In Vitro Reconstituted Dictyostelium discoideum Early Endosome Fusion Is Regulated by Rab7 but Proceeds in the Absence of ATP-Mg2+ from the Bulk Solution*

We characterized the in vitro fusion of endosomal compartments from Dictyostelium discoideum. Fusion activity was restricted to early compartments, was dependent on cytosolic proteins, and was activated by GTP and guanosine 5′-O(3-thio)triphosphate (GTPγS). This stimulation suggests the involvement of a small G protein, which we propose to be Rab7 on the basis of the strong inhibitory effect of anti-Rab7 antibodies. It is noteworthy that in the presence of GTPγS, the concentration of ATP-Mg2+ could be reduced to less than 1 nm without loss of fusion activity. Under these conditions, competing residual ATP with adenosine 5′-O-(3-thio)triphosphate-Mg2+ also failed to inhibit endosome fusion. The presence of an ATP-depleting system alone blocked fusion probably because endogenous GTP was removed by coupling through NDP kinase. Moreover, whether ATP was present or not, GTPγS-activated fusion was equally sensitive to anti-Rab7 antibodies or N-ethylmaleimide and was restricted to early compartments. These results show that soluble ATP-Mg2+ is not needed for endosome fusion. Since homotypic fusion of endosomes inD. discoideum has been shown to depend on the ATPaseN-ethylmaleimide-sensitive factor (Lenhard, J. M., Mayorga, L., and Stahl, P. D. (1992) J. Biol. Chem.267, 1896–1903), the nucleotide exchange on theN-ethylmaleimide sensitive factor must take place before GTPγS activation in this system.

Endocytosis is the process by which eukaryotic cells internalize, sort, and digest extracellular molecules (pinocytosis and receptor-mediated endocytosis) or particles (phagocytosis). In wild-type Dictyostelium discoideum, living on soil bacteria, phagocytosis is very active, but fluid phase uptake is almost undetectable. In contrast, axenic mutant cell lines such as Ax-2 exhibit intense pinocytosis in parallel with normal phagocytic behavior. Pinocytosis in D. discoideum Ax-2 strain should therefore be considered as a deregulated phagocytosis, whose defect is as yet unknown. Although initial findings argued in favor of clathrin-coated vesicle-mediated internalization (1), it has recently been shown that most of the fluid ingested by Ax-2 is contained in intracellular structures similar to phagosomes (2). Thus, D. discoideum is a very attractive organism to study the intracellular fate of phagosomes, because of the ease of biochemical experiments, the existence of a whole series of endocytosis mutants, and the possibility of genetic engineering.
In mammalian cells, phagosomes become acidic and progressively acquire hydrolytic enzymes from primary lysosomes to eventually constitute phagolysosomes (3). In D. discoideum, the internalized material similarly passes into an acidic, hydrolase-rich compartment, but, contrary to mammalian lysosomes (which can retain the ingested material for expanded time periods (4)), undigested material transits through a less acidic postlysosomal compartment before egestion (5,6). An understanding of how the material is carried along these successive compartments could be brought about by the in vitro study of the different steps of the phagocytic pathway and by the identification of the proteins responsible for membrane recognition and fusion.
Of central importance in this respect has been the discovery by Rothman and co-workers of the N-ethylmaleimide-sensitive factor (NSF). 1 First isolated on the basis of its capability to restore intra-Golgi transport after treatment by N-ethylmaleimide (NEM), this ATPase has now been shown to participate in almost all intracellular vesicle fusion events (7). In its ATPbound form, NSF binds to membranes via soluble NSF attachment proteins (SNAPs) and acts upon a complex made by integral membrane proteins specific for the donor and acceptor compartments called v-and t-SNAREs (SNAP receptors). It is currently proposed that hydrolysis of ATP by NSF drives the rearrangement of SNAREs, which actually permits membrane fusion and the dissociation of the SNARE complex (8). The binding of ATP by NSF is therefore a prerequisite of membrane fusion.
