Peroxisomal Targeting, Import, and Assembly of Alcohol Oxidase in Pichia pastoris

Yeast strains used in this study are listed in Table I. Shake-flask cultures were incubated for 10–15 h at 30° (P. pastoris strains) or 37°C (H. polymorpha strains) in selective minimal YND or YNM medium (0.17% [wt/ vol] yeast nitrogen base without amino acids {Difco Laboratories Inc., Detroit, MI} supplemented with 0.5% [wt/vol] glucose [dextrose] or 0.5% [vol/vol] methanol). For growth of auxotrophic strains, amino acids were added to a final concentration of 50 μg/ml. Transformations of P. pastoris (Becker and Guarente, 1991) or H. polymorpha (Faber et al., 1994) were performed by electrotransformation. Cultivation of Escherichia coli strain DH5α and standard recombinant DNA techniques were performed essentially as described (Sambrook et al., 1989).

Plasmids encoding chimeric proteins composed of selected AOX amino acid sequences fused to a modified E. coli β-lactamase protein were constructed. The modified β-lactamase was composed of amino acid residues H²⁴–W²⁸⁶ (Sutcliffe, 1978) preceded by the amino acids methione-serine-glycine (MSG) as previously described (Waterham et al., 1994, 1997). DNA sequences encoding the COOH termini of wild-type AOX (AOX-WT), AOX–RSC, and AOX–HA (see below) were ligated in reading frame to the 3′ end of the bla gene using either the StuI site located at 1,924 bp or the AgeI site (made blunt ended with the Klenow fragment of DNA polymerase I) located at 1,972-bp downstream of the AOX1 start codon (Koutz et al., 1989). The proper DNA sequence at each fusion junction was verified by DNA sequencing. The primary sequence of each fusion protein is shown in Table II. All β-lactamase fusion proteins were expressed under transcriptional control of the constitutive GAP promoter (P_GAP) from the P. pastoris GAP gene (Waterham et al., 1997) in vector pHWO10K. This vector was created from the HIS4-based vector pHWO10 by replacing the ampicillin-resistance gene with the kanamycine-resistance gene. The vectors were integrated into the genomic HIS4 locus of P. pastoris strains GS115 or pex5Δ (Table I) after linearization with SalI, a unique site in the HIS4 gene of each vector.

The P. pastoris AOX1 gene was amplified from plasmid pPG5.4 (Cregg et al., 1989) by the PCR using a forward primer, composed of the first 18 bp of the AOX1 open reading frame preceded by an EcoRI site, and a reverse primer, composed of 19 bp located 247–266-bp downstream of the AOX1 stop codon which includes a genomic HindIII site (Koutz et al., 1989). After digestion with EcoRI and HindIII, the AOX1 gene was first subcloned in EcoRI–HindIII–digested pBS-SK (Stratagene, La Jolla, CA), and subsequently as an EcoRI–ClaI fragment under transcriptional control of the P. pastoris AOX1 promoter (P_AOX1) into EcoRI- and ClaI-digested vector pHIL-AI (Invitrogen, San Diego, CA; Waterham et al., 1997). The resulting vector was named pHWO40. To modify sequences encoding the COOH terminus of AOX-WT, vector pHWO40 was digested with AgeI (located 1,972-bp downstream of the AOX1 start codon) and then the resulting termini were made blunt with Klenow fragment of DNA polymerase I and religated. This resulted in a frame shift changing the COOH-terminal residues of AOX-WT leucine-alanine-arginine-phenyl-alanine_COOH (LARF) to arginine-serine-cysteine_COOH (RSC). The resulting vector that expressed AOX–RSC was named pHWO41. A second vector encoding a COOH-terminal modified AOX was created by inserting two complementary oligonucleotides, which changed the four COOH-terminal residues from LARF to that for the human influenza virus epitope (HA) tag at the AgeI site of pHWO40. These residues are recognized by the mouse monoclonal antibody 12CA5 (Boehringer Mannheim Biochemicals, Indianapolis, IN). The resulting modified AOX protein is called AOX–HA (vector pHWO42). Finally, a vector was created that encodes a modified AOX (AOX-Δ22) in which the last 22 amino acids of the protein are deleted. This vector (pHWO43) was generated by inserting a blunt AgeI adaptor encoding an in-frame stop codon between the StuI and AgeI sites of AOX1 in vector pHWO40. The primary sequences of the COOH termini of the modified AOX proteins are shown in Table II. All constructs were verified by DNA sequencing. The constructs were integrated into the genomic HIS4 locus of P. pastoris strains MC100-3, GS115, and pex5Δ (Table I) by linearization at the unique EcoNI site in the HIS4 gene of each vector.

