Phagocytosed live Listeria monocytogenes influences Rab5-regulated in vitro phagosome-endosome fusion.

Survival or destruction of a pathogen following phagocytosis depends, in part, on fusion events between the phagosome and the endosomal or lysosomal compartments. Here we use an in vitro assay to show that phagosome-endosome fusion is regulated by the small GTPase rab5 and that fusion events are influenced by an internalized live organism, Listeria monocytogenes (LM). We compare the in vitro fusion of phagosomes containing heat-killed organisms (dead LM) with that of phagosomes containing a live nonhemolytic mutant (live LMhly−). Unlike the wild-type organism, LMhly− remains trapped inside the phagosome. Phagosome-endosome fusion was reconstituted using biotinylated organisms and endosomes containing horseradish peroxidase conjugated with avidin. With both live LMhly− and dead LM preparations, in vitro phagosome-endosome fusion was time-, temperature-, and cytosol-dependent. Live LMhly− phagosomes exhibited a faster rate of fusion. Fusion in both preparations was regulated by rab5 and possibly by other GTPases. Anti-rab5 antibodies and immunodepletion of cytosolic rab5 inhibited fusion. Addition of glutatione S-transferase-rab5 in the GTP form stimulated phagosome-endosome fusion, whereas addition of a dominant negative mutant of rab5 blocked fusion. Purified live LMhly− phagosomal membranes were enriched in rab5 as revealed by Western blotting, compared with dead LM phagosomes. Fusion of endosomes with dead LM containing phagosomes required ATP and was inhibited by ATP depletion and by N-ethylmaleimide (NEM) and anti-NEM-sensitive factor (NSF) antibodies. Unexpectedly, phagosome-endosome fusion with live LMhly−-containing phagosomes was not inhibited by ATP depletion nor by NEM or anti-NSF antibodies. Western blot analysis revealed that live LMhly−-containing phagosomes were enriched for membrane-bound NSF, while dead LM containing phagosomes contained low or undetectable quantities. Washing live LMhly−-containing phagosomes with 0.5 M KCl removed NSF associated with the membranes and rendered them NEM, ATP, anti-NSF antibody sensitive for fusion. We conclude that rab5 regulates phagosome-endosome fusion and that live microorganisms can up-regulate this process by recruiting rab5 to the membrane.

conclude that rab5 regulates phagosome-endosome fusion and that live microorganisms can up-regulate this process by recruiting rab5 to the membrane.
Phagocytosis, an important element of the host-defense system, is initiated by the attachment of a particle or organism to the phagocyte via cell surface receptors followed by ingestion and formation of a phagosome. Subsequently, a series of sequential intracellular membrane fusion and budding events occur that accommodate maturation of the phagosome and the delivery of hydrolytic enzymes and other molecules important for host defense. Proper orchestration of these events is thought to lead to the destruction of the pathogen and to the initiation of appropriate immunological responses.
After internalization, phagosomes containing LM are thought to fuse with lysosomes to form phagolysosomes (7)(8)(9)(10)(11). Studies with heat-killed LM (dead LM) indicate that phagosomes receive material from the trans-Golgi network and from lysosomes (10). Studies with Staphylococcus aureus phagosomes have rendered similar conclusions (12). Virulent LM has been shown to lyse the phagolysosomal membrane and to escape into the cytoplasm where growth flourishes. Escape from the phagosome is mediated by listeriolysin O (also known as hemolysin). Listeriolysin O-defective mutants, i.e. nonhemolytic mutants (LM hlyϪ ) (13)(14)(15), are unable to escape from the phagolysosome and are degraded by the phagocyte. In this respect, LM hlyϪ mutants behave as heat-killed bacteria (dead LM). Our goal was to analyze early phagosome-endosome fusion events with LM containing phagosomes using an in vitro fusion assay, where the role of different factors and the influence of the live microorganism could be examined.