The small GTP-binding proteins of the Rab subfamily also play a crucial role in the fusion process. The most widely held model states that for each intracellular membrane fusion event, activation of a specific Rab protein is needed (9). Several reports indicate that Rab proteins regulate the rate of formation of the SNARE complexes (10 -12). In mammalian cells, Rab4, Rab5, Rab7, and Rab 9 are known to be involved in the endocytic pathway and regulate the shuttling of recycling vesicles, early endosome fusion, early endosome to lysosome transport, and lysosome to Golgi transit (13). In D. discoideum, only Rab7 and to a lesser extent Rab4 have been shown to be present in the endocytic pathway (14,15).
Like mammalian endosomes, D. discoideum endosomes are able to fuse in vitro, reproducing two characteristic features of mammalian endosome fusion, namely sensitivity to GTP␥S and inhibition by NEM (16). Furthermore, the addition of mammalian NSF could reverse this inhibition, which showed that the general fusion machinery described above is conserved in the D. discoideum endocytic pathway. In this study, we established an in vitro fusion assay using partially purified D. discoideum endocytic vesicles and cytosol and determined the nucleotide requirements of early endosome fusion.
Protein concentrations were determined by the BCA assay (Pierce) with bovine serum albumin as a standard. Polypeptides were separated by SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose for immunostaining. Rab7 was detected by purified polyclonal anti-Rab7 antibodies described in this study and HRP-coupled secondary antibodies (Bio-Rad), revealed by enhanced chemiluminescence reagents (Amersham Corp.) on Kodak X-Omat AR films.
Unless specified, all experiments were performed on ice. All data are representative of at least two independent experiments. Preparation of Avidin-or b-HRP-loaded D. discoideum Endosomes-Avidin-or b-HRP-loaded endosomes were prepared in parallel from the same batch of cells. Amoebas (1 ϫ 10 8 cells⅐ml Ϫ1 ) were incubated for 5 min at 21°C in axenic medium containing either b-HRP or avidin (1 mg⅐ml Ϫ1 ). Internalization of the markers was stopped by the addition of 5 volumes of ice-cold washing buffer (200 mM sucrose, 0.5 mM EGTA-KOH, 10 mM HEPES-KOH, pH 7.4).
Cells were washed twice in washing buffer and resuspended in breaking buffer (1 mM dithiothreitol (DTT), 5 g⅐ml Ϫ1 leupeptin, and 5 g⅐ml Ϫ1 pepstatin in washing buffer) at 3⅐10 8 cells⅐ml Ϫ1 . After breaking by six strokes in a ball bearing cell cracker (19), a postnuclear supernatant (PNS) was prepared by centrifugation (1000 ϫ g, 5 min, 4°C). When fusion between whole endocytic compartments was assayed, PNSs at this stage were used.
To prepare early endosomes, 3 ml of PNS was loaded onto a discontinuous sucrose gradient formed by layering 1 ml of 54%, 4 ml of 40%, and 4 ml of 30% sucrose (w/w) in breaking buffer. After centrifugation in a Beckman SW41 rotor (100,000 ϫ g, 1 h, 4°C), the early endosomes were collected at the 30 -40% interface of the gradient, quickly frozen in liquid nitrogen, and stored at Ϫ80°C.
Preparation of D. discoideum Cytosol-10 10 cells were harvested and broken as described above except that the breaking buffer was supplemented with 500 mM KCl to extract peripheral membrane-bound proteins. The resulting PNS was then centrifuged at 200,000 ϫ g (1 h, 4°C) to remove the membrane components. The supernatant was filtrated through a gauze, and (NH 4 ) 2 SO 4 was added to 70% saturation. After 1 h at 4°C, the precipitated proteins were collected by centrifugation and resuspended in 6 ml of ice-cold cytosol buffer (10 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA-KOH, 1 mM DTT, 20 mM HEPES-KOH, pH 7.4). The cytosol was dialyzed twice against cytosol buffer, cleared by centrifugation (400,000 ϫ g, 10 min, 4°C), and stored at Ϫ80°C.
Preparation of Anti-avidin Antibody-coated Plates-Enzyme-linked immunosorbent assay plates (Labsystem) were coated for 3 h at 37°C with 3 g/well of monoclonal anti-avidin WC19.10 antibodies (Sigma) in 100 l of 100 mM sodium carbonate, pH 9.0. The plates were washed with PBS-Tween (150 mM NaCl, 2 mM NaH 2 PO 4 , 10 mM Na 2 HPO 4 , 0.1% Tween 20, pH 7.4) and blocked with 300 l/well of 0.1 mg⅐ml Ϫ1 bovine serum albumin in PBS-Tween for 1 h at 37°C. The processed plates were washed again and used within 2 h.
