ARF1 regulates pH-dependent COP functions in the early endocytic pathway.

Coat proteins of the COP family were recently shown by us and others to be involved in membrane transport in the endocytic pathway, in addition to their known functions in the biosynthetic pathway. We have also shown that membrane association of endosomal COPs depends on the acidic endosomal pH, in contrast to biosynthetic COPs. In this paper, we report that both membrane recruitment of endosomal COPs and in vitro biogenesis of transport intermediates destined for late endosomes, depend on a cytosolic factor, which we identified as the small GTP-binding protein ARF1. Our data indicate that ARF1 does not act via activation of an endosomal phospholipase D. We also find that ARF1 membrane association is regulated by the endosomal pH, and that this controls the pH-dependent association of endosomal COPs. These studies thus show that ARF1 regulates COP functions in the endocytic pathway, and indicate that ARF1 acts as the cytosolic component of a transmembrane pH-sensing mechanism.

In animal cells, the dynamic flow of proteins and lipids between the plasma membrane and early endosome is maintained by rapid internalization and recycling processes. Most internalized cell surface molecules are recycled, whereas molecules which are destined to be degraded, including all downregulated cell surface receptors, are sorted within early endosomes, and then transported toward late endosomes and lysosomes (1,2). Transport from early to late endosomes is mediated by relatively large carrier vesicles (0.4 -0.5 m diameter) with a typical multivesicular appearance (3), which will be referred to here as endosomal carrier vesicles/multivesicular bodies (ECV/MVBs). 1 Once formed on early endosomal membranes, ECV/MVBs move toward late endosome, and this movement depends on intact microtubules and motor proteins (3)(4)(5). Eventually, ECVs dock onto and fuse with late endosomes, in a process which depends on ␣ soluble N-ethylmaleimide sensitive factor (NSF) attachment protein, NSF, and perhaps another member of the triple A ATPase family (6), as well as presumably the small GTPase rab7 (7).
In vivo and in vitro studies have shown that ECV/MVB formation on early endosomes depends on some, but not all, components of the COP-I coat complex (8 -10), which is known to be also involved in the early secretory pathway (11). Recent studies, in fact, indicated that endosomal COPs contribute to the down-regulation of the Nef-CD4 complex, via direct interactions between Nef and ␤COP (12). Similarly, the AP3 adaptor complex also appears to be involved in more than one pathway, namely transport toward late endosomes/lysosomes (13) and synaptic vesicle formation (14). In addition to endosomal COPs, ECV/MVB formation also depends on the acidification properties of early endosomes (15). These pH-and COP-dependent processes are related functionally and biochemically, since COP association to endosomal membranes, but not to biosynthetic membranes, is itself pH-sensitive (8,9). These observations lead us to propose that a transmembrane pH-sensor regulates COP association to endosomes, thereby signaling the onset of the degradation pathway on early endosomal membranes.
In the present paper, we have further dissected the molecular process which regulates membrane association of endosomal COPs. Our data show that the small GTP-binding protein ARF1 is required for COP recruitment onto endosomes, and for ECV/MVB biogenesis from donor early endosomal membranes in vitro. We find that PLD and phosphatidic acid are not involved in this process, indicating that ARF1 does not act via PLD on early endosomal membranes, in contrast to biosynthetic COPs (16). Moreover, our data show that ARF1 recruitment onto endosomes depends on the acidic lumenal pH, in agreement with previous studies on ARF binding to microsomes (17), and that this mechanism accounts for the pH dependence of endosomal COPs association to membranes. Our data thus show that ARF1 mediates COP binding to endosomes and ECV/MVB biogenesis, in a pH-dependent, but PLD-independent, process. ARF1 thus appears to act as the cytosolic component relaying lumenal pH variations to endosomal COPs during ECV/MVB biogenesis.