Over the past few years, there has been a substantial increase in our knowledge of vesicle budding and membrane fusion events along the endocytic pathway and the role of GTP-binding proteins as regulators of these processes (16 -20). The GTPase rab5 plays a central role in early endocytic events. Although rab5 was found to be present on phagosomes isolated following phagocytosis of latex beads (21), a functional role for rab5 in phagocytosis has not yet been described. Here, we show that in vitro fusion of phagosomes, containing both live and dead bacteria, with endosomes is regulated by rab5. Both fusion events require cytosolic proteins. However, the ATP requirement and the sensitivity of fusion to NEM are quite different for phagosomes containing live and dead organisms. These differences lead us to postulate that rab5-regulated early phagosome fusion is dependent on the status of the microorganism. Moreover, the data indicate that live L. monocytogenes directly affects and modulates the fusion process by recruiting rab5 to the phagosomal membranes.

MATERIALS AND METHODS
Biological Reagents-J774E clone, a mannose receptor-positive macrophage cell line, was grown as described previously (12). Monoclonal antibodies (mAb) used include: 4F11, a mouse IgG2a k monoclonal antibody that recognizes mouse rab5 (22); HDP-1, an anti-DNP mouse mAb was employed as an irrelevant antibody (23); 6E6, a mouse IgG mAb that recognizes the D1-D2 domains of NSF (a gift from Sidney W. Whiteheart (University of Kentucky, Lexington, KY) (24) was used for Western blotting; 4A6, a mouse IgM monoclonal antibody that recognizes native NSF, was kindly provided by James E. Rothman (Sloane-Kettering Memorial Hospital, NY) and used in the in vitro fusion assays. An irrelevant IgM monoclonal antibody was included as a control (Sigma). Polyclonal antibodies specific for rab7 and rab5 were generated by immunizing rabbits with GST-rab fusion proteins. These antibodies recognize human and mouse rab5 and rab7, respectively. 2 A polyclonal antibody against chicken immunoglobulins made in rabbits was obtained from Sigma and included as an irrelevant antibody. Horseradish peroxidase (HRP) conjugated with avidin (HRP-avidin) was obtained from Pierce. Cytosol was prepared as described elsewhere (16). Cytosol was gel-filtered through 1 ml of Sephadex G-25 spin column just before use in the fusion assay. Protein concentrations were measured as described previously (25) using bovine serum albumin as the standard. All other chemicals were obtained from Sigma. Purified GDI was obtained from B. Goud (Pasteur Institute, Paris, France) (26).
Purification of Recombinant Rab5 Fusion Proteins, Prenylation, and Solubility-The rab5 mutants used in this study were described previously (27,28) and included rab5:S34N and rab5:Q79L. Recombinant rab5 proteins were expressed in large quantity as glutathione S-transferase fusion proteins in Escherichia coli strain JM 101 and were affinity-purified by glutathione-Sepharose resin (GST-rab5). Before prenylation, rab-geranylgeranyl transferase (REP-1) was partially purified following a protocol previously reported (29,30). Recombinant rabs were prenylated as described previously (29,31,32). Briefly, rab5 substrate (10 M) was incubated with the REP-1 in 50 l of 50 mM Hepes/KOH, pH 7, 5 mM MgCl 2 , 0.5 mM Nonidet-P40, 1 mM DTT (buffer A) containing 12 M geranylgeranyl pyrophosphate (GGPP) for 30 min at 37°C. After the reaction, the prenylated rab5 proteins were aliquoted at 4 mg/ml and stored at Ϫ80°C. Under these conditions, 70% of the rab5 was prenylated as measured by the incorporation of [ 3 H]GGPP. Rab5 copurifies with rab escort protein as described by Andres et al. (33). Rab escort protein has been shown to mediate rab5 binding to endosomal membranes (34). Solubility of the purified proteins was checked as follows. Recombinant purified proteins (1 g) were resuspended in 200 l of the following solutions: (a) buffer B (50 mM Hepes/ KOH, pH 7, 1 mM MgCl 2 , 1 mM DTT); (b) buffer B containing 1 mM Nonident P-40; and (c) buffer B containing 0.1% BSA. The solubility of rab5 proteins was evaluated by the presence of a visible aggregate after 40 min at 37°C incubation under each of the conditions described above. Aggregates were found where the protein was resuspended in buffer B without detergent or BSA. Purified protein stocks contained 1-3% BSA, which resulted in a relatively small amount of BSA in our fusion reactions (less than 100 mol of BSA/mol of prenylated rab5). BSA may affect the solubility and the appropriate targetting of prenylated rabs in the absence of appropriate escort proteins (35).
