Topogenesis of Peroxisomal Membrane Protein Requires a Short, Positively Charged Intervening-loop Sequence and Flanking Hydrophobic Segments

Human 34-kDa peroxisomal membrane protein (PMP34) consisting of 307 amino acids was previously identified as an ortholog of, or a similar protein (with 27% identity) to the, 423-amino acid-long PMP47 of the yeast Candida boidinii. We investigated membrane topogenesis of PMP34 with six putative transmembrane segments, as a model peroxisomal membrane protein. PMP34 was characterized as an integral membrane protein of peroxisomes. Transmembrane topology of PMP34 was determined by differential permeabilization and immunofluorescent staining of HeLa cells ectopically expressing PMP34 as well as of Chinese hamster ovary-K1 expressing epitope-tagged PMP34. As opposed to PMP47, PMP34 was found to expose its N- and C-terminal parts to the cytosol. Various deletion variants of PMP34 and their fusion proteins with green fluorescent protein were expressed in Chinese hamster ovary-K1 and were verified with respect to intracellular localization. The loop region between transmembrane segments 4 and 5 was required for the peroxisome-targeting activity, in which Ala substitution for basic residues abrogated the activity. Three hydrophobic transmembrane segments linked in a flanking region of the basic loop were essential for integration of PMP34 to peroxisome membranes. Therefore, it is evident that the intervening basic loop plus three transmembrane segments of PMP34 function as a peroxisomal targeting and topogenic signal.

The peroxisome is a model system for addressing protein traffic and membrane biogenesis, especially where it is linked to human neurological and metabolic diseases, called peroxisome biogenesis disorders. Import of matrix proteins into peroxisomes has been investigated to a greater extent at the molecular and cellular levels (1,2). Peroxisomal matrix proteins are synthesized on free polysomes, and most contain peroxisome targeting signals type 1 or 2, called PTS1 1 or PTS2.
Cytosolic receptors for PTS1 and PTS2, Pex5p and Pex7p, respectively, function in the transport of cargo proteins to peroxisomes, by docking with the convergent membrane peroxin, Pex14p (3)(4)(5). In contrast, molecular mechanisms involved in membrane protein transport and membrane assembly of peroxisomes are not well understood. Peroxisomal integral membrane proteins (PMPs), such as 22-kDa PMP (PMP22) (6), PMP70 (7), and Pex2p (8) of rat liver are also synthesized on free polysomes. Post-translational import of PMP22 and PMP70 were shown in vitro (9,10). It is noteworthy that Pex16p and Pex2p are N-glycosylated in the yeast Yarrowia lipolytica (11). PTS of membrane protein, termed mPTS, was previously suggested for PMP47 with six putative transmembrane segments of the yeast Candida boidinii (12). The loop region between the fourth transmembrane segment (TM4) and TM5, which was enriched in positively charged amino acids was functional as a PTS when fused to cytosolic proteins (12). It is also noteworthy that an internal region, including the predicted TM3, of rat PMP70, a six-TM protein, is essential for the peroxisomal localization (10). The N-terminal amino acid residues at positions 1-40, which are relatively enriched in positive charged residues, were recently shown to be required for translocation of Pex3p, an integral membrane peroxin, to peroxisomes in mammals (13)(14)(15) and yeast (16,17). Accordingly, to delineate mPTS and to establish a general paradigm for topogenesis of PMPs, more information is required.
In the present study, we have chosen PMP34 as a model protein, to address these issues. PMP34 was recently cloned by expressed sequence tag data base search using C. boidinii PMP47 and was shown to be localized to peroxisomes (18). We found that PMP34 is an integral membrane protein of peroxisomes. In contrast to PMP47, PMP34 exposes both N-and C-terminal parts to the cytosol. We also identified the loop region between TM4 and TM5 as a potential mPTS. This loop plus three TMs were essential for targeting and integration of PMP34. These results provide the first evidence for a functional mPTS for the topogenesis of a membrane protein that spans membranes multiple times in mammalian peroxisomes.