Endosome Fusion Assay-Fusions between purified endosomes were conducted in 10 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA-KOH, 200 mM sucrose, 1 mM DTT, 10 mM HEPES-KOH, pH 7.4. Avidin and b-HRPloaded D. discoideum endosomes (5 l each) were mixed in a 50-l total volume with biotinylated insulin (0.1 mg⅐ml Ϫ1 ), an ATP-regenerating (200 M ATP-Mg 2ϩ , 10 mM creatine phosphate, 85 units⅐ml Ϫ1 creatine phosphate kinase, pH 7.4) or ATP-depleting (10 mM glucose, 2.2 units⅐ml Ϫ1 hexokinase or 10 units⅐ml Ϫ1 apyrase) system, and D. discoideum cytosol (10 l). Except where otherwise stated, the final concentration of cytosolic or membrane-bound proteins was 0.8 mg⅐ml Ϫ1 or 0.4 -0.8 mg⅐ml Ϫ1 , respectively, and the ATP-depleting system consisted of hexokinase and glucose. Fusions between whole endosomal compartments contained in PNS were conducted in the same way except that 40 l of both b-HRP-and avidin-loaded D. discoideum PNS were mixed in a total reaction volume of 120 l, KCl was omitted, and MgCl 2 was raised to 3 mM.
After 60 min of incubation at 21°C, the assay was stopped by the addition of one volume of ice-cold 2% Triton X-100 and incubated at 4°C for 30 min. The samples were then loaded into the wells of an antiavidin-coated enzyme-linked immunosorbent assay plate. The b-HRPavidin complexes were allowed to bind for 2 h at 21°C or overnight at 4°C, and the wells were washed twice in PBS-Tween. To measure the amount of immobilized b-HRP, 100 l of HRP substrate (0.1 mg⅐ml Ϫ1 tetramethylbenzidine, 0.6% H 2 O 2 , in 0.05 M sodium citrate, pH 5.0) was added to each well. The reaction was stopped by the addition of 20 l of 2 M H 2 SO 4 , and the optical density was measured at 450 nm. Fusion efficiency was defined as the ratio of the amount of b-HRP immobilized in the fusion assay to the amount of potentially immobilizable b-HRP, obtained in a separate reaction where biotinylated insulin was omitted.
Determination of the Free ATP Concentration in Endosome Fusion Assays-After 2 min of incubation at 21°C, a portion of an endosome fusion assay was clarified by centrifugation (100,000 ϫ g, 15 min, 4°C), and free ATP in the supernatant was separated from protein-bound ATP by ultrafiltration on a 10-kDa cut-off Ultrafree Millipore filter. The ATP concentration in the filtrate was directly assayed with an LKB 1250 luminometer using a luciferase-luciferine assay (Sigma).
Preparation of His 6 -Rab7 Recombinant Protein-D. discoideum rab7 was amplified from a vegetative cell library (kindly given by Dr. Herb Ennis, New York) using Taq DNA polymerase. The oligonucleotide primers were designed to contain a BamHI site on the 5Ј-end. The PCR product was cloned into pGEM-T (Promega), introduced into the pQE-9 expression vector (Qiagen), and transformed in Escherichia coli XL1blue cells (Stratagene). The pQE9-rab7 plasmid was checked by sequencing and used for protein expression.
Preparation and Affinity Purification of Anti-Rab7 Antibodies-To raise antibodies against Rab7, two peptides from the effector (amino acids 37-51) and the C-terminal hypervariable domains (amino acids 176 -191) were coupled to rabbit serum albumin and coinjected in rabbits (Elevage Scientifique des Dombes, Romans, France). An affinity column was prepared by coupling the above peptides to glutaraldehydeactivated Affi-Gel 102 beads (Bio-Rad) and used to purify anti-Rab7 antibodies from rabbit serum as described in Ref. 20. The antibody fractions eluted in acidic or alkaline conditions were immediately neutralized, dialyzed twice against PBS, concentrated by ultrafiltration (Centricon, Amicon), and stored at 4°C. Both pools of anti-Rab7 antibodies inhibited the endosome fusion assay equally.

