Targeting of the tail-anchored peroxisomal membrane proteins PEX26 and PEX15 occurs through C-terminal PEX19-binding sites

Tail-anchored (TA) proteins are type II membrane proteins that harbor a single transmembrane domain (TMD) close to their C-termini and are thus equipped with a characteristic short luminal domain. Members of this class of transmembrane proteins are found in the endoplasmic reticulum (ER) and its related endomembranes, the outer mitochondrial membrane, peroxisomes and in plants also in the outer membrane of plastids (Borgese et al., 2003). The methods by which TA proteins are correctly targeted to the ER and the mitochondrial outer membrane, the principal sites of membrane integration, has attracted significant attention since it became clear that the posttranslational import of these proteins must occur by novel routes (High and Abell, 2004; Steel et al., 2002; Yabal et al., 2003). Machineries that would enable the insertion of TA proteins into the target membranes remain elusive and the question as to whether membrane insertion requires proteinaceous membrane components such as the signal recognition particle (SRP) is currently disputed (Abell et al., 2004; Brambillasca et al., 2005; Steel et al., 2002).

The targeting signals of TA proteins are generally located within their C-termini and include the TMD. From several studies addressing the differential targeting of TA proteins to mitochondria and the ER, a concept emerged that explains targeting specificity by general physicochemical rather than sequence-specific features of the targeting signal. These include the length and hydrophobicity of the TMD and the net charges of its flanking residues. A mitochondrial targeting signal, for instance, exhibits a rather short TMD with only moderate hydrophobicity and flanking regions that are positively charged (Beilharz et al., 2003; Borgese et al., 2003; Hwang et al., 2004; Rapaport, 2003). Insertion of a tail anchor into the ER membrane is thought to occur by default, i.e. in the absence of a mitochondrial (or plastid) targeting signal (Borgese et al., 2003), although SRP might support guidance of at least some TA proteins (Abell et al., 2004). A number of TA proteins are initially targeted to the ER and subsequently sorted to their final destination within the endomembrane system such as the Golgi or the nuclear envelope through additional motifs that are located within their cytosolic domain (Beilharz et al., 2003). TA proteins destined for the peroxisomal membrane were also reported to use a bipartite system of targeting signals; an ER targeting signal to bring the protein to the ER, and a peroxisomal signal, which delivers the protein from the ER to the peroxisome (Elgersma et al., 1997; Mullen et al., 1999; Mullen and Trelease, 2000; Nito et al., 2001). In this case, however, the tail-anchor itself rather than the cytosolic domain contains the peroxisome-specific targeting information.

Although PEX3 (Hoepfner et al., 2005) and probably a few other peroxisomal membrane proteins (PMPs) also first target to the ER (Geuze et al., 2003; Titorenko and Rachubinski, 1998), most PMPs are likely to be inserted into peroxisomes directly from the cytosol (Lazarow and Fujiki, 1985). This latter process is facilitated by PEX19, a predominantly cytosolic protein that is also found at the peroxisomal membrane (Götte et al., 1998; Matsuzono et al., 1999). PEX19 interacts with virtually all PMPs (Fransen et al., 2001; Sacksteder et al., 2000; Snyder et al., 2000) through a common motif that is present at least once (Halbach et al., 2005; Rottensteiner et al., 2004). This interaction on the one hand prevents the PMP from aggregation before membrane insertion and on the other serves to guide the PMP to the peroxisomal membrane (Jones et al., 2004; Shibata et al., 2004). PEX19-binding sites are sufficient for peroxisomal targeting provided that one or more TMDs are additionally present to anchor the fragment within the membrane (Rottensteiner et al., 2004). A distinct mode of PEX19 interaction applies for PEX3 (Fransen et al., 2005; Jones et al., 2004; Mayerhofer et al., 2002), consistent with the proposed function of PEX3 as the docking factor for PEX19 at the peroxisomal membrane (Jones et al., 2004; Muntau et al., 2003). Thus, current knowledge suggests the presence of at least two distinct import pathways for PMPs: a PEX19-dependent one and one that involves the ER (Heiland and Erdmann, 2005).

