Binding and functions of ADP-ribosylation factor on mammalian and yeast peroxisomes.

We have analyzed in vitro the binding characteristics of members of the ADP-ribosylation factor (ARF) family of proteins to a highly purified rat liver peroxisome preparation void of Golgi membranes and studied in vivo a role these proteins play in the proliferation of yeast peroxisomes. Although both ARF1 and ARF6 were found on peroxisomes, coatomer recruitment only depended on ARF1-GTP. Recruitment of ARF1 and coatomer to peroxisomes was significantly affected both by pretreating the animals with peroxisome proliferators and by ATP and a cytosolic fraction designated the intermediate pool fraction depleted of ARF and coatomer. In the presence of ATP, the concentrations of ARF1 and coatomer on peroxisomes were reduced, whereas intermediate pool fraction led to a concentration-dependent decrease in ARF and increase in coatomer. Brefeldin A, a fungal toxin that is known to reduce ARF1 binding to Golgi membranes, did not affect ARF1 binding to peroxisomes. In Saccharomyces cerevisiae, both ScARF1 and ScARF3, the yeast orthologs of mammalian ARF1 and ARF6, were implicated in the control of peroxisome proliferation. ScARF1 regulated this process in a positive manner, and ScARF3 regulated it in a negative manner.

The intracellular communication of peroxisomes with other subcellular compartments is only poorly explored. Questions such as how peroxisomes accept their membrane phospholipids or distribute their ether lipid precursors are still unanswered. The endoplasmic reticulum (ER) 2 as the central station of phospholipid synthesis might play a particular role in these processes. Recently, it was shown that the ER is implicated in the biogenesis of peroxisomes (1,2). In Saccharomyces cerevisiae Pex3p, an integral peroxisomal membrane protein (PMP) was demonstrated to first assemble in the ER before budding in a Pex19pdependent manner (3). On the other hand, fusion of distinct precursor vesicles to generate functional peroxisomes was suggested, although it is not clear from which membrane these precursor vesicles are derived (4).
Immunofluorescence experiments in mammalian cells have shown that under appropriate conditions, PMPs segregate into membrane patches and proliferate, involving dynamin-like protein 1 (DLP1). This member of the dynamin family of large GTPases was recently localized to yeast and mammalian peroxisomes, and expression of a dominant negative mutant inhibited peroxisomal division (5)(6)(7). The mechanisms underlying these intercommunicative processes are not clear; however, they might include the generation and fusion of vesicles. In this context, the observation is of great interest that isolated peroxisomes recruit ARF and the COPI coat complex (coatomer) from the cytosol. Peroxisomal coatomer recruitment was dependent on GTP and sensitive to protease pretreatment of the organelles (8,9).
ARF-dependent coatomer recruitment plays a dominant role in Golgi transport processes (10). Binding of ARF and coatomer to peroxisomes thus may indicate that peroxisomes are functionally linked to another cellular compartment, in order to enable transport and exchange of materials. ARF molecules are widely implicated in membrane traffic within eukaryotic cells functioning in processes of the secretory pathway (10 -12) and the endocytic transport mediating recruitment of the COP I coat and the clathrin adaptor proteins AP-1, AP-3, and AP-4. Whereas the COP I coat is involved in both the retrograde Golgi to ER and the anterograde Golgi transport, AP-1 and AP-4 are recruited to the trans-Golgi network and AP-3 to endosomal membranes (see Refs. 13 and 14). Besides vesicular transport, ARF1 seems also to be implicated in the assembly of apolipoprotein B-containing very low density lipoproteins by activating phospholipase D and in regulating in its GDP-bound state the association of adipocyte differentiation-related protein to lipid droplets (15,16). There may be still other processes implicating ARF molecules, since in mammals six ARF subtypes occur, and to date the role of many of them is not really understood (17,18). In S. cerevisiae, by comparison, three ARF subtypes exist (17,19). ScARF1 and ScARF2 functionally complement each other in regulating Golgi transport, whereas the role of ScARF3 that is neither implicated in Golgi transport nor essential for viability (20) still remains to be elucidated.
In the present study, we further explored peroxisomal ARF and coatomer binding. Our results demonstrate that (i) ARF-GTP binding to peroxisomes in vitro is essential for subsequent binding of coatomer, (ii) it is the subtype 1 of ARF that favors coatomer recruitment, (iii) ARF and coatomer binding is affected by ATP and a cytosolic factor but is not inhibited by brefeldin A (BFA), and (iv) besides ARF1, ARF6 is also found on peroxisomal membranes. Utilizing various ARF mutants of the yeast S. cerevisiae, we show that ScARF1 is essential for oleateinduced peroxisome proliferation and that deletion of ScARF3, the yeast ortholog of mammalian ARF6, significantly stimulates proliferation. The results suggest that ARF1 is essential for coatomer recruitment to peroxisomes and that, besides ARF1, ARF6 may also be involved in the control of peroxisomal processes.

EXPERIMENTAL PROCEDURES
Materials-AMP-PNP, GMP-PNP, and GDP␤S were purchased from Sigma, and ATP and creatine kinase were from Roche Applied Science. Creatine phosphate was from Calbiochem.
