Peroxisomal Membrane Ascorbate Peroxidase Is Sorted to a Membranous Network That Resembles a Subdomain of the Endoplasmic Reticulum

Correspondence to: Richard N. Trelease, trelease.dick{at}asu.edu (E-mail), 480-965-6899 (fax)

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

The peroxisomal isoform of ascorbate peroxidase (APX) is a novel membrane isoform that functions in the regeneration of NAD⁺ and protection against toxic reactive oxygen species. The intracellular localization and sorting of peroxisomal APX were examined both in vivo and in vitro. Epitope-tagged peroxisomal APX, which was expressed transiently in tobacco BY-2 cells, localized to a reticular/circular network that resembled endoplasmic reticulum (ER; 3,3'-dihexyloxacarbocyanine iodide–stained membranes) and to peroxisomes. The reticular network did not colocalize with other organelle marker proteins, including three ER reticuloplasmins. However, in vitro, peroxisomal APX inserted post-translationally into the ER but not into other purified organelle membranes (including peroxisomal membranes). Insertion into the ER depended on the presence of molecular chaperones and ATP. These results suggest that regions of the ER serve as a possible intermediate in the sorting pathway of peroxisomal APX. Insight into this hypothesis was obtained from in vivo experiments with brefeldin A (BFA), a toxin that blocks vesicle-mediated protein export from ER. A transiently expressed chloramphenicol acetyltransferase–peroxisomal APX (CAT-pAPX) fusion protein accumulated only in the reticular/circular network in BFA-treated cells; after subsequent removal of BFA from these cells, the CAT-pAPX was distributed to preexisting peroxisomes. Thus, plant peroxisomal APX, a representative enzymatic peroxisomal membrane protein, is sorted to peroxisomes through an indirect pathway involving a preperoxisomal compartment with characteristics of a distinct subdomain of the ER, possibly a peroxisomal ER subdomain.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Peroxisomes are found in virtually all eukaryotic cells and are delineated by a single boundary membrane. These organelles typically are involved in the generation and degradation of toxic hydrogen peroxide and the ß oxidation of fatty acids; they also house a diversity of enzymes that participate in various other metabolic processes specific to the organism, cell/tissue type, or environmental conditions (Huang et al. 1983 ; Baker 1996 ; Gietl 1996 ).

Nuclear genes encode all of the protein and enzyme constituents of the peroxisomal matrix. After synthesis on polysomes in the cytosol, these proteins are targeted post-translationally to the organelle in a regulated manner. At least two types of evolutionarily conserved peroxisomal targeting signals (PTSs) are capable of directing proteins to the peroxisomal matrix (Olsen 1998 ; Subramani 1998 ). The type 1 PTS (PTS1) is an uncleaved C-terminal tripeptide motif—that is, small basic hydrophobic residues—or variants thereof (Mullen et al. 1997a , Mullen et al. 1997b ), that is found in the many peroxisomal matrix–destined proteins. The type 2 PTS (PTS2) is a nonapeptide motif (R/K-X₆-H/Q-A/L/F [where X indicates any amino acid]) (Flynn et al. 1998 ) located in the N terminus of another set of matrix proteins that are proteolytically processed after import into peroxisomes. Proteinaceous receptors (those recognizing PTS1- and PTS2-containing proteins) and several components of the peroxisome translocation machinery have been identified in various organisms. These proteins are termed peroxins, given as Pex to describe the protein and PEX to describe the gene (Distel et al. 1996 ), and have been identified in a number of species. For instance, PTS1 receptors encoded by PEX5 genes have been isolated from watermelon (Wimmer et al. 1998 ), Arabidopsis (Brickner et al. 1998 ), and tobacco (Kragler et al. 1998 ), and PEX10 and PEX16 genes have been isolated more recently from Arabidopsis (Lin et al. 1999 ; Schumann et al. 1999 ).

Details related to targeting signals and trafficking pathways for peroxisomal membrane proteins (PMPs) are beginning to emerge. PMPs are synthesized in the cytosol (with one exception, reported by Bodnar and Rachubinski 1991 ) and are post-translationally targeted to peroxisomes directly or indirectly. Support for the direct insertion of PMPs comes mainly from in vitro import studies with mammalian and plant peroxisomes (Imanaka et al. 1996 ; Just and Diestelkotter 1996 ; Tugal et al. 1999 ) and from in vivo targeting studies in which a deletion of the membrane peroxisomal targeting signal in Candida boidinii PMP47 resulted in the protein being localized exclusively in the cytosol (Dyer et al. 1996 ).

Other PMPs appear to be targeted to peroxisomes indirectly, that is, by way of the endoplasmic reticulum (ER), a formerly discredited pathway for peroxisome biogenesis that has now been resurrected (albeit significantly modi-fied) by some investigators (Erdmann et al. 1997 ; Kunau and Erdmann 1998 ; Subramani 1998 ; Titorenko and Rachubinski 1998a ). Evidence for this pathway includes the pulse–chase analyses of radiolabeled rat PMP50, which revealed that this membrane protein was synthesized on membrane-bound polysomes and localized in the ER before insertion into peroxisomes (Bodnar and Rachubinski 1991 ). Moreover, human PMP Pex3p and a chimeric protein composed of the N terminus of Pex3p from Hansenula polymorpha fused to a reporter protein both caused a profound proliferation of ER membranes when overexpressed in vivo (Baerends et al. 1996 ; Kammerer et al. 1998 ). Proliferation of ER membranes also was observed when the Saccharomyces cerevisiae PMP Pex15p was overexpressed. In addition, the C-terminal domain of Pex15p was O-glycosylated, indicating that this portion of the protein protruded, at least transiently, into the lumen of the ER (Elgersma et al. 1997 ). The indirect trafficking pathway for PMPs seems to involve participation of distinct ER-derived vesicles that contain components of coat protein II (COP II) (Titorenko and Rachubinski 1997 ) and at least two Pexs (Pex1p and Pex6p) that show sequence similarity to N-ethylmaleimide-sensitive factor-like ATPases. All of these proteins may function during vesicle fusion with the peroxisome boundary membrane (Titorenko and Rachubinski 1997 , Titorenko and Rachubinski 1998b ; Faber et al. 1998 ).

Our knowledge of the sorting of PMPs to the peroxisomal boundary membrane in plant cells is rudimentary. Only three authentic plant PMPs have been identified, none of which has been demonstrated to function as a Pex, that is, as a protein involved in peroxisome biogenesis. Authentic plant PMPs include a 28-kD polypeptide of unknown function (Yamaguchi et al. 1995 ), a putative 73-kD molecular chaperone (Corpas and Trelease 1997 ), and a 30- to 31-kD isoform of ascorbate peroxidase (APX; Yamaguchi et al. 1995 ; Bunkelmann and Trelease 1996 ). Peroxisomal APX, a constitutive component of the boundary membrane in oilseed seedling glyoxysomes and leaf (and leaf-type) peroxisomes, functions to protect the cell from toxic reactive oxygen species generated through the ß oxidation of fatty acids within the organelle (Mullen and Trelease 1996 ; Corpas and Trelease 1998 ). cDNAs encoding peroxisomal APX have been isolated from cotton (Bunkelmann and Trelease 1996 ), Arabidopsis (Zhang et al. 1997 ), and spinach (Ishikawa et al. 1998 ). The deduced amino acid sequence of peroxisomal APX reveals a high degree of identity with cytosolic APX but it has, in addition, a C-terminal 41–amino acid extension that contains a single, putative membrane-spanning region.