RESULTS
An in vitro assay for homotypic fusions between fluid phase loaded D. discoideum endocytic compartments was derived from the one based on the formation of a complex between b-HRP and avidin described for mammalian cells (17). In preliminary experiments, the uptake rate, intracellular transit time, and exit rate of b-HRP and avidin were found similar to those of fluorescein isothiocyanate-dextran, a fluid phase marker in D. discoideum (21). In various mammalian cells like macrophages and hepatocytes, HRP has been shown to enter the cells by a mannose receptor pathway, characterized by saturation of the uptake rate at low HRP concentrations (0.02 mg⅐ml Ϫ1 ) (22). In contrast, HRP uptake by D. discoideum did not exhibit any saturation from 0.05 mg⅐ml Ϫ1 up to the maximum concentration tested of 5 mg⅐ml Ϫ1 . Avidin uptake was also linear with concentration in the same range. It is therefore likely that these markers are internalized in the fluid phase and not bound to a receptor. This validates the use of b-HRP and avidin as fluid phase markers in D. discoideum.
Kinetic Characterization and Partial Purification of Fusogenic Endosomal Compartments in D. discoideum-Avidin and b-HRP were internalized by D. discoideum cells for 5 min and chased for various times. PNS were prepared, and those having the same chase time were combined in a fusion assay. All PNS preparations contained the same amount of internalized markers; however, the efficiency of the fusion reaction decreased exponentially with chase duration (t 1 ⁄2 ϭ 5 min, Fig. 1A). Furthermore, markers contained in PNS with 0-min chase time were unable to fuse with markers contained in PNS with 15min chase time (data not shown), which shows that only homotypic and no heterotypic fusion of early endocytic compartments occurs in D. discoideum.
To separate the fusogenic endosomes from other membranes and from the cytosol, a PNS preparation from D. discoideum cells loaded with b-HRP for 5 min was fractionated by centrifugation through a discontinuous sucrose gradient. Cytosolic components were retained on the upper 3 ml of the gradient, as indicated by the high protein concentration, whereas the dense lysosomes, assessed by the presence of cathepsin B, acid phosphatase, ␣-mannosidase, and ␤-glucosidase activities, sedimented to the bottom 40 -54% sucrose interface. Most of the alkaline phosphatase activity, a marker of the plasma membrane and the contractile vacuolar system in D. discoideum, was present at the 8 -30% interface, as expected (23). Little HRP activity was found at the top of the gradient, indicating that most of the endosomal compartments remained sealed during the purification procedure. Half of the HRP activity was recovered at the 30 -40% interface, and half sedimented to the 40 -54% interface (Fig. 2). When fluid phase markers were chased for 15 min, the whole HRP activity was found in a single peak at the 40 -54% interface, which indicated that the light compartments were occupied before the dense ones by the markers and that the transit time from the lighter to the denser compartment was less than 15 min. Identical results were obtained with fluorescein isothiocyanate-dextran as a fluid phase marker.
After a 5-min pulse, when endocytic compartments recovered at each of the interfaces were tested for homotypic fusion ac- FIG. 1. Fusion between D. discoideum endocytic compartments is restricted to the early ones. A, amoebas were pulsed in parallel with avidin or b-HRP at 21°C for 5 min and chased for the indicated times. Preparations of PNS from avidin-or b-HRP-loaded cells of the same chase time points were allowed to fuse in the presence of an ATP-regenerating system. Fusion efficiency is defined under "Experimental Procedures." The solid line is the best exponential fit of the data. B, amoebas were incubated with avidin or b-HRP at 21°C for 5 min, and PNS were prepared and fractionated in parallel on discontinuous sucrose gradients as described under "Experimental Procedures." After centrifugation, the light 30 -40% and dense 40 -54% interfaces were recovered (fractions 6 and 2 in Fig. 2), and two of them were combined in a fusion assay performed in the presence of an ATPregenerating system with (ϩ) or without (Ϫ) 50 M GTP␥S. Avidin and b-HRP refer to avidin-or b-HRP-loaded endosomes, and L and D stand for the light and dense fractions.