Antibodies-ARF1-and ARF6-specific antibodies were raised in rabbits. For ARF1C1 antibody, animals were injected with the C-terminal peptide of ARF1 (CDWLSNQLRNQK) coupled to keyhole limpet hemocyanin (Sigma). ARF6C1 antibody was raised against the C-terminal peptide CGLTWLTSNYKS of ARF6. The antibody 2048 raised against recombinant ARF1 (21) and recognizing different ARF isotypes was a gift from Dr. Bernd Helms (Utrecht University, The Netherlands). Unless otherwise stated, ARF1C1 and ARF6C1 were used to detect ARF1 or ARF6, respectively. Polyclonal antibody specific for ␤ЈCOP (22) was kindly provided by Dr. Felix Wieland (Universität Heidelberg, Germany). Mouse monoclonal anti-GM130 was purchased from BD Biosciences, and monoclonal mouse anti-HA (HA.11, clone 16B12) was from Babco. Polyclonal antibodies recognizing peroxisomal proteins were raised in rabbits. The anti-PMP36 antibody recognizes a 36-kDa peroxisomal membrane protein. The anti-PMP69 and anti-Pex11␣p antibodies were described elsewhere (8). Secondary antibodies antimouse peroxidase and anti-rabbit peroxidase were from Dianova (Hamburg, Germany).
Isolation of Subcellular Fractions-Cytosol and peroxisomes were isolated from rat liver of either control or clofibrate-treated (0.5% clofibrate added to the chow for 5-7 consecutive days) male Wistar rats. Animals were starved 16 h before the preparation.
Cytosol was prepared according to Ref. 8. Briefly, all steps were carried out at 4°C. Livers from six rats were collected in buffer H (50 mM Hepes, pH 7.55, 165 mM KOH, 2 mM MgAc 2 , 1 mM dithiothreitol) supplemented with protease inhibitors (1 M leupeptin, 10 M antipain, 100 M phenylmethylsulfonyl fluoride), and a homogenate (1:10) was prepared. After removing nuclei and mitochondria (10 min at 6,000 ϫ g), the supernatant was subjected to centrifugations at 100,000 ϫ g for 60 min followed by 180,000 ϫ g for 90 min (TFT 55.38-rotor; Kontron Instruments, Hanau, Germany). The resulting supernatant was concentrated about 10-fold in a MINITAN ultra concentration unit (molecular weight cut-off 10,000; Millipore, Eschborn, Germany). The concentrated cytosol was finally cleared at 100,000 ϫ g for 60 min and stored in aliquots at Ϫ80°C. The cytosol had a final concentration of 45-50 mg/ml. Fractionation of stimulated cytosol was performed on a Superdex 200 column (HiLoad XK 16/60; Amersham Biosciences) in buffer H applying 2 ml of cytosol, collecting 20 fractions of 4 ml each. The coatomerand ARF-containing fractions were identified by Western blotting. Pooled fractions were concentrated to a final volume of 500 -800 l using the Vivaspin 20 units according to the manufacturer's protocol (Vivascience, Hannover, Germany) to obtain the coatomer pool fraction (CPF), the intermediate pool fraction (IPF) and the ARF pool fraction (APF), respectively.
Highly purified peroxisomes were isolated as described previously (23). The isolated peroxisomes usually had a concentration of 2.5-3.5 mg/ml and were stored in small aliquots at Ϫ80°C.
Rat liver Golgi was prepared according to Ref. 24 and was kindly provided by Dr. Britta Brügger (Universität Heidelberg, Germany).
Incubation of Peroxisomes and Golgi-Peroxisomes were incubated with cytosol as described in Ref. 8, applying slight modifications. Peroxisomes were resuspended in sucrose buffer (10 mM glycylglycin, pH 7.4, 250 mM sucrose, 1 mM EDTA) at a concentration of 1.25 mg/ml, and cytosol was added in a 40 -50-fold excess of protein over peroxisomes. For incubations with cytosolic fractions, the protein amount was calculated equivalently, and the volume was topped up with buffer H. Incubations were carried out in the presence of a 50 -100 M concentration of either GMP-PNP or GDP␤S. Unless otherwise mentioned, all incubations were done in the presence of an ATP regenerating system (ARS) containing 2 mM ATP, 20 mM creatine phosphate, 250 g/ml creatine kinase. GMP-PNP, GDP␤S, and ARS were prepared as 40-fold stocks in 50 mM Hepes/KOH, pH 7.5. The organelles were incubated for 30 min at 37°C and reisolated after 30 min on ice and washing in sucrose buffer (10 min, 15,000 ϫ g, 4°C).
Incubations with recombinant ARF proteins were done in the same way, by incubating 50 g of peroxisomes with 1 g of the ARF protein.
CPF was used at a concentration equivalent to 2 mg of complete cytosol.
Incubations of Golgi membranes were done essentially as described for peroxisomes, except for stripping the Golgi membranes (300 mM KCl in sucrose buffer) prior to incubation, in order to remove bound ARF and coatomer.
Incubations with BFA were done according to Ref. 27. The organelles were preincubated with 160 M BFA for 10 min at 37°C. BFA was dissolved in methanol (8 mM). Controls were performed by adding an equal amount of the solvent, the final methanol concentration in all assays not exceeding 2% (v/v). After adding 50 M GMP-PNP and the appropriate amount of APF, the incubation was continued for 20 min at 37°C. All samples were floated up through 43% (w/v) Nycodenz, the organelles were reisolated, and samples were analyzed by Western blotting.
Lipid Determination-The phospholipid (PL) content of isolated peroxisomes and Golgi was determined as described (24).