Interactions between the P. pastoris PTS1 receptor protein Pex5p and selected sequences from AOX were studied with the yeast two-hybrid system (Matchmaker; Clontech Laboratories, Inc., Palo Alto, CA). The P. pastoris PEX5 open reading frame was released from pSP72 (a gift from S. Subramani, University of California at San Diego, San Diego, CA) by digestion at a SacII site located 42-bp downstream of the PEX5 start codon, treated with mung bean exonuclease, and then digested at a PstI site located downstream of the PEX5 stop codon. This fragment was ligated into SmaI–PstI–digested pGAD424 which resulted in the expression of Pex5p fused in reading frame with the GAL4 activation domain (pHWO51). The COOH termini of AOX-WT, AOX–RSC, and AOX– HA were released from vectors pHWO40, pHWO41, and pHWO42, respectively, by digestion with StuI (located 1,924-bp downstream of the AOX1 start codon) and PstI (located downstream of the AOX1 stop codon). These fragments were then ligated into pGBT9 that had been cut with XmaI, filled with Klenow, and then digested with PstI. The resulting vector expressed the GAL4–DNA binding domain protein fused in-frame to each of the AOX-derived peptides described above. The COOH-terminal sequences of the resulting constructs named pHWO48 through pHWO50 are shown in Table II. All constructs were verified by DNA sequencing. These constructs and control vectors were transformed by electrotransformation into the S. cerevisiae reporter strain SFY526 (Becker and Guarente, 1991). β-Galactosidase activity in each strain was determined qualitatively by the filter assay method and quantitatively by the cell-free activity assay method described in the Clontech technical manual.

AOX–HA was coexpressed with AOX-WT by inserting vector pHWO42 in GS115 (GS-HWO42), and expressed alone by insertion of the vector into the AOX1 and AOX2 deletion strain MC100-3 (MS-HWO42). Octamerization of AOX proteins was examined using a modified version of the velocity sedimentation method described by Goodman et al. (1984). Cells (50 OD₆₀₀ U) were lysed by vortexing for 15 min at 4°C with 1 vol of glass beads in 0.3 ml TENT buffer (10 mM Tris-HCl, pH 8, 5 mM EDTA, 50 mM NaCl, 1% Triton X-100) and then centrifuged for 10 min at maximum speed in a microcentrifuge (Eppendorf model 5415C; Brinkman Instruments, Westbury, NY).

Supernatant samples of 0.1 ml were separated through 5–30% (wt/vol) sucrose gradients (6 layers of 1.5 ml sucrose in TENT) by centrifugation for 5 h at 40,000 rpm and 2°C in a rotor (SW41Ti; Beckman Instruments, Palo Alto, CA). Fractions of 0.5 ml were removed from the top and analyzed for sucrose density, AOX, and CAT activities, and for AOX-WT or AOX–HA protein by immunoblotting using polyclonal antibodies against AOX or monoclonal antibodies against the HA epitope.

The H. polymorpha MOX gene was expressed under transcriptional control of its own promoter by integration of KasI-linearized vector pHWO56 into the HIS4 locus of the P. pastoris AOX1 and AOX2 deletion strain MC100-3 and by integration of SacI-linearized vector pHWO55 into the P_AOX1 locus of the P. pastoris pex2Δ aox1Δ aox2Δ strain MC200. The MC200 strain was generated by disruption of the PEX2 gene in MC100-3 using BamHI-digested pUZ12 (Waterham et al., 1996). The P. pastoris AOX1 gene was expressed under transcriptional control of its own promoter (pHWO57) and the H. polymorpha P_MOX (pHWO58) in a H. polymorpha moxΔ pex10-1 strain. This strain was constructed by disruption of the MOX gene in the H. polymorpha pex10-1 mutant strain using AlwNI– SalI–digested pHWO52.

P. pastoris cells (25 OD₆₀₀ U) were washed twice with water and fixed in 1 ml of 40 mM KPO₄ buffer, pH 6.5, with 3.7% (vol/vol) formaldehyde for 1 h at room temperature. After two washes with 1 ml MOPS⁺ (5 mM 3-[morpholino]propane-sulfonic acid, 10 mM Na₂SO₃, and 0.5 M KCl), fixed cells were converted into spheroplasts in 0.5 ml MOPS⁺ supplemented to 85 mM β-mercaptoethanol and 0.08 mg/ml Zymolyase-100T for 1 h at 30°C. The spheroplasts were washed twice with 1 ml MOPS⁺, incubated for 5 min at −20°C in 100% methanol, washed twice with 1 ml PBS⁺ (PBS, pH 7.2, 1 M sorbitol, and 1% bovine serum albumin), and finally suspended in 1 ml PBS⁺. Suspended cells (0.2 ml) were incubated for 2 h at room temperature with 0.5 μl of polyclonal antibodies raised against CAT or β-lactamase (5 Prime, 3 Prime, Boulder, CO). After three washes with 1 ml PBS⁺, cells were suspended in 0.2 ml PBS⁺ and incubated for 1 h at room temperature in the dark with 0.5 μl FITC-conjugated goat anti–rabbit antiserum (Boehringer Mannheim, Indianapolis, IN). After three washes with 1 ml PBS⁺, cells were suspended in 0.1 ml PBS⁺. For microscopic observation, small samples of the cell suspensions were mixed with equal volumes of 0.1 M n-propyl gallate in 90% glycerol. Cells were examined with a microscope equipped for indirect immunofluorescence (Leitz Laborlux S; Wild Leitz, Rockleigh, NJ) at 1,000× and photographed (T-MAX 400 film; Eastman Kodak Co., Rochester, NY).