(Tokyo, Japan), and Takara (Tokyo, Japan). Fetal calf serum and Ham's F-12 were from Life Technologies, Inc. Anti-PMP34 antibody was raised in rabbits by immunizing with synthetic peptide comprising the Cterminal, 19-amino acid sequence of human PMP34, supplemented with cysteine at the N terminus that had been linked to keyhole limpet hemocyanin (19). Rabbit antibody against influenza virus hemagglutinin (HA) was likewise raised using synthetic peptide, CYPYDVPD-YASLRS-NH 2 . Guinea pig antibody to human catalase (Sigma) was raised by conventional subcutaneous injection. We also used rabbit antibodies to PTS1 peptide (20), acyl-CoA oxidase (AOx) (21), and C-terminal 19-amino acid residues of rat Pex14p (4). Rabbit antibody to green fluorescent protein (GFP) (CLONTECH) and mouse monoclonal antibodies to flag (M2) and HA (12CA5) (Sigma) were purchased.
Isolation and Epitope Tagging of HsPMP34 -Human PMP34 cDNA (HsPMP34) was cloned by PCR-based technique on the human liver cDNA library in pCMVSPORT (Life Technologies, Inc.) (22), using a set of primer, HsPMP34F and HsPMP34R (Table I). Six independent cDNA clones were sequenced, all showing that nucleotide residue at position 315 in a codon (CTG) for Leu 105 was G instead of CTC reported by Wylin et al. (18). HsPMP34 cDNA was cloned into the NotI-SalI site in pUcD2HygSR␣ (23). Tagging of epitopes, flag, and tandem HA (HA-HA) to the N and C terminus, respectively, of human PMP34 was conducted as follows. The full length of HsPMP34 was amplified using a pair of primers, HsPMP34F and PMP34RNhe. The PCR product was digested with NotI and NheI, then ligated into the NotI-NheI sites, upstream of a double-HA tag sequence, of pUcD2HygSR␣⅐PEX16-HA (24). NcoI-SalI fragment of PMP34-HA was replaced into the NcoI-SalI sites of pBluescript SK(Ϫ)flag-PEX13 (25), to construct flag-PMP34-HA. The NotI-SalI fragment of pBluescript SK(Ϫ)flag-PMP34-HA was subcloned into pUcD2HygSR␣, termed pUcD2HygSR␣⅐flag-HsPMP34-HA. All plasmid constructs were assessed by sequence analysis.
Construction of PMP34 Variants-N-terminal truncation mutants of PMP34 were generated by PCR, using as forward primers, ⌬N30F, ⌬N125F, ⌬N186F, or ⌬N204F, and a reverse primer PMP34RNhe. C-terminal deletion mutants were likewise constructed by PCR, using reverse primers, 204TerR and 186TerR, and a forward primer HsPMP34F. HA tagging at their C terminus was done as described above. To construct enhanced green fluorescent protein (GFP)-fusion proteins, SpeI site was introduced to the initiation codon of GFPexpression plasmid phGFP105-C1, a generous gift from T. Tsukamoto and T. Osumi (Himeji Institute of Technology, Himeji, Hyogo, Japan) by PCR using primers, GFPSpeF and GFPshortR. The resulting PCR fragment was digested with SpeI and SalI and introduced into the SpeI-SalI sites of pBluescript SK(Ϫ), termed pBluescript SK(Ϫ)GFP. SpeI-SalI fragment of pBluescript SK(Ϫ) GFP and NotI and NheI fragment of ⌬N125, ⌬N186, 204Ter, or 186Ter, cloned in pGEM/T-easy vector were introduced into the NotI-SalI site of pUcD2HygSR␣. Chimera constructs for 30/204GFP and 86/204GFP were generated by PCR, using a forward primer ⌬N30F or ⌬N86F, and a reverse primer 204TerR. Fusion for 86/273GFP and 125/273GFP were likewise constructed, using a forward primer ⌬N86F or ⌬N125F, and a reverse primer 273TerR; GFP tagging was performed as above. Expression plasmid for 187/204GFP, the fourth loop domain of PMP34 fused to the N terminus of GFP, was constructed by PCR, using as a template pUcD2HygSR␣⅐204GFP and as primers ⌬N186F and GFPshortR. For construction of GFP187/204HA, BglII site was introduced by PCR on pUcD2HygSR␣⅐204HA using BglII187F and pUcD2HygSR␣ reverse (D3R) as forward and reverse primers, respectively. The BglII and SalI fragments of 187/204HA and NotI-BglII fragment of pBluescript SK(Ϫ) GFP were introduced into the NotI-SalI site of pUcD2HygSR␣.
Alanine substitution in the fourth loop domain between TM4 and TM5 of PMP34 was done by PCR using as a template pGEM/T-easy⅐204Ter and a forward primer HsPMP34F and a reverse primer 6Ar, 4Ar, 3Ar, or Ar. Construction 2A was done by PCR using as a template pGEM/T-easy⅐6A204Ter and a forward primer HsPMP34F and a reverse primer 2Ar. GFP tagging was performed as described above.