Mass Spectrometry and Edman Sequencing-Peroxisomal membranes were prepared by carbonate treatment (28) and incubated with APF and GMP-PNP as described above for intact peroxisomes. Samples were treated with 1 M KCl for 30 min on ice, and the membranes were reisolated by centrifugation. The pellet was dissolved in sample buffer and run on a 12% SDS-PAGE. The protein band of 20 kDa was excised from the gel (stained with Coomassie R-250), digested with trypsin, and analyzed by matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) on a Reflex II spectrometer (Bruker-Daltonic, Bremen, Germany). Alternatively, the excised protein band was digested in gel using Asp-N (Roche Applied Science), and the peptides obtained were separated by HPLC. Distinct peptides were analyzed by Edman sequencing as described previously (29).
Yeast Strains and Culture Conditions-The S. cerevisiae strains used in this study are listed in TABLE ONE. Unless otherwise stated, strains without plasmids were grown in complete YPD medium (1% yeast extract, 2% peptone, 2% glucose), whereas strains containing plasmids were selected on SD minimal medium (0.67% yeast nitrogen base without amino acids, 2% glucose) supplemented with the appropriate amino acids. Standard growth temperature was 25°C. Sporulation, tetrad dissection, and scoring of genetic markers were performed as described (30). Transformation of yeast was accomplished by the lithium acetate method (31). SSY145 (arf1-wt, arf2⌬) and SSY146 (arf1-ts, arf2⌬) were kindly provided by Dr. Stephan Schröder-Köhne (Würzburg, Germany). SCMIG631 (arf3⌬) was obtained by substituting the chromosomal ARF3 open reading frame with LEU2 marker in SCMIG22. SCMIG561 (arf1-ts, arf2⌬, arf3⌬) was obtained by crossing strain SCMIG631 to SSY146, diploids were selected and sporulated, tetrads were dissected, and segregants were scored for the appropriate markers. To visualize the peroxisomal compartment, cells were transformed with a GFP construct containing a C-terminal peroxisomal targeting signal 1 (GFP-PTS1). For transformation of SSY145 and SSY146, the plasmid pRS315 was used. Whereas SCMIG22, SCMIG631, PC70 (32), and RH1436 (33) were transformed using pRS416, pRS414 was introduced into SCMIG561. Oleic acid-induced peroxisome proliferation was performed by growing the preculture to late log phase in SD medium. Subsequently, the cells were kept continuously in the log phase for 3 ϫ 6 h by repeated inoculations and growth of the cultures. For the temperature shift experiments, the appropriate strain was transferred to the restrictive temperature (35°C) 30 min prior to incubations in the oleic acid-containing medium. The precultured cells were transferred to temperature-equilibrated YNO medium (0.67% yeast nitrogen base without amino acids, 0.1% yeast extract, 0.1% (w/v) oleic acid, 0.02% (w/v) Tween 40, amino acids as required) and grown for 6 h at the temperature indicated.
To determine the number of peroxisomes per cell, cells were immobilized in calcium alginate gel on coverslips (34) and immediately analyzed by confocal laser-scanning microscopy using a Leica TCS SP (Leica Microsystems, Mannheim, Germany). For each viewing field, nine consecutive layers were taken together with the brightfield image. The brightfield image and the layers 3, 5, and 7 were merged, and the numbers of both cells and peroxisomes were determined. For each con-dition tested, at least 200 cells were analyzed to calculate the average number of peroxisomes per cell.

Specificity of ARF Binding to Isolated Rat Liver Peroxisomes-Several
peroxisomal membrane proteins including Pex11␣p are increased in concentration more than 10-fold by eliciting peroxisome proliferation. This process implicating peroxisome proliferator-activated receptor ␣ is induced by a series of drugs and rather specifically affects the peroxisomal compartment (35). We investigated ARF and coatomer binding using preparations of highly purified peroxisomes and cytosol from proliferation-stimulated and -unstimulated rat livers (Fig. 1). Binding of ARF and coatomer, visualized with antibodies specific for ARF and ␤ЈCOP, was strictly dependent on GTP and did not occur with GDP. Irrespective of the use of stimulated or unstimulated cytosol that both contained the same amounts of ARF and coatomer (Fig. 1A), stronger ARF binding was observed on stimulated than on unstimulated peroxisomes (Fig. 1B). In contrast to that, about double the amount of coatomer was recruited from stimulated compared with unstimulated cytosol, irrespective of the stimulation state of the peroxisomes. Although the amount of Pex11␣p was much higher in stimulated than in unstimulated peroxisomes, the amount of PMP22p, a peroxisomal marker not induced by peroxisome proliferators, was the same in all assays, demonstrating equal amounts of peroxisomes applied. Care was taken in these experiments to avoid artifactual precipitation of cytosolic proteins in the absence of organelles that easily occurred in the assays containing GTP␥S or GMP-PNP. For that reason, prior to SDS-PAGE and Western blotting, peroxisomes were recovered by flotation instead of pelleting.