In this study, we describe the intracellular trafficking pathway of cottonseed peroxisomal APX in suspension-cultured tobacco BY-2 cells, a well-characterized in vivo peroxisomal import system (Banjoko and Trelease 1995 ; Trelease et al. 1996a ; Lee et al. 1997 ; Mullen et al. 1997a , Mullen et al. 1997b ; Flynn et al. 1998 ). Immunofluorescence microscopic analyses revealed that epitope-tagged cottonseed peroxisomal APX is sorted both to peroxisomes (glyoxysomes) and to a distinct membrane compartment with characteristics of a subdomain of the ER. Direct in vitro evidence showed that peroxisomal APX is inserted post-translationally into highly purified ER membranes but not into peroxisome membranes. Collectively, the results suggest that peroxisomal APX is sorted indirectly to peroxisomes by way of the ER, implicating this endomembrane compartment in the biogenesis of peroxisomes in plant cells.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Transiently Expressed, Epitope-Tagged Peroxisomal APX Is Localized to BY-2 Peroxisomes and to a Reticular/Circular 3,3'-Dihexyloxacarbocyanine-Stained Subcompartment
As with peroxisomes (glyoxysomes) in oilseed cotyledons and endosperm, peroxisomes in suspension-cultured tobacco BY-2 cells possess an endogenous membrane-bound peroxisomal APX (see below). Because immunodetection of endogenous peroxisomal APX would complicate experiments designed to elucidate the subcellular location of introduced wild-type or mutated cottonseed peroxisomal APX, a hemagglutinin (HA) epitope-tagged version of this peroxisomal APX (HA-pAPX; Figure 1) was constructed with a single copy of the HA epitope tag at the N-terminal end of the peroxisomal APX . The N terminus of peroxisomal APX was deemed a suitable location for the HA tag for several reasons: (1) preliminary data indicated that the targeting information, along with the transmembrane spanning domain, were located at the C terminus of the protein; (2) HA-tagged peroxisomal APX was inserted into membranes in vivo and in vitro (J.A. Miernyk and R.T. Mullen, unpublished data); and (3) as shown in the crystalline structure of recombinant cytosolic pea APX, the N terminus protruded from the tertiary core of the polypeptide (Patterson and Poulos 1995 ).

View larger version (11K): [in this window] [in a new window]   Figure 1. Schematic Representation of Cottonseed HA-pAPX. A sequence encoding a single HA epitope tag (MGYPYDVPDYAG- underlined; depicted by solid box) was appended to the 5' end of the cottonseed peroxisomal APX open reading frame by using the polymerase chain reaction (see Methods). The polypeptide sequence shown is for a unique C-terminal 41–amino acid extension that is not found in cytosolic APXs. This extension contains a single, putative transmembrane domain (TMD, underlined and depicted by the hatched box).

Figure 2A illustrates a representative punctate immunofluorescence pattern observed within nontransformed BY-2 cells incubated with anti-cucumber peroxisomal APX IgGs. This pattern is characteristic of antigenic proteins localized to peroxisomes (Lee et al. 1997 ; Mullen et al. 1997a , Mullen et al. 1997b ; Flynn et al. 1998 ). The subcellular location of HA-pAPX expressed transiently (its expression driven by the cauliflower mosaic virus 35S promoter) for 20 hr in a BY-2 cell is shown in Figure 2B. Immunodetection of the HA epitope on peroxisomal APX overexpressed within transformed cells revealed a substantially different immunofluorescence pattern, that is, both punctate and reticular/circular patterns. The circular structures commonly observed within this reticular pattern also are observed in nontransformed living cells (see Figure 3) and thus are not formed solely as a consequence of transient overexpression of HA-pAPX. Figure 2C illustrates in the same transformed cell the punctate immunofluorescence pattern attributable to endogenous peroxisomal catalase. In Figure 2D, the yellow/orange color of the merged images (Figure 2B and Figure 2C) reveals that only the punctate portion of HA-pAPX fluorescence colocalizes entirely with the punctate catalase fluorescence pattern. This obvious colocalization indicates that at least some of the introduced HA-pAPX is sorted to endogenous peroxisomes. However, Figure 2D also shows that a significant portion of the introduced HA-pAPX—that in the reticular/circular structures (Figure 2B and Figure 2D)—does not colocalize with catalase in peroxisomes. Thus, after expression for ~20 hr, introduced HA-pAPX becomes localized to at least two distinguishable subcellular compartments in BY-2 cells, namely, peroxisomes and a reticular/circular network.

View larger version (35K): [in this window] [in a new window]   Figure 2. Immunofluorescence Localization of Transiently Expressed HA-pAPX and Endogenous Organelle Marker Proteins in BY-2 Cells. Nontransformed cells, or cells transiently transformed with HA-pAPX by biolistic bombardment, were fixed in formaldehyde (~20 hr after bombardment), treated with pectolyase Y-23, incubated in 0.3% Triton X-100 (to permeabilize all cellular membranes), and then incubated with primary and dye-conjugated secondary antibodies. (A) Nontransformed cells exhibiting a punctate Cy2 fluorescence attributable to endogenous peroxisomal APX after application of anti-cucumber peroxisomal APX IgGs. (B) to (D) HA-pAPX–transformed cell showing the colocalization (D) of only a portion of expressed HA-pAPX (B) with endogenous peroxisomal catalase (C). In (B), both a punctate Cy2 fluorescence (solid arrows) and reticular/circular Cy2 fluorescence (outlined arrows) of HA-pAPX (anti-HA IgGs) are apparent. In (C), only a punctate Cy5 fluorescence (solid arrows) attributed to endogenous catalase (anti-cottonseed catalase IgGs) in peroxisomes is apparent. The merged image in (D) shows that the green punctate HA-pAPX fluorescence, but not the reticular/circular fluorescence, colocalizes (yellow/orange) with the red punctate fluorescence of endogenous catalase. Solid arrowheads indicate obvious colocalizations; outlined arrowheads indicate non-colocalizations. (E) to (G) HA-pAPX–transformed cells illustrating the absence of colocalization of expressed HA-pAPX with resident Golgi body (reversibly glycosylated polypeptide) (E), mitochondrial (ß-ATPase) (F), and vacuolar ( isoform of tonoplast intrinsic protein) (G) proteins. Outlined arrowheads indicate obvious non-colocalizations between green reticular/circular structures of HA-pAPX and the red organelle markers. Juxtaposition of the HA-pAPX–labeled reticulum (green) and the mitochondria (red) sometimes results in a misleading appearance of yellow/orange colocalization. (H) Nontransformed cell showing absence of Cy2 fluorescence when primary IgGs are omitted. Bar in (A) = 10 µm for (A) to (H).

View larger version (70K): [in this window] [in a new window]   Figure 3. DiOC6 Staining and Immunofluorescence Labeling of ER in BY-2 Cells. Cells transiently transformed with HA-pAPX (except [D] and [E]) were fixed (except [E]) in formaldehyde (~20 hr after bombardment), treated with pectolyase Y-23, incubated in 0.3% Triton X-100, and then incubated with primary and dye-conjugated secondary antibodies. Cells shown in (A) to (E) were stained with DiOC6. (A) to (C) HA-pAPX–transformed cell showing the colocalization (C) of reticular/circular HA-pAPX (A) with a portion of DiOC6-stained membranes (ER) (B). In (A), both the punctate Cy5 fluorescence (solid arrows) and reticular/circular Cy5 fluorescence (outlined arrows) of expressed HA-pAPX (anti-HA IgGs) are apparent. (B) shows reticular structures (outlined arrows) attributable to DiOC6 staining in the transformed cell; punctate staining of peroxisomes, indicated by solid arrows in (A), is not apparent. The merged image in (C) shows the red reticular/circular HA-pAPX fluorescence, but not the punctate fluorescence (individual peroxisomes), colocalizing (yellow/orange) with a portion of the green reticular DiOC6 fluorescence in the transformed cell. Solid arrowheads indicate obvious colocalizations. (D) and (E) DiOC6 fluorescence in nontransformed cells. Both bombarded ([D], two adjacent cells) and living (E) nontransformed cells contain reticular/circular structures (outlined arrows). (F) to (H) HA-pAPX–transformed cell showing the absence of colocalization (H) of expressed HA-pAPX (F) with endogenous calreticulin (G). The merged image in (H) shows that the green reticular/circular Cy2 fluorescence of HA-pAPX does not colocalize with the red reticular Cy5 fluorescence of endogenous calreticulin (anti-castor calreticulin antiserum). Outlined arrowheads indicate obvious structures at which the proteins did not colocalize. (I) to (K) HA-pAPX–transformed cell showing the absence of colocalization (K) of expressed HA-pAPX (I) with endogenous BiP (J). The merged image in (K) shows that the green reticular/circular Cy2 fluorescence of HA-pAPX does not colocalize with the red reticular Cy5 fluorescence of endogenous BiP (anti-maize BiP antiserum). Outlined arrowheads indicate obvious non-colocalizations. (L) Merged image showing the non-colocalization of expressed HA-pAPX (green) with endogenous calnexin (red, anti-castor calnexin antiserum). Outlined arrowheads indicate obvious non-colocalizations. Faint yellow coloration in (H), (K), and (L) is the result of juxtaposition of red and green fluorescent structures. Bar in (A) = 10 µm for (A) to (L).

