.
In wild-type cells, Pex20p is localized to the outer surface of peroxisomes, as demonstrated by its susceptibility in vitro to the action of external proteases (17). To determine whether in the absence of Pex8p, peroxisome-associated Pex20p and thiolase are protected from the action of proteases by the peroxisomal membrane, protease protection analysis was performed on a 20KgP fraction isolated from cells of the pex8-KA strain. Equal portions of the 20KgP fraction were incubated with increasing amounts of trypsin in the absence or presence of the nonionic detergent, Triton X-100. The proteins in equal portions of each digest were separated by SDS-PAGE and analyzed by immunoblotting (Fig. 8C). Both Pex20p and thiolase showed greater resistance to trypsin in the absence of detergent than in its presence, suggesting that they are afforded protection from the action of proteases by the peroxisomal membrane in pex8-KA cells.

DISCUSSION
Pex20p is a peroxin that binds thiolase in the cytosol and is necessary for its oligomerization and targeting to peroxisomes (17). Pex8p is an intraperoxisomal peroxin required for the import of a number of peroxisomal proteins, including thiolase (11)(12)(13)(14). Using the yeast two-hybrid system, we identified an interaction between Pex8p and the amino terminus of Pex20p. The Pex8p-Pex20p interaction was confirmed by the isolation of a complex containing Pex8p and Pex20p from yeast lysates by coimmunoprecipitation and by in vitro binding studies with recombinant proteins showing that this interaction is specific, direct, and autonomous. Although Pex20p and Pex8p are in different cellular compartments, two readily possible scenarios for their interaction can be proposed. One scenario is that Pex20p binds Pex8p in the cytosol and is required for the oligomerization and/or peroxisomal targeting of Pex8p. The second scenario is that Pex8p interacts with Pex20p at the peroxisomal membrane at a step in the import of thiolase into peroxisomes. We localized Pex8p and Pex20p in cells of PEX20 and PEX8 disruption strains, respectively, to elucidate the mechanism of interaction of these two proteins.
Pex20p-dependent Targeting and/or Oligomerization of Pex8p-The possibility of a role for Pex20p in the targeting and/or oligomerization of Pex8p is consistent with evidence that suggests that Pex8p of the yeast Hansenula polymorpha may be an oligomer (11). It is also compatible with the fact that Pex8p is intraperoxisomal and in most yeasts contains PTS1 (12) or both PTS1 and PTS2 motifs (11,14). Since Pex8p shares PTS with peroxisomal matrix enzymes, it is conceivable that it may also be targeted to peroxisomes by the same peroxins involved in targeting these enzymes. Although such an explanation is plausible, two lines of evidence suggest that Pex20p is not necessary for the targeting and/or oligomerization of Pex8p. First, assuming that mislocalized or nonoligomerized Pex8p molecules are nonfunctional and that Pex20p is necessary for the oligomerization and/or the targeting of Pex8p, then a PEX20 disruption strain should present a phenotype that is as severe as or more severe than that of a PEX8 disruption strain. However, this is not the case, since several matrix proteins are FIG. 7. Pex20p is peroxisomal in the PEX8 disruption strain, pex8-KA. The PEX8 disruption strain pex8-KA (top panels) and the wild-type strain E122 (bottom panels) were grown overnight in YEPD medium and transferred to YPBO medium and grown for an additional 9 h. Cells of each strain were processed for immunofluorescence microscopy. Cells were double-labeled with guinea pig anti-Pex2p antibodies (left panels) and rabbit anti-Pex20p antibodies (middle panels) or labeled with guinea pig anti-thiolase antibodies (right panels). Guinea pig primary antibodies were detected with rhodamine-conjugated donkey anti-guinea pig IgG secondary antibodies. Rabbit primary antibodies were detected with fluorescein-conjugated donkey anti-rabbit IgG secondary antibodies.