ARF1-dependent recruitment of coatomer is a process highly active in the Golgi apparatus (10, 36 -38). Thus, the absence of Golgi membranes from the peroxisomal preparation is critical. In order to assess the Golgi content of our peroxisome preparation, we analyzed ARF and coatomer binding in parallel to both peroxisomes and rat liver Golgi membranes (Fig. 2). In each assay, comparable amounts of membranes were analyzed based on their phospholipid rather than protein content. From the results of these experiments, we concluded that (i) the GMP-PNP-dependent ARF and coatomer binding to Golgi is dependent on the amount of membranes; (ii) compared with peroxisomes, Golgi membranes recruit about half the amount of ARF but double the amount of coatomer; (iii) Golgi, but not peroxisomal membranes, consistently recruit small but significant amounts of coatomer also in the presence of GDP␤S. GM130, a Golgi matrix protein and putative vesicle docking receptor (39), was used as a Golgi marker to monitor Golgi contamination in the isolated peroxisome fraction. Even after prolonged exposition of the blot (Fig. 2, last line), no GM130 signal was recognized in the peroxisome preparation. Identification of the ARF Subtype Recruited to Peroxisomal Membranes-By mass spectroscopic analysis, we determined which of the six ARF subtypes is recruited to peroxisomal membranes. We analyzed the binding of cytosolic ARF to untreated peroxisomes or to peroxisomes treated with either 1 M KCl or 100 mM carbonate. These treatments should minimize matrix and peripherally attached membrane proteins that could interfere with the mass spectroscopic analysis of the bound ARF proteins. As seen in Fig. 3A, the high salt and high pH treatment did not noticeably affect ARF and coatomer binding, and therefore carbonate membranes were used in the following experiment. ARF-and coatomer-containing cytosol fractions were obtained by separating rat liver cytosol by gel chromatography into three main fractions designated CPF, IPF, and APF (Fig. 3B). Following incubation with APF, peroxisomes were recovered, washed, and subjected to SDS-PAGE. Coomassie-stained polypeptides were digested in gel using either trypsin or Asp-N cleavage, and the isolated peptides were analyzed by MALDI-MS or HPLC separation and Edman sequencing. The peptide masses and sequences obtained are listed in TABLES TWO and THREE. Due to the high sequence homology within the ARF subtypes, many peptide sequences, as indicated in TABLE TWO, matched more than one subtype. However, five peptide sequences were obtained that were specific for ARF6. By Asp-N digestion and Edman degradation of HPLC-separated peptides, we identified another three sequences specific for ARF6 and one sequence specific for ARF1 (TABLE THREE), suggesting that isolated rat liver peroxisomal membranes specifically bind ARF1 and ARF6 from a cytosolic ARF-containing fraction. How-ever, these data do not exclude the possibility that still other ARF subtypes may bind to peroxisomes.
Peroxisomal Coatomer Binding Is Mediated by the ARF1 Subtype-These results raised two major questions. (i) Is ARF binding to peroxisomes essential for subsequent recruitment of coatomer? (ii) Which ARF subtype is involved in this process? To answer the first question, peroxisomes were incubated in the presence of GMP-PNP with either cytosol, individual pool fractions, or combinations of these pool fractions. As shown in Fig. 4A, recruitment of coatomer to peroxisomal membranes required the presence of ARF. When the APF was omitted from the incubations, no coatomer recruitment was detected. The IPF principally was not necessary but significantly enhanced coatomer binding and reduced binding of ARF to an extent similar to that observed with complete cytosol. Using ARF subtype-specific antibodies, recruitment of both ARF1 and ARF6 to peroxisomes was noticed. Interestingly, only ARF1 binding was affected by the IPF (Fig. 4A).
In order to investigate the ARF involved in coatomer binding, we analyzed in parallel the activities of recombinant myristoylated ARF1 and ARF6, both HA-His-tagged. As shown in Fig. 4B, ARF1 produced a significant GMP-PNP-dependent signal for ␤ЈCOP, whereas with ARF6, a nucleotide-independent background signal was observed. Tagging the ARF1 protein slightly altered its properties. Whereas the tagged version also weakly associated in its GDP form with the peroxisomal membrane and was accompanied by a faint coatomer signal, this effect was observed neither with the cytosolic nor with the recombinant wild type ARF1. These two proteins were only active in their GTP-bound state mediating distinct coatomer recruitment (Fig. 4C). The concentration of PMP36p was used as a marker for the amount of peroxisomes loaded onto the gel. In summary, these data indicate that the ARF1 FIGURE 1. Peroxisome proliferators affect ARF and coatomer binding to rat liver peroxisomes. Cytosol and highly purified peroxisomes were isolated from normal rats (U, unstimulated) and rats that have been treated with the peroxisome proliferator clofibrate (S, stimulated). A, cytosol from clofibrate-treated animals (S) as compared with normal cytosol (U) did not show altered concentrations of ␤ЈCOP (arrowhead) and ARF (antibody 2048). B, peroxisomes (250 g of protein) were incubated with cytosol in the presence of ARS and either GMP-PNP (lanes 1-4) or GDP␤S (lanes 5-8) prior to floating through a layer of 48% Nycodenz (5 h at 150,000 ϫ g). The recovered organelles were precipitated and analyzed by Western blotting. The signal for Pex11␣p was used to visualize peroxisome proliferation, since Pex11␣p is known to become induced by peroxisome proliferators. The signal for PMP22p, a PMP not induced by peroxisome proliferators, indicates an equal load of each lane. Note that increased ARF binding (antibody 2048) is observed on stimulated peroxisomes, whereas increased coatomer binding is seen in the presence of stimulated cytosol. No nonspecific flotation is observed in cytosol incubations omitting the organelles (lanes 9 and 10). The given data are representative of three independent experiments. subtype is the physiologically relevant regulator of peroxisomal coatomer binding.