Control experiments included bombardment of cells with different gene constructs. In one experiment, peroxisomal APX (lacking an N-appended HA epitope tag) was introduced biolistically; the transformed cells that were labeled with anti-peroxisomal APX IgGs and conjugated secondary antibodies exhibited both a punctate and a reticular/circular fluorescence pattern similar to that illustrated in Figure 2B (results not shown), indicating that the observed sorting of HA-pAPX was not due to the appended HA tag. In a mock transformation control, application of anti-HA IgGs and conjugated secondary antibodies yielded no immunofluorescence image (e.g., Figure 4F in Lee et al. 1997 ).

View larger version (32K): [in this window] [in a new window]   Figure 4. Immunofluorescence Images Illustrating Sorting of CAT-pAPX to Preexisting Peroxisomes and Reticular/Circular Structures. Transiently transformed cells were fixed in formaldehyde (~20 hr after bombardment), treated with pectolyase Y-23, incubated in 0.3% Triton X-100, and then incubated with primary and dye-conjugated secondary antibodies. In (E) to (G), cells also were incubated in DiOC6. (A) shows a composite image (projection) of multiple optical sections. (A) CAT-transformed cell exhibiting Cy2 fluorescence throughout the cytosol (anti-CAT IgGs). (B) to (D) CAT-pAPX–transformed cell showing the colocalization (D) of a portion of expressed CAT-pAPX (B) with endogenous peroxisomal catalase (C). In (B), both globular Cy2 fluorescence (asterisks) and reticular/circular Cy2 fluorescence (outlined arrows) of CAT-pAPX (anti-CAT IgGs) are apparent. In (C), nearly all Cy5 fluorescence attributed to endogenous catalase (anti-cottonseed catalase IgGs) is globular (asterisks). The merged image in (D) shows that the green globular CAT-pAPX fluorescence, but not the reticular/circular fluorescence, colocalizes (yellow/orange) with the red globular fluorescence of endogenous catalase in aggregated (enlarged) peroxisomes. Solid arrowheads indicate obvious colocalizations; outlined arrowheads indicate non-colocalizations. (E) to (G) CAT-pAPX–transformed cell showing the colocalization (G) of reticular/circular CAT-pAPX (E) with a portion of DiOC6-stained membranes (ER) (F). In (E), both a globular Cy5 fluorescence (asterisk) and reticular/circular Cy5 fluorescence (outlined arrows) of CAT-pAPX are apparent. (F) shows a reticular fluorescence pattern (outlined arrows) attributable to DiOC6 staining in the transformed (and surrounding nontransformed) cell. The merged image in (G) shows that the red reticular/circular CAT-pAPX fluorescence, but not the globular fluorescence (aggregated peroxisomes), colocalizes (yellow/orange) with a portion of the green reticular DiOC6 fluorescence. Solid arrowheads indicate obvious colocalizations. (H) Representative image illustrating lack of fluorescence in a CAT-pAPX–transformed cell after application of Cy2-conjugated anti-mouse IgGs only (anti-CAT IgGs were omitted). Bar in (A) = 10 µm for (A) to (H).

Elucidation of the HA-pAPX immunofluorescence pattern led to follow-up in vivo experiments aimed at determining the identity of the reticular/circular network. Specifically, HA-pAPX–transformed cells were double labeled with fluor-conjugated secondary antibodies bound to anti-HA IgGs and to IgGs raised against resident enzymes/proteins that specifically mark various plant cell organelles. Figure 2E through 2G illustrate merged confocal images of these results. In Figure 2E, reversibly glycosylated polypeptide (Dhugga et al. 1997 ) obviously is not colocalized with HA-pAPX, indicating that Golgi bodies are not part of the reticular/circular network to which HA-pAPX is localized. Figure 2F reveals that mitochondrial ß-ATPase (Luethy et al. 1993 ) is distinct from the HA-pAPX reticular/circular network, as is the isoform of tonoplast intrinsic protein (Figure 2G) that marks one of several types of vacuoles in plant cells (Neuhaus and Rogers 1998 ). Plastids detected with anti-stearoyl–acyl carrier protein -9 desaturase antibody applied to BY-2 cells also did not resemble the HA-pAPX–stained network (Trelease et al. 1996a ). Although the results from this series of studies provide negative evidence, the data are quite compelling that Golgi bodies, mitochondria, a vacuolar compartment, and plastids are not part of the reticular/circular network to which HA-pAPX becomes localized.

3,3'-Dihexyloxacarbocyanine iodide (DiOC₆) is commonly used to visualize ER, even though it often stains other membranes as well (see Discussion). Figure 3A illustrates a HA-pAPX immunofluorescence pattern, including punctate (peroxisomes) and reticular/circular structures (similar to Figure 2B), but imaged with cyanine 5 (Cy5). Figure 3B shows the results of DiOC₆ staining of the same cell. A reticular/circular fluorescence pattern, but not a punctate peroxisomal pattern, is observed in the transformed (and surrounding nontransformed; Figure 3D) cell. The yellow/orange color in the merged image (Figure 3C) shows substantial, but not complete, colocalization of the HA-pAPX and DiOC₆ patterns. This and previous results suggest that in addition to being localized to peroxisomes, HA-pAPX is localized to some part of the ER that constitutes virtually all of the reticular/circular network.

One of our concerns was that the reticular/circular network constituted an anomalous membrane compartment within biolistically transformed cells. Several observations, however, indicated that this was not the case. In Figure 3D, a representative image of two DiOC₆-stained, nontransformed cells in a population of bombarded cells, the reticular/circular pattern is clearly evident. As a control for the effects of physical and chemical handling of cells involved in the biolistic bombardment procedure, we also stained living, unbombarded cells with DiOC₆. Figure 3E illustrates that a reticular/circular fluorescence pattern similar to that observed in HA-pAPX–transformed (cf. Figure 3B) and in nontransformed neighboring (cf. Figure 3D) cells occurs in normal, living BY-2 cells. The control experiments thus confirm that the reticular/circular network observed with anti-HA antibodies and DiOC₆ staining in transformed cells is not a consequence of transient overexpression (or fixation with formaldehyde) of HA-pAPX.

Evidence to help support or refute our interpretation that HA-pAPX was localized to the ER, or to a portion thereof, was sought with antibodies to calreticulin, calnexin, and the lumenal binding protein (BiP), three proteins well-established as resident in the ER. Although we expected that the reticular/circular staining pattern of HA-pAPX would coincide with one or more of the three ER marker proteins, the results were surprisingly different. Figure 3F shows a HA-pAPX reticular/circular immunofluorescence image, and Figure 3G shows in the same cell the immunofluorescence image of calreticulin. As seen in the merged image (Figure 3H), the two antigens are not colocalized. Close inspection of the yellow/orange coloration revealed it to be the vivid fluorescence of juxtaposed antigens rather than true colocalization. A similar three-image comparison for a transformed cell double labeled for HA-pAPX (Figure 3I) and BiP (Figure 3J) is shown as a merged image in Figure 3K. Again, close observation reveals that essentially no colocalization is apparent for these two antigens. Finally, Figure 3L shows only the merged image for the double-labeling experiments with anti-calnexin and anti-HA IgGs; colocalization is not apparent. One interpretation of these results, for which supporting experimental evidence is presented (see below), is that HA-pAPX is localized to the ER, but primarily in a distinct domain of the ER that does not possess the majority of the resident calreticulin, calnexin, or BiP. Alternatively, the data presented so far could be interpreted to indicate that the HA-pAPX is localized to an unidentified, DiOC₆-staining compartment that seems to be closely associated with the ER.