The mammalian PEX26 belongs to the class of peroxisomal TA proteins. Despite the lack of sequence similarity, PEX26 might represent the functional orthologue of Pex15p in that both proteins recruit PEX6 to the peroxisomal membrane (Birschmann et al., 2003; Matsumoto et al., 2003a). Patients carrying loss-of-function alleles of PEX26 suffer from the typical symptoms ascribed to the Zellweger spectrum. On a molecular level, the absence of PEX26 causes an import defect of peroxisomal matrix proteins of type II (Matsumoto et al., 2003b) and to a large extent also of type I (Weller et al., 2005).

Here we have studied the targeting of the mammalian TA protein PEX26 and analyzed whether this occurs through the PEX19-dependent pathway. We show that the luminal domain of PEX26 contained a typical PEX19-binding site that was capable of peroxisomal targeting. Based on these observations, we re-examined targeting of yeast Pex15p and found a similar functional composition of its targeting signal. We discuss our findings in terms of peroxisomal TA protein sorting that is enabled through an appended C-terminal PEX19-binding site and that does not necessarily require an interim transit through the ER.

PEX26 from Homo sapiens is a tail-anchored protein of the peroxisomal membrane. It has been proposed that this class of PMPs might be targeted to peroxisomes by a pathway that involves specialized regions of the ER (Elgersma et al., 1997; Mullen and Trelease, 2000). Most PMPs, however are inserted into the peroxisomal membrane directly from the cytosol. This process requires PEX19, which interacts with all PMPs tested so far. Interestingly, by means of our developed prediction program for PEX19-binding sites (Rottensteiner et al., 2004), we were able to identify two potential PEX19-binding sites also in PEX26 (Halbach et al., 2005). The predicted scores for the entire PEX26 are shown in Fig. 1A. One of the two peak-scoring peptides is located within the predicted single TMD comprising amino acids 253-267 (Matsumoto et al., 2003a) and the other one (amino acids 276-290) C-terminally adjoins it (see Fig. 1D for a graphic representation of PEX26). To test whether the predicted sites are indeed capable of binding PEX19, we synthesized synthetic 15-mer peptides covering the entire PEX26 protein in an overlapping arrangement on a membrane. This membrane was then incubated with a purified GST-fusion protein of human PEX19. As control, a duplicate membrane was incubated with GST alone. Bound PEX19 was visualized immunologically with a combination of monoclonal anti-GST antibodies and appropriate HRP-conjugated secondary antibodies. Strikingly, two arrays of spots were recognized by GST-PEX19 (Fig. 1B), whereas the GST control did not bind significantly to any of the spots on the membrane (not shown). The serial spots covered amino acids 251-269 and 277-297, respectively, demonstrating the existence of two neighboring PEX19-binding sites in the C-terminus of PEX26. Noticeably, the experimentally determined PEX19-binding sites in PEX26 comprise peptides that lie within the highest-scoring peaks in the respective regions (BS I, 248-270; BS II, 276-296, see Fig. 1A), and thus reflect the conciseness of the prediction made. Quantification of bound PEX19 revealed that the individual peptides from the second binding site showed an interaction that was weaker by a factor of about three to four (Fig. 1C). Since four of these peptides interacted comparably, whereas only three did so for the first binding site, it appears that the second PEX19-binding site is slightly larger than the first. It is also worth noting that the first PEX19-binding site was devoid of the typical basic amino acids within its core region.

Binding of PEX19 to PEX26 was also analyzed in vivo by using a yeast two-hybrid assay. Full-length PEX26 was fused to the GAL4 DNA-binding domain (BD) and expressed together with a GAL4 activation domain (AD) fusion of HsPEX19. The transformed strain grew clearly on histidine-adenine double-dropout plates, whereas coexpression of Gal4-BD-PEX26 with the GAL4-AD alone did not lead to growth (Fig. 2). This result showed that PEX19 also interacted with PEX26 in vivo, as reported previously (Fransen et al., 2005). We then analyzed two fragments of PEX26 for interaction with PEX19; one that comprised the large cytosolic domain (amino acids 2-244) and one that included both PEX19-binding sites (amino acids 245-305). Fig. 2 shows that only the C-terminal fragment was tested positive in the assay. Further dissection of the C-terminus into the two individual binding sites revealed that the more C-terminal, luminal PEX19-binding site (PEX26_275-305) did, albeit weakly, interact with PEX19, whereas the neighboring, TMD-containing binding site (PEX26_245-274) did not. These data suggest that in vivo PEX19 might bind cooperatively to the two sites in PEX26, although it is also possible that solubility or folding problems, particularly of the fragment harboring the TMD, have hampered the binding efficiency. Nonetheless, the two-hybrid data correlated well with the result of the peptide scan.