MATERIALS AND METHODS
Cell Culture and Immunological Reagents-Monolayers of baby hamster kidney cells (BHK-21) were grown and maintained as described (3). For large scale endosome preparation, 24 ϫ 24-cm dishes were used. The M3A5 monoclonal antibodies against ␤COP peptide was a gift of T. Kreis (University of Geneva, Geneva, Switzerland). Antibodies against all other COP-I subunits were a gift of C. Harter (University of Heidelberg, Heidelberg, Germany). The monoclonal anti-ARFs antibody was from Affinity Bioreagents (ABR, Golden, CO). The specific rabbit antiserum against ARF6 was obtained from V. Hsu (Harvard Medical School, Boston, MA). The affinity-purified rabbit antibodies which specifically recognize ARF1, as well as those against ARF5 were a gift from S. Robinson (University of Cambridge, Cambridge, United Kingdom). PLD produced by fermentation of Actinomadura sp. No. 362 was a generous gift from Meito Sangyo Co., Ltd. (Tokyo, Japan). PLD from Streptomyces chromofuscus (C12:0) PA and (C8:0) diacylglycerol were from Sigma (Buchs, Switzerland). GTP␥S was from Roche Molecular Biochemicals (Rotkreuz, Switzerland). Subcellular Fractionation of Endosomes-Endosomal fractions were prepared from cells pretreated with 20 g/ml brefeldin A for 1 h at 37°C to eliminate Golgi contamination from endosome fractions (9,18), except when PLD function was tested. Subcellular fractionation was carried out using a step flotation gradient, as described (4,8,9,19). Briefly, a postnuclear supernatant was prepared, adjusted to 40.6% sucrose, 3 mM imidazole, pH 7.4, loaded at the bottom of an SW60 tube, and overlaid sequentially with 35 (1.5 ml) and 25% sucrose (1 ml) in 3 mM imidazole, and finally with homogenization buffer (HB, 250 mM sucrose, 3 mM imidazole, pH 7.4) to fill the tube (in large scale preparations, postnuclear supernatant obtained from four 24 ϫ 24-cm dishes were loaded into six SW40 tubes). Gradients were centrifuged at 35,000 rpm for 60 min (SW60 rotor) or 90 min (SW40 rotor). Early endosomal fractions were collected at the 35/25% interface and both ECV/MVBs and late endosomes at the 25%/HB interface.
Rat Liver Cytosol Preparation, Fractionation, and ARF Depletion-Two livers were removed from rats, washed in HB, weighted, and homogenized using an electrical mixer in a volume (ml) of HB corresponding to 5ϫ weight of livers, and containing 10 M leupeptin, 10 g/ml aprotinin, 1 M pepstatin, 1 mM phenylmethylsulfonyl fluoride, and 1 mM benzamidine (20). The mixture was centrifuged at 5,000 rpm in a SS-34 rotor for 10 min at 4°C. The supernatant was further centrifuged at 15,000 rpm for 10 min and then at 35,000 rpm for 1 h in an SW40 rotor. The final supernatant was aliquoted, and then frozen and stored in liquid nitrogen. COP-and ARF-enriched fractions were prepared from 500 l of thawed rat liver cytosol, which was loaded on the top of a continuous gradient (8 to 50% sucrose in 3 mM imidazole, pH 7.4) in an SW60 tube. The gradient was centrifuged for 18 h at 35,000 rpm. Fractions (450 l) were collected, and either used in experiments or analyzed by SDS gels and Western blotting. For ARF-depleted cytosol, 5 mg of thawed rat liver cytosol was incubated with 10 mg of a membrane fraction obtained at the 35/40.6% sucrose in BHK subcellular fractionation gradient in the presence of 20 M GTP␥S at 37°C for 30 min. Membranes were then sedimented by ultracentrifugation at 200,000 ϫ g for 30 min. The supernatant was then dialyzed against 8.5% sucrose in 3 mM imidazole, and complemented with 250 g of COP enriched fraction. By Western blotting, 80 -90% of ARF was depleted this way.
Expression and Purification of Recombinant Myristoylated ARF1-Recombinant myristoylated wild-type and mutant ARF1 were prepared from BL21 Escherichia coli transformed with plasmids encoding for yeast N-myristoyl transferase and human wild-type ARF1, or T31N ARF1 (obtained from V. Faundez, University of California, San Francisco, CA). Transformed cells were induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside in the presence of 200 M myristic acid. Cells were lysed and the expressed protein was purified using DEAE-Sephacel and Ultrogel AcA 54 columns as described in Ref. 21.
COP and/or ARF Binding Assay-COP binding to endosomes was analyzed as described (9). As a source of COPs and ARFs we used either 50 l (1 mg) of complete cytosol, or 100 l of COP-enriched fraction supplemented with 100 l of ARF-enriched fraction or 15 g of recombinant myristoylated ARF. The extracts were mixed on ice with 50 g of early endosomal fraction and complemented with 12.5 mM HEPES, pH 7.4, 1.5 mM MgOAc 2 , 1 mM dithiothreitol, 65 mM KCl, and 15 l of an ATP-regenerating system (22), or 1 mM ATP when no complete cytosol was present, in a total volume of 200 -350 l. The mixture was incubated at 37°C for 15 min, adjusted to 40.6% sucrose, 3 mM imidazole, pH 7.4, loaded on the bottom of a TLS55 tube, overlaid first with 500 l of 35% sucrose, 3 mM imidazole, pH 7.4, and then with HB. The step gradient was centrifuged at 45,000 rpm for 45 min. Early endosomes were collected at the 35%/HB interface, and analyzed by SDSpolyacrylamide gel electrophoresis and Western blotting.