Subcellular fractionation of P. pastoris and H. polymorpha cells was performed as described previously (Waterham et al., 1996) except that H. polymorpha cells were converted to spheroplasts in the presence of 3 M sorbitol and homogenized in the presence of 2 M sorbitol. Cell-free extracts were made using a glass bead method (Waterham et al., 1997). Peroxisomal AOX and CAT, mitochondrial cytochrome c oxidase, cytosolic glyceraldehyde-3-phosphate dehydrogenase, and β-lactamase activities were assayed as described previously (Waterham et al., 1996, 1997). Protein concentrations were determined with the bicinchoninic acid protein assay kit (Pierce Chemical Co., Rockford, IL) using bovine serum albumin as a standard. The transfer of proteins to nitrocellulose after SDS-PAGE electrophoresis was performed using the Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad Laboratories, Hercules, CA) as directed by the manufacturer. Immunoblotting experiments were performed using the Western Light Kit (Tropix Inc., Bedford, MA) with specific polyclonal antibodies against AOX and monoclonal antibody 12CA5 against the hemagglutinin epitope (Boehringer Mannheim). Immunoprecipitations were performed as described in Rehling et al. (1996) except that protein A–Sepharose (Pharmacia Biotech. Inc., Piscataway, NJ) was used instead of DynaBeads. Native (nonreducing nondenaturing discontinuous) gels were electrophoresed at 200 v for 2 h at 4°C through 5% polyacrylamide using a Mini-Protean II apparatus (BioRad Laboratories) as described in Ausubel et al. (1996).

Sequence data for the P. pastoris AOX1 and AOX2 genes, as published by Koutz et al. (1989), have been submitted to the databases and are available from GenBank/EMBL/DDBJ under accession numbers U96967 (AOX1) and U96968 (AOX2).

Our previous studies with P. pastoris pex mutants indicated that methanol-induced pex cells contain little or no AOX activity, although significant amounts of AOX protein are present in the cells (Liu et al., 1992, 1995; Waterham et al., 1996). The level of residual AOX activity in an individual pex mutant allele strongly correlated with the severity of the peroxisome-deficient phenotype of that strain. For example, in methanol-induced cells of pex2-1 and pex8-1 (two slightly leaky mutants generated by chemical mutagenesis), numerous small peroxisomal remnants were observed that retain the ability to import small amounts of peroxisomal enzymes (Liu et al., 1995; Waterham et al., 1996). In both of these mutants, small but significant amounts of active AOX were present and these appeared to be exclusively within the peroxisomal remnants as judged by subcellular fractionation studies (Table III). In contrast, methanol-induced cells of pex2 and pex8 deletion strains (pex2Δ and pex8Δ), which have a more severe peroxisome-deficient phenotype with few, if any, peroxisomal remnants and no measurable import ability, contained no active AOX, although AOX protein was present (Fig. 1). The inactive AOX protein was concentrated in cytoplasmic protein aggregates which sedimented with organelles during subcellular fractionation (Fig. 1; Liu et al., 1995; Waterham et al., 1996). These results indicate that active AOX exists only in the peroxisomal matrix and suggest that AOX precursors require the presence of functional peroxisomes to assemble into active octamers.

Unlike P. pastoris, AOX from H. polymorpha (MOX) efficiently assembles into an active octameric enzyme in the cytoplasm of H. polymorpha pex mutants (van der Klei et al., 1991). To determine whether this difference is a characteristic of the AOX proteins themselves or the cellular environment provided by their hosts, we constructed pex strains of each yeast species that expressed AOX/MOX from the heterologous species. For H. polymorpha, a vector containing the P. pastoris AOX1 gene was inserted into the genome of an H. polymorpha pex10 moxΔ strain. AOX activity could not be detected in the AOX-expressing pex10 moxΔ strain CW111 at either 30° or 37°C (Table IV). Immunoblot analysis performed on crude organelle pellet and cytosolic supernatant fractions from methanol-induced cells of this strain showed that AOX protein was synthesized and located primarily in pellet fractions (Fig. 2, group 2 and 3 lanes). The pellet location was most likely a consequence of AOX protein aggregation as seen in P. pastoris pex mutants (Fig. 1). Identical results were obtained when AOX1 was expressed under control of the H. polymorpha MOX promoter (data not shown). Thus, in contrast to endogenous H. polymorpha MOX, P. pastoris AOX is not assembled into active protein in the cytosol of an H. polymorpha pex mutant.

The converse result was obtained when H. polymorpha MOX was expressed in a P. pastoris pex mutant. As expected, MOX was able to complement for methanol growth in a P. pastoris strain on MC100-3 that is deleted for both AOX genes (aox1Δ aox2Δ). When the P_AOX-MOX vector was then inserted into the genome of a P. pastoris pex2Δ aox1Δ aox2Δ strain (MC200), methanol-induced cells of the transformed strain contained active MOX (Table IV). Upon subcellular fractionation, both MOX activity and protein were found in the cytosolic supernatant (Fig. 2, 1 lanes). The same result was obtained when MOX was expressed in a P. pastoris pex5Δ aox1Δ aox2Δ strain (data not shown). Since H. polymorpha MOX assembles into an active enzyme in the cytoplasm of either H. polymorpha or P. pastoris pex strains, the inability of P. pastoris AOX to properly assemble in these yeasts must be a characteristic of the protein itself and not of the environment provided by the P. pastoris cytoplasm.