FIG. 2. Subcellular fractionation of D. discoideum endocytic compartments. Amoebas were pulsed with b-HRP at 21°C for 5 min and chased for 0 or 15 min. PNSs were prepared and loaded in parallel onto discontinuous sucrose gradients (see "Experimental Procedures"). After centrifugation, 1-ml fractions were collected from the bottom of the tube. Protein and sucrose concentrations are represented by the solid and dotted lines, respectively. HRP activity in the conditions without chase (q) or after 15 min of chase (E) was determined using tetramethylbenzidine (see "Experimental Procedures"). Alkaline phosphatase activity (OE) was determined using p-nitrophenyl phosphate. Cathepsin B activity (Ⅺ) was determined using an MNA peptide. ␣-Mannosidase, ␤-glucosidase, and acid phosphatase activities were also measured, and their profiles along the gradient were found to be identical to that of cathepsin B (data not shown). tivity, only the low density endosome fraction present at the 30 -40% interface was active. Fusion efficiency was comparable with that obtained with PNS. The high density endocytic compartments present at the 40 -50% interface were never found to be self-fusogenic, regardless of the cytosol concentration and the presence or absence of guanine and adenine nucleotides (Fig. 1B). Furthermore, the endocytic compartments recovered at the 30 -40% interface and at the 40 -54% interface, despite containing similar amounts of fluid phase markers, never showed heterotypic fusion activity (Fig. 1B). These results are most simply interpreted by the existence of two compartments, with the internalized fluid passing from one compartment (light and self-fusogenic) into another compartment (denser and nonfusogenic). The dense, lysosomal enzyme-rich compartment found at the 40 -54% interface corresponds to the "lysosomes" (14). By analogy with mammalian endocytosis, the newly described light fusogenic compartment present at the 30 -40% interface will hereafter be designated as "early endosomes." GTP and Nonhydrolyzable Analogues Activate in Vitro Homotypic Fusions between Partially Purified Early Endosomes in D. discoideum-Guanine nucleotide-binding proteins are well known to regulate endosomal membrane fusion (24). To test whether in vitro fusion of D. discoideum early endosomes was also regulated by GTP-binding proteins, the effects of GDP, GDP␤S, GTP, or GTP␥S were tested (Fig. 3A). The three nucleotides GTP (K 50 ϭ 10 M), GTP␥S (K 50 ϭ 5 M), and GDP (K 50 ϭ 5 M) all activated fusion, 3-fold for GTP␥S and 2-fold for GTP and GDP, while GDP␤S was slightly inhibitory. The activating effect of GDP, but not GDP␤S, can be explained by the presence in the cytosol of a strong NDP kinase activity (25) catalyzing the regeneration of GTP from GDP and ATP but unable to catalyze the transphosphorylation of GDP␤S.
In mammalian cells, endosome fusion exhibits a complex response to GTP␥S, which activates fusion at low cytosol con-centrations and inhibits fusion at higher ones (26). Similarly, it has been reported that GTP␥S stimulates the fusion of pelleted D. discoideum endosomes in the absence of added cytosol and gradually inhibits fusion as the cytosol concentration is raised (16). In contrast, partially purified endosomes exhibit a clear requirement for soluble proteins (Fig. 4). Furthermore, the relative effect of the various guanine nucleotides tested above is conserved over a large range of cytosol concentrations (Fig.  4). This shows that at the cytosol concentration used within this study (0.8 mg⅐ml Ϫ1 ), the stimulatory effect of GTP␥S predominates, since the onset of GTP␥S inhibition starts only at cytosol concentrations higher than 2 mg⅐ml Ϫ1 (Fig. 4).
Anti-Rab7 Antibodies Inhibit Fusion of Early Endosomes in D. discoideum-In D. discoideum, the small G-protein Rab7 is associated with the endocytic pathway (27). The distribution of Rab7 along the endocytic pathway was examined on magnetically purified iron-dextran-fed endocytic compartments, prepared under different pulse/chase conditions: early compartments (Fig. 5A, lane 1), lysosomes (lane 2), and postlysosomes (lane 3). Much less Rab7 was found per g of loaded protein material on postlysosomes than on lysosomes or early compartments. Rab7 is therefore enriched on early compartments and lysosomes as compared with postlysosomes.