The C Terminus of Peroxisomal APX Sorts Chloramphenicol Acetyltransferase to Peroxisomes and to the Reticular/Circular Network
Chloramphenicol acetyltransferase (CAT) was shown in previous targeting studies to be a suitable passenger protein for appended polypeptides (e.g., Mullen et al. 1997a ). Figure 4A shows that transiently expressed CAT accumulates throughout the cytosol. However, when the 36 C-terminal residues of peroxisomal APX, including the single transmembrane domain (Figure 1), were appended to the C terminus of CAT, the fusion protein (CAT-pAPX) was sorted from the cytosol to at least two distinct subcellular sites. Figure 4B shows that after 20 hr of expression, CAT-pAPX localized to both globular and reticular/circular structures within the same cell; in contrast, endogenous catalase within this cell was located only in the globular structures (Figure 4C). In the merged image (Figure 4D), some of the CAT-pAPX colocalized (yellow color) with the endogenous catalase in the globular structures. Similar results were obtained with isocitrate lyase, another endogenous peroxisomal enzyme that was observed only in the globular structures (data not shown). All CAT-pAPX–transformed cells possessed an altered endogenous catalase fluorescence; that is, approximately one to 12 globular catalase-containing structures were observed per cell. Aggregated (enlarged) peroxisomes also were observed in a small percentage (~15%) of HA-pAPX–transformed cells; in both cases, the aggregation appeared to result from "zippering" of the organelle boundary membranes, a consequence of the oligomerization of cytosolic-facing, membrane-bound subunit proteins. Space and focus do not allow inclusion of further details or evidence in this article; they will be presented elsewhere.

Experimental results illustrated in Figure 4E to 4G show that the C terminus (36 amino acid residues) of peroxisomal APX also directs CAT to a portion of the DiOC₆-stained compartment (putative ER subdomain). The CAT-pAPX localizated in the reticular/circular network (Figure 4E) also colocalized with a portion of the DiOC₆ fluorescence (Figure 4F), as shown by the yellow/orange color in Figure 4G (merged image). These results are essentially the same as those in Figure 3C, that is, partial colocalization of HA-pAPX with DiOC₆-fluorescent structures. Also, as was found for HA-pAPX (Figure 3F to 3L), the CAT-pAPX reticular staining pattern did not colocalize with the immunofluorescence attributable to calreticulin, BiP, or calnexin (data not shown). Figure 4H is a representative image of a control experiment in which anti-CAT IgGs were omitted but secondary IgGs conjugated to Cy2 were added. Together, these results indicate that the C-terminal extension of peroxisomal APX contains sufficient topogenic information for sorting CAT to peroxisomes and to a putative subdomain of the ER.

Brefeldin A Interferes with the Sorting of CAT-pAPX and CAT-SKL to Peroxisomes
To elucidate some of the dynamics of the sorting pathway for peroxisomal APX within BY-2 cells, we examined the subcellular location of CAT-pAPX microscopically in cells treated with the fungal toxin brefeldin A (BFA). BFA, among other cellular effects, blocks protein export from the ER by preventing the formation of ER-derived coated vesicles (Klausner et al. 1992 ; Satiat-Jeunemaitre et al. 1996 ; Staehelin and Driouich 1997 ). We selected CAT-pAPX, rather than HA-pAPX, for this series of experiments because observations of CAT-pAPX in peroxisomal aggregates would be convincing evidence that CAT-pAPX localized to preexisting peroxisomes. If transiently expressed CAT-pAPX were sorted indirectly to preexisting peroxisomes by way of a putative ER subdomain, then incubation of transformed cells in BFA during the postbombardment sorting period should interfere with sorting to these peroxisomes. Alternatively, if CAT-pAPX were sorted concurrently to peroxisomes and the putative ER subdomain, then BFA treatment should not affect sorting to either of these compartments.

Figure 5A illustrates a reticular network attributable to CAT-pAPX localization in a transformed cell maintained 8 to 10 hr after bombardment in BFA. The punctate rhodamine fluorescence shown in Figure 5B illustrates the localization of endogenous catalase in the same cell. Fluorescent globular structures (aggregated peroxisomes) were not observed. The failure of CAT-pAPX to colocalize (Figure 5A) with endogenous catalase (Figure 5B) and the absence of globular (aggregated) peroxisomes revealed that CAT-pAPX was not sorted to preexisting peroxisomes in these BFA-treated cells.

View larger version (50K): [in this window] [in a new window]   Figure 5. Epifluorescence Images Illustrating the Interference of BFA with the Sorting of CAT-pAPX to BY-2 Peroxisomes. Cells were biolistically bombarded in transformation medium with or without 100 µg/mL BFA. After transient expression for 8 to 10 hr, cells were fixed in formaldehyde, treated with pectoylase Y-23, incubated in 0.3% Triton X-100, then incubated with primary and dye-conjugated secondary antibodies. For the recovery experiments in (C) and (D), bombarded cells were incubated in transformation medium plus BFA for 8 to 10 hr and then washed and incubated in transformation medium without BFA for an additional 8 to 10 hr. In (E), only DMSO (no BFA) was added to the transformation medium. In (F) to (J), CAT-SKL and CATß60 were expressed for 8 to 10 hr prior to fixation. (A) and (B) CAT-pAPX–transformed cell (treated with BFA) exhibiting a reticular Cy2 fluorescence attributable to overexpressed CAT-pAPX (anti-CAT IgGs) (A) and a punctate rhodamine fluorescence attributable to endogenous catalase (anti-cottonseed catalase IgGs) that is not colocalized in individual peroxisomes (outlined arrowheads) (B). (C) and (D) CAT-pAPX–transformed cell (no BFA) showing the colocalization (solid arrowheads) of a portion of expressed CAT-pAPX (C) with endogenous catalase (D) in globular peroxisomes. In (C), both globular and reticular/circular structures (outlined arrows) exhibit Cy2 fluorescence from CAT-pAPX. In (D), all rhodamine fluorescence attributable to endogenous catalase is globular. (E) CAT-pAPX–transformed cell, bombarded in transformation medium containing 1% (v/v) DMSO, exhibiting reticular/circular (outlined arrows) and globular Cy2 fluorescence. (F) CAT-SKL–transformed cell (no BFA) exhibiting a punctate Cy2 fluorescence (solid arrows) attributable to CAT-SKL localized in individual peroxisomes. (G) CAT-SKL–transformed cell (treated with BFA) exhibiting cytosolic and punctate (solid arrows) Cy2 fluorescence attributable to CAT-SKL localized in the cytosol and in individual peroxisomes, respectively. (H) and (I) CATß60-transformed cell (no BFA) showing the colocalization (solid arrowheads) of punctate Cy2 fluorescence attributable to expressed CATß60 (H) with the punctate Cy3 fluorescence attributable to endogenous mitochondrial ß-ATPase E (I). (J) CATß60-transformed cell (treated with BFA) exhibiting a punctate Cy2 fluorescence pattern attributable to CATß60 localized to mitochondria. Bar in (A) = 10 µm for (A) to (J).