Determination of PLD Activity-In the COP/ARF binding assay, PLD produced by fermentation of Actinomadura was used at 8 g/ml and exogenous (C12:0) PA, and (C8:0) diacylglycerol were added at 200 M final concentrations. PLD activity was controlled by measuring the hydrolysis of PC to PA, using 0.1 Ci of 14 C-labeled PC (Amersham Pharmacia Biotech) in 100 M PC liposomes, as substrate. Liposomes were incubated with 5 g/ml S. chromofuscus PLD or 8 g/ml Actinomadura PLD, in order to obtain similar PC hydrolysis activity. Conditions were the same as used in binding assay, in the presence or absence of 20% ether for 15 min at 37°C. Lipids were then extracted in CHCl 3 / methanol, spotted on TLC plates, and developed in CHCl 3 /methanol/ NH 3 (32%): 65/25/5. TLC plates were exposed using Kodak Bio-Max films. Migration of PC and PA was revealed using standards. Endogenous PLD activity on early endosome membranes was measured by transphosphatidylation using 1-butanol as a substrate (23). Twelve 10-cm dishes of BHK cells were radiolabeled with 10 Ci of [ 14 C]oleic acid (Amersham Pharmacia Biotech). 10 Ci of [ 14 C]oleic acid solution was evaporated with argon, resuspended in phosphate-buffered saline/ bovine serum albumin (5 mg/ml), added to BHK cells for 16 h and early endosome membranes were prepared as above. Endosomes were incubated with 0.3% 1-butanol and recombinant myristoylated ARF1, in the presence or absence of GTP␥S, using conditions supporting COP binding as above. To provide a positive control for PC cleavage into PA, exogenous Actinomadura PLD was added in the controls. The mixture was incubated at 37°C for 1 h. Lipids were extracted, spotted onto TLC plates, and developed in isooctane/ethyl acetate/acetic acid/water (50: 110:20:100) as a solvent system. TLC plates were exposed and 14 Clipids revealed as above. Migration of phosphatidylbutanol was determined with a standard.
In Vitro Formation of ECV/MVBs from Early Endosomes-Formation of ECV/MVBs from donor early endosomal membranes was measured as described (8,9). Briefly, HRP was internalized into BHK cells by fluid phase endocytosis, after incubation for 5 min with 5 mg/ml HRP at 37°C, to provide a marker of the early endosomal content. Early endosome fractions were prepared as above. In the assay, 300 g of early endosomes in a final volume of 1.5 ml were incubated for 30 min at 37°C in 12.5 mM HEPES, pH 7.0, 1 mM dithiothreitol, 1.5 mM MgOAc, 60 mM KCl, and supplemented with an ATP-regenerating system and 1 mg/ml rat liver cytosol. When indicated, 0.1 mg/ml recombinant WT ARF1 or T31N mutant form of ARF1 were added. After the assay, the mixture was adjusted to 25% sucrose, 3 mM imidazole, pH 7.4, loaded at the bottom of an SW60 tube and overlaid with HB. After 1 h of centrifugation at 35,000 rpm, donor early endosomes and ECV/ MVBs formed in vitro were recovered from the pellet and the 25% sucrose/HB interface, respectively. ECV/MVBs was then sedimented after centrifugation for 30 min at 100,000 ϫ g. HRP activity in the endosomal and ECV/MVB fraction was quantified. The efficiency of ECV/MVB formation was calculated as a percentage of the total HRP activity present in donor endosomal membranes and ECV/MVBs formed in vitro.

COPs Recruitment onto Endosomes Requires Cytosolic Factor(s)-
In order to investigate the mechanisms of COP recruitment onto endosomes, early endosomal fractions were prepared by subcellular fractionation using a well established protocol (4,24). To ensure that endosomal fractions were not contaminated with biosynthetic membranes, cells were pretreated with brefeldin A, which causes Golgi proteins to redistribute to the endoplasmic reticulum (25). This treatment does not affect the fractionation of endosomes, but markers of the Golgi complex and the intermediate compartment/cis-Golgi network, which contains the bulk of biosynthetic COPs, then co-fractionate with the endoplasmic reticulum (6,9,18). Fractions enriched in COP-I coatomer were prepared separately by centrifugation of rat liver cytosol on a continuous sucrose gradient, in order to separate COPs from other cytosolic components. All COP subunits were then recovered in 30% sucrose, as a well defined peak (Fig. 1A, fraction 6).
When mixing purified endosomes with the COP-enriched fraction in the binding assay, COP recruitment onto endosomes was very inefficient, even in the presence of GTP␥S, when compared with recruitment from complete rat liver cytosol, as a source of COPs (Fig. 1B). In these experiments, cytosol and COP-enriched fractions were normalized so that equal amounts of COPs were tested in the assay. COP binding, however, could be fully restored when the assay was supplemented with the lightest fraction (10% sucrose, fraction 1) from the same continuous gradient, which is devoid of COPs (Fig. 1A). These experiments demonstrate that efficient recruitment of COPs onto endosomes requires additional cytosolic factor(s), which are not present in the COP-enriched fraction.