Expression of PMP34 and Its Derivatives in HeLa and CHO-K1 Cells-DNA transfection was done using LipofectAMINE (Life Technologies, Inc.), as described (23). The cells were cultured for 36 h after transfecting the plasmid into the cells. Cells were fixed with 4% paraformaldehyde and permeabilized with 1% Triton X-100, and peroxisomes were visualized by indirect immunofluorescence light microscopy, as described (26). Antigen-antibody complexes were detected under a Carl Zeiss Axioskop FL microscope, using fluorescein isothiocyanate (FITC)-labeled sheep anti-mouse antibody (Amersham Pharmacia Biotech, Tokyo, Japan), FITC-labeled sheep anti-rabbit immunoglobulin (Ig) G antibody (Cappel), or Texas Red-labeled goat antibodies to guinea pig IgG (Vector Laboratories) and rabbit IgG (Leinco Technologies). GFP was directly observed by fluorescent microscopy with the use of the same filter for FITC after fixation (27). Flag-PMP34-HA was detected using rabbit anti-HA antibody and mouse anti-flag antibody, in cells that had been fixed as above and then permeabilized with either 25 g/ml digitonin or 1% Triton X-100 (23,28).
HeLa cells were transfected with pUcD2HygSR␣⅐HsPMP34. Cells were stained with rabbit anti-PMP34 C-terminal peptide antibody and guinea pig anti-human catalase antibody, after permeabilization for 5 min using either 25 g/ml digitonin or for 2 min with 1% Triton X-100.
Protease Sensitivity Assay-HsPMP34-transfected CHO-K1 cells (1 ϫ 10 7 ) were homogenized in 0.5 ml of a homogenizing buffer: 0.25 M sucrose, 25 mM ammonium carbonate, pH 7.4, 20 g/ml each of leupeptin and antipain, and 500 units/ml aprotinin, by 10 strokes of an Elvehjem-Potter homogenizer. A postnuclear supernatant (PNS) fraction was prepared by centrifugation of homogenates at 750 ϫ g for 5 min. The PNS from 1 ϫ 10 6 cells was treated with 2 g of Staphylococcus aureus V8 protease (Roche Molecular Biochemicals) at 25°C for 1.5 h in 0.5 ml of the homogenizing buffer, in the absence or presence of 1% Triton X-100. The reaction was terminated by precipitation using trichloroacetic acid, and whole proteins were then analyzed by SDSpolyacrylamide gel electrophoresis (PAGE) and immunoblot.

TGGTTCTTTCCGCCTCAG
Other Methods-Liver peroxisomes were isolated from a normal rat, as described (29). Western blot analysis was done using primary antibodies and a second antibody, donkey anti-rabbit IgG antibody conjugated to horseradish peroxidase (Amersham Pharmacia Biotech). Antigen-antibody complex was visualized with ECL Western blotting detection reagent (Amersham Pharmacia Biotech).

Characterization of HsPMP34
Protein-The intracellular localization of PMP34 was determined by ectopic expression of epitope-tagged HsPMP34 and immunofluorescence microscopy. In wild-type CHO-K1 cells expressing PMP34 tagged with HA at its C terminus, PMP34 was detected as a punctate staining pattern using anti-HA antibody. The staining pattern was superimposable with that obtained using anti-Pex14p antibody, thereby suggesting that PMP34-HA was targeted to peroxisomes (Fig. 1A). Similar results were obtained when HsPMP34 was expressed in HeLa cells and stained with anti-PMP34 peptide antibody (see Fig. 1C). The rabbit antibody raised to the C-terminal peptide of human PMP34 specifically reacted with an endogenous, single protein with an apparent molecular mass of ϳ34 kDa in immunoblot of rat liver peroxisomes ( Fig.  1B, left), confirming that PMP34 is a bona fide protein of peroxisomes. By sequence similarity to C. boidinii PMP47 (12,18), HsPMP34 is likely to be an integral membrane protein with six putative transmembrane segments. The integrity of PMP34 in peroxisomal membranes was verified by extraction with 0.1 M sodium carbonate, pH 11.3 (30). PMP34-HA expressed in CHO-K1 cells was not extracted with sodium carbonate (Fig. 1B, right) and was recovered in membrane fraction at a similar level as in postnuclear supernatant fraction (data not shown). Pex14p, a peroxisomal membrane peroxin, was likewise in membrane pellet, whereas a matrix enzyme AOx was in soluble fraction. These data thereby indicated that PMP34 is integrated into membranes.