We then analyzed whether the PEX19-binding sites of PEX26 are involved in the topogenesis of this tail-anchored PMP. Expression in human fibroblasts of PEX26 that had been fused to the C-terminus of GFP resulted in a punctate staining pattern that coincided with that of the peroxisomal integral membrane marker protein PEX14 (Fig. 3A), thereby demonstrating the peroxisomal localization of the GFP-PEX26 fusion protein. When PEX26_2-244, a fragment that lacked both PEX19-binding sites, was fused to GFP, a diffuse fluorescent staining pattern was obtained, indicating that this fusion protein was localized to the cytosol. By contrast, the PEX26_245-305 fragment devoid of the entire cytosolic domain but containing both PEX19-binding sites and the TMD predominantly localized to peroxisomes (Fig. 3A). Thus, the peroxisomal membrane protein-targeting signal (mPTS) of PEX26 was narrowed down to a fragment that contains both PEX19-binding sites as well as its TMD. When the luminal PEX19-binding site was clipped from PEX26, the truncated protein (PEX26_2-274) was no longer peroxisomal but was localized to internal membranes instead. The observed mistargeting of PEX26_2-274 suggests that the luminal PEX19-binding site contains peroxisomal targeting information. To prove this assumption, this binding site was examined for its ability to direct an otherwise nonperoxisomal fragment to peroxisomes. For this, a fragment of the mPTS of the adrenoleukodystrophy protein (ALDP_87-164) that contains two TMDs but lacks its PEX19-binding site was chosen (Halbach et al., 2005; Landgraf et al., 2003). Expression of a fusion of the luminal PEX19-binding site of PEX26 with this ALDP fragment revealed a clear peroxisomal localization of the chimera, whereas the PEX19-binding site (PEX26_275-305) alone was almost entirely localized to nonperoxisomal structures (Fig. 3B), only in a few cells were peroxisomes also discernible (not shown). Thus, PEX26_275-305 provoked a redirection of ALDP_87-164 to peroxisomes in human cells.

The fluorescence studies left unclear whether this chimera, as well as the identified mPTS fragment of PEX26, are only targeted to peroxisomes or are also inserted into the membrane. Cells expressing GFP fusions of the respective protein fragments or, as control, full-length PEX26 were therefore lysed in hypotonic buffer and membranes were separated from the soluble proteins by centrifugation at 100,000 g. The resulting membrane pellet was treated with alkaline sodium carbonate buffer to extract all proteins barring integral membrane proteins. Finally, extracted proteins were separated from the membranes by centrifugation at 100,000 g. Immunoblot analysis of samples of each fraction revealed that the soluble peroxisomal matrix enzyme catalase was released after hypotonic lysis in all cells, whereas the integral PMP PEX14 was exclusively found in the membrane fraction (Fig. 4). Likewise, PEX26_245-305, PEX26_275-305-ALDP_87-164, as well as full-length PEX26 were also resistant to carbonate extraction. It was therefore concluded that the analyzed PEX19 binding site-containing fragments behaved like integral membrane proteins that were indeed sufficient for peroxisomal membrane insertion.

So far, the results showed that the luminal PEX19-binding site of PEX26 represents a bona fide PMP-targeting motif. At the same time, the question of the role of the other PEX19-binding site in PEX26 arose. Since PEX26_2-274 was mislocalized, it was conceivable that binding of PEX19 to the TMD-containing PEX26_245-274 might be solely required to protect the protein from aggregation in the cytosol. To determine, nonetheless, whether this PEX19-binding site can in principle also serve as a targeting motif, the ability of PEX26_245-274 to target GFP to peroxisomes was analyzed. This was indeed the case, even though significant cytosolic fluorescence was also visible (Fig. 3C). PEX26_245-274 can thus be regarded as a minimal mPTS that is composed of a PEX19-binding site coinciding with a TMD. In the physiological context of the full-length protein, however, the luminal PEX19-binding site is additionally required for the correct targeting of PEX26.