To further investigate the targeting and import of AOX, it was first necessary to define the AOX PTS. The three COOH-terminal amino acids of AOX, ARF, are similar in sequence to the PTS1 motif and, therefore, were a good candidate for the AOX PTS. Preliminary evidence in support of this was provided by Hansen et al. (1992) who showed that the last four amino acids of H. polymorpha MOX (LARF) are capable of targeting a nonperoxisomal protein to peroxisomes. To examine targeting of P. pastoris AOX, we first expressed a protein composed of the bacterial reporter enzyme β-lactamase fused to the four COOH-terminal amino acids of P. pastoris AOX (LARF) and determined whether this fusion protein (β-lac–LARF) was targeted to P. pastoris peroxisomes. Methanol-grown cells expressing β-lac–LARF were fractionated into an organelle pellet, consisting mainly of peroxisomes and mitochondria, and a cytosolic supernatant (Fig. 3,B). Biochemical analysis of these fractions revealed that a significant portion of β-lactamase activity colocalized with CAT activity in the organelle pellet fraction, suggesting that β-lac– LARF was targeted to peroxisomes. (Typically, a significant amount of peroxisomal matrix protein is also found in the supernatant fraction due to breakage of these fragile organelles.) Peroxisomal targeting of β-lac–LARF was confirmed by indirect immunofluorescence using β-lactamase– and CAT-specific antibodies (Fig. 3,B) and by further fractionation of the organelle pellet through sucrose density gradients (Fig. 4). In these gradients, β-lac–LARF was present at the same density as CAT and not with mitochondrial cytochrome c oxidase. As a control, β-lactamase without LARF expressed in P. pastoris fractionated to the cytosolic supernatant along with the cytoplasmic marker enzyme glyceraldehyde-3-phosphate dehydrogenase and appeared to be cytoplasmic in immunofluorescence assays (Fig. 3 A). We concluded a PTS is located within the LARF sequence.

We next investigated the necessity of LARF for targeting and import of AOX. In one set of experiments, the 22 COOH-terminal amino acids of AOX were fused to the COOH terminus of β-lactamase (β-lac–AOX). As expected, this fusion protein was efficiently targeted to peroxisomes of P. pastoris (Fig. 3,D). We then deleted the four COOH-terminal residues from this construct and replaced them with either the non-PTS amino acid sequence RSC (β-lac–AOX–RSC) or the hemagglutinin–epitope-tag sequence (β-lac–AOX–HA). When expressed in methanol-grown P. pastoris cells, both of these chimeric proteins were localized to the cytoplasm as judged by indirect immunofluorescence and subcellular fractionation assays (Fig. 3, E and F).

The necessity of LARF for AOX targeting was further investigated by expressing mutant versions of AOX in which the tetrapeptide sequence had been removed and replaced with either RSC (AOX–RSC) or the HA tag (AOX–HA). Each mutant protein was expressed under control of the P. pastoris AOX1 promoter in a P. pastoris aox1Δ aox2Δ strain. Both the AOX–RSC and AOX–HA constructs partially complemented the strain for growth on methanol (18 and 22 h/generation, respectively, versus 3.5 h/generation for wild type), and both contained low but significant levels of AOX activity (16 and 3% of wild type; Table V). Subcellular fractionation studies of the AOX–RSC and AOX–HA expression strains showed that the majority of AOX activity was in the organelle pellet (Table V), suggesting that the active portions of the mutant AOXs were properly targeted to peroxisomes. Further fractionation of the organelle pellet from the AOX–HA strain through a sucrose density gradient confirmed that the AOX activity was peroxisomal (Fig. 5 B). Thus, a portion of both AOX–RSC and AOX–HA synthesized in P. pastoris cells is properly imported into peroxisomes where it assembles into an active enzyme.

To estimate the proportion of total AOX–RSC and AOX–HA protein that assembled into an active enzyme, total cell extracts from methanol-induced cells of the AOX–RSC and AOX–HA-expressing strains were prepared by mechanical disruption and subjected to sucrose velocity gradient sedimentation (Goodman et al., 1984). This glass bead-based procedure was chosen because it avoids the long incubation period needed to prepare yeast protoplasts for subcellular fractionation, during which unstable inactive AOX protein is rapidly degraded relative to stable active AOX octamers. As a marker for the velocity gradients, we assayed activity for CAT, a tetrameric protein of ∼300 kD. In a control gradient with total lysates prepared from strains expressing AOX-WT (Fig. 6,C), AOX activity and protein comigrated at one position through the velocity gradient, the same position as purified octameric AOX with a molecular mass of ∼600 kD (Fig. 6,A). Gradients prepared from cells of strains expressing either AOX–RSC or AOX–HA alone contained small amounts of AOX activity at the normal octameric AOX position (Fig. 6, D and E). However, most of the AOX protein was spread over fractions representing lower molecular masses. Control gradients containing chemically denatured (mainly monomeric) AOX (Fig. 6,A, AOX*) or extracts prepared from a pex5Δ mutant strain expressing AOX–HA (Fig. 6 B) indicated that the inactive AOX proteins at these lower mass positions were most likely monomeric and aggregated forms of AOX.