Next, we determined transmembrane topology of PMP34 by the differential permeabilization/immunofluorescence microscopy method (28,31). HeLa cells were transfected with HsPMP34. Upon treatment with 1% Triton X-100, which solubilizes all cellular membranes, a punctate staining pattern was observed using anti-PMP34 C-terminal peptide antibody, in a superimposable manner with catalase presumably representing peroxisomes (Fig. 1C, a and c). Similar morphological pattern, PMP34-staining, was observed after permeabilization with 25 g/ml digitonin, which selectively permeabilized plasma membranes (28, 31) ( Fig. 1C, b), whereas none of the cells were stained with anti-catalase antibody under the same condition (Fig. 1C, d), indicating that peroxisomal matrix proteins are not accessible to antibody. Taken together, C-terminal part of PMP34 is most likely to be exposed to the cytosol. Moreover, the C-terminal part of PMP34 in HsPMP34-overexpressing CHO-K1 was likewise detected by the digitonin treatment, although requiring much higher concentration of digitonin under which conditions matrix proteins still remained inaccessible to antibodies (data not shown). We do not know the reason why PMP34 was visible only at higher concentration of digitonin in this particular case, as compared with that normally permeabilizing plasma membranes (4,23). Membrane topology of PMP34 was also determined using N-terminally flag-and C-terminally HA-tagged PMP34. When flag-PMP34-HA-transfected CHO-K1 cells were permeabilized with 25 g/ml digitonin, a punctate staining pattern was observed using both anti-flag and anti-HA antibodies (Fig. 1D, b and f). In contrast, PTS1 proteins were not detectable under the same condition, indicating that lumenal proteins are not accessible to the antibody (Fig. 1D, d). Upon treatment with Triton X-100, PTS1 proteins as well as PMP34 were discernible in numerous particles, peroxisomes (Fig. 1D, a, c, and e). Therefore, it is apparent that N-and C-terminal portions are exposed to the cytosol. Of note, the topology of PMP34 contrasts with that of PMP47, which exposes both terminal parts to matrix of peroxisomes (12).
Furthermore, PMP34 expressed in CHO-K1 cells was resistant to the treatment with exogenously added S. aureus protease V8, as verified using PNS fraction, under which condition Left, endogenous PMP34 in rat liver. Rat liver peroxisomes (50 g) were analyzed by SDS-PAGE and immunoblot using anti-PMP34 peptide antibody. Arrowhead indicates ϳ34-kDa protein band specifically detected using anti-PMP34 antibody. Asterisk designates a nonspecific band. Right, membrane (P) and soluble (S) fractions of CHO-K1 cells expressing PMP34-HA were prepared from postnuclear supernatant fraction by the sodium carbonate method (30). Immunoblotting was done using antibodies specific for HA, AOx, and Pex14p. C, membrane topology of PMP34; morphological analysis. Transmembrane topology of PMP34 was determined by the differential cell-permeabilization method. HeLa cells were transfected with HsPMP34. Cells were fixed and then treated with 1% Triton X-100 (TX-100; a and c) or 25 g/ml digitonin (b and d). A set of cells were stained for PMP34 using antibodies to PMP34 C-terminal peptide (a and b) and catalase (c and d).
HsPMP34-expressing cells were indicated by arrowheads (a). Note that C-terminal part of PMP34, not a matrix enzyme catalase, was detected after both types of treatments. Original magnification, ϫ630; bar, 20 m. D, transmembrane topology of flag-PMP34-HA. CHO-K1 cells transfected with flag-HsPMP34-HA were fixed, then treated with 1% Triton X-100 (a, c, and e) or 25 g/ml digitonin (b, d, and f). Pex14p was digested (Fig. 1E, left panel). Matrix enzyme AOx was also fully protected from the digestion. Both PMP34 and AOx were no longer discernible after protease treatment in the presence of Triton X-100 (Fig. 1E, lane 3). These results were interpreted to mean that multiple sites, C-terminal side of Glu, for cleavage by V8 protease locate inside of peroxisomes (Fig.  1E, right panel). It is conceivable that three potential cleavable Glu sites may not be readily attacked by the protease, probably owing to their location close to the transmembrane-spanning segments. The data together support the notion described above that the N-and C-terminal parts of PMP34 are exposed to the cytosol.