A Small GTP-binding Protein Is Required for Endosomal COP Binding-We had previously observed that COP binding onto early endosomes was stimulated by GTP␥S, a non-hydrolyzable analog of GTP, indicating that a GTP-binding protein was involved in the process (8,9). Since COP binding to biosynthetic membranes requires the small GTP-binding protein ARF1, and since ARFs were detected in endosomal fractions (10), we analyzed the distribution of ARFs after cytosol fractionation. As shown in Fig. 1A, ARFs were not detected in the COP-enriched fraction, but were recovered exclusively within the top fraction, which was competent to restore efficient COP binding to endosomes (Fig. 1B).
To further investigate the possible involvement of a GTPbinding protein, COP binding to endosomes was tested after addition of GTP␥S. Whereas GTP␥S had no effect when the COP-enriched fraction was used alone (see Fig. 1B), addition of the ARF-enriched fraction restored GTP␥S-mediated stimulation of COP binding to endosomes (Fig. 2B). These experiments suggested that an ARF protein was responsible for efficient COP membrane association. Five human ARF proteins have been identified, which are highly homologous and can be subdivided into 3 classes by homology, with ARF 1 and 3 in class I, ARF 4 and 5 in class II, and ARF6 alone in class III (26,27). Class I and II seem to contain isomeric ARF forms (Ͼ90% identical in each class). Moreover, class I and II ARFs are also highly homologous to each other (Ϸ80% identity), whereas ARF6 is somewhat more distantly related to other ARFs (65-70% identify).
Previous studies have already shown that ARF1 is involved in different transport steps and shared by several protein coats (28 -33). In addition, both activated ARF1 and -5 seem to be able to mediate AP1 and COPI recruitment onto Golgi membranes, but via different activation mechanisms (34), while only ARF1, but not ARF5, is involved in AP3 recruitment (33). When purified early endosomes were incubated with excess cytosol, ARF5 could not be recruited onto membranes even in the presence of GTP␥S, in contrast to ARF1 ( Fig. 2A), suggesting that ARF1, but not ARF5, may be involved in membrane association of endosomal COPs. We then investigated whether ARF6 was possibly involved in the process, since the protein was reported to distribute to early endosomal membranes (35,36). These observations, however, remain controversial (37,38), and the precise function of ARF6 is being debated (35,36,39,40). We thus tested whether ARF6 was selectively recruited onto endosomes upon COP binding. As shown in Fig. 2B, no correlation was observed between amounts of ARF6 and COPs, demonstrating that ARF6 distribution was not coupled to COP recruitment. In contrast, the binding of ARF1 correlated well with COP recruitment revealed by a specific antibody against ARF1 (Fig. 2B), and by GTP overlay (not shown). These results suggested that an ARF protein, perhaps ARF1 but not ARF 5 or -6, was required for COP binding onto endosomes.
ARF1 Mediates COP Recruitment onto Endosomal Membranes-In order to test directly whether ARF1 was involved in endosomal COP recruitment, we prepared and purified recom-  (1, bottom; 10, top) and analyzed by SDS-gel electrophoresis and Western blotting using antibodies against each COP subunit or ARF (this antibody recognizes several ARF isoforms). For comparison, lane "cyt" was loaded with 30 g of rat liver cytosol. The profile of total protein is also shown. Fraction 6 contained the COPI coatomer, whereas fraction 1 contained ARFs (revealed using a monoclonal antibody which recognizes all ARF isoforms, except ARF6). B, in order to measure COP membrane recruitment, endosomes were mixed in the presence of 10 M GTP␥S with 0.5 mg of rat liver cytosol (cyt), or 100 l of fraction 6 containing COPI (F6), or 100 l of both fraction 6 and 100 l of fraction 1 (F6 ϩ F1). Cytosol and fractions were normalized so that equal amounts of ␤-COP were added in each assay. Early endosomal membranes were then recovered after flotation on gradient, loaded on SDS gels, and analyzed by Western blotting with the M3A5 monoclonal antibody against ␤COP. Experiments shown in A and B were performed 10 and three times, respectively, and were highly reproducible. Representative experiments are shown.

FIG. 2. ARF-enriched fraction restores GTP␥S-dependent COP binding.