FIG. 8. Peroxisomal Pex20p from the PEX8 disruption strain is membrane-associated and protected from the action of externally added proteases. A, immunoblot analysis of subcellular fractions of cells of the wild-type strain E122 and of the PEX8 disruption strain pex8-KA. Proteins in equal portions of each subcellular fraction were separated by SDS-PAGE and subjected to immunoblotting with antibodies to Pex20p and thiolase. B, organelles in the 20KgP fraction were disrupted by incubation with dilute Tris, and membranes and soluble proteins were separated by differential centrifugation. Proteins in equal portions of the 20KgP-D, the membrane pellet (Ti8P), and supernatant (Ti8S) containing soluble proteins were processed for immunoblotting as described for A. C, protease protection analysis. Organelles in the 20KgP fraction from cells of the pex8-KA strain were incubated with 0, 2, 4, 8, 16, or 32 g of trypsin in the absence (Ϫ) or presence (ϩ) of 0.1% (v/v) Triton X-100 for 20 min on ice. Samples were subjected to immunoblotting with antibodies to thiolase and Pex20p. mislocalized in PEX8 disruption strains (11)(12)(13)(14), while only thiolase is mislocalized in a PEX20 disruption strain (17). Second, an epitope-tagged version of Pex8p was localized to subcellular fractions from cells of pex20 mutant strains enriched for peroxisomes. Although these data strongly suggest that Pex20p is not required for the targeting and/or oligomerization of Pex8p, we cannot rule out the possibility that redundant systems exist for these functions, with only one requiring Pex20p.
A Role for Pex8p in Pex20p-dependent Import of Thiolase-The alternative scenario that Pex8p and Pex20p interact at the peroxisomal membrane in a step in the targeting or import of thiolase is consistent with previously published findings. Pex8p has been shown to be associated with the peroxisomal membrane in three different yeasts, including Y. lipolytica (12)(13)(14). It has also been shown that in cells of a Y. lipolytica PEX8 disruption strain, thiolase is primarily cytosolic, although a small amount of thiolase associates with peroxisomes (13). The localization of Pex20p and thiolase in cells of a PEX8 disruption strain also points to a direct role for Pex8p in the Pex20p-dependent import of thiolase. While a small amount of total Pex20p has previously been shown to be associated with peroxisomes in wild-type cells (17), we find that a large fraction of Pex20p associates with peroxisomal membranes in cells devoid of Pex8p relative to wild-type cells. These data point to a role for the interaction between Pex8p and Pex20p in the import of thiolase at a stage following the docking of Pex20p-thiolase complexes at the peroxisomal membrane. The exact role of Pex8p in thiolase import is unknown; however, since Pex20p has been shown to interact with Pex8p in the absence thiolase (or any other protein), it is tempting to speculate that Pex8p has a role either in dissociating thiolase from Pex20p or in recycling Pex20p after dissociation. Future experiments will be aimed at studying the interactions between Pex8p and Pex20p bound to thiolase.
Interestingly, we find that the peroxisome-associated Pex20p and thiolase in cells of the PEX8 disruption strain are protected from the action of external proteases. These data suggest that in the absence of Pex8p, both Pex20p and thiolase are protected by membranes. This, together with the fact that Pex20p directly interacts with Pex8p, an intraperoxisomal peroxin, suggests that Pex20p could enter peroxisomes with its cargo. This event may not be readily detectable in wild-type cells (17), either because the steady-state level of the population of intraperoxisomal Pex20p is very low or because Pex20p may exit the peroxisome through a putative translocation channel during fractionation of wild-type cells.
A Role for Pex8p in Pex5p-dependent Import of PTS1-containing Proteins-As mentioned previously, thiolase is mislocalized in cells lacking Pex20p, while several matrix proteins, including PTS1-containing proteins, are mislocalized in cells lacking Pex8p (13). One possible reason for this more extensive mislocalization of matrix proteins in pex8 mutant strains is that in these strains, Pex20p becomes trapped in the peroxisomal membrane, thereby clogging the translocation apparatus. Since the PTS1-and PTS2-dependent import pathways are apparently convergent (9,10), this obstruction may prevent PTS1-containing proteins from being imported into the perox-isome. Alternatively, Pex8p may be directly involved in both Pex5p-dependent and Pex20p-dependent import. Our identification of an interaction between Pex8p and Pex5p in the yeast two-hybrid system is consistent with this possibility. In addition, Pex8p and Pex5p of S. cerevisiae have recently been shown to interact, even in the absence of the PTS1 of Pex8p (14). Although Pex8p and Pex5p have not been shown to interact directly, Pex8p and Pex5p may interact at their amino termini, as we have shown that Pex8p and Pex20p interact at their amino termini, and Pex20p and Pex5p show sequence similarity at their amino termini (17).