The Fourth Loop Domain of PMP34 Contains Peroxisome Membrane Targeting Signal-Although transmembrane topology of PMP34 is opposite to that of CbPMP47, amino acid sequence of the fourth intervening-loop region of PMP34 is similar to the one containing mPTS of PMP47 (12) (see Fig. 3A). To search for peroxisome targeting signal of PMP34, various mutants with deletion either from N or C terminus were constructed ( Fig. 2A). Mutants ⌬N30HA, ⌬N125HA, ⌬N186HA and ⌬N204HA, lacking the sequence from N terminus to the first, third, and fourth transmembrane segments, and to the fourth loop domain, respectively, were HA-tagged at the C terminus. These truncation mutants were transfected into wild-type CHO-K1 cells and analyzed for intracellular localization by immunofluorescence microscopy using anti-HA antibody (Fig. 2B). In the cells expressing ⌬N30HA, punctate immunofluorescence pattern was observed and superimposable to that using anti-Pex14p antibody, thus indicating that ⌬N30HA was localized to peroxisomes (Fig. 2B, a and b). ⌬N125HA was likewise detected in a superimposable manner to Pex14p, demonstrating peroxisomal localization (Fig. 2B, c and d). In the case of ⌬N186HA, punctate pattern was obtained, however, some signal did not correspond to that obtained from anti-Pex14p antibody (Fig. 2B, e and f, arrowheads). To confirm the intracellular localization of these two mutants, we also expressed GFP-tagged ⌬N125 and ⌬N186 and their peroxisomal localization were analyzed. The full-length PMP34 fused with GFP was localized to peroxisomes, as seen for PMP34-HA, when expressed in CHO-K1 cells (data not shown). Punctate signals were observed in the cells expressing ⌬N125GFP, and colocalized with those noted using anti-Pex14p antibody (Fig.   FIG. 2. Functional and topogenic regions of PMP34. C-terminally HAtagged or GFP-fused PMP34 and its variants were verified for intracellular localization in CHO-K1. A, constructs of deletion mutants of PMP34. ⌬N30HA, PMP34-HA with deletion of N-terminal residues from 1 to 30; 204HA, HA-tagged PMP34 with residues 1-204; 204GFP, PMP34 comprising residues 1-204 fused with GFP. Others likewise representing respective constructs were indicated. 2B, g and h). In the case of ⌬N186GFP, two types of punctate structures were found, where one type was superimposable with those seen using anti-Pex14p antibody and the other was not (Fig. 2B, i and j, arrowheads). ⌬N204HA was apparently localized to endoplasmic reticulum-like structures in addition to peroxisomes (Fig. 2B, k), thereby inferring a decrease in the topogenic activity. Together, these results suggest that 79amino acid sequence between the loop domains 3 (L3) and 4 (L4) is required for localization of PMP34 to peroxisomes.
Next, we constructed two other types of PMP34 mutants truncated in the C-terminal region and analyzed their intracellular localization. A deletion mutant, 204HA, truncated from the C terminus to the fifth transmembrane segment was localized to peroxisomes in CHO-K1, as assessed by colocalization with Pex14p (Fig. 2B, m and n). Moreover, the expressed 204HA protein showed the membrane topology exposing the HA-tagged C terminus to the cytosol and was resistant to the sodium carbonate treatment (data not shown), hence indicating that 204HA was properly targeted and integrated to peroxisome membranes. In contrast, cytoplasmically diffused staining was observed in the cells expressing 186HA truncated from C terminus to fourth loop segment (Fig. 2B, l). Essentially the same results were obtained using GFP as a reporter fusion protein; a fusion protein, 204GFP, was found in peroxisomes (Fig. 2B, o and p), while 186GFP was not localized to peroxisomes (data not shown).
We further investigated transmembrane topology of C-terminal portion of ⌬N125GFP and 204GFP, using anti-GFP antibody. Fig. 2C shows the topology of their C-terminal portion. GFP of the both fusion proteins was recognized by anti-GFP antibody at 25 g/ml digitonin, suggesting that both mutant proteins were integrated into peroxisomal membranes and exposed the C-terminal part to the cytosol, as wild-type PMP34. These results demonstrate that the fourth cytosolically faced hydrophilic loop of 18 amino acids is necessary for targeting of PMP34 to peroxisomes.