Since peroxisomal TA proteins were suggested to harbor both a peroxisomal- as well as an ER-targeting signal (Borgese et al., 2003; Elgersma et al., 1997), we also determined to which organelle the nonperoxisomal PEX26 fragments are targeted. Mislocalized PEX26_2-274 was targeted to mitochondria as evidenced by a staining pattern that was superimposable with that of the mitochondrial matrix protein TRAP1 (Fig. 5A). In addition, the mPTS fragment PEX26_245-305 was detectable in mitochondria to some extent, whereas full-length PEX26 was exclusively peroxisomal in most cells. It should be noted, however, that in some cells which exhibit a very strong GFP-derived fluorescence, even full-length PEX26 was localized to mitochondria. The fluorescence pattern of PEX26_275-305 did not coincide with the mitochondrial marker, but revealed some overlap with the ER, visualized with an antibody directed against calreticulin (Fig. 5B). None of the other PEX26 fragments revealed any congruent staining with the ER (not shown).

The combined results strongly indicate that the TA protein PEX26 follows the PEX19-dependent targeting route that is used by most PMPs. To underscore this conclusion, the subcellular localization of GFP-PEX26 was studied in a PEX19-deficient cell line. Since these cells lack morphologically detectable peroxisomes, PEX26 should accumulate in the ER if it indeed had to pass the ER en route to peroxisomes. However, colocalization studies with the mitochondrial marker TRAP1 clearly demonstrated that in the absence of PEX19, PEX26 was localized to mitochondria (Fig. 6A). The same fate was also reported to be met by other PMPs including PEX14, PEX12 or ALDP (Sacksteder et al., 2000). Thus, either the removal of the luminal PEX19-binding site or the absence of PEX19 resulted in a mitochondrial localization of PEX26. The results obtained are therefore consistent with the hypothesis that the luminal PEX19-binding site prevents PEX26 from mitochondrial mistargeting.

To determine whether the remote possibility that the targeting motif of PEX26, which does bind PEX19, remains functional in the absence of PEX19 as long as peroxisomes are present, the subcellular location of PEX26 was analyzed after the expression of PEX19 had been transiently knocked down by siRNA. Transfection of human fibroblasts with a published PEX19 duplex siRNA (Jones et al., 2004) led to a drastic decrease in the protein concentration of PEX19 at day two after transfection and remained low until day five, whereas levels of the PMP PEX14 were hardly affected throughout the duration of the experiment (Fig. 6B, lower panel). A control knock down with PEX5-specific siRNA (Jones et al., 2004) that selectively inhibits peroxisomal matrix protein import did not lead to a significant decrease in protein levels of both PEX19 and PEX14 (Fig. 6B, upper panel).

Sixty hours after the siRNA treatment, cells were transfected with a GFP-PEX26-expressing plasmid. After another 24 hours, localization of the newly synthesized fusion protein was inspected and compared with the fluorescence pattern of endogenous PEX14. Visualization of PEX14 revealed that peroxisomes persisted even after 4 days of incubation with the PEX19-specific siRNA in most cells, which was consistent with a previous observation (Jones et al., 2004). Cells with a normal appearance of peroxisomes were then scrutinized for the localization of PEX26. About 60% of the cells showed at least a partial mislocalization of PEX26 to mitochondria as exemplified in Fig. 6C,D, whereas in the PEX5 siRNA control, this was the case for only 35% of the cells. The reason for the large number of control cells with mislocalized PEX26 needs to be addressed in future work but can in part be explained by the fact that strong overexpression of PEX26 resulted in its targeting to mitochondria even in untreated cells. Nonetheless, the significant increase of mislocalized PEX26 in the PEX19 siRNA-treated cells adds further support for the function of PEX19 as a soluble import receptor for PEX26.

We then analyzed whether the targeting signal of Pex15p, the only known tail-anchored PMP of the yeast S. cerevisiae, also contains PEX19-binding sites. The mPTS of Pex15p has been localized to its C-terminal 80 amino acids, which includes the TMD and its luminal domain (Elgersma et al., 1997). Since deletion of the latter domain (30 amino acids) resulted in a mislocalization of Pex15p to the ER, the authors concluded that amino acids 302-353 of Pex15p contain an ER-targeting signal, and that the luminal tail (amino acids 353-383) directs Pex15p from the ER to peroxisomes (Fig. 7A). Our prediction for PEX19-binding sites within the mPTS of Pex15p suggested the existence of two binding sites that are distributed similarly to PEX26: one of them comprises the C-terminal 15 amino acids, the second predicted site roughly coincides with the putative TMD (Fig. 7B). Yeast Pex19p indeed interacted with a fragment of Pex15p that contained both predicted binding sites (Pex15p_315-383) in a yeast two-hybrid assay, whereas the large cytosolic domain (Pex15p_2-314) did not (Fig. 7C). The luminal tail of Pex15p alone (amino acids 350-383) interacted with Pex19p even more strongly, probably as a result of the absence of the TMD in this construct. The fragment Pex15p_315-358, which only contained the PEX19-binding site proposed to coincide with the TMD, did not give rise to a positive signal when tested for interaction with Pex19p (Fig. 7C). Thus, the two-hybrid assay did not clarify whether the TMD would bind Pex19p, but did confirm the presence of a PEX19-binding site within the luminal domain of Pex15p.