A, endosomes were mixed in the presence of 10 M GTP␥S with 0.5 mg of rat liver cytosol, and then recovered after flotation, loaded on SDS gels and analyzed by Western blotting, as in Fig. 1B, using specific antibodies against ARF1 or ARF5 (43). Endosomes (memb) and input cytosol (cytosol) are shown. B, COP membrane association was carried out as described in the legend to Fig. 1, using the COP-enriched fraction 6 alone (F6), or complemented with the ARFenriched fraction 1 (F1) in the absence or presence of 10 M GTP␥S. COP binding was measured as described in the legend to Fig. 1. ARF1 and ARF6 binding were revealed using specific antibodies against each protein (36,55). Experiments shown were performed three times, and were highly reproducible. Representative experiments are shown. binant, myristoylated human ARF1. When tested in our assay, purified ARF1 alone was sufficient to fully restore the capacity of COPs to bind endosomal membranes, in the absence of the ARF-enriched fraction (Fig. 3A). In addition, recombinant ARF1, much like endogenous ARF, was efficiently recruited onto endosomes in the presence of GTP␥S, together with COPs (Fig. 2). Non-myristoylated ARF1 remained completely inert in the assay; it did not bind endosomes and did not support COP membrane association (not shown). These experiments demonstrate that recombinant ARF1 alone can support COP binding onto early endosomal membranes.
As a next step, we investigated whether the action of recombinant ARF1 was promiscuous, or whether the protein could mediate membrane association of the proper endosomal COP complex. Previous studies showed that the ␦and ␥-COP subunits are not detected on endosomes (8 -10), indicating that the composition of endosomal and biosynthetic COPs is different. As shown in Fig. 3B, ␦ COP subunit was not recruited to any significant extent onto endosomal membranes, when using COP-enriched fractions and recombinant ARF1 in the assay. This experiment demonstrates that recombinant ARF1 did not lead to spurious COP membrane association, and restored membrane recruitment of the endosomal COP complex.
To further characterize the role of recombinant ARF, we made use of our previous observations that COP recruitment onto early endosomes is inhibited after neutralization of the endosomal pH, even in the presence of GTP␥S (8, 9) and see Fig. 4A). These order of addition type experiments led us to propose that an endosomal pH sensor regulates COP binding to endosomes by acting prior to the GTP␥S-dependent step. As shown in Fig. 4B, preneutralization of the endosomal pH with 50 M of the protonophore nigericin reduced COP binding to endosomes, when recombinant ARF1 and the COP-enriched fraction were added together with GTP␥S. We had previously shown that the same drug concentrations prevented COP association to endosomes in vivo, and caused both early endosome fragmentation and inhibition of early to late endosome transport (8,15). Association of ARF1 to endosomes was itself reduced when compared with controls, suggesting that ARF1 is also under control of the endosomal pH-sensing mechanism. From these experiments, we conclude that ARF1 is required for the specific, pH-dependent membrane association of endosomal COPs.
Sequential Association of ARF1 and COPs to Endosomes-Membrane association of both ARF1 and COPs is pH-dependent, when proteins are added together onto endosomal membranes. We then determined whether ARF1 binding to endosomes was itself pH-sensitive in the absence of COPs, or whether pH sensitivity was conferred to ARF1 upon COP binding. In these experiments, early endosomes were sequentially incubated with ARF1 in the absence of COPs (Fig. 4C, step 1), and then with COPs in the absence of ARF1 (Fig. 4C, step 2). Briefly, early endosomal fractions were first incubated with recombinant myristoylated ARF1 in the presence of GTP␥S. Then, endosomes (with bound recombinant ARF1) were re-FIG. 3. ARF1 supports endosomal COP binding. A, COP binding was measured as described in the legend to Fig. 1, using 100 l of COP fraction 6 together with 10 g of purified, recombinant myristoylated ARF1, rARF1ϩF6 (COP) in the absence or presence of 10 M GTP␥S. As a control, COP binding was also measured using 0.5 mg of rat liver cytosol as a COP source (complete cytosol). Analysis was as described in the legend to Fig. 1. The figure shows that amounts of recombinant or cytosolic ARF1 recruited by endosomal membranes are similar, as are amounts of COPs, when using complete cytosol or recombinant ARF1 together with fraction 6. B, experiment was as in A in the presence of GTP␥S. The figure shows that ␦COP was present in the cytosol used as starting materials in the assay, before the experiment (cytosol before expt), but was not recruited onto endosomes. Experiment was analyzed using antibodies against ␤or ␦COP. Experiments shown were performed three times, and were highly reproducible. Representative experiments are shown.
FIG. 4. Sequential binding of ARF1 and COPs. COP binding to endosomes was studied as described in the legend to Fig. 1. A, as a COP source, we used 0.5 mg of complete cytosol, and the assay was carried out in the absence (C), or presence of 10 M GTP␥S (G). Alternatively, the early endosomal pH was preneutralized with 50 M nigericin at room temperature for 5 min, before adding 10 M GTP␥S (NG). B, as a COP source, we used 100 l of COP-enriched fraction 6, In the assay, the fraction was combined with 10 g of recombinant ARF1, as described in the legend to Fig. 3, in the presence of 10 M GTP␥S (G), or endosomes were preneutralized before GTP␥S addition as in A (NG). C, pH-regulated sequential binding of recombinant ARF1 and COPI. Early endosomal membranes were first incubated with 10 g of recombinant myristoylated ARF1 and 10 M GTP␥S, in the absence (G) or presence (NG) of 50 M nigericin, as in B. The mixture was then re-fractionated by flotation on a sucrose gradient to remove excess ARF1 (in the absence of nigericin), and analyzed by Western blotting using the monoclonal anti-ARF antibody (step 1). In the second step, endosomes were reincubated with 100 l of COP enriched fraction 6 in the presence of 1 mM ATP, but in the absence of GTP␥S, nigericin, and ARF1, repurified (to remove excess COPs), and then analyzed by Western blotting using antibodies against ␤COP (step 2). Experiments shown A were repeated many times (see also Refs. 8 and 9), and those in B and C were performed three times and were highly reproducible. Representative experiments are shown.