Positively Charged Region of the Fourth Loop Functions as a Peroxisome Targeting Signal-The fourth loop domain of PMP34 has positively charged amino acids at positions 190, 191, 195-197, and 199 (Fig. 3A). The positively charged amino acid cluster in the fourth loop of CbPMP47 was shown to be sufficient for localizing soluble reporter proteins to peroxisomes in yeast (12). To investigate whether the sequence enriched in basic amino acids in the fourth loop of PMP34 functions as an mPTS, we first replaced all basic amino acids to Ala of this region in 204GFP, termed 6A204GFP (Fig. 3B). In the cells expressing 6A204GFP, GFP fluorescence was diffused to the cytosol, suggesting that positively charged amino acids are required for peroxisome targeting (Fig. 3C, a). To determine which amino acid is important for the peroxisome-targeting function, we subdivided the fourth loop domain into two parts and changed basic amino acids in respective parts to Ala. Mutants 4A204GFP and 3A204GFP were partially localized to peroxisomes as well as in the cytoplasm (Fig. 3C, b and c). In contrast, 2A204GFP and A204GFP was localized to peroxisomes (Fig. 3C, d and e), as efficiently as 204GFP. Four chimera proteins, 4A204GFP as well as 3A204GFP localized to peroxisomes, 2A204GFP, and A204GFP, showed the same membrane topology, exposing C-terminal GFP to the cytoplasm (data not shown). Therefore, it is more likely that KR in the first portion and KKRMK in the second part function as mPTS and that the latter half of the loop is the most important.
The Fourth Loop Is Not Sufficient for Peroxisomal Localization-Dyer et al. (12) reported that the last 12 amino acids of the fourth loop domain of CbPMP47 was sufficient for targeting to peroxisomes. So we fused the fourth loop domain of PMP34 to N-or C terminus of GFP and analyzed its sufficiency for peroxisome localization. To determine whether the fourth loop functions as a sufficient information for peroxisomal targeting, a fusion protein 187/204GFP, the loop domain fused to the N terminus of GFP was expressed in CHO-K1, and found not to be localized to peroxisomes (Figs. 4A and 4B). Contrary to this, another fusion protein, GFP187/204HA, was targeted to mitochondria (Fig. 4B), although the positively charged loop in the fusion construct was located at position distinct from a general, N-terminal mitochondrial targeting sequence. These results demonstrated that the fourth loop domain of PMP34 is necessary for transport of PMP34 to peroxisomes, but not sufficient for integration into peroxisomal membranes.
Three Transmembrane Segments of PMP34 Are Required for Integration-204GFP and ⌬N125GFP which possess four and three transmembrane segments, respectively, were localized to peroxisomes as efficiently as wild-type PMP34, suggesting that the region(s) required for integration to peroxisomal membranes is in both N-and C-terminal domains (see above). ⌬N125GFP with three TMs containing the fourth TM, was localized to peroxisomes, while ⌬N186GFP with two TMs, but lacking the fourth TM, was localized to not only peroxisomes but also another organelles. These results suggest two possibilities for integration into peroxisomal membrane. One is that fourth TM is important for integration into peroxisomal membranes. The other is that at least three TMs are required. To address this issue, two mutants, 86/204GFP and 125/273GFP, containing the fourth TM plus one TM, were constructed (Fig.  5A). Their intracellular localization is shown in Fig. 5B. Both 86/204GFP and 125/273GFP were not localized to peroxisomes (Fig. 5B, a and e). It is of interest to note that 86/204GFP was instead transported to mitochondria, as verified by staining using anti-malate dehydrogenase antibody (Fig. 5B, b). These results suggest that at least three TMs, not the fourth TM, are required for integration into peroxisomal membranes.
To ask if three TMs of PMP34 are also required in the N-terminal portion of the fourth loop, we expressed 30/204GFP carrying three TMs, TM2-TM4, and analyzed its intracellular localization. The 30/204GFP was detected mostly in particles (Fig. 5B, c). In the GFP-positive cells, punctate structures were only partly superimposable with those stained using Pex14p antibody (Fig. 5B, c and d). These results suggest that the presence of simply three TMs upstream of the loop is not sufficient for localization, rather implying that TMs from the first to the third or the fourth to the sixth are required for integration to peroxisomal membranes. To confirm this results, a fusion protein 86/273GFP containing the third to fifth TMs of PMP34 was expressed in CHO-K1. Localization of 86/273GFP was not in peroxisomes (Fig. 5B, f). A chimera 86/230GFP, where GFP was fused to immediately downstream of TM5, showed very similar GFP fluorescence pattern as 86/273GFP (data not shown). Taken together, it is most likely that the fourth loop and three TMs, comprising either TM1-TM3 or TM4 -TM6, coordinately function as peroxisomal membranetargeting and integration information.