Factors Influencing ARF and Coatomer Binding to Peroxisomes-Searching for factors and conditions that influence ARF1 binding to peroxisomes, we first investigated a role of ATP. Peroxisomal incubations with cytosol and GMP-PNP were performed in the absence and presence of an ARS (Fig. 5A). Presence of ATP reduced both ARF1 and coatomer binding to peroxisomes,     . Note that ARF is essential for coatomer binding, and the addition of IPF reduces the concentration of bound ARF1 but not ARF6. Under these conditions, ARF6 binding was only slightly influenced. Pex11␣p was used to monitor the amount of peroxisomes used for the assays. B, to discriminate which of the two ARF subtypes mediates coatomer binding, peroxisomes (50 g of protein) were incubated with recombinant HA-and His-tagged ARF1 and ARF6 in the presence of CPF, ARS, and either GMP-PNP or GDP␤S, as indicated. ARFs were detected by an anti-HA antibody. Note that only ARF1 mediates a GTP-dependent recruitment of coatomer (lanes 1 and 2). C, comparison of ARF1 and coatomer recruitment on incubations of peroxisomes (50 g of protein) with CPF and ARS in the presence of either APF (lanes 1 and 3) or recombinant untagged ARF1 (recARF1wt, lanes 2 and 4). The PMP36 and ␤ЈCOP signals visualized peroxisome loads and coatomer binding, respectively. The signals were quantified by densitometry using the Quantity One software. Values were calculated as percentage of maximum. The error bars designate the range of experimental values from at least two independent trials.

Identification of ARF subtype-specific peptides by Asp-N digestion and Edman degradation
with ARF1 binding being more distinctly affected. This effect of ATP was dependent on the hydrolysis of ATP, since the addition of AMP-PNP, a nonhydrolyzable ATP analog, did not influence the amount of ARF1 bound to peroxisomes (Fig. 5A). In contrast to this, ARF6 recruitment was even enhanced by both ATP and its nonhydrolyzable analog, suggesting distinct mechanisms of binding for each of these ARF molecules. Another factor that significantly affected ARF1 but not ARF6 binding to peroxisomes was found to be present in the IPF. Adding IPF to incubations containing APF and CPF reduced the amount of peroxisome-bound ARF1 to the level observed in incubations with complete cytosol (Fig. 4A). To investigate these effects of ATP and IPF in more detail, we compared ARF1/coatomer binding from both complete cytosol and the cytosolic fractions either in the absence or the presence of ATP. The results of these experiments are shown in Fig. 5B. Whereas IPF reduced ARF1 binding and enhanced coatomer binding to peroxisomes, ATP, irrespective of the presence of IPF, clearly lowered both ARF1 and coatomer bound to peroxisomes (Fig. 5B, lanes 1-4). These data suggest that the activity contained in the IPF is ATP-independent and that IPF and ATP work synergistically.
Control incubations with cytosol/GMP-PNP omitting the organelles did not generate any ARF1 or ␤ЈCOP signals. By using peroxisomal membranes obtained by carbonate treatment, we observed that unlike intact peroxisomes, these membranes no longer showed the ATP-dependent reduction in ARF and coatomer binding when incubated with APF/CPF (Fig. 5, B and C, lanes 1 and 2). However, adding IPF to the membranes restored the ATP effect that was also observed in the incubations with complete cytosol (Fig. 5C, lanes 5 and  6). Thus, we conclude that the ATP-dependent activity resides on the peroxisomal membrane, is removed from the membrane by carbonate treatment and can be recruited from the IPF.
In order to further characterize the IPF-contained activity, we analyzed the stoichiometric relationship between IPF and ARF1/ coatomer binding in the presence and absence of ATP. The concentrations of IPF applied were normalized to the original cytosol (Fig.  6A, ϩcytosol), and the Pex11␣p signals were used to normalize the concentrations of ARF1 and ␤ЈCOP for equal amounts of peroxi- were incubated with GMP-PNP and cytosol or the indicated pool fractions in the presence or absence of ARS. Note that less ARF1 and coatomer is recruited in the presence of ATP, whereas IPF reduced ARF1 binding and enhanced coatomer recruitment. C, ARF1 and coatomer recruitment to peroxisomal membranes void of peripheral proteins. Peroxisomal membranes (10 g of protein) were incubated under the same conditions as described for intact peroxisomes (B). Note that ARF1 and coatomer binding are not dependent on ATP in the absence of IPF (lanes 1 and 2). Pex11␣p represents peroxisome load, and ␤ЈCOP monitors coatomer. The ARF1 and ␤ЈCOP signals were quantified by using the Quantity One software. Data were normalized to the Pex11␣p signal and given as a percentage of maximum. Similar results were obtained in three independent runs of the experiment. One characteristic set of data is shown. somes in the assays. As a result, the IPF, regardless of whether ATP was present or not, reduced in a concentration-dependent manner the amount of ARF1 bound to peroxisomes up to a constant level (Fig. 6A). The binding characteristics in the presence and absence of ATP were quite similar, except that without ATP, the initial amount of bound ARF1 was ϳ2.5-fold higher. As the amount of bound ARF1 decreased by titrating in IPF, the concentrations of ␤ЈCOP increased until a level of saturation. Thus, the IPF activity facilitates recruitment of coatomer by either preventing association of excess ARF1 or clearing the peroxisomal membrane from excess ARF1.