To address the role of the luminal PEX19-binding site in targeting of Pex15p, a fragment including this site as well as the TMD (amino acids 315-383) was fused to GFP. The fragment, which was slightly shorter than the previously published mPTS of Pex15p (Elgersma et al., 1997), gave rise to a fluorescence pattern that was superimposable with that of the peroxisomal marker protein PTS2-DsRed (Fig. 8A), indicating that Pex15p_315-383 was correctly targeted to peroxisomes. As expected for a fragment lacking a TMD, the C-terminal PEX19-binding site alone (amino acids 361-383) localized to the cytosol and to nonperoxisomal structures (Fig. 8B). However, a chimera composed of this binding site and the nonperoxisomal ALDP_87-164 fragment regained the ability to target to peroxisomes (Fig. 8C,D). Likewise, the C-terminal PEX19-binding site of PEX26 also redirected this ALDP fragment to yeast peroxisomes (Fig. 8E). Thus, the luminal PEX19-binding site of both TA proteins Pex15p and PEX26 function as peroxisomal targeting motifs in yeast.

Finally, we analyzed whether Fis1p (Mozdy et al., 2000), a TA protein of the outer mitochondrial membrane, is also present in yeast peroxisomes, because the apparent mammalian orthologue was recently reported to be dually localized to mitochondria and peroxisomes (Koch et al., 2005). A GFP-Fis1p fusion colocalized with the synthetic mitochondrial marker PrF₀ATP9-DsRed, but only very rarely, if at all, with peroxisomal PTS2-DsRed (Fig. 9), indicating that yeast Fis1p is probably not a peroxisomal TA protein. Nonetheless, because mitochondrial TA proteins are necessarily inserted directly from the cytosol, we asked whether the luminal PEX19-binding site of Pex15p would be able to redirect this mitochondrial TA protein to peroxisomes. Appending the luminal PEX19-binding site of Pex15p indeed provoked an almost exclusive localization of Fis1p to peroxisomes (Fig. 9). The PEX19-binding site was therefore able to enforce peroxisomal targeting of an otherwise predominantly mitochondrial protein.

Our study on the targeting of two peroxisomal TA proteins, human PEX26 and yeast Pex15p, provides unequivocal evidence for a protein-based sorting machinery for TA proteins. The mPTS of PEX26 turned out to comprise its C-terminal 61 amino acids, consistent with the fact that targeting signals for TA proteins are generally located at the C-termini and include the single TMD. Noticeably, however, whereas the luminal tails of most TA proteins are usually very short, that of PEX26 comprises 36 amino acids. We could show that this luminal tail of PEX26 harbors a typical peroxisomal targeting motif. PEX26_275-305 was not only necessary for the correct targeting of PEX26 but also sufficient to redirect an otherwise nonperoxisomal ALDP fragment to peroxisomes. This chimera as well as the mPTS of PEX26 also correctly inserted into the peroxisomal membrane, because both fragments proved to be resistant to sodium carbonate extraction. Importantly, the sequence of the luminal domain of PEX26 was predicted to harbor a PEX19-binding site, which we could indeed demonstrate both in vivo and in vitro. The targeting signal of yeast Pex15p also contained a PEX19-binding site within its luminal tail. Alone, this PEX19-binding site failed to target GFP to peroxisomes, in line with the previous observation that the entire luminal domain of Pex15p (amino acids 350-383) is not sufficient for targeting (Elgersma et al., 1997). However, the C-terminal 23 amino acids of Pex15p provoked a redirection to peroxisomes of the human ALDP fragment and of the mitochondrial TA protein Fis1p (Mozdy et al., 2000).