ARF1 Regulates pH-dependent COP Functions
purified by flotation on the gradient, to remove free ARF1. In the second step, endosomal membranes were reincubated with the COP-enriched fraction, but in the absence of ARF1 and GTP␥S. Using this sequential assay, the efficiency of ARF and COP membrane association was identical to that measured when all components were mixed in the same incubation (data not shown).
When the first incubation (ARF1 and endosomes in the absence of COPs) was carried out in the presence of nigericin, ARF1 binding was inhibited when compared with untreated controls (Fig. 4C, step 1). This experiment indicated that ARF1 association to endosomes is itself under the control of the endosomal pH-sensing mechanism, and that this mechanism could be fully reconstituted in vitro with purified recombinant ARF1. When these membranes were then reincubated with COPs (step 2) in the absence of ARF1 and nigericin but in the presence of ATP (to allow endosome re-acidification), COP binding was reduced in parallel with decreased amounts of ARF1 associated with endosomes. In contrast, COP membrane association was not affected when the pH was neutralized during the second incubation only, after ARF1 had already been recruited onto endosomes (data not shown). These experiments indicate that the binding of ARF1, rather than COPs, is sensitive to the endosomal pH, and that ARF1 is the single cytosolic factor required for pH-dependent COP association. From these data, we conclude that a pH-sensing mechanism regulates ARF1 association to endosomes, which in turn mediates endosomal COP recruitment.

Association of COPs to Endosomal Membranes Depends on ARF1 but Not on PLD-mediated Production of Phosphatidic
Acid-The fact that ARF1 was necessary for the pH-dependent, membrane association of endosomal COPs prompted us to test whether it acted on endosomal membranes via PLD activation. Indeed, it has been reported that ARF1 mediates COP assembly on biosynthetic membranes via PLD activation and production of phosphatidic acid (PA) (16).
Endosomes were first mixed with 200 M of the short chain forms of (C12) PA or (C8) diacylglycerol, an immediate metabolic product of PA, so that lipids could be spontaneously incorporated into endosomes. As shown in Fig. 5A, COP binding was not significantly stimulated by these treatments (1.5-2fold), when compared with the effects of GTP␥S (Ϸ50-fold), indicating that weak interactions with lipids may contribute to coat association and/or stabilization, but that these cannot account for the observed efficiency of COP recruitment. Similarly COP binding was only marginally stimulated (1.5-fold) after pretreating endosomes with Actinomadura PLD in order to generate PA from endogenous phosphatidylcholine (PC) (Fig.  5A), although the enzyme was clearly active, as measured in parallel by PA production from endogenous PC (not shown) or by hydrolysis of 14 C-labeled PC incorporated into liposomes (Fig. 5B). Another PLD from S. chromofuscus exhibited a somewhat lower activity on liposomal PC under our conditions (Fig.  5A), and also failed to stimulate COP binding to endosomes (data not shown). Finally, we determined whether a putative endosomal PLD activity may become activated upon GTP␥Smediated stimulation of ARF1. Endosomal fractions were prepared after metabolic labeling of cells with [ 14 C]oleic acid (in order to label endogenous PC and other lipids), and used in the COP binding assay in the presence of ARF1 and GTP␥S. PLD activity was measured by phosphatidylbutanol production via trans-phosphatidylation, using 1-butanol as substrate (23). As shown in Fig. 5C, endosomal membranes did not catalyze the production of phosphatidylbutanol in the presence of ARF1 and GTP␥S, although phosphatidylbutanol production could readily be observed after addition of exogenous Actinomadura sp. PLD. Altogether, these observations show that ARF1 does not act on early endosomal membranes via PLD activation and PA production, and therefore that PA production is not involved in endosomal COP membrane association, ARF1 Is Involved in the ECV/MVB Biogenesis in Vitro-Our data indicate that ARF1 is required for the pH-dependent, PLD-independent association of endosomal COPs to membranes. To further illustrate the role of ARF1 in the endocytic pathway, we made use of an assay we have established which reconstitutes the biogenesis of ECV/MVBs from donor early endosomal membranes in vitro (8,9). In this assay, the content of early endosomes is labeled in vivo after fluid phase endocytosis of HRP for 5 min at 37°C. Early endosomal fractions, which are typically devoid of ECV/MVBs present in the cell at steady state, are then prepared by flotation on gradient, and used as donor membranes in the assay (4,19). Then, ECV/ MVBs which were generated in vitro are separated from the donor membranes on a second gradient, and efficiency of ECV/ MVB formation is measured by quantifying HRP in the fractions. Analysis was as in Fig. 1. B, the activity of exogenous PLD was measured in the assay. 