ARF binding to peroxisomes was dependent on GTP, suggesting ARF-GEFs being implicated in peroxisomal ARF functioning. Responsible for the nucleotide exchange activity on ARF are two classes of Sec7-GEFs, the high (Ͼ100 kDa) and the low molecular mass (Ͻ100 kDa) GEFs. In addition to the size difference, the two classes of GEFs respond differently to the fungal toxin BFA that specifically blocks the high molecular weight ARF-GEFs inhibiting ARF activation (40). In order to characterize the ARF-GEF activity on peroxisomes, we investigated in vitro its BFA sensitivity and compared it with that on Golgi membranes (Fig. 6B). Whereas ARF1 recruitment was completely blocked on Golgi membranes, BFA had no effect on peroxisomal ARF binding, suggesting distinct classes of ARF1-GEFs to be active on these organelles.
Functions of ARF1 and ARF3 on Peroxisome Proliferation in Yeast-Similar to peroxisomes of rat liver, yeast peroxisomes proliferate in response to various stimuli. Depending on the strain, fatty acids, methanol, or methylamine as the sole carbon source induce proliferation (41). Thus, the yeast system offers a possibility to investigate peroxisome proliferation in vivo. S. cerevisiae contains three ARF proteins: ScARF1, the ortholog of the mammalian ARF1; ScARF2, which functionally overlaps with ScARF1; and ScARF3, corresponding to mammalian ARF6 (19,20). Using deletion (⌬) or temperaturesensitive (ts) mutants of the ARF proteins, we studied a role of ScARF1 and ScARF3 in peroxisome proliferation. The genotype of the strains used in these experiments is described in TABLE ONE. Analysis was facilitated by transforming the cells with a construct that codes for a GFP molecule containing a C-terminal PTS1 (42). Fig. 7 shows the GFP fluorescence merged with the Nomarski image of control cells grown on glucose at the permissive temperature of 25°C and the arf1-ts/arf2⌬, arf3⌬, and ␣COP-ts strains 6 h after growth on oleate. The temperature-sensitive mutants (arf1-ts/arf2⌬ and ␣COP-ts) were grown at the nonpermissive temperature of FIGURE 6. Two independent activities, ATP and a cytosolic IPF-contained factor, control peroxisomal ARF1 and coatomer binding; BFA does not affect this ARF1 binding. A, peroxisomes (50 g of protein) were incubated with GMP-PNP, APF/CPF, and increasing concentrations of IPF in the presence (ϩATP) and absence (ϪATP) of ARS. The amount of IPF contained in the cytosol incubations represents 100%. Whereas maximal ARF1 binding (100%) was observed in the absence of both IPF and ATP, ␤ЈCOP binding was maximal (100%) with IPF concentrations equivalent to 200% cytosol in the absence of ATP. Note the ATPindependent but concentration-dependent effect of IPF to reduce the amount of recruited ARF1 and to enhance coatomer binding. Signal intensities are quantified as described in the legend to Fig. 5. Error bars designate the range of experimental values from at least two independent trials. B, peroxisomes (100 g of protein) and Golgi (15 g of protein) were preincubated with 160 M BFA as indicated, followed by incubation with APF/GMP-PNP and by flotation through a layer of 43% Nycodenz (2 h at 150,000 ϫ g). The reisolated organelles were analyzed by Western blotting. Pex11␣p and p23, a Golgi membrane marker, were used to visualize the amount of either peroxisomal or Golgi protein loaded onto the gel. The representative results of four independent experiments are given. BFA that blocks ARF1 recruitment to the Golgi has no effect on the association of ARF1 to peroxisomes.
35°C. Whereas under glucose, the cells of the three strains contain on the average 2-3 peroxisomes/cell, after 6 h on oleate, only cells of the arf3⌬ strain have proliferated their peroxisomes. Interestingly, cells carrying a temperature-sensitive mutation in the ␣COP gene occasionally revealed strikingly elongated peroxisomes, a possible indication for impaired division. The quantitative evaluation of these experiments is shown in Fig. 8. To further analyze the role of ScARF1 on peroxisome proliferation, we used the strain arf1-ts/arf2⌬, deleted in ARF2 carrying a temperature-sensitive mutation in the ARF1 gene. The isogenic strain arf1-wt/arf2⌬ carrying wtARF1 in an arf2⌬ background served as a control. Cells were imaged by transmission light microscopy and confocal laser-scanning microscopy, and the number of both cells and peroxisomes was determined. The number of peroxisomes per cell in cells grown on glucose at 25°C was set to 0% proliferation. Transfer of the cells to oleate for 6 h at the permissive temperature increased the number of peroxisomes per cell significantly by 17 and 13% in the control and the arf1-ts/arf2⌬ strain, respectively. Thus, at the permissive temperature, these strains proliferate peroxisomes at comparable rates. At the nonpermissive temperature, control cells (arf1-wt/arf2⌬) on oleate increased the number of peroxisomes per cell by 33%, whereas in the arf1-ts/arf2⌬ strain, the number of peroxisomes remained unchanged, suggesting that ScARF1 is required for the oleate-induced peroxisome proliferation in S. cerevisiae.