Phagocytic Probes-The nonhemolytic L. monocytogenes mutant strain used in this study (DP-L2161) (LM hlyϪ ) (15) derived from the wild-type strain (10403S) was kindly provided by D. A. Portnoy (Uni-versity of Pennsylvania, Philadelphia, PA). The bacteria were grown in brain-heart infusion broth (Difco) at 37°C with aeration in the presence of 150 g/ml streptomycin. The bacteria were harvested in the logarithmic phase of growth and were heat-killed by treatment at 60°C for 1 h. Heat-killed bacteria (dead LM) were confirmed dead by plating in blood agar plates. Bacteria were extensively washed and stored at 4°C. LM hlyϪ and dead LM (10 9 bacteria) were biotinylated with ss-NHSbiotin (Pierce) according to the manufacturer's recommendations and as described previously (36). Briefly, bacteria were washed three times in PBS (pH 7.4) and resuspended in PBS (pH 8.0) at 4°C. The bacteria were treated with ss-NHS-biotin at 0.5 mg/ml for 2 min with gentle shaking. They were washed sequentially in PBS, 50 mM NH 4 Cl, PBS, 0.1 mM CaCl 2 , PBS, 1 mM MgCl 2 to quench free biotin, and resuspended in PBS. Biotinylation did not affect bacteria viability as determined by plating in blood agar plates. The number of biotin molecules per organism was estimated in every experiment using a colorimetric assay with the dye 2-(4Ј-hydroxyazobenzene)benzoic acid (Pierce) according to the manufacturer's instructions. While the numbers vary from experiment to experiment, routinely ϳ40 mol of ss-NHS-biotin/10 9 bacteria were incorporated, and no significant differences were found between dead and live LM. ss-NHS-biotin/dead LM were stored at 4°C and ss-NHS-biotin/live LM hlyϪ were stored at Ϫ70°C until use.
To identify biotinylated bacterial proteins, 10 9 biotinylated bacteria were extracted (36) by resuspension in 100 l of PBS containing 1% SDS. Following centrifugation, SDS-PAGE, and transfer to nitrocellulose, blotting with avidin-HRP indicated that multiple proteins in both preparations of bacteria were biotinylated. The blots were essentially identical for the two preparations.
Assay for Bacterial Uptake-The assay was performed essentially as previously reported (37,38). Briefly, LM hlyϪ bacteria were grown overnight in brain-heart infusion broth. After extensive washing, bacteria were resuspended in RPMI (minus methionine and cysteine) with 1 mCi of [ 35 S]methionine (ICN Tran 35 S-label). The bacteria were incubated with rotation for 6 h, washed with cold PBS, and resuspended in PBS. 35 S-Labeled dead LM were prepared by heating as described above. Labeled bacteria incorporated 1-2 cpm/bacterium. Dead LM were also surface-labeled with 125 I by the chloramine T method at 4°C (37). For uptake assays, live 35 S-labeled LM hlyϪ and 125 I-or 35 S-labeled dead LM were added (3 ϫ 10 5 cpm/well) to 2 ϫ 10 6 cells (E-clone) in 24-well plates and centrifuged (2,000 rpm, 5 min, 4°C) to speed adherence and to synchronize the infection. Following incubation at 37°C for different times, cells were solubilized with 1% Triton X-100. Experiments were performed in duplicate.