These results corroborate the requirement for a TMD to anchor a PMP in the peroxisomal membrane. More importantly, they show that the luminal tails of PEX26 and Pex15p contain targeting motifs that remain functional in a heterologous context. Both ALDP as a typical PMP and Fis1p as a mitochondrial TA protein are thought to be inserted into the target membrane directly from the cytosol (Lazarow and Fujiki, 1985; Purdue and Lazarow, 2001). It therefore also follows that the motifs are not dependent on a precedent integration of the target protein into the ER membrane.

The current model for PEX19 function stipulates that PEX19 acts as a soluble import receptor for a number of PMPs including ALDP (Halbach et al., 2005; Jones et al., 2004). PEX19 thus recognizes its substrate in the cytosol and delivers it to the peroxisomal membrane, where it binds to its docking factor, PEX3 (Fang et al., 2004; Götte et al., 1998; Muntau et al., 2003). Our data can therefore be reconciled with a model in which the luminal domains of peroxisomal TA proteins are immediately recognized by PEX19 in the cytosol upon which the proteins are directly delivered to peroxisomes. In line with this concept, transient inhibition of PEX19 by siRNA resulted in a significant increase in newly synthesized PEX26 that was targeted to mitochondria. It should be noted, however, that because of the fact that PEX19 siRNA-treated cells often showed a dual distribution of PEX26 to peroxisomes and mitochondria, this latter experiment might also be interpreted to suggest that PEX19 is not essential for peroxisomal targeting of PEX26, but considerably increases the specificity of that process. This would be achieved by competition of PEX19 with other cytosolic chaperones to bind to the luminal tail of PEX26. However, because this fragment proved capable of redirecting a nonperoxisomal ALDP fragment to peroxisomes, a direct role for PEX19 in targeting appears more likely. That PEX26 is a typical PMP binding partner for PEX19 is also supported by the fact that its interaction depends on the same residues of PEX19 that are required for the interaction with multiple other PMPs (Fransen et al., 2005). Moreover, recent results from Matsuzono and Fujiki (Matsuzono and Fujiki, 2005) that are complementary to our approach revealed evidence for a PEX19-dependent in vitro insertion of PEX26 into isolated peroxisomes.

Peroxisomal TA proteins are anticipated to traverse the ER en route to peroxisomes. Support for this assumption was provided by an in vitro import experiment that suggested preferential import of Pex15p into ER versus peroxisomal membranes (Mullen et al., 1999). Furthermore, overexpressed Pex15p appeared to be O-glycosylated, accumulated in the ER and caused a massive proliferation of the compartment (Elgersma et al., 1997). However, it is also possible that in both cases the artificial excess of Pex15p prevented insertion into the correct target membrane because the amount of PEX19 had become limiting. Another argument in favour of a bipartite targeting signal was the observed ER localization of a Pex15p devoid of its luminal domain (Elgersma et al., 1997). We found here that a similarly truncated PEX26 was also mistargeted, but to mitochondria. Moreover, in the absence of PEX19, even full-length PEX26 inserted into the mitochondrial membrane. Thus, it seems more likely that binding of the luminal domains of peroxisomal TA proteins by PEX19 prevents their insertion into spurious membranes. The difference in targeting of the two truncated PMPs is probably due to the different extreme C-terminal residues that emerged from clipping the PEX19-binding site. A number of studies have shown that modifying a few C-terminal amino acids of TA proteins can suffice to alter their subcellular localization (Borgese et al., 2003; Hwang et al., 2004).

Recent work by Hoepfner et al. (Hoepfner et al., 2005) showed that some PEX19 is also localized to distinct foci emanating from the ER. These foci contain PEX3 and are thought to represent peroxisomes early in their development. It is therefore entirely possible that TA proteins are targeted to these PEX3-enriched ER subdomains rather than to mature peroxisomes, but this would not change our concept of a direct and PEX19-dependent delivery of peroxisomal TA proteins to peroxisomes. In fact, if peroxisomal TA proteins were first targeted to the ER by a PEX19-independent mechanism, then PEX19 would be able to interact with its cognate luminal site only from within the ER, which would require translocation of PEX19 across the ER membrane.