The representative images shown in Figure 5C and Figure 5D illustrate that CAT-pAPX–transformed cells could recover from BFA treatment. Bombarded cells were incubated in BFA for 8 to 10 hr as described above, washed several times, and then incubated in transformation medium minus BFA for an additional 8 to 10 hr. Figure 5C shows that CAT-pAPX was sorted to peroxisomal aggregates and to the reticular/circular network (putative ER subdomain). Endogenous catalase in the same cell (Figure 5D) was localized to peroxisomal aggregates that colocalized with CAT-pAPX (cf. Figure 5C and Figure 5D). Because the endogenous catalase in neighboring nontransformed cells (Figure 5D) was localized within individual, or "unzippered," preexisting peroxisomes, the peroxisomal aggregation did not result from withdrawal of the BFA. We are therefore confident that the peroxisomal aggregation observed in transformed cells is a consequence of CAT-pAPX sorting to preexisting peroxisomes after removal of BFA. Figure 5E is a representative image from a control experiment illustrating that DMSO, which was included in the BFA solution, does not prevent sorting of CAT-pAPX to peroxisomes (aggregates) or to the putative ER subdomain (reticular/circular structures).

In a positive control experiment, BFA surprisingly also interfered with the sorting of a CAT fusion protein that had a prototypic type 1 PTS, Ser-Lys-Leu, appended to its C terminus (CAT-SKL). At 8 to 10 hr after bombardment in the absence of BFA, CAT-SKL was localized almost exclusively to individual peroxisomes, as expected (Figure 5F). In the presence of BFA, however, CAT-SKL accumulated throughout the cytosol, being only partially localized to individual peroxisomes (Figure 5G). This apparently was not a result of temporal expression because similar images were observed when CAT-SKL was expressed transiently for 20 hr in BFA (data not shown).

To evaluate the possible nonspecific interference of BFA on post-translational sorting of proteins to organelles, we examined as a control the targeting of a CAT fusion protein to mitochondria in the presence of BFA. Chaumont et al. 1994 have shown previously that the 60–amino acid presequence on the tobacco mitochondrial F₁-ATPase ß subunit is sufficient to target CAT to mitochondria in transgenic tobacco. Figure 5H illustrates that at 8 to 10 hr after bombardment the CAT fusion protein (CATß60) had sorted in transformed cells to mitochondria, which were observed as numerous rod-shaped immunofluorescent structures. These CATß60-containing structures colocalized with the rod-like immunofluorescence in the same cells that could be attributed to endogenous mitochondrial ß-ATPase E (Figure 5I). Figure 5J shows that when CATß60-bombarded cells were maintained in BFA for ~9 hr, the CAT fusion protein was targeted to mitochondria in a normal fashion. Thus, the results presented in Figure 5A to 5J indicate that BFA interferes with the post-translational sorting of proteins to the boundary membrane and matrix of peroxisomes but not with the sorting of a protein to the matrix of mitochondria.

Peroxisomal APX Is Inserted Post-Translationally into ER Membranes in Vitro
In vitro membrane insertion/association experiments were performed to help interpret results of the in vivo microscopic sorting experiments. To properly test our working hypothesis that proteins were sorted indirectly to peroxisomes by way of the ER, it was particularly important to use highly purified ER membranes in the in vitro assays. Table 1 shows that these microsomal-derived ER membranes (MEMs) fit this criterion. The microsomal fraction (the pellet obtained after ultracentrifugation at 100,000g for 60 min) was enriched with the two functionally independent ER markers UDP–GlcNAc transferase and NADH–cytochrome c reductase. As shown by the presence of various enzymes, it also contained significant amounts of mitochondrial (succinate dehydrogenase), peroxisomal (catalase), and plastidic (ribulose bisphosphate carboxylase oxygenase) proteins, with some Golgi body (acid phosphatase) proteins. Equilibrium density sucrose gradient centrifugation of these microsomes, followed by gel permeation chromatography of the ER band from this gradient, yielded highly purified MEMs that possessed some mitochondria and considerably less peroxisomal contamination (Table 1).

For most of the experiments we used a coupled transcription/translation rabbit reticulocyte lysate system programmed with DNA coding for cottonseed peroxisomal APX (Figure 6A and Figure 6B). Lane 1 (Figure 6A) shows that a single product of the expected molecular mass for the peroxisomal APX polypeptide (31 kD) was recovered in the soluble fraction of the reticulocyte lysate. Lane 2 illustrates that when MEMs were added 1 hr after initiation of the transcription/translation reactions, a 31-kD product (constituting 94% of the total radiolabeled pAPX, as determined by radioanalytical imaging) was in the membrane pellet collected after centrifugation through a sucrose cushion. Resuspension of these membranes in 100 mM sodium carbonate, pH 11.5, and recentrifugation through an alkaline sucrose cushion resulted in recovery of all (116%) of the peroxisomal APX in the pellet fraction (lane 3), indicating that peroxisomal APX was integrated stably into the MEMs. Washing an MEM preparation with 300 mM KCl (to remove signal recognition particles) before adding it to the translation mixture did not inhibit integration of peroxisomal APX into membranes (lane 4), the measured value being 107% that of the unwashed sample. Collectively, these data indicate that peroxisomal APX integrates in vitro into ER membranes in a post-translational (signal recognition particle–independent) manner.

View larger version (22K): [in this window] [in a new window]   Figure 6. Insertion of Peroxisomal APX into ER Membranes in Vitro. For data shown in (A) to (C), radiolabeled proteins were acid precipitated, separated by SDS-PAGE in 12.5% acrylamide gels, and visualized with an AMBIS radioanalytical imaging system. (A) pAPX integrates post-translationally into ER membranes. Lane 1 contains radiolabeled peroxisomal APX that was synthesized for 1 hr at 25°C from a cDNA in a coupled in vitro transcription/translation assay in the presence of 35S-methionine; lane 2, translation reactions (terminated after 1 hr by the addition of emetine) were incubated with MEMs for 1 hr at 25°C, after which membranes were collected by centrifugation through a sucrose cushion; lane 3, the MEM pellet described for lane 2 was resuspended and extracted with 100 mM Na2CO3, pH 11.5, to remove soluble and peripheral membrane-associated proteins from the remaining integral membrane proteins, and the MEMs were repelleted by centrifugation through an alkaline sucrose cushion; lane 4, same as lane 2 except that MEMs were prewashed in 300 mM KCl to remove membrane-bound ribosomes (including signal recognition particles); lane 5, same as lane 2, except that the completed translation reactions were preincubated with apyrase for 30 min at 25°C to remove ATP; lane 6, MEMs were pretreated with trypsin at 4°C for 30 min before incubation with translation reactions and later were extracted in Na2CO3 as described for lane 3; and lane 7, MEM translation reaction mixtures were incubated with trypsin at 4°C for 30 min and then extracted in Na2CO3 as described for lane 3. Addition to the reaction mixtures is indicated as (+), omission as (-). (B) Comparison of the insertion of peroxisomal APX and other membrane proteins into various organelle membranes. Peroxisomal APX (lanes 1 to 5), HA-pAPX (lanes 6 and 7), S. cerevisiae Pex15p (lanes 8 and 9), mouse VAMP2 (lanes 10 and 11), and the 47-kD peroxisomal membrane protein PMP47 from C. boidinii (lanes 12 and 13) were synthesized from their corresponding cDNAs by using a coupled in vitro transcription/translation system in the presence of 35S-methionine. Various membrane vesicles (as indicated above lanes) were incubated in terminated reaction mixtures for 1 hr at 25°C. Membranes were collected by centrifugation through a sucrose cushion, resuspended in 100 mM Na2CO3, and recentrifuged through an alkaline sucrose cushion. (C) Inhibition of peroxisomal APX integration into ER membranes by immunodepletion of molecular chaperones from translation mixtures. Peroxisomal APX RNA was translated in vitro by using a wheat germ–derived translation system (lane 1). Wheat germ extracts were incubated with antibodies before the translation reactions, and MEMs were added after translation for 1 hr at 25°C. The following antibodies were added to pre-translation reactions: anti-Hsp70 antibodies (lane 2); anti-Hsp70 plus anti-AtJ2 antibodies (lane 3); anti-Hsp70, anti-AtJ2, and anti-AtE1 antibodies (lane 4); anti-pyruvate dehydrogenase E1 subunit antibodies (lane 5). MEMs were collected by centrifugation through a sucrose cushion, resuspended in 100 mM Na2CO3, and recentrifuged through an alkaline sucrose cushion. Addition to the reaction mixtures is indicated as (+), omission as (-).