Preparation of Phagosomes and Endosomes for in Vitro Fusion-J774E clone macrophages (10 8 cells) were incubated with ss-NHSbiotin/dead-LM or ss-NHS-biotin/live LM hlyϪ (10 9 bacteria) (1 h, 4°C) and centrifuged to synchronize the infection (2,000 rpm, 5 min, 4°C). Uptake was initiated by addition of prewarmed HBSA (Hank's balanced salt solution buffered with 10 mM HEPES and 10 mM TES, pH 7.4, and supplemented with 1% BSA). After 10 min at 37°C, uptake was stopped by addition of ice-cold HBSA. Cells were sequentially washed with HBSA, PBS-EDTA, and HBE (250 mM sucrose, 0.5 mM EGTA, 20 mM HEPES-KOH, pH 7.2) by centrifugation (300 ϫ g for 4 min). Cells were resuspended with HBE (2 ϫ 10 8 cells/ml) and homogenized in a ball bearing homogenizer (12). Homogenates were centrifuged at 400 ϫ g for 3 min to eliminate nuclei and intact cells. The postnuclear supernatants (PNS) were quickly frozen in liquid nitrogen and stored at Ϫ80°C. Phagosomal fractions were obtained by diluting a quickly thawed PNS aliquot in 1 ml of HBE and centrifuging at 12,000 ϫ g in a Microfuge for 10 s at 4°C, as previously reported (12,39,40). The supernatant was kept at 4°C and the pellet resuspended with HBE and centrifuged again. The resulting supernatants were combined and centrifuged at 12,000 ϫ g for 6 min at 4°C. This pellet (phagosomal fraction) containing 70% of the total phagosomes in the sample was used for in vitro fusion assay. Early endosomes containing HRP-avidin conjugate were prepared as described previously with several modifications (41). J774E clone (10 8 cells) was incubated with HRP-avidin (200 g/ml) for 1 h at 4°C. Under these conditions, HRP-avidin is essentially endocytosed via the mannose receptor (42). Uptake was initiated by the addition of prewarmed medium (1 ml) at 37°C and, after 10 min, terminated by adding ice-cold medium. Cells were washed and homogenized, and the PNS fraction was quickly frozen. To prepare endosomes, a thawed PNS (200 l) was diluted up to 3 ml with HBE and pelleted for 1 min at 37,000 ϫ g at 4°C. The supernatant was centrifuged for 5 min at 50,000 ϫ g at 4°C. The pellet of the second centrifugation, enriched in 10-min endosomes (endosomal fraction) was used for in vitro fusion assays. The HRP-avidin was contained within intact vesicles by the criterion that greater than 90% of the HRP activity was sedimented by centrifugation 2 G. Li, personal communication.
Fusion Reaction-Phagosomal fractions containing biotinylated dead LM or biotinylated live LM hlyϪ and endosomal fractions containing HRP-avidin were mixed in fusion buffer (250 mM sucrose, 0.5 mM EGTA, 20 mM HEPES-KOH, pH 7.2, 1 mM dithiothreitol, 1.5 mM MgCl 2 , 100 mM KCl, including an ATP-regenerating system, 1 mM ATP, 8 mM creatine phosphate, 31 units/ml creatine phosphokinase, and 0.25 mg/ml avidin as scavenger) supplemented with gel-filtered cytosol. After incubation for 60 min at 37°C, the reaction was stopped by chilling on ice. The HRP-avidin/biotin-bacteria complexes were recovered by centrifugation (10,000 ϫ g, 6 min, at 4°C) after solubilization of membranes with solubilization buffer (PBS, 0.5% Triton X-100 containing 0.25 mg/ml avidin as scavenger). The enzymatic activity of HRP-avidin associated with the bacteria was then measured. The fusion activity was quantified as absorbance units/mg of protein (the HRP assay described below). Protein concentrations were determined in each experiment for all samples. Two controls were included in each experiment corresponding to (a) total activity and (b) background activity. Total activity is that which is particle associated when endosomes and phagosomes are mixed and then lysed with detergent. Total activity corresponds to the total fusion activity that could be achieved per experiment. Values around 480 absorbance units/mg were routinely recorded for both dead and live LM preparations. Background activity corresponds to particle-associated HRP when endocytic and phagocytic vesicles were mixed in fusion buffer lacking cytosolic proteins. These values were quite low and were subtracted from all other values. For measuring the fusion activity, each sample is compared to the fusion observed in the presence of complete fusion buffer containing the highest cytosol concentration which was normalized to a value of 1 after subtracting the background activity. The highest effective cytosol concentration was standarized for both fusion assays (with dead and live LM) and observed to be around 1 mg/ml for both assays. This fusion value corresponds to as much as 70% of the total activity depending on the experiment. The values obtained in each experiment are listed in table and figure legends. All reactions were incubated 60 min at 37°C.
Antibodies (mAb 4F11 anti-rab5, anti-rab7, mAb 4A6 anti-NSF, mAb HDP-1 anti-DNP, and controls) were added at the concentration indicated and incubated for 45 min on ice before the fusion assay. For kinetic studies, samples were incubated for different times at 37°C and then cooled on ice.