We concluded from these studies that a PTS sufficient for targeting proteins to peroxisomes is present within LARF (most likely ARF), and that this PTS is essential for the efficient targeting and import of AOX. However, since small but significant amounts of AOX are imported in the absence of LARF, a second less efficient PTS exists in another part of the protein. This second PTS may be located within the next 18 COOH-terminal amino acids of AOX, since a mutant protein in which the 22 COOH-terminal amino acid residues were deleted was completely inactive (Table V).

The similarity of ARF to the consensus PTS1 motif, SKL, suggests that AOX may be imported via the PTS1 pathway. Furthermore, import of AOX or β-lac–AOX is blocked in a pex5Δ mutant which is specifically defective in the PTS1 receptor Pex5p (Fig. 3,C; McCollum et al., 1993). Additional evidence that AOX is targeted to peroxisomes by the PTS1 pathway was obtained in yeast two- hybrid system assays in which the 22 COOH-terminal amino acids of AOX were expressed as a fusion with the GAL4 DNA-binding domain (GAL4_B–AOX) in combination with the P. pastoris PTS1 receptor protein Pex5p fused to the GAL4 activation domain (GAL4_A–Pex5p). This combination produced a strong response in the system (Fig. 7). The interaction was specific since expression of GAL4_B–AOX with unfused GAL4_A or GAL4_A–Pex5p with unfused GAL4_B produced little response. We then tested GAL4_B–AOX variants in which the COOH-terminal LARF sequence was replaced by either RSC or the HA tag. Neither of these constructs produced a specific response in combination with GAL4_A–Pex5p.

We examined whether the import efficiency of AOX without its PTS1 motif was increased when coexpressed with AOX-WT. To distinguish between AOX without its PTS1 motif and AOX-WT, we used the AOX–HA construct in which the four COOH-terminal amino acids had been replaced with the HA-epitope tag. Immunoblot analysis of total extracts prepared from P. pastoris strains expressing either AOX-WT or AOX–HA alone demonstrated that the anti-HA monoclonal antibodies exclusively recognized the AOX–HA protein, whereas polyclonal antibodies against AOX protein recognized both AOX-WT and AOX–HA (Fig. 5,A). To estimate the proportion of imported versus cytoplasmic AOX–HA, we again made use of the sucrose velocity sedimentation method to separate imported, assembled, and active octameric AOX from cytoplasmic inactive monomers and aggregated forms of the protein (Goodman et al., 1984). Velocity gradients prepared with lysates from a strain expressing only AOX–HA showed a small amount of AOX activity at the position expected for octameric AOX (Fig. 6,E). However, immunoblot analysis revealed that most AOX–HA protein migrated to lower molecular mass positions in the gradient. This distribution was not because of the HA tag since a similar result was obtained with a strain expressing AOX–RSC (Fig. 6,D). As described above, control gradients performed with purified and then chemically denatured AOX (Fig. 6,A, AOX*), or with a lysate prepared from a pex5Δ strain expressing AOX–HA (Fig. 6 B), indicated that the AOX protein at the lower positions represented monomeric and/or aggregated forms of AOX.

A P. pastoris strain coexpressing both AOX–HA and AOX-WT was first examined by subcellular fractionation. After the organelle pellet from methanol-grown cells of the strain were centrifuged through a sucrose density gradient, a significant portion of AOX–HA was found to be peroxisome-bound as indicated by the comigration of AOX–HA protein with activities for AOX and CAT (Fig. 5,C). When a lysate prepared from this strain was analyzed by sucrose velocity gradient sedimentation, large amounts of AOX activity migrated to the position of octameric protein (Fig. 6,F). Immunoblot analysis with anti-AOX antibodies that recognize both AOX-WT and AOX–HA confirmed that the major portion of AOX protein comigrated with the AOX activity. However, immunoblot analysis with the anti-HA antibodies did not show a shift of AOX–HA protein to the position of octameric AOX (compare the distribution of AOX–HA in Fig. 6, F and E). Thus, the import efficiency of AOX–HA, as measured by octamerization, was not improved by coexpression with AOX-WT, indicating that AOX–HA could not oligomerize with AOX-WT in the cytoplasm before import into peroxisomes.