We also detected a second PEX19-binding site within the C-terminal region of PEX26. Binding of PEX19 to this second binding site, which coincides with the TMD of PEX26, could be demonstrated in vitro by using a PEX26 peptide scan. Interestingly, a GFP fusion of this binding site (PEX26_245-274) was clearly detectable also in peroxisomes, thereby proving function of this second PEX19-binding site as a peroxisomal targeting motif. That the TMDs of a TA protein can possess sequence-specific sorting information is not without precedence, as has been demonstrated recently for a plant mitochondrial cytochrome b5 isoform (Hwang et al., 2004). On the other hand, the site was not sufficient for targeting in the presence of the N-terminal cytosolic domain of PEX26. It might be that in this construct (PEX26_2-274) the binding site is not accessible for PEX19 and that appending the luminal PEX19 binding site is required to achieve an efficient association with PEX19. It should be noted that our studies left open whether binding of PEX19 to the second site is essential for targeting of full-length PEX26, because this site overlaps with the TMD, which by itself is an essential part of the targeting signal. This issue could be resolved in future work by exchanging the TMD of PEX26 for one that does not bind PEX19.

In summary, we show in this work that peroxisomal TA proteins are targeted to peroxisomes in a PEX19-dependent manner and that targeting specificity is achieved by the presence of a C-terminal PEX19-binding site that largely prevents insertion into other intracellular membranes. Thus, our data lend credence to the existence of protein-based targeting machineries for TA proteins. Since the PEX19-binding site of PEX26 and Pex15p functioned as typical targeting motifs that insert PMPs directly from the cytosol, we also suggest that topogenesis of peroxisomal TA protein does not normally require an additional ER-specific targeting signal.

Escherichia coli strain DH5α was used for all plasmid amplifications and isolations. E. coli strain BL21(DE3) (Merck, Darmstadt, Germany) was used for heterologous expression of recombinant GST-HsPEX19. Two-hybrid assays were performed with Saccharomyces cerevisiae strain PJ69-4A (P. James, Madison, USA). The S. cerevisiae wild-type strain UTL7-A and its derivative yHPR251 harboring an integrated copy of a PTS2-DsRed construct were used for the expression of enhanced green fluorescent protein (GFP) fusions. Standard media for the cultivation of yeast and bacterial strains were prepared as described (Sambrook et al., 1989).

PEX26, the adrenoleukodystrophy protein (ALDP), and all fragments thereof were amplified from the commercially obtained cDNA clones IMAGp998J1311625Q3 and IRAKp961I0514Q2, respectively (RZDP, Berlin, Germany). FIS1 and all PEX15 fragments used in this study were amplified from genomic DNA of S. cerevisiae UTL-7A. The primer pairs used in these PCR reactions, as well as the vectors and restriction sites used for cloning the PCR fragments are listed in Table S1 in supplementary material. The sequences of the primers are shown in Table 2 in supplementary material. The sequences of all PCR-generated fragments of this study were verified by automated sequencing (MWG Biotech, Ebersberg, Germany).

Live yeast cells were analyzed for enhanced green fluorescent protein (GFP) and DsRed fluorescence as described (Rottensteiner et al., 2004). Human fibroblast cells were cultured at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100,000 U/l penicillin and 100 mg/l streptomycin at 8% CO₂. The human skin fibroblast cell line GM5756T was grown for 1 day in 12-well plates before it was transfected with the pEGFP-C1-derived plasmids. Transfection was performed with 0.5 μg plasmid DNA and 1.5 μl FUGENE 6 according to the manufacturer's instruction (Roche Diagnostics, Mannheim, Germany). The patient-derived PEX19-deficient skin fibroblasts (primary cell line RW/mf/0854872, kindly donated by Ron Wanders, AMC, Amsterdam, The Netherlands) were similarly transfected, but were grown for 3 days before transfection. 2 days after the transfection, cells were fixed on cover glasses with 3% formaldehyde in phosphate-buffered saline (PBS), permeabilized with 1% Triton X-100 in PBS and subjected to immunofluorescence microscopy using polyclonal rabbit anti-PEX14 antibodies (Will et al., 1999) in conjunction with Alexa Fluor 594-conjugated antibodies. Samples were also inspected for GFP fluorescence.

To visualize the ER and mitochondria, rabbit polyclonal anti-calreticulin and mouse monoclonal anti-TRAP1 antibodies (Affinity Bio Reagents, Golden, CO), respectively were used in combination with the appropriate Alexa Fluor 594- or Alexa Fluor 568-conjugated secondary antibodies. All micrographs were recorded on a Zeiss Axioplan 2 microscope with a Zeiss Plan-Apochromat 63×/1.4 oil objective and an Axiocam MR digital camera and were processed with AxioVision 4.2 software (Zeiss, Jena, Germany).