Removal of ATP from translation mixtures by the addition of apyrase greatly reduced (90%) the post-translational insertion of peroxisomal APX into MEMs (Figure 6A, cf. lanes 5 and 3). Pretreatment of MEMs with trypsin also substantially reduced (97%) membrane insertion of peroxisomal APX (Figure 6A, cf. lanes 6 and 3), indicating that a protease-sensitive membrane component is required for peroxisomal APX insertion. Intact peroxisomal APX remained in the supernatant after treatment with alkaline sodium carbonate (data not shown), indicating that residual protease did not digest peroxisomal APX and thus gave results different from those shown in lane 6. Figure 6A, lane 7, shows that membrane-inserted peroxisomal APX, obtained in experiments described for lane 3, was digested by trypsin afterwards, indicating that most of the polypeptide chain faces outward on membrane vesicles. This is consistent with the topological orientation of peroxisomal APX determined by us in in vivo experiments and by others for peroxisomes in other cell types (Yamaguchi et al. 1995 ; Ishikawa et al. 1998 ).

Results presented in Figure 6B show that peroxisomal APX integrates specifically into MEMs (93%; lane 5) and not into membranes of mitochondria (0%; lane 1), chloroplasts (0%; lane 2), or plasma membranes (0%; lane 3). Relatively little peroxisomal APX integrated into peroxisomal membranes (5%; lane 4). Data on the purity of these membranes have been published elsewhere (see Methods). When in vitro reaction mixtures were programmed with the DNA encoding HA-pAPX, virtually the same results were obtained as with peroxisomal APX DNA: HA-pAPX was integrated efficiently into MEMs (85%; lane 7) but not into peroxisomal membranes (2%; lane 6). Positive controls (lanes 8 to 11) show that S. cerevisiae Pex15p (Elgersma et al. 1997 ) and mouse synaptobrevin/vesicle-associated membrane protein 2 (VAMP2) (Weber et al. 1998 ) also were integrated into ER membranes (43% and 94%, respectively) but were not efficiently integrated into peroxisomal membranes (5% and 3%, respectively), as expected. In contrast, PMP47 from C. boidinii (McCammon et al. 1994 ) inserted into peroxisomal membranes (66%; lane 12) but not into ER membranes (5%; lane 13), also as anticipated.

A wheat germ system programmed with peroxisomal APX RNA was used to assess the results of immunodepleting molecular chaperones from the translation mixtures (Figure 6C). Comparing lanes 1 and 2 shows that immunodepletion of the 70-kD heat shock chaperone protein (Hsp70) (Miernyk et al. 1992 ) resulted in inefficient (32%) insertion into ER membranes. Lanes 3 and 4 show that immunodepletion of Hsp70 plus the chaperone AtJ2, a homolog of Escherichia coli DnaJ (Zhou et al. 1995 ) (lane 3), or immunodepletion of Hsp70, AtJ2, and the nucleotide exchange factor AtE1, a functional homolog of E. coli GrpE (B. Kroczynska and J.A. Miernyk, unpublished data) (lane 4), resulted in little (9%) or no (0%) detectable insertion of peroxisomal APX into ER membranes, respectively. Lane 5 shows results of a control reaction in which peroxisomal APX insertion (94%) into MEMs was relatively unaffected when wheat germ extracts were incubated with the unrelated anti-pyruvate dehydrogenase E1 subunit antibody. These results indicate that the Hsp70 reaction cycle, or the recently renamed Hsp70 chaperone machine (Bukau and Horwich 1998 ), is involved in the insertion of peroxisomal APX into ER membranes in vitro.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

We used two approaches to elucidate the intracellular sorting of cottonseed peroxisomal APX to its functional site in the peroxisome boundary membrane. Immunofluorescence microscopy of gene products transiently expressed in BY-2 cells provided novel in vivo subcellular localization information, and results from in vitro protein association/insertion assays complemented those from the microscopic analyses. In BY-2 cells, HA-epitope–tagged peroxisomal APX localized to two distinct subcellular sites, peroxisomes and a reticular/circular network, the latter postulated to be a distinct subdomain of the ER. Results from in vitro experiments also showed that peroxisomal APX was inserted directly into purified ER membranes, but not into peroxisomal membranes, in a post-translational manner that depended on the presence of ATP and molecular chaperones. Collectively, our findings support the notion that, as with several mammalian and yeast PMPs, plant peroxisomal APX is sorted to peroxisomes by way of the ER.

Peroxisomal APX Sorts to a Reticular/Circular Compartment in Vivo
As shown by the immunofluorescence data presented here, overexpression of peroxisomal APX in BY-2 cells provides the opportunity to discern and dissect the sorting pathway of this important enzymatic PMP. The localization of expressed HA-pAPX to a nonperoxisomal compartment(s) raised the question as to why endogenous peroxisomal APX was observed only in peroxisomes in normal BY-2 cells (Figure 2A) and in oilseed cotyledons (Corpas et al. 1994 ; Yamaguchi et al. 1995 ; Bunkelmann and Trelease 1996 ; Ishikawa et al. 1998 ). Although it seems reasonable that the greatest proportion of peroxisomal APX at its steady state is within the peroxisomal boundary membrane, whereas in overexpressing cells, the protein is detectable throughout its sorting pathway (including the putative ER subdomain), one could also argue that the reticular/circular network was formed as a result of peroxisomal APX overexpression, cell fixation conditions, or both.

Results from several sources, however, indicate that neither of these possibilities is likely. For example, the network is readily observed in living, unbombarded cells stained with DiOC₆ (Figure 3E), and similar structures are seen consistently in both nontransformed (Figure 3D) and transformed cells (Figure 3C and Figure 4G) within bombarded samples fixed in formaldehyde and in unbombarded cells fixed in glutaraldehyde (not shown). In other cells, such as moss caulonemata (McCauley and Hepler 1990 ) and cultured mammalian Madin–Darby canine kidney cells (Hannan and Edidin 1996 ), DiOC₆ also labels a reticular/circular network. Furthermore, immunofluorescence localization of calreticulin in mature leaves of maize (Napier et al. 1995 ) and cotyledons of germinated tobacco seeds (Denecke et al. 1995 ) has provided morphological evidence that a similar network occurs in plant tissues as well as in cultured cells. Thus, numerous observations and literature examples have convinced us that the reticular/circular network described in this study represents an authentic membranous compartment(s) within BY-2 cells.

Intriguing evidence for the apparent sorting of peroxisomal APX to the ER comes from the colocalization of HA-pAPX (and CAT-pAPX) with a portion of the reticular/circular network that stains with DiOC₆ (solid arrowheads in Figure 3C and Figure 4G). As referenced above with specific examples, DiOC₆ commonly is used to label ER fluorescently in a variety of fixed and living cells (see also Terasaki 1993 ; Sabnis et al. 1997 ), but it also has been shown to label mitochondria and other organelles. Despite this caveat, we varied the concentration of DiOC₆ in preliminary experiments to achieve optimal staining of the ER network. Although some punctate (mitochondrial) fluorescence was observed even under optimized conditions, we discounted this background, given that HA-pAPX did not sort to mitochondria, Golgi, plastids, or vacuoles (Figure 2E to 2G). Thus, the reticular/circular network that could be immunostained in cells expressing HA-pAPX or CAT-pAPX (Figure 2D and Figure 4D) and that colocalized with DiOC₆ (Figure 3C and Figure 4G) seems to be morphologically and logically part of the ER. The virtual lack of colocalization of this network with common ER resident proteins (Figure 3H, Figure 3K, and Figure 3L) indicates that if peroxisomal APX is indeed localized to the ER, then a distinct subdomain that somehow is involved in the eventual distribution of PMPs to (preexisting) peroxisomes (discussed below) must exist.