100 M 14 C-labeled PC liposomes (0.1 Ci) were incubated with either 5 g/ml PLD from S. chromofuscus (PLD Strep) or 8 g/ml PLD (PLD Acti) for similar activity at the same pH, salts, and energy conditions as in the assay. In each case, 20% ether was added as a positive control. PA production was revealed by TLC analysis and exposed to x-ray film. C, the activity of a possible endogenous PLD present on endosomal membranes was measured according to Ref. 23. Cells were metabolically labeled with [ 14 C]oleic acid, fractionated, and then 50 g of early endosome membranes were incubated with 15 g of recombinant myristoylated ARF1, in the presence or absence of GTP␥S. In order to allow transphosphatidylation catalyzed by PLD, 3% 1-butanol was added. Production of phosphatidylbutanol (PB), the transphosphatidylation product, was revealed by TLC and exposure to x-ray film. 8 g/ml exogenous Actinomadura sp. PLD was added as a positive control. Experiments in A were repeated three times with different endosome preparations and were highly reproducible, as were those in B and C. Representative experiments are shown.

ARF1 Regulates pH-dependent COP Functions
In a first series of experiments, the cytosol used in the assay, which contains all desired factors including COPs and endogenous ARF1, was supplemented with purified recombinant myristoylated dominant-negative T31N ARF1 mutant, which preferentially binds GDP (41) or with WT ARF1. As shown in Fig. 6A, addition of T31N ARF1 significantly inhibited ECV/ MVB formation. Inhibition, however, was not complete, presumably because endogenous ARF1 is abundant in the cytosol. Then, the cytosol was depleted of ARF1 by GTP␥S-mediated ARF recruitment onto an excess of total microsomal membranes, and then cytosol was separated from ARF-loaded membranes by centrifugation. The cytosol was then dialyzed to remove excess GTP␥S, and used in the assay. As shown in Fig.  6B, ARF-depleted cytosol inhibited ECV/MVB formation, and addition of WT ARF1 to the ARF-depleted cytosol partially restored ECV/MVB biogenesis in vitro. These experiments show that ARF1 is directly involved in the biogenesis of ECV/ MVBs from early endosomal membranes, and thus confirm the role of ARF1 in endosomal COP recruitment onto membranes.

ARF1 Regulates COP Functions in the Endocytic Pathway-
The small GTP-binding protein ARF1 was originally identified as a cofactor for cholera toxin-catalyzed ADP-ribosylation of the G s subunit of the trimeric G protein (42). Since then, ARF1 was shown to be involved in the selective association of different types of coat complexes to the membranes of different organelles, including COPI to the Golgi complex (28), AP1 to the trans-Golgi network (29), clathrin/AP1 to immature secretory granules (31), AP3 to synaptic vesicles (32), and nonneuronal membranes (33). ARF1 has not been shown until now to play a role in the binding of any coat protein to endocytic membranes, except for endosomal targeting of AP2 (43). However, it has previously been suspected that ARF1 may be involved in the endocytic pathway. Brefeldin A, which blocks a Golgi ARF exchange factor (44), causes the formation of endosomal tubules, which are reminiscent of the drug-induced Golgi tubules (45). In addition, both endocytic and biosynthetic pathways are affected after expression of mutant ARF1 (46), and ARF is required for maintenance of yeast Golgi and endosome structure and function (47). However, it is not clear to what extent these effects may be indirect, and/or reflect the multiple roles of ARF1.
Our observations that recombinant ARF1 is necessary and sufficient to support efficient, and GTP␥S-sensitive COP binding to endosomes demonstrate that this small GTP-binding protein regulates membrane association of endosomal COPs. Moreover, the fact that recombinant ARF1 is required for ECV/ MVB biogenesis using an in vitro transport assay, also demonstrate that ARF1 directly controls the formation of these transport intermediates destined for late endosomes. We can thus conclude that ARF1 regulates COP functions in the endocytic pathway.
Endosomal COPs-One of the characteristic features of endosomes is that their lumen is acidified by the action of the vacuolar ATPase (48). Whereas early endosomes are mildly acidic (pH Ϸ 6.2), the pH drops at later stages of the pathway, and reaches values in the 5.5-5.0 range in late endosomes/ lysosomes. Several endocytic processes are known to depend on acidification, including receptor-ligand uncoupling and activity of hydrolases. In addition, an active vacuolar ATPase is also required for transport from early to late endosomes (15,49), and our previous studies showed that acidification regulates ECV/MVB biogenesis (15). In addition to this pH-dependent mechanism, both transport from early to late endosomes (8 -10, 12, 50) and ECV/MVB biogenesis (8,9) depend on proteins of the COP-I coat, which are also involved in the early secretory pathway (11).