DISCUSSION
To elucidate the molecular mechanisms involved in topogenesis of PMPs, we selected PMP34 as a model PMP in the present study. We first determined the transmembrane topology of PMP34 by expressing the full-length human PMP34 in HeLa cells as well as epitope-tagged PMP34 in CHO-K1 cells. From several lines of morphological assessment using a combination of ectopic expression and the differential permeabilization method, we concluded that PMP34 exposes both N-and C-terminal portions to the cytoplasm. The orientation of PMP34 contradicts that of PMP47, a potential yeast ortholog of PMP34, of which both terminal regions reside in the matrix side of peroxisomes (12). One possible interpretation of these observations would be that PMP34 is not an ortholog of Cb-PMP47, at least with respect to the transmembrane topology; another one is that it is a homolog but the membrane topology is distinct. Otherwise, PMP34 is closely related to PMP47 but a distinct protein. Physiological function of PMP34 has not been elucidated, while PMP47 was recently suggested to be a transporter of ATP required in activating middle-chain fatty acids with CoA (32). We noted that PMP34 contains in the first loop region a highly conserved sequence, PX(D/E)XX(K/R)(X 20 -30)(D/E)G(X 4 )X(K/R)G, found in the extramembrane loop of mitochondrial carrier proteins (33,34). PMP34 may also be a transporter like others, such as a Ca 2ϩ -dependent peroxisomal transporter exposing both terminal regions to the cytoplasm (35).
We demonstrated that the hydrophilic loop between TM4 and TM5 of PMP34, similar to that of PMP47, is required for transport to peroxisomes but is not sufficient for integration. Instead, this basic loop plus at least three TMs, such as TM1-TM3, are likely to be responsible for the proper localization of PMP34. In the assay system we used, it was very difficult to discriminate between the activities of targeting and integration. We concluded, however, that the TM4-TM5 loop domain functions as mPTS, based on the following observations. 1) A PMP34 mutant 186HA, including TMs 1-4 but devoid of the fourth loop domain, was not transported to peroxisomes; 2) a mutant PMP34-GFP fusion, 6AGFP, in which all of the six basic amino acids were substituted by Ala, was completely mislocalized to the cytosol; 3) the fourth loop domain faces to the cytosol; but it is unlikely that this loop can translocate through the membrane into the matrix side and finally retranslocate back to the cytoplasmic face. Together, these findings strongly suggest that the fourth loop domain functions as an mPTS.
The fourth loop region of PMP47 is enriched in positively charged amino acids and functions as a mPTS (12). The region responsible for peroxisomal targeting of other membrane proteins, including Pex3p and PMP70, has been searched for. N-terminal region, residues at positions 1-40, of the membrane peroxin Pex3p was shown to be necessary and sufficient for peroxisomal targeting in mammals (13)(14)(15) and yeast, Pichia pastoris (16) and Hansenula polymorpha (17). The highly conserved residues at 9 -15, LKRHKKK, of human and H. polymorpha Pex3p was recently suggested as an mPTS (14,17). This sequence is very similar to that of the basic loop in CbPMP47 (12), thereby suggesting that the mechanisms of targeting Pex3p and PMP47 may resemble. Imanaka et al. (10) demonstrated that the N-terminal sequence encompassing one third of the full-length rat PMP70 was necessary for in vitro targeting and integration. In this regard, Sackesteder et al. (36) very recently showed the N-terminal residues at 1-124 of human PMP70 are targeted to peroxisomes in vivo. It is also noteworthy that positively charged amino acid residues are noted immediately downstream of the TM1 (residues at 28 -42) as well as upstream of the TM3 (residues at 117-124).
Between PMP34, PMP47, PMP70, and Pex3p, the cluster enriched in relatively positively charged amino acids which located flanking region of hydrophobic segment is a common feature. However, no conserved amino acid sequence is observed. In our Ala-scanning analysis of the basic loop of PMP34, the peroxisomal targeting activity decreased as the number of mutated residues in the basic amino acids increased. A mutant PMP34-GFP fusion, 4AGFP, in which four of the six basic amino acids were substituted by Ala, was partially localized to peroxisomes. Another mutant, 204Rev-GFP, where five amino acid residues at 195-199 were reversed, was localized only to peroxisomes (data not shown). Thus, the fourth basic loop domain is most likely to function as mPTS, despite there being apparently no need to specify the positively charged amino acids. Whether such basic amino acids are prerequisite for targeting of other PMPs remains to be determined. Similarly, a conformational requirement is postulated for the mPTS of PMP47 (12). In contrast to PMP47, the fourth loop region of PMP34 was predicted to form an ␣-helix structure (data not shown). Moreover, Ala is not a helix breaker. Therefore, it is most likely that the positively charged loop works as mPTS.