Treatment of Phagocytic Vesicles with GDI-The protocol was essentially as previously reported (43,44). Briefly, phagocytic vesicles (100 g) were preincubated in fusion buffer plus a protease inhibitor mixture (1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 20 g/ml aprotinin) for 20 min at 30°C in the presence or absence of either GTP␥S (1 mM) or GDP (1 mM). This incubation was followed by addition of 6 g of purified GDI for 10 min at 30°C. Samples were washed twice in fusion buffer and resuspended either in sample buffer to be analyzed by SDS-PAGE or in fusion buffer and tested in a fusion reaction assay (the vesicles to be evaluated for fusion were pretreated with GDI in the presence of GDP).
Horseradish Peroxidase Assay-The enzyme assay was conducted in 96-well microplates (Costar Co.) using o-dianisidine as the chromogenic substrate (45). Briefly, the reaction was started by adding 20 l of each sample to 100 l of 0.5 N sodium acetate (pH 5.0) containing 0.342 mM o-dianisidine and 0.003% H 2 O 2 . The reaction was conducted at room temperature for 20 min and stopped by adding 100 l of 0.1 N H 2 SO 4 . Absorbance was measured at A 450 nm in a Bio-Rad microplate reader. Each value was expressed as absorbance units/mg of protein. For quantification, a standard curve with different concentrations of HRP-avidin was included in each experiment.
Plastic Embedding and Cryosection Staining with Anti-rab5 Antibody-J774E clone cells were grown to confluence on 35-mm tissue culture dishes. 5 ϫ 10 7 dead or live LM were added per dish and centrifuged to synchronize the infection. Following 12 min of incubation at 37°C, unbound bacteria were washed out with PBS, 5 mM EDTA. Cells were fixed in 1% glutaraldehyde, 0.1 M Na-cacodylate buffer and prepared for electron microscopy as described previously (16,41). For cryosection staining experiments, cells grown on 35-mm tissue culture dishes were incubated for 6 min with BSA-gold (10 nm), washed, and infected with dead or live LM hlyϪ for 10 min, as described above. The cells were washed, scraped off, and fixed in suspension in 2% paraformaldehyde, 0.2% glutaraldehyde in PBS, pH 7.2, for 2 h at room temperature. Tissue was embedded in 10% gelatin. After pelleting the tissue, the gelatin was solidified on ice. Blocks for ultracryotomy were prepared and immunolabeled as described by Slot et al. (46) with the following modifications: 10% goat serum was used in the blocking buffer in place of 1% BSA. Immunolabeling with primary antibody mAb 4F11 (anti-rab5) (65 g/ml) was carried out overnight at 4°C. Incubation with secondary antibody (Jackson ImmunoResearch Labs., West Grove, PA) 18 nm of goat anti-mouse IgG/gold (1:15) was carried out for 1 h. Sections were stained with uranyl acetate and embedded in methyl cellulose according to a modification of the Tokuyasu method (47,48). All electron microscopy specimens were viewed and photographed using a Zeiss 902 electron microscope.
Preparation of Highly Purified Phagosomes-The phagosomal fractions described above were resuspended in HBE containing 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 20 g/ml aprotinin, and filtered through a 5-m pore filter. The flowthrough was loaded on top of a 12% sucrose cushion. Samples were centrifuged (1,700 rpm, 45 min, 4°C), and purified phagosomes were recovered in the last 100 l at the bottom of the tube. This protocol is similar to one described previously (49).