As presented in the Introduction, the sorting of PMPs to peroxisomes by way of the ER is not without precedent in the literature, at least in mammalian and yeast cells. For example, yeast cells expressing either ER-targeted PMPs (Baerends et al. 1996 ; Elgersma et al. 1997 ; Kammerer et al. 1998 ) or ER membrane proteins (Wright et al. 1988 ; Vergeres et al. 1993 ) showed a marked elaboration of ER membranes. In particular, Pex15p specifically accumulated in proliferated ER in S. cerevisiae cells that were overexpressing this protein (Elgersma et al. 1997 ). Furthermore, Pex2p and Pex16p were N-glycosylated within the ER lumen while en route to peroxisomes in Yarrowia lipolytica mutants that were defective both in protein egress from the ER and in peroxisome biogenesis (Titorenko et al. 1997 ). Our finding that peroxisomal APX sorts to a compartment that morphologically resembles the ER suggests this organelle may be involved in directing plant PMPs as well.

Peroxisomal APX Is Inserted into Isolated ER Membranes in Vitro
A comprehensive series of in vitro experiments was performed to obtain further evidence for or against involvement of the ER in the post-translational sorting of peroxisomal APX. Well-established procedures were used (Miernyk et al. 1992 ), and the purity (Table 1) and uptake competency of the highly purified ER vesicles (MEMs) were exemplary. The membranes from other organelles were prepared from plant species that provide some of the purest membranes that can be prepared from these particular organelles; the species source was not considered a factor in interpreting the data. Three lines of in vitro evidence supported post-translational sorting/insertion of peroxisomal APX to or into the ER: the insertion of peroxisomal APX into MEMs after translation was arrested by the addition of emetine and RNase (our standard protocol); the efficiency of peroxisomal APX insertion was not enhanced when MEMs were added to reaction mixtures during the first hour of translation (data not shown); and the peroxisomal APX was inserted into MEMs in the absence of signal recognition particles (stripped from MEMs with KCl; Figure 6A, lane 4). Additional evidence came from our unpublished time-course studies in vivo. At ~4 to 5 hr after bombardment, at least part of the immunofluorescence attributable to the expressed HA-pAPX was observed in the cytosol and was not associated with reticular/circular or punctate structures. If peroxisomal APX were inserted cotranslationally in vivo, then peroxisomal APX most probably would have been detected only in the ER and/or peroxisomes at all post-bombardment times.

The insertion of PMPs (rat peroxisomal assembly factor-1, PMP22, PMP70, and Arabidopsis PMP22) into peroxisomal membranes in vitro has been reported to be independent of ATP (Imanaka et al. 1996 ; Just and Diestelkotter 1996 ; Tugal et al. 1999 ). These results seemingly are in contrast to those obtained here for peroxisomal APX; that is, the addition of apyrase to reaction mixtures markedly reduced insertion of peroxisomal APX into MEMs (Figure 6A, lane 5). Considering that importing proteins directly into the peroxisomal matrix also requires ATP (Horng et al. 1995 ; Olsen 1998 ; Subramani 1998 ), it may be inappropriate to make comparisons of the ATP-dependent insertion of peroxisomal APX with the ATP-independent insertion of the above-listed PMPs. However, peroxisomal APX is inserted into ER membranes and not into peroxisomal membranes, whereas all of the PMPs listed above were shown previously to have been inserted directly into peroxisomal membranes in vitro.

Molecular chaperones have been implicated in the acquisition of peroxisomal matrix proteins; however, data for interactions with a membrane protein have been reported in only one study. Pause et al. 1997 found that rat PMP22 coimmunoprecipitated separately with two proteins in rabbit reticulocyte lysate mixtures, namely, the chaperonin TRiC and a 40-kD polypeptide of unknown function. Frydman et al. 1994 also reported the apparent involvement of TRiC but only for the import of matrix proteins. Both groups proposed that the TRiC interacted with other chaperones, such as Hsp70, Hsp90, and DnaJ, as reported for uptake of matrix proteins (Walton et al. 1994 ; Crookes and Olsen 1998 ; Hettema et al. 1998 ).

In our study, apparent interaction with the three components of the Hsp70 chaperone machine (Bukau and Horwich 1998 ) substantially influenced insertion of peroxisomal APX into MEMs. Immunodepletion of Hsp70 alone, which is a relatively abundant component in wheat germ extracts (Miernyk et al. 1992 ), reduced peroxisomal APX insertion by ~70% (Figure 6C, lane 2). Additional immunodepletion of the cytosolic DnaJ homolog and the nucleotide exchange factor, two other members of the Hsp70 machine, cooperatively abolished insertion of peroxisomal APX into MEMs. Whether removal of ATP with apyrase, which resulted in less efficient insertion, affected chaperone function is not known. Nonetheless, the cooperative interactive necessity of all of these components for acquisition of any matrix or membrane protein into peroxisomes or the ER has not been reported previously. Our data, and those related to TRiC interactions, provide substantial evidence that post-translational sorting of peroxisomal APX to the ER (or of other PMPs directly to peroxisomes) involves the cooperative action of cytosolic molecular chaperones and the presence of ATP. This does not preclude subsequent interaction(s) with proteinaceous components in the boundary membranes, such as a putative membrane-bound chaperone (Corpas and Trelease 1997 ) or a putative PMP receptor (Ballard et al. 1998 ; Figure 6A, lane 6).

One realistic concern related to our interpretations of in vitro insertions of membrane proteins was the possibility of nonspecific, hydrophobic interactions between peroxisomal APX and any added target membrane. Such was not the case, however, because nascent peroxisomal APX was consistently inserted at high efficiencies into MEMs but not to any appreciable extent into other added membranes derived from mitochondria, plastids, plasma membranes, or peroxisomes (Figure 6B). An important control was that the peroxisome membranes used in this study were indeed capable of acquiring a PMP in vitro. C. boidinii PMP47 inserted specifically into peroxisome membranes but not into MEMs (Figure 6B, lanes 12 and 13), confirming the findings of Dyer et al. 1996 , who showed that PMP47 did not sort to ER in vivo but was targeted directly from the cytosol to peroxisomes. Two other experiments performed as positive controls revealed that mouse VAMP2 and S. cerevisiae Pex15p were inserted into MEMs and not into peroxisomes, as was predicted from published results (Kutay et al. 1995 ; Elgersma et al. 1997 ). These controls add credibility to our contention that newly synthesized peroxisomal APX selectively sorts post-translationally to, and inserts into, the ER in plant cells. We were not able, however, to learn from the in vitro experiments whether peroxisomal APX sorts to a subdomain of ER, as had been postulated from the immunofluorescence studies.

Does Peroxisomal APX Sort to Peroxisomes by Means of ER Transport Vesicles?
Virtually all preexisting peroxisomes become aggregated as CAT-pAPX sorts to these organelles, a phenomenon that provided an added dimension for interpreting the results of BFA treatments because it enabled us to distinguish between events that involved putative preperoxisomal compartments or preexisting peroxisomes. Detailed results describing the progressive aggregation of CAT-pAPX–bound peroxisomes (apparently by oligomerization of CAT polypeptides on the cytosolic face of these peroxisomes) will be presented in another paper (R.T. Mullen, C.S. Lisenbee, and R.N. Trelease, in preparation).