To test whether Rab7 could mediate the activating effect of GTP and GTP␥S on early endosome fusion, we used affinitypurified anti-Rab7 antibodies that recognize a 23-kDa band in Western blots of whole cell extracts or endosomal membranes. This protein was identified by microsequencing to be Rab7 (15). The addition of these antibodies to the endosome fusion assay resulted in more than 70% inhibition (Fig. 5B), with a K 50 of 30 nM. The inhibitory effect of the antibodies was reversed by the addition of a 3-fold excess of purified recombinant His 6 -Rab7 (Fig. 5B). Another abundant protein at the surface of endocytic compartments in D. discoideum is the vacuolar ATPase (14,15). No inhibition was observed when similar amounts (100 nM) of purified monoclonal antibodies raised against the C subunit of the vacuolar ATPase (kindly given by Dr. G. Gerisch) were added to the fusion assay (data not shown). This rules out a steric effect of antibodies binding to the endosomes being the cause of this inhibition. These data strongly indicate that Rab7 activates endosome fusion in D. discoideum.
In the Presence of GTP␥S, ATP-Mg 2ϩ from the Bulk Solution Is Not Required for in Vitro Fusion of Partially Purified Early Endosomes in D. discoideum-In the absence of added guanine nucleotides, the level of homotypic fusions between partially purified early endosomes was reduced to less than 10% when an ATP-depleting system was substituted to the ATP-regenerating system. However, the addition of GTP␥S restored fusion activity to the maximum level obtained in the presence of ATP  ( Fig. 3B). When GTP replaced GTP␥S, very little stimulation was observed. This was due to the conversion of GTP into GDP by the NDP kinase activity coupled to the ATP depletion system. The same explanation holds true for endogenous GTP.
Free ATP in the endosome fusion assays was separated by ultrafiltration and quantified (see "Experimental Procedures"). The residual ATP concentration was 0.6 Ϯ 0.4 nM or 0.05 Ϯ 0.04 nM with an ATP-depleting system consisting of hexokinase and glucose or apyrase, respectively, and was unaffected by the presence of GTP␥S. Furthermore, ATP-depletion was very fast, dropping to the described levels in less than 2 min, while endosome fusion continued steadily over 30 min. Therefore, it can be concluded that more than 90% of the observed fusions occurred in the presence of less than 1 nM free ATP-Mg 2ϩ .
The Addition of ATP␥S Did Not Inhibit in Vitro Early Endosome Fusions in D. discoideum-To rule out the possibility that an ATP requirement for fusion was in fact being substituted by GTP contaminating the commercial GTP␥S preparations, we examined the effect of ATP␥S on the fusion activity measured in the presence of an ATP-depleting system and GTP␥S. Although the low ATP concentration favored the competition of ATP␥S versus ATP even more, no inhibition was observed (Fig. 6). Moreover, in the presence of 100 M GTP, ATP␥S even activated endosome fusions up to the level of GTP␥S. This might be due to the transfer of the thiophosphate from ATP␥S to contaminant GDP by D. discoideum NDP kinase, a reaction already described in other cells (28).

D. discoideum Endosome Fusions Performed in the Absence of ATP Are Specific to Early Compartments and Sensitive to N-Ethylmaleimide and Anti-Rab7 Antibodies-Most in vitro
reconstituted vesicular transport and membrane fusion systems that have been studied so far show no activity when the ATP concentration is reduced to the nanomolar level. Furthermore, endosome fusion in mammalian cells clearly requires ATP in addition to Rab protein activation (29). We therefore checked that the absence of ATP did not change (i) the specificity of D. discoideum endosome fusion for early compartments, (ii) their sensitivity to anti-Rab7 antibodies, and (iii) their inhibition by the NEM analogue A-485. As shown in Fig.  7, D. discoideum endosome fusion conducted in the absence of ATP resembles the fusions conducted in the presence of ATP with respect to all of the above criteria.