The purity of the organelles was monitored by two criteria: (i) electron microscopy observation to assess contamination with other organelles and (ii) biochemical analysis to check contamination with other cellular components. Plasma membrane contamination was assayed as previously reported (50). Briefly, after internalization of bacteria and extensive washings, cell surface was labeled with HRP (300 g/ml) for 30 min at 4°C. J774E clone, a mannose-receptor cell line, binds and internalizes HRP, a mannosylated protein. Mannan inhibits both HRP binding and uptake in these cells. 3 After several washings in PBS, homogenization and phagosome isolation was performed as described above. HRP was then measured in the final purified phagosomes. The galactosyltransferase activity was checked as a marker for Golgi contamination using [ 3 H]UDP-galactose (50). The endosome contamination was recorded by mixing an aliquot of PNS after bacterial uptake and an aliquot of a PNS after 5 min of uptake of ␤-glucuronidase or HRP. Phagosomes were isolated as above, and endosome contamination was measured as the percentage of ␤-glucuronidase (41), or HRP activity recovered in the phagosomes was compared to the total activity present in PNS containing the endosomal probe. Endosomes were marked by allowing a separate set of cells to internalize either HRP or ␤-glucuronidase for 5 min. An aliquot of the PNS obtained from such cells was included in the phagosome purification protocol. Recovery of 0.32 and 0.25% for ␤-glucuronidase and HRP, respectively, from a total recovery of 70% of the bacteria indicated low enrichment of endosomes in the preparation.
Immunodepletion of Cytosolic Rab5-The assay was performed as described previously (44). In brief, quick thawed cytosol aliquots (200 l) (5 mg protein/ml) were incubated for 4 h at 4°C with 10 g/ml mAb 4F11 (anti-rab5) or control antibody mAb HDP-1 (anti-DNP). Both cytosols were separately incubated (1 h, 4°C) with protein A-Sepharose beads (a 50% solution prepared in PBS containing 2% BSA). After centrifugation (12,000 ϫ g, 5 min, 4°C) supernatants were treated for another cycle with protein A-Sepharose beads, and 50 g of each sample were separated by SDS-PAGE in 15% polyacrylamide gels transferred onto nitrocellulose membranes. The membranes were incubated with a rabbit anti-rab5 polyclonal antibody (1:3,000) followed by incubation with an HRP-conjugated goat anti-rabbit IgG (1:10,000). Blots were developed by ECL (Amersham Corp.). Fig. 1A shows a typical in vitro fusion experiment carried out with phagosomes containing dead LM (filled circles) or live LM hlyϪ (open circles) with endosomes containing HRP-avidin. These data indicate that phagosome fusion, using both live and dead organisms, is cytosol-dependent. The two assays were quite similar in their cytosol dependence. The average of a large number of experiments indicated essentially no difference between the two preparations. To further delineate possible influences of the live organism on fusion, we performed a detailed study of the parameters regulating these fusion events (Table I). GTP␥S, when added in the presence of low levels of cytosol, stimulated fusion implicating one or more GTPases as regulators of phagosome fusion, irrespective of whether the microorganism was live or heat-killed. Previous work revealed the presence of rab5 and rab7 on the phagosomal membranes prepared following latex bead ingestion (21). To test whether in vitro fusion could be regulated by these small GTPases, antibodies specific for rab5 and rab7 were added to the fusion reactions. Neither of the phagosome-endosome fusion assays was affected by anti-rab7 antibodies; however, anti-rab5 antibodies strongly inhibited both assays. As indicated in Table I, the presence of anti-rab5 in the fusion reactions clearly blocked phagosomeendosome fusion in both assays. The inhibition was observed at all cytosol concentrations tested (data not shown). Inhibition was dependent on antibody concentration, and concentrations as low as 10 ng/ml blocked 75-90% of the fusion reaction (data not shown). Interestingly, the fusion of live LM hlyϪ -containing phagosomes was particularly sensitive to the rab5 antibody.

Reconstitution of Phagosome-Endosome Fusion with Dead LM-and Live LM hlyϪ -containing Phagosomes-
Both phagosome-endosome fusion assays were temperaturesensitive and required salt (e.g. KCl) and both cytosolic and membrane-bound proteins. Trypsin treatment of both phagosomal and endosomal membranes impaired the fusion process.