We conclude from the following observations that three TMs 1-3 or 4 -6 are essential for integration of PMP34 to peroxisomes. 1) PMP34 mutants with only two TMs were not localized to peroxisomes; 2) fusion proteins, 204GFP and ⌬N125GFP, were specifically localized to peroxisomes as efficiently as the full-length PMP34-GFP, although 86/273GFP and 86/230GFP, both despite with three TMs 3-5, were not localized to peroxisomes; 3) 30/204GFP with deletion of TM1 of TMs 1-4, or ⌬N186GFP likewise deleted in TM4 of TMs 4 -6 was significantly reduced in the level of peroxisomal localization. Thus, we infer that three TMs are necessary to integrate into peroxisome membrane. This conclusion was confirmed by Ala substitution mutants. Since Ala substitution mutants were integrated into peroxisomal membrane in the same manner as wild-type PMP34, although their targeting activity is weak, the implication is that integration of PMP34 is dependent on its transmembrane segment but does not depend on the fourth loop domain. It is of interest to note that the fourth basic loop domain fused to GFP was transported to mitochondria, presumably recognized as a mitochondrial targeting signal. Hence, three TMs are more likely to play a role in allowing the basic loop to be readily recognized by a putative mPTS recognition factor (see Fig. 6, X) and not by a mitochondrial signal receptor, as was seen in the case of 204GFP and ⌬N125GFP specifically localized to peroxisomes.
Several peroxisomal proteins have been suggested to be glycosylated, including N-glycosylated Pex2p and Pex16p of Y. lipolytica (11). We investigated this issue extensively using mammalian cells. Mutation of two potential N-glycosylation sites, including N 167 GT to DGT or N 241 RT to DRT did not alter the mobility of PMP34 in SDS-PAGE (data not shown), suggesting that PMP34 is not N-glycosylated. Moreover, we observed no staining of endoplasmic reticulum, using anti-PMP34 antibody, in CHO-K1 cell overexpressing PMP34 (data not shown). These results imply that endoplasmic reticulum is less likely to be involved in biogenesis of PMP34.
Given the findings described here, we propose a hypothetical model of the topogenesis of PMP34 (Fig. 6). After the synthesis on cytoplasmic polysomes, PMP34 is transported to peroxisomes in an mPTS-dependent manner, and inserted into membranes with the aid of at least three hydrophobic TMs, such as TMs 1-3 and 4 -6. Each of three TMs may form a targetingcompetent conformation, a module-like structure, that enables PMP34 to be recognized by a cytosolic factor (Fig. 6, X) and/or to be readily integrated. The complex then binds to a putative membrane protein import receptor, and PMP34 is finally localized into peroxisome membranes. The lower hydrophobicity of the TMs of PMP34, as compared with other PMPs, such as Pex3p, may require three TMs to maintain such conformation. Similar mechanisms are involved in topogenesis of the ADP/ ATP carrier in mitochondrial inner membrane (37). This topogenesis mechanism is compatible with two previous observations on the in vitro import of peroxisomal membrane proteins, PMP22 (9) and PMP70 (10). The peroxin Pex19p is a partially farnesylated, acidic peroxin responsible for peroxisome biogenesis disorders of complementation group J (38). Pex19p shows a dual intracellular localization, on peroxisome membranes and in the cytosol. It is noteworthy that Pex19p was very FIG. 6. A schematic view of PMP34 topogenesis. After synthesis on cytoplasmic free polyribosomes, PMP34 is recognized by a soluble factor such as putative "mPTS receptor" (designated by X) at the loop region between TM4 and TM5, transported to the surface of peroxisome membrane, imported by a "membrane protein import receptor," then finally localized to the membrane by vectorial translocation. At least three TMs are also responsible for the targeting and/or integration. recently reported to weakly interact with PMP34 (36). The region containing the basic loop of PMP34 may mediate binding to Pex19p. Other cytosolic factors may also be involved in the transport of PMP34.