We finally tested whether NSF requirements could be fulfilled before Rab7 activation, using A-485 to inactivate NSF and GTP␥S to trigger Rab7 activation in the presence of an ATP-depleting system. When GTP␥S was added at the beginning of the 21°C incubation, the time course of resistance to A-485 closely followed the time course of the fusion reaction, indicating that NSF was required up to a late step in the fusion (data not shown). After 20 min of preincubation at 21°C in the presence of GTP␥S, 50% of the fusion signal had become A-485insensitive (Fig. 8, a and b). However, when GTP␥S and A-485 were added simultaneously after a 20-min preincubation, no fusion activity was observed (Fig. 8d). In this experiment, NSF was active during the preincubation, and Rab7 was only activated when NSF had been inactivated. The complete absence of fusion activity shows that the NSF requirements for endosome fusion in D. discoideum cannot be fulfilled before Rab7 activation. Adding GTP␥S alone after 20 min of preincubation gave almost maximum levels of fusion activity (Fig. 8c), showing that no factor required for the fusion activity was lost during the preincubation. DISCUSSION In this study, we show that only early endosomal compartments of D. discoideum are capable of homotypic fusion in vitro. In addition, no heterotypic fusion between early endosomal and lysosomal or postlysosomal compartments is observed. The early endosomes probably correspond to the prelysosomal vesicles (30) and are separated from the lysosomes and other dense organelles on a sucrose density gradient. In additional experiments, we observed that these dense compartments inhibited the fusion reaction of early endosomes (data not shown). Conversely, the addition of early endosomes to a fusion reaction conducted with the dense compartments did not promote lysosome fusion. The ability of light endosomes to fuse is therefore an intrinsic property of these membranes, and the FIG. 5. Involvement of Rab7 in early endosome fusion. A, distribution of Rab7 along the endocytic pathway. Amoebas were loaded with iron-dextran for the indicated pulse and chase times, and endocytic vesicles were magnetically purified (15). Similar amounts of magnetic vesicles were loaded on a 12% SDS-polyacrylamide gel electrophoresis gel (5 g). After transfer to nitrocellulose, the presence of Rab7 was detected by affinity-purified anti-Rab7 antibodies. The band of lower molecular weight in lanes 1 and 2 is probably a Rab7 degradation product and accounts for about 20% of the 23-kDa band, as quantified by densitometry on a film exposed for a shorter time. B, the addition of anti-Rab7 antibodies inhibits early endosome fusion, and the addition of His 6 -Rab7 prevents the effect of anti-Rab7 antibodies. In vitro fusion of partially purified early endosomes was conducted in the presence of an ATP-regenerating system, 50 M GTP␥S and either the indicated amount of affinity-purified anti-Rab7 antibodies (q) or 130 nM of anti-Rab7 antibodies and the indicated amount of purified recombinant His 6 -Rab7 (E). Buffers and His 6 -Rab7 alone were without effect. Our results show that GTP and GTP analogues activate endosome fusion. The presence of an activated G protein, but not its deactivation, is therefore required for the fusion process. This stimulatory G protein probably does not belong to the Arf subfamily, because impairment of GTP hydrolysis on Arf proteins is known to block homotypic or heterotypic fusions (31)(32)(33). In contrast, impairment of GTP hydrolysis on Rab5 protein activates early endosome fusion in mammalian cells (29,33,34). Rab proteins are therefore good candidates to account for the activating effect of GTP and GTP␥S. Furthermore, affinitypurified anti-Rab7 antibodies inhibit early endosome fusion, the inhibition being reverted by purified recombinant Rab7. Rab7 is indeed present on magnetically purified endocytic compartments corresponding to early endosomes and lysosomes but not postlysosomes. It has recently been proposed that Rab7 is involved in the recycling of lysosomal enzymes from the postlysosomal compartment back to the lysosomes (35). Based on our results, we propose to extend this hypothesis and assume that Rab7, perhaps along with other Rab proteins, also controls the fusion between these recycling vesicles and incoming new material. In this context, homotypic endosome fusion would appear as a side effect of the physiologically relevant fusion between endosomes and recycling vesicles. The hypothesis is further supported by the phenotype of an overexpressed mutant rab7 (35). Rab7 T22N , defective in GTP binding, exhibited a highly reduced rate of fluid phase entry and acidification, the opposite effect being observed with Rab7 Q67L , a mutant protein impaired in GTP hydrolysis. Rab7 could therefore couple both ends of the endocytic pathway in D. discoideum.