The observation that mammalian peroxisomes bind ARF6 prompted us to further investigate the role of its S. cerevisiae ortholog ARF3 in peroxisome proliferation. Since ScARF3 is not essential for growth (20), an arf3 deletion strain (arf3⌬) was used. The isogenic wild type strain (arf-wt) served as the control. After 6 h on oleate, the arf-wt strain had increased the number of peroxisomes by 42%, whereas deletion of ScARF3 resulted in an even higher rate of proliferation, increasing the number of peroxisomes by 82%. In line with this, the triple mutation strain arf1-ts/arf2⌬/arf3⌬ increased the peroxisome number at the permissive and nonpermissive temperature by 100 and 29%, respectively. For the two strains arf1-ts/arf2⌬ (Fig. 8B) and arf1-ts/arf2⌬/arf3⌬ (Fig.  8C), the results were also presented in the form of a histogram profiling the distribution of peroxisomes per cell. By this evaluation, the percentage of cells containing more than two peroxisomes/cell increased by the oleate medium at the permissive temperature by 16% (arf1-ts/arf2⌬) and 120% (arf1-ts/arf2⌬/arf3D).
To exclude effects on peroxisome proliferation by just impairing vesicular transport processes due to ScARF1 inactivation, we analyzed peroxisome proliferation in a sec23-ts mutant strain blocked in protein export from the ER (33). As seen in Fig. 8A, these cells are not at all affected in oleate-mediated peroxisome proliferation, since their peroxisome number compared with the glucose-grown controls increased by 26 and 44% at the permissive and nonpermissive temperature, respectively. In summary, these results provide evidence that (i) ScARF1 is required for the oleate-induced proliferation of peroxisomes; (ii) deletion of ARF3 in the presence of a functional ARF1 remarkably increased proliferation, whereas in its absence proliferation is still maintained although to a much lower extent; and (iii) block of protein exit from the ER does not affect peroxisomal proliferation.

DISCUSSION
Specificity of Peroxisomal ARF Binding-So far, recruitment of ARF1 has been described to the Golgi, the trans-Golgi network and endosomes (10,13,36,43) mediating coatomer, clathrin/adaptor protein 1, 3, and 4 (AP-1, -3, and -4), and clathrin/GGA1-3 binding. Major activities are concentrated at the Golgi as the main sorting station. Based on marker enzyme assays, these membranes are virtually absent from our peroxisomal preparations that exhibit a degree of purification as high as 95-97%. Contaminations found are derived from ER and to a lesser degree from mitochondria and lysosomes (23) that all do not recruit ARF and coatomer (8,44). Due to the dominant role of the Golgi in ARF1/coatomer-dependent vesiculation processes, we particularly focused on the comparison of peroxisomes with the Golgi. Differences were observed in the binding characteristics of ARF1 and coatomer to equal amounts of peroxisomes and Golgi. Peroxisomes bound less coatomer but more ARF1 and did not recruit coatomer by ARF-GDP, as was consistently observed with Golgi membranes (Fig. 2).
Further indication for a peroxisomal specificity of ARF1 binding came from experiments utilizing peroxisomes and cytosol from rats fed clofibrate, a hypolipidemic drug and peroxisome proliferator (45). By this treatment, the capacity of both peroxisomes to recruit ARF1 and cytosol to promote coatomer binding was increased compared with untreated peroxisomes and cytosol (Fig. 1B). Both activities act synergistically suggesting two distinct factors to be involved, one localized to the peroxisomal membrane and the other to the cytosol. The pharmacological activity of peroxisome proliferators is triggered by their ligand interaction with peroxisome proliferator-activated receptor ␣ that is mainly expressed in liver, kidney, and intestinal mucosa and known to be involved in regulating fatty acid metabolism and peroxisome proliferation (35). Finally, the experiments utilizing mutant yeast strains defective in ARF functions also demonstrated a dependence of peroxisome proliferation on ARFs. Moreover, these studies were performed on intact cells, demonstrating the in vivo functioning of ARF on peroxisomes (see below).
Factors Involved in Peroxisomal ARF Binding-Previously, we have shown that protease pretreatment of peroxisomes reduced ARF binding and abolished coatomer recruitment, suggesting that protein(s) attached to the peroxisomal membrane are involved in this process (8,9). The present experiments further revealed that ATP and a cytosolic factor contained in the IPF affect ARF1 and coatomer binding to peroxisomes. ATPases that were confined to peroxisomal membranes include members of the family of both peroxisomal ABC transporters, such as PMP70p and the ALD protein, and AAA proteins, such as Pex1p and Pex6p. Whereas the peroxisomal ABC transporters function in the shuttling of long-chain and very long-chain fatty acyl-CoAs into peroxisomes (46), Pex1p and Pex6p have been implicated in late steps of per- oxisomal matrix protein import (47,48). Thus, clear metabolic and biogenetic roles were attributed to these proteins, making additional regulatory functions unlikely. Which kind of ATP-dependent process on peroxisomes could then be implicated? In a simplified explanation, we may interpret these results as the involvement of a kinase activity recruited from cytosol and bound to the peroxisomal membrane. Upon isolation of peroxisomes, this kinase activity at least partially remained associated with the organelle membrane. Recently, the casein kinase I␦ has been suggested to regulate the ARF-GAP1 binding to Golgi membranes, leading to both accelerated GTP hydrolysis on ARF1 and its membrane release (49). Although we could identify ARF-GAP1 in the IPF, we did not observe its recruitment to peroxisomes (data not shown). Considering that we used nonhydrolyzable GTP analogs in our experiments, GTP hydrolysis would not have been possible. Alternatively, the kinase implicated might be a phosphoinositide kinase, consistent with our recent observation that isolated peroxisomes synthesize phosphatidylinositol mono-and bisphosphates. 3 On Golgi membranes, ARF1 has been shown to directly associate in a PId-dependent manner involving both phosphatidylinositol 4,5-bisphosphate and phosphatidic acid (50 -53). Interestingly, the BFA-insensitive low molecular weight ARF-GEFs contain a pleckstrin homology domain within their common domain structure. This pleckstrin homology domain appears to mediate membrane association by binding to phosphatidylinositol 4,5-bisphosphate or phosphatidylinositol 3,4,5-trisphosphate (40). Thus, ARF might be recruited to membranes in a PId-dependent manner, stimulating various activities, including PId kinases (54), PLD (55,56), and vesicle coat protein association (57)(58)(59).