These pH-and COP-dependent mechanisms are coupled functionally, since COP association to early endosomal membranes is itself dependent on proper acidification of the endosomal lumen (8,9). In addition, both neutralization of the endosomal pH and COP inactivation prevent ECV/MVB formation in vivo and in vitro, and both conditions cause similar dramatic changes in early endosome ultrastructure (8,9,15). Finally, both conditions also affect recycling of the transferrin receptor, without major effects on bulk internalization and recycling of a fluid phase tracer (9,15,50,51). Altogether, these observations suggest that early endosome structure and functions are coupled to ECV/MVB biogenesis, and thus that COP functions in endosomes may extend beyond their generally accepted role in vesicle formation. Since COP membrane association is pH-dependent, we proposed that lumenal acidification of endosomes is translated across the bilayer by the pHdependent conformational change of a transmembrane pH sensor (8). Acidification may thus regulate early endosomal functions, including the onset of the degradation pathway and ECV/MVB biogenesis, by regulating assembly of the COP coat, This process differs from COP association to biosynthetic membranes, which is not pH-dependent (8,9), although both retrograde transport in the secretory pathway and subcellular distribution of biosynthetic COPs are affected by inhibitors of the vacuolar ATPase (52). Endosomal and biosynthetic COPs

FIG. 6. Involvement of ARF1 in ECV/MVB biogenesis in vitro.
A, the in vitro formation of ECV/MVBs from donor early endosomes was measured as described previously (8,9). Early endosomes prelabeled with HRP internalized for 10 min at 37°C were prepared and used as donor membranes. These were then incubated with 1 mg/ml rat liver cytosol in the presence of 0.1 mg/ml WT rARF1 or T31N rARF1 (dominant negative) and a ATP regenerating system. As a control, ATP was depleted in the assay, using an ATP-depleting system. ECV/MVBs formed in vitro were then separated by flotation in a second sucrose gradient, and the HRP content of both donor early endosomes and vesicles formed in vitro were quantified. Efficiency was measured as the percentage of HRP present in vesicles formed in vitro, reflecting the volume entrapped within newly formed ECV/MVBs, and corresponded to Ϸ10% of the early endosomal volume, consistent with in vivo observations (8). In the figure, this value was normalized to 100% in the presence of WT ARF1. B, ECV/MVB formation in the assay was measured as in A, except that 1 mg/ml ARF-depleted cytosol (cyt. ARF-) was used ("Materials and Methods"), instead of complete cytosol, and 0.1 mg/ml WT rARF1 were added. Analysis was as in A. Experiments in A and B were repeated five and three times, respectively, using different endosome preparations. also exhibit other differences. Whereas all COP subunits are found on biosynthetic membranes, ␥ and ␦ COP are not present on endosomes (8 -10). In addition, biosynthetic COPs can interact with the cytoplasmic domains of many proteins which typically contain the KKXX endoplasmic reticulum retrieval motif (11), but such proteins are not present in endosomes. In contrast, endosomal ␤COP appears to interact with a diacidic motif in the Nef protein, during Nef-mediated CD4 down-regulation (12). Finally, COP recruitment in the biosynthetic pathway appears to be regulated by ARF1-mediated PLD activation and subsequent production of PA (16), although the possibility that the activation of coat assembly by ARF is purely catalytic has been questioned (53). PLD activity was also shown to be involved in endoplasmic reticulum to Golgi transport (54), release of nascent secretory vesicles from the trans-Golgi network (23), and AP2 targeting to an endosomal compartment (43). In contrast, we find that PLD and PA are not involved in COP association to endosomal membranes, as was previously reported for AP1 binding to the trans-Golgi network (43). Altogether, these differences point at the existence of somewhat plastic interactions between COP subunits and membrane constituents during coat recruitment and/or assembly, and suggest that COP functions may be differentially modulated on different sets of membranes.
ARF1 Is a Cytosolic Component of the Endosomal pH-sensing Device-We now find that ARF1 association to endosomes is itself sensitive to the endosomal pH, in agreement with previous studies on ARF binding to microsomes (17). Moreover, our observations show that membrane association of ARF1 and endosomal COPs can occur sequentially, and that ARF1 association, and not COP recruitment, is sensitive to the endosomal pH. While the mechanism responsible for translating pH changes across the bilayer remain to be established, our observations now show that ARF1 is a molecular target of this mechanism on the cytoplasmic face of endosomal membranes, which is both necessary and sufficient for COP recruitment, suggesting that ARF1 is a key cytoplasmic element of this pH-sensing device.