In BFA-treated cells, the accumulation of CAT-pAPX in the ER (circular/reticular network; Figure 5A) and its absence in preexisting peroxisomes (peroxisomes not aggregated; Figure 5B) are interpreted to result from BFA preventing the formation/exit of vesicles from the ER that ordinarily would sort to preexisting peroxisomes. Supporting evidence for this interpretation was obtained from immunofluorescence images of cells that had been relieved from the effects of BFA. In these cells, CAT-pAPX was observed in aggregated preexisting peroxisomes (Figure 5C and Figure 5D) apparently because the CAT-pAPX was allowed to exit the ER. BFA affects vesicular formation/transport from the ER (Klausner et al. 1992 ; Staehelin and Driouich 1997 ), and BFA treatment in another system has been interpreted to block ER vesicle formation and subsequent transport to preexisting peroxisomes (Salomons et al. 1997 ); consequently, we suggest that CAT-pAPX exits ER within vesicles.

Two sets of positive controls were performed as part of the BFA experiments, namely, the post-translational targeting of a mitochondrial protein (CATß60) and a peroxisomal matrix protein (CAT-SKL) in the presence of BFA. The protein destined to insert into mitochondria was acquired by mitochondria without any notable effect of BFA (Figure 5H to 5J). The same result was expected for CAT-SKL because peroxisomal matrix proteins generally are thought to be targeted directly from the cytosol to preexisting peroxisomes (Olsen 1998 ; Subramani 1998 ). However, our results with the sorting of CAT-SKL, and the results of Salomons et al. 1997 with H. polymorpha alcohol oxidase (a peroxisomal matrix enzyme), in BFA-treated cells shed a different light on this widely accepted view. CAT-SKL accumulated mostly in the cytosol (Figure 5G), whereas alcohol oxidase accumulated in the ER.

Salomons et al. 1997 suggested that H. polymorpha matrix proteins are delivered to their target organelle together with membrane components by accumulating with PMPs either in the ER or in transport vesicles that migrate to preexisting peroxisomes. In BY-2 cells, the mostly cytosolic localization of CAT-SKL seems to be a strong indication that CAT-SKL (representing most or all of the matrix-destined proteins with a PTS1) is not sorted directly to peroxisomes or to the ER. Why is some of the CAT-SKL located in peroxisomes (Figure 5G)? An explanation similar to one given for alcohol oxidase sorting in H. polymorpha (Salomons et al. 1997 ) is that CAT-SKL normally sorts to ER-derived transport vesicles and that some of the expressed CAT-SKL was sorted to such vesicles that had formed before uptake of BFA. Or, perhaps CAT-SKL accumulated in the cytosol as an indirect consequence of the failure to assemble membrane-bound components of the translocation machinery. Of course, some matrix proteins may sort directly to preexisting peroxisomes.

Salomons et al. 1997 further speculated that in yeast cells, only a subset of PMPs, that is, "functional" membrane proteins such as transporters (e.g., PMP47) and presumably enzymes (e.g., peroxisomal APX), sorts directly from the cytosol to preexisting peroxisomes. In contrast, they suggested that PMPs involved in early stages of peroxisome biogenesis, that is, "early peroxins" such as Pex15p, sort indirectly to peroxisomes by way of the ER and ER-derived vesicles. Kunau and Erdmann 1998 make similar predictions for these types of proteins. Our in vitro data support the postulates for Pex15p and PMP47 (Figure 6B, lanes 8 and 9, and 12 and 13) but not for peroxisomal APX (Figure 6B, lanes 6 and 7). Thus, there appear to be various pathways for sorting PMPs, but at this early stage of investigation, our data do not support the generalization that is emerging from studies of yeast peroxisomal biogenesis that "functional" PMPs sort through a different pathway from that involving "biogenesis" PMPs.

Overview of a Sorting Pathway through a Putative Peroxisomal ER Subdomainm
The ER is a dynamic membranous compartment, constituted functionally as an assemblage of specialized regions or subdomains (Hepler et al. 1990 ; Sitia and Meldolesi 1992 ; Okita and Rogers 1996 ; Staehelin 1997 ). The distinct ER subdomain that seems to be involved in the sorting of peroxisomal APX to BY-2 peroxisomes is distinguishable as a portion of the reticular/circular ER to which HA-pAPX is localized and reticuloplasmins somehow are not (Figure 3H, Figure 3K, and Figure 3L). Reticuloplasmins (e.g., BiP, calnexin, and calreticulin) are believed to exist within the same subdomain of the ER (Crofts and Denecke 1998 ), although they (and their subdomains?) are not necessarily distributed uniformly throughout the ER (Okita and Rogers 1996 ). Clearly, much needs to be learned about the localizations of resident and cargo proteins within the ER. Our non-colocalization results for peroxisomal APX and the reticuloplasmins are consistent with current generalizations that include spatial and functional separations of these proteins.

The "privileged site budding model" (Kuehn and Schekman 1997 ) and modifications thereof (Aridor et al. 1998 ; Hobman et al. 1998 ) seem to offer a reasonable working explanation for the lack of colocalization of HA-pAPX and reticuloplasmins and for the putative vesicular exit of PMPs from specific subdomains of the ER. In general, the models suggest that cargo proteins sequestered within the lumen or membrane of the ER are concentrated within specific subdomains or "privileged sites" before export by COPII vesicles. "Gating proteins" are postulated to allow cargo to enter the privileged sites while restricting the access of most ER-resident proteins, including reticuloplasmins. Those resident proteins that inadvertently become removed from the ER are retrieved through retrograde Golgi vesicles (Kuehn and Schekman 1997 ).

Figure 7 presents a hypothetical model illustrating the possible post-translational sorting pathway(s) of peroxisomal membrane (peroxisomal APX) and matrix (CAT-SKL) proteins in plant cells. Rather than using the general term "privileged site," we propose the term "peroxisomal ER subdomain" (pER). In Figure 7, cytosolic molecular chaperones are shown associated with nascent polypeptides because these chaperones seem to be involved in organelle acquisition of PMPs (Figure 6C) and matrix proteins (Crookes and Olsen 1998 ; Pool et al. 1998 ). Although our data are meager for suggesting an initial sorting of matrix proteins mainly into putative transport vesicles rather than directly into preexisting peroxisomes, this alternative sorting pathway should be considered seriously in view of recent studies implicating ER and vesicular transport in peroxisomal biogenesis in plants and other organisms. Whether PMPs sort from the cytosol initially to an ER that possesses reticuloplasmins or to pER is not known. Our expectation is that they sort first to ER subdomains with the reticuloplasmins because these residents are involved in both post-translational and cotranslational uptake of proteins into the ER (Denecke et al. 1991 ; McClellan et al. 1998 ), and because Pex2p and Pex16p colocalize with Kar2p (a BiP homolog) in the ER of Y. lipolytica cells (Titorenko and Rachubinski 1998b ). If PMPs insert into reticuloplasmin ER, then they presumably move to and become localized within pER. Reticuloplasmins probably would be excluded from pER because they could not be retrieved by a peroxisomal retrograde vesicle system. The mechanisms for these latter processes are completely unknown.

View larger version (35K): [in this window] [in a new window]   Figure 7. Model for the Sorting of Membrane and Matrix Proteins to Plant Peroxisomes. Membrane and matrix proteins synthesized on cytosolic polysomes appear to interact with molecular chaperones before post-translational organelle sorting. PMPs (such as peroxisomal APX) may sort first to the reticuloplasmic-containing ER or to a peroxisomal ER (pER) subdomain (reticular/circular structure). Exit of PMPs from pER seems to involve vesicles that are subsequently sorted to preexisting peroxisomes (evidence from BFA experiments). Matrix proteins (with PTS1 or PTS2) also could sort indirectly (possibly to vesicles), or in some instances directly (dashed line), to preexisting peroxisomes.

Properly conceived and executed time-course studies clearly are needed to help resolve potential reticuloplasmin–peroxisomal APX interactions and to provide direct evidence for vesicular transport of PMPs from pER to peroxisomes. Such studies should provide meaningful modifications of the model in Figure 7. Meanwhile, that model serves not only as a base on which to test the validity of sorting pathways for peroxisomal proteins but also as a useful starting point for considering how peroxisomes differentiate (enlarge) and proliferate in plant cells.

	ME