Human fibroblasts (GM5756T) were cultivated in T75 flasks, transfected as described above and incubated for two additional days. For whole cell lysates, cells were washed with Hank's buffered salts (BSS; PAA Laboratories, Pasching, Austria), treated with trypsin, resuspended in 5 ml Hank's BSS containing 1 mM PMSF and harvested by a 200 g centrifugation step. After washing the cells with PBS, pH 7.3, samples were resuspended in standard Laemmli SDS-PAGE buffer. For the membrane localization analysis, cells were disintegrated in hypotonic lysis buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, Complete™ protease inhibitor cocktail from Roche Diagnostics, Mannheim, Germany) and by passing the suspension through a 20G syringe needle five times. Membranes were separated from the soluble fraction by a 100,000 g ultracentrifugation step for 1 hour. Subsequently, membranes were resuspended in 100 mM Na₂CO₃ for 30 minutes at 4°C with gentle agitation and the extracted proteins were separated from membranes by repeating the ultracentrifugation step. Finally, aliquots of each fraction, adjusted to equal amounts of cells, were resuspended in standard Laemmli SDS-PAGE buffer. Samples were analyzed by immunoblotting for the distribution of endogenous PEX14, catalase, and the transiently expressed GFP fusions of PEX26 fragments. Monoclonal anti-GFP (JL-8) antibodies were obtained from BD Biosciences, Pharmingen, Germany; polyclonal antibodies against human catalase were purchased from The Binding Site, Schwetzingen, Germany.

The sequences of the PEX19 and PEX5-specific siRNA oligonucleotides were adopted from Jones et al. (Jones et al., 2004). The oligonucleotides were obtained as duplex RNA from Perbio/Dharmacon RNA Technologies, Bonn, Germany (PEX19: 5′-GAG AUC GCC AGG AGA CAC U dTdT-3′ and 5′-AGU GUC UCC UGG CGA UCU C dTdT-3′; PEX5: 5′-AGA AGC UAC UCC CAA AGG C dTdT-3′ and 5′-GCC UUU GGG AGU AGC UUC U dTdT-3′).

For analysis of siRNA-dependent repression of PEX19, human fibroblasts (GM5756T) were grown on coverslips in a 12-well format for 24 hours. Transfection with siRNA was done with 50 μl of duplex PEX19 siRNA (2 μM) and 2 μl Dharmafect 4 reagent (Perbio/Dharmacon) according to the manufacturer's instructions. Cells were cultivated for 5 days without addition of antibiotics and every 24 hours a sample was withdrawn and prepared for immunoblot analysis. Cells were washed with Hank's BSS, trypsinized, suspended in Hank's BSS plus 1 mM PMSF and collected by centrifugation. Cells were washed in PBS pH 7.3 and transferred in hypotonic lysis buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, Complete™ protease inhibitor cocktail, 0.5% Triton X-100). The protein concentration was determined by Coo Protein Assay Reagent (Uptima-Interchim, Montlucon, France). Equal amounts of protein were analyzed by immunoblotting using anti-PEX14 and monoclonal anti-PEX19 antibodies (BD Biosciences, Pharmingen, Germany).

To investigate PEX19-dependent peroxisomal import of PEX26, cells were transfected with pAH13 (GFP-PEX26_2-305) 60 hours after cells had been treated with siRNA. 24 hours after transfection, cells were prepared for immunofluorescence microscopy using anti-PEX14 or anti-TRAP1 antibodies as described above.

The following procedures were carried out essentially as described: prediction of PEX19-binding site and the purification of GST-HsPEX19 from E. coli (Halbach et al., 2005; Rottensteiner et al., 2004); the PEX19 in vitro-binding assay with peptide arrays on membrane blots (Landgraf et al., 2004); and the yeast two-hybrid assays (Rottensteiner et al., 2004).

We thank T. Schievelbusch for help with cell culture, R. Wanders for the PEX19-deficient fibroblast cell line, W.-H. Kunau for plasmids HH15/1 and HH15/3 and J. M. Shaw for plasmid pMitoRFP. Special thanks go to W. Schliebs for critical reading of the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft (grants SFB480 and RO 2494/1 to H.R., SFB449 to R.V.-E., and SFB642 to R.E.), by the FP6 European Union Project `Peroxisome' (LSHG-CT-2004-512018), and by the Fonds der Chemischen Industrie.