Differential Accumulation of Rab5 on Phagosomal Membranes of Dead and Live LM-containing Phagosomes-Although only rab5 seems to regulate both phagosome-endosome fusion assays and rab7 plays no role (as reflected in the data obtained with specific antibodies), we checked for the presence of these proteins on the membranes of both dead LM-and live LM hlyϪ -containing phagosomes. Highly purified phagosome membranes were analyzed by SDS-PAGE, and the presence of rab5 or rab7 was detected by Western blot analysis using specific antibodies. Measurement of a Golgi membrane marker, galactosyltransferase activity, in this highly purified phagosome preparation detected less than 0.2% of the total (relative to the marker present in the PNS). HRP bound to the E-clone cell surface (as a measure of cell surface marker) was always less than 0.8%. Contamination of phagosome preparations with endosomes was never more than 0.35%, as measured by the ␤-glucuronidase or HRP activities recovered in these isolated phagosomes (see "Materials and Methods"). As shown in Fig.  2A, rab5 was detected on both types of phagosomes; however, the time course of rab5 recruitment to the membranes (7 and 20 min) was significantly faster for phagosomes containing live LM hlyϪ . Small amounts of rab7 were found compared to rab5 levels but, whereas rab7 was found on the live LM hlyϪ phagosomes, it was virtually absent from the dead LM phagosomes. Since differences in the enrichment of rab5 could be a consequence of faster fusion kinetics with phagosomes containing live bacteria, we compared the fusion kinetics of both assays. As shown in Fig. 2B, the kinetics obtained with the live LM hlyϪ phagosomes revealed a much faster fusion rate with significant fusion occurring in just 5 min, while phagosomes containing dead LM required at least 30 min to reach a similar relative fusion activity. Differences in the uptake of dead or live bacteria cannot explain these results because the uptake rate for the two preparations of bacteria was the same (data shown in Fig.  2C). The levels of biotinylation of dead or live bacteria were similar, 42 mol of ss-NHS-biotin/10 9 for dead LM and 41 mol ss-NHS-biotin/10 9 for live LM hlyϪ . Moreover, the same proteins were biotinylated in the dead and live LM hlyϪ preparations (data not shown) and the maximal signal recovered in the phagosomes was similar for both preparations, around 70% of the total signal measured in the PNS. Analysis of the number of bacteria per phagosome compared to total numbers of bacteria per cell (data shown in Fig. 3; panel A corresponds to cells infected with dead LM, and panel B with live LM) showed that, independent of the status of the bacteria (dead or live), 17% of internalized bacteria were in phagosomes containing single bacteria, 26% in phagosomes containing two to three bacteria, and the remainder, 56%, in phagosomes containing four or more bacteria. It is possible that in our phagosome purification procedure we may have selected those phagosomes containing one, two, or three bacteria instead of those containing more than four that might be more sensitive to membrane disruption during cell lysis.
The presence of rab5 in the phagosomes containing live LM hlyϪ or dead LM was also analyzed using the cryosection immunogold technique. Endosomes were labeled with BSAgold (10 nm) (indicated by large arrowheads, Fig. 4) and rab5   Fig. 1) were resuspended in complete fusion buffer and 1 mg/ml cytosolic proteins, or as otherwise indicated, and fusion was measured as described under "Materials and Methods." Background activities were 54 and 57 and control values (1 mg/ml cytosolic proteins) were 274 and 287 absorbance units/mg of protein for the dead and live fusion assays, respectively.
b The following conditions were tested: Ϫcytosol, cytosol was omitted; low cytosol ϩGTP␥S, cytosolic proteins at 0.125 mg/ml and 20 M GTP␥S added to the fusion assay (values were 356 and 345 absorbance units/mg protein for dead and live LM phagosomes); ϪKCl, the salt was not included in the buffer; trypsin, both sets of vesicles were incubated (1 h, 4°C) with 10 g/ml trypsin. To quench the excess trypsin, the samples were incubated (30 min, 4°C) with 20 g/ml soybean trypsin inhibitor, 1 mM phenylmethylsulfonyl fluoride, and 1 M leupeptin; anti-rab7, polyclonal anti-rab7 antibody (0.1 g/assay) was added to the fusion reaction (control antibody was included); anti-rab5, mAb 4F11 anti-rab5 antibody (0.1 g/assay) (as a control mAb HDP-1) was added to the fusion reaction. Values are the mean of at least four determinations Ϯ SD. was detected by immunogold staining (18 nm) (indicated by arrows, Fig. 4). As shown in Fig. 4, B and D, rab5 is localized in those phagosomes containing live LM hlyϪ . Fusion of these phagosomes with endosomes was detected by the presence of BSAgold (large arrowheads). Phagosomes containing dead LM (panels A and C) had