An unexpected feature of the in vitro fusion of D. discoideum early endosomes is the absence of a requirement for free ATP-Mg 2ϩ , provided that GTP␥S is present. Similarly, fusion assays performed with whole PNS were also insensitive to ATP depletion in the presence of GTP␥S (data not shown). Therefore, no factor requiring external ATP was lost during the endosome purification procedure. In the absence as well as in the presence of ATP, endosome fusion exhibited the same specificity to early compartments and was sensitive to inhibition both by anti-Rab7 antibodies and NEM. Considering these criteria, endosome fusions performed in the absence of ATP are similar to other well described homotypic fusion systems. Finally, the addition of ATP␥S in the presence of an ATP-depleting system and GTP␥S does not inhibit fusion. Altogether, these results show that either no ATP-binding protein is needed in endosome fusion in D. discoideum or that the fusion proceeds with tightly bound endogenous ATP.
The first hypothesis is not tenable because the addition of mammalian NSF can restore endosome fusion activity to NEMinactivated D. discoideum PNS (16). Furthermore, evidence presented above suggests that Rab7 plays a role in endosome fusion. The fusion machinery seems therefore very close to the one described in mammalian cells and yeast that involves NSF, SNAPs, SNAREs, and Rab proteins (7). We therefore favor the FIG. 7. Early endosome fusion performed in the absence of ATP is specific to early compartments and sensitive to anti-Rab7 antibodies and to the NEM analogue A-485. A, fusion assays were conducted exactly as in Fig. 1B, except that the ATP-regenerating system was replaced by an ATP-depleting system. B, fusion assays were conducted exactly as in Fig. 5B, except that the ATP-regenerating system was replaced by an ATP-depleting system. C, partially purified endosomes prepared in the absence of DTT were treated with 1 mM A-485 for 10 min on ice, and then 2 mM DTT was added (A-485 treated). As a control, the order of A-485 and DTT additions was reversed (mock treated). These endosomal membranes were then assayed for fusion activity in the presence of 50 M GTP␥S and an ATP-regenerating (ϩ) or -depleting system (Ϫ). Use of a membrane-impermeable NEM analogue was necessary, because the NEM treatment inhibits the formation of b-HRP-avidin complexes.
FIG. 8. NSF activity is required up to a late step in the fusion of early endosomes performed in the absence of ATP. Four fusion assays were prepared with partially purified early endosomes and an ATP-depleting system as described under "Experimental Procedures," except that no DTT was added. Samples a and b were also supplemented with 50 M GTP␥S. After 20 min of incubation at 21°C, all samples were set on ice. Samples b and d were treated with 1 mM A-485 for 10 min followed by 2 mM DTT, and samples a and c were mocktreated by reversing the order of A-485 and DTT additions. Furthermore, samples c and d were supplemented with 50 M GTP␥S. All samples were then returned to 21°C for 40 min, and the amount of b-HRP-avidin complexes was thereafter quantified as described under "Experimental Procedures." hypothesis that the nucleotide site of the ATP-binding proteins needed in this fusion assay, possibly D. discoideum NSF or an NSF-like protein, is inaccessible at this stage. Since the fusion activity originates from endosomes prepared from two different cell populations, this implies that ATP binding has occurred before the preparation of the PNS and therefore before the docking of the membranes. The order of the steps leading to membrane fusion would therefore differ from the initial model, accounting for synaptic vesicle fusion (8), but be consistent with more recent findings, that NSF is already present at the surface of isolated clathrin-coated vesicles (36) and synaptic vesicles (37) and that a predocking attachment site for NSF exists on endosomes (38) and on synaptic vesicles (39). Interestingly, it has been observed that phagocytosed live Listeria monocytogenes recruit Rab5 and NSF on the membranes of the phagosomes and make the fusion of the phagosomes with early endosomes insensitive to ATP depletion, NEM inhibition, and anti-NSF antibodies (40). It is therefore possible that in the case of live L. monocytogenes-containing macrophage phagosomes and early D. discoideum endosomes, a similar intermediate state in membrane fusion is isolated. Finally, endosome fusion in D. discoideum shows some similarity to homotypic fusion of yeast vacuoles where Sec17, Sec18, and ATP requirements can be fulfilled prior to vacuole docking (41), while mixing of vacuole membranes is required to proceed beyond the Ypt7-dependent step (42). However, in contrast with the conclusions of these studies (43), NSF function is required for the completion of endosome fusion in D. discoideum, and this requirement cannot be fulfilled in the absence of Rab7 activation.