ARFs Regulate Peroxisome Proliferation in Yeast-The experiments utilizing yeast mutants defective in ARF signaling confirmed our findings of ARF1 and ARF6 (ScARF3) binding to the peroxisomal membrane. The simultaneous occurrence of these two ARF isoforms seems to be unique for peroxisomes. However, a functional explanation for this is lacking. Our results on oleate-induced proliferation of peroxisomes suggested an antagonistic effect of ScARF1 and ScARF3. So far we have envisaged two models that may help to explain these observations. Model 1 predicts that ScARF1 is involved in the up-regulation of 3 B. Jeynov, D. Lay, F. Schmidt, S. Tahirovic, and W. W. Just, manuscript in preparation. FIGURE 8. ScARF1 and ScARF3 regulate the oleate-induced peroxisome proliferation in S. cerevisiae. Cells of the indicated strains were grown on either glucose (SD medium) at 25°C or for 6 h on oleate at 25 or 35°C, respectively. Representative aliquots of at least 200 cells were analyzed for the number of GFP-labeled peroxisomes per cell following confocal laser-scanning imaging. A, the peroxisome proliferation (increase in the number of peroxisomes per cell) is expressed as a percentage, setting the proliferation in cells grown on glucose at 25°C to 0%. Note that ScARF1 regulates peroxisome proliferation positively, whereas ScARF3 regulates it negatively. B and C, for the two strains arf1-ts/arf2⌬ (B) and arf1-ts/ arf2⌬/arf3⌬ (C), the frequency distribution of cells containing a distinct number of peroxisomes is shown.
peroxisome proliferation, whereas ScARF3 triggers peroxisome degradation. This model explains many of the results obtained. However, under conditions where proliferation is blocked but degradation is still active (arf1-ts/arf2⌬), we might expect the number of peroxisomes to drop below that of normal wild type conditions rather than remaining at the normal level. In model 2, ScARF1 and ScARF3 both regulate peroxisome proliferation. Whereas ScARF1 activates a pathway or factor leading to enhanced proliferation, ScARF3 regulates this pathway negatively. This view explains most of the results obtained so far. For example, in the triple mutant after 6 h on oleate at the permissive temperature, the stimulating activity of ScARF1 still was preserved, and the loss of the inhibiting activity of ScARF3 additionally enhanced proliferation. At the nonpermissive temperature, the ScARF1-stimulating activity was eliminated; however, stimulation by loss of ScARF3-inhibiting activity still allowed a 29% increase in the number of peroxisomes per cell.
Peroxisomal Coatomer Binding-The existence of a functional dilysine motif at the C-terminal end of Pex11␣p initially led us to assume that Pex11␣p might function as a coatomer receptor in analogy to the p24 family of Golgi-localized transmembrane proteins (8,9). In a recent model of ARF1 recruitment to the Golgi membrane (60), two binding sites for coatomer were proposed. One is contributed by membranebound ARF1-GTP, which interacts with the ␤and ␥-COP subunits of coatomer, and the other is contributed by p23 oligomers that were shown to interact with ␥COP (61, 62). So far, we have not obtained evidence for a similar role of Pex11␣p, neither by co-immunoprecipitation experiments nor by confocal immunofluorescence studies (results not shown). The drastic increase in the concentration of Pex11␣p by hypolipidemic drugs, for example, was not accompanied by a proportional increase in coatomer recruitment (Fig. 1B). Although Pex11␣p might be implicated in peroxisome biogenesis in mammals, its precise role still has to be established. The situation in mammals becomes even more complex by the recent identification of two additional Pex11 proteins, Pex11␤p and Pex11␥p, that are highly homologous to the ␣ form (63)(64)(65). Since only Pex11␣p is stimulated by hypolipidemic drugs, this isoform was supposed to be involved in the induced proliferation, whereas Pex11␤p might regulate the constitutive process (66). Similarly, in S. cerevisiae, a family of Pex11p proteins comprising Pex11p, Pex25p, and Pex27p was identified that is required for peroxisome biogenesis regulating peroxisome abundance (67,68). Recently, an interaction between Pex25p and the small GTPase Rho1p was suggested to play a role in controlling yeast peroxisome membrane dynamics (69).
So far, we have further characterized peroxisomal ARF and coatomer binding and identified distinct factors implicated in this process. ARF1-GTP is necessary for the recruitment of coatomer. In its GDP-bound state, ARF1 is completely inactive. ARF6 that also associates with peroxisomes cannot replace ARF1 in this process. In S. cerevisiae, ARF1 and ARF3 (the ortholog of mammalian ARF6) show antagonistic effects on oleate-induced peroxisome proliferation. Whereas ARF1 is essential for proliferation, the presence of ARF3 inhibits this process.