An alternative interpretation of the coexpression results was that AOX–HA could not oligomerize with AOX-WT in the cytoplasm because of interference by the HA tag in AOX–HA/AOX-WT assembly or to the rapid assembly kinetics of AOX-WT, as suggested for MDH by Elgersma et al. (1996). If this were true, then the small amount of AOX–HA that does assemble into active enzyme would be expected to be present as AOX–HA homooctamers and not as AOX–HA/AOX-WT heterooctamers. To examine whether or not active AOX–HA oligomerized with AOX-WT in cells coexpressing the two proteins, two experiments were performed. In the first, sucrose velocity gradient fractions containing octameric AOX from the coexpression strain (Fig. 6,F, fraction 9) and from the strain expressing AOX–HA alone (Fig. 6,E, fraction 9) were immunoprecipitated using anti–HA antibodies. The immunoprecipitates were subsequently examined via immunoblot for the relative amounts of HA and AOX cross- reacting protein (Fig. 8,A). If active AOX–HA in the coexpression strain was a homooctamer, the relative amount of HA- and AOX-reacting protein should be the same as in the AOX–HA fractions. In contrast, results showed that the HA antibodies precipitated significantly more AOX protein from the coexpression strain fractions, indicating that AOX-WT protein was coprecipitated along with AOX–HA. As a control for the specificity of the HA antibody preparation, no AOX protein was immunoprecipitated from AOX-WT gradient fractions (Fig. 6,C, fraction 9; Fig. 8 A, lane 4). These results suggested that AOX-WT and AOX–HA subunits exist as heteromers in octameric fractions from the coexpression strain.

In the second experiment, the presence of AOX–HA and AOX-WT protein in the same species of AOX molecule was examined directly by subjecting active octameric AOX (Fig. 6, from the gradients in fraction 9) to native PAGE and immunoblotting (Fig. 8 B). In these native gels, AOX-WT homooctamers migrate slower than AOX–HA homooctamers (despite the fact that AOX–HA is slightly higher in predicted molecular mass than AOX-WT). In fractions from the coexpression strain, AOX–HA migrated at approximately the same rate as homooctameric AOX-WT. Thus, most of the AOX–HA was present as heterooctamers with AOX-WT. Furthermore, the fact that heteromers migrated at about the same rate as AOX-WT homomers indicated that the heteromers were composed mostly of AOX-WT with only one or two AOX–HA subunits, indicating that the small amount of AOX–HA that reaches the peroxisomal matrix had no difficulty oligomerizing with the large amount of AOX-WT present in the organelle.

In this paper, we examined the targeting and assembly of peroxisomal AOX in the yeast P. pastoris. In particular, we were interested in determining whether newly synthesized AOX is capable of oligomerizing in the cytoplasm before import, as recently reported for certain other peroxisomal proteins. Previous results from our laboratory suggested that P. pastoris AOX could not assemble outside the peroxisome. Methanol-induced pex mutants of this yeast contain little or no activity for AOX, although substantial amounts of AOX protein are present in the cytoplasm (Liu et al., 1992, 1995; Waterham et al., 1996). Here we show that the amount of residual AOX activity in the pex mutants closely correlates with the severity of the peroxisomal biogenesis defect with chemically induced (and slightly leaky) pex mutants typically containing some residual AOX activity, while most pexΔ strains contain no detectable AOX activity. In addition, we show that residual active AOX in the pex mutants is located inside the few small peroxisomes or peroxisomal remnants present in these cells. Although these results do not eliminate the possibility that small amounts of AOX can octamerize into an active enzyme in the cytoplasm of pex mutants, they strongly suggest that the bulk of AOX protein must reach the peroxisome to efficiently assemble.

The inability of AOX to properly assemble in the cytoplasm of pex mutants contrasts with other peroxisomal enzymes which appear to be fully active in the cytoplasm and, therefore, have little difficulty assembling there. These enzymes include CAT from P. pastoris, an enzyme that must properly fold, incorporate a heme cofactor, and tetramerize to become active (Gould et al., 1992; Liu et al., 1992). Thus, the inability of AOX to assemble is not the result of some general inability of peroxisomal enzymes to incorporate cofactors or form homooligomers in the cytoplasm. The assembly problem does not appear to be because of the inability of AOX to oligomerize outside the peroxisome, since Evers et al. (1995) were able to efficiently reassemble active octameric AOX in vitro from FAD-containing monomeric subunits obtained from either P. pastoris or the related yeast H. polymorpha. They further showed that, in a riboflavin auxotrophic mutant of H. polymorpha, AOX protein is found in misfolded aggregates, indicating that insertion of FAD into AOX is an essential step in the assembly of H. polymorpha AOX (Evers et al., 1994, 1996). They postulated that the FAD insertion step may be mediated by an unknown assembly factor since, once removed, FAD could not be reinserted into AOX in vitro (Evers et al., 1995). Thus, in the absence of peroxisomes, the inability of P. pastoris AOX to assemble in the cytoplasm may also be related to a failure of FAD insertion (e.g., FAD and/or FAD insertion factor concentration in the cytoplasm may be too low to support efficient FAD binding).

Interestingly, AOX (MOX) from H. polymorpha efficiently assembles into an active enzyme in the cytoplasm of H. polymorpha pex mutants (Cregg et al., 1990; van der Klei et al., 1991). We found that MOX is active in the cytoplasm of P. pastoris pex mutants while AOX is inactive in the cytoplasm of pex mutants of H. polymorpha. Thus, this distinction between P. pastoris and H. polymorpha AOXs must be due to differences between their AOX polypeptides and not to the environment provided by their hosts. This is a surprising result given the high degree of sequence similarity shared by these polypeptides (∼85% identical). Perhaps, for proper assembly, P. pastoris AOX requires a higher concentration of FAD or FAD-insertion factor, or is more dependent on the acidic environment of the peroxisome than H. polymorpha AOX.

Recent studies suggest that at least some peroxisomal proteins are not only capable of cytoplasmic assembly but are also imported into peroxisomes in a preassembled oligomeric state. The key experiment in these studies is the coexpression of the peroxisomal protein without its PTS along with the wild-type protein. In each case, although the PTS-less protein expressed alone is not imported, coexpression resulted in the efficient coimport or piggybacking of the PTS-less polypeptides. Coimport has been reported for the human PTS1 enzyme, alanine/glyoxylate amino transferase (Leiper et al., 1996), a chloramphenicol acetyl transferase PTS1 fusion protein (McNew and Goodman, 1994), a yeast PTS1 enzyme, MDH3 (Elgermsa et al., 1996), and the yeast PTS2 enzyme thiolase (Glover et al., 1994). The ability of each of these proteins to piggyback into peroxisomes strongly suggests that these proteins can oligomerize in the cytoplasm before import.

To perform the coexpression study with P. pastoris AOX, it was first necessary to identify and characterize its PTS. We found that critical information for efficient peroxisomal targeting of P. pastoris AOX is located within its four COOH-terminal amino acids, LARF. Previously, a similar conclusion was drawn for the same four COOH-terminal amino acids of H. polymorpha MOX (Hansen et al., 1992). This conclusion was based on immunocytochemical data which indicated that these four amino acids were capable of targeting β-lactamase to peroxisomes in H. polymorpha. We show that these same four COOH-terminal amino acids are also sufficient to target β-lactamase to P. pastoris peroxisomes. In addition, we show that these four amino acids are critical for targeting since their removal and substitution with either RSC or an HA-epitope tag results in proteins that are only inefficiently imported into peroxisomes. Interestingly, a small portion of both AOX– RSC and AOX–HA is properly targeted to peroxisomes. Cell fractionation studies confirmed that these activities represent AOX protein that is correctly imported in peroxisomes, although the efficiency of import is not sufficient to support normal methanol-growth rates. One explanation for the residual import of LARF-less AOX is that, in addition to a COOH-terminal PTS, AOX contains a second independent but less efficient PTS. A second PTS has been reported for other peroxisomal matrix proteins including S. cerevisiae catalase A (Kragler et al., 1993) and H. polymorpha Per1p (Waterham et al., 1994). This second PTS may be located within the 18 amino acids immediately adjacent to LARF in AOX, since a mutant AOX deleted for the 22 COOH-terminal amino acids is not imported at all. Furthermore, a β-lactamase fusion containing the 22 COOH-terminal amino acids of AOX is targeted more efficiently than one fused to just the LARF sequence. A second explanation for the residual import of AOX without LARF is that the protein oligomerizes in the cytoplasm with another methanol pathway protein, such as dihydroxyacetone synthase, and is coimported. The preimport coassembly of AOX and dihydroxyacetone synthase was previously suggested from the results of ionophore experiments with the yeast C. boidinii (Bellion and Goodman, 1987). We observed that, in glucose-grown cells that do not synthesize dihydroxyacetone synthase, AOX– RSC (expressed under control of the constitutive GAP promoter) is not imported, although AOX-WT is imported (not shown), a result that is consistent with the notion that residual import of AOX without LARF may be dependent on another methanol pathway protein.

AOX is clearly targeted to peroxisomes via the PTS1 import pathway. Previous results demonstrated that AOX behaves in a manner similar to luciferase, the prototypical PTS1 protein, in P. pastoris wild-type and pex mutant cells (McCollum et al., 1993; Spong et al., 1993; Liu et al., 1995; Waterham et al., 1996). Here, we extend these results by demonstrating that neither AOX nor β-lactamase fused to LARF is targeted to peroxisomes in a pex5Δ strain that is specifically defective in the import of PTS1 proteins (Spong et al., 1993). In addition, two-hybrid system results indicate that AOX strongly interacts with the PTS1 receptor Pex5p, and that this interaction is dependent upon the last four amino acids of AOX. We conclude that the primary targeting signal for AOX is a PTS1 located within LARF and that the PTS most probably consists of the COOH-terminal tripeptide sequence ARF. The first two amino acids of the motif in AOX, alanine and arginine, are known functional variants of the prototypical PTS1 sequence (SKL; Gould et al., 1989; Swinkels et al., 1992). The substitution of a phenylalanine for leucine at the ultimate position was shown to abolish peroxisomal targeting in mammalian cells (Swinkels et al., 1992). However, while this work was in progress, Elgersma et al. (1996) reported that PTS1 motifs ending in phenylalanine were imported in S. cerevisiae.

To investigate the effect of coexpressing AOX-WT and AOX without its primary PTS, it was not possible to examine samples by differential or sucrose density gradient centrifugations, as done in previous studies of this kind. These techniques require prolonged incubation of cells to remove cell walls, conditions that result in degradation of improperly folded AOX (Liu et al., 1992, 1995; and Waterham et al., 1996). Furthermore, improperly folded AOX forms into protein aggregates which sediment upon differential centrifugation and, thus, appear in organelle pellets along with properly imported and assembled AOX. To observe import of AOX, cell extracts were prepared by the glass bead disruption method at a low temperature and were subjected to sucrose velocity sedimentation (Goodman et al., 1984). The proportion of AOX protein present at the position of fully assembled AOX octamers was then noted in the gradients. Previous studies (Liu et al., 1992, 1995) as well as those presented here indicate that active octameric AOX is only found in the peroxisomal matrix. Thus, the presence of AOX–HA at the position of octameric AOX in velocity gradients represents AOX–HA that was properly imported into the organelle.

The coexpression studies showed that the efficiency of AOX–HA import is not improved by AOX-WT. Thus, it appears that AOX–HA cannot initiate cytoplasmic assembly with AOX-WT in the cytoplasm and be piggybacked into the peroxisome. An alternative explanation for the failure of AOX–HA to efficiently oligomerize with AOX-WT in the coexpression experiment was that the HA tag somehow interfered with its assembly. Elgersma et al. (1996) proposed that the apparent inability of S. cerevisiae PTS1 enzyme MDH to coimport PTS1-less MDH was because of a reduction in the assembly kinetics of the mutant polypeptide. As a result, the mutant subunits were outcompeted by the wild-type subunits that rapidly homodimerize in the cytoplasm. Our results indicate that AOX–HA is not impaired in its ability to assemble with AOX-WT. When we examined the small portion of AOX–HA in the coexpression strain that does reach the peroxisomal matrix and assemble into active octameric enzyme, we found that virtually all of the mutant protein was present as heterooligomers with AOX-WT. Thus, it appears that despite the presence of large amounts of AOX-WT, most AOX–HA remains in the cytoplasm because AOX cannot initiate oligomerization in this compartment and, without its primary PTS1, the mutant AOX is inefficiently targeted to and imported into the peroxisome. In summary, although peroxisomes are capable of importing some proteins as preassembled structures, AOX must be targeted to peroxisomes as monomers for import and assembly to occur properly.

Despite recent progress, basic features of protein import into peroxisomes remain largely unknown (Waterham and Cregg, 1997). Although the organelle appears to be morphologically simple with a single membrane and uncomplicated matrix, the import mechanism is unexpectedly complex. Peroxisomes have evolved at least two matrix protein import pathways and a third independent pathway specific for peroxisomal integral membrane proteins. As shown by several recent studies, the matrix protein import machinery is capable of importing large preassembled structures ⩾9 nm in diameter (Walton et al., 1995), a size much greater than AOX octamers at ∼600 kD. Therefore, the necessity of importing AOX as unassembled monomeric polypeptides is probably not related to its size but may reflect the necessity to incorporate FAD before octamerization, a process proposed to occur in the peroxisomal matrix (Evers et al., 1996).

Although this is the first description of a yeast peroxisomal protein that cannot assemble in the absence of functional peroxisomes, this phenomenon may be relatively common in patients afflicted with the lethal peroxisomal biogenesis disorder Zellweger syndrome (Lazarow and Moser, 1994). In cells of these patients, peroxisomes are absent and peroxisomal matrix enzymes are left in the cytoplasm. Some of these enzymes are stable and active, whereas several others (including the plasmalogen biosynthetic enzymes, acyl-CoA:dihydroxyacetone phosphate acyltransferase and alkyl-dihydroxyacetone phosphate synthase, and the peroxisomal β-oxidation pathway enzymes, acyl-CoA oxidase, bifunctional enzyme, and 3-oxoacyl-CoA thiolase) are deficient (Lazarow and Moser, 1994). Pulse chase studies on acyl-CoA oxidase and thiolase demonstrated that the proteins are synthesized normally but remain largely inactive in the cytosol where they are rapidly degraded (Schram et al., 1986; Suzuki et al., 1986). The inability of these and other enzymes to assemble into stable active enzymes may be the reason that specific peroxisomal metabolic pathways are defective in Zellweger patients. Further investigations into the fate of AOX in P. pastoris mutants may shed light on the molecular details of this phenomenon and the molecular etiology of this aspect of Zellweger syndrome.

We thank S. Subramani (University of California at San Diego) for the gift of vector pSP72 containing the P. pastoris PEX5 gene and the pex5 deletion strain. We thank T. Hadfield (Oregon Graduate Institute, Portland, OR) for assistance in preparing the manuscript.

AOX

alcohol oxidase

AOX-WT

wild-type AOX

ARF

alanine-arginine-phenylalanine_COOH

CAT

catalase

FAD

flavine adenine dinucleotide

LARF

leucine-alanine-arginine-phenylalanine_COOH

MDHC

malate dehydrogenase 3

MOX; AOX from Hansenula polymorpha; MSG; methione-serine-glysine; pex

peroxisome biogenesis defective

PTS

proximal targeting signal

SKL

arginine-serine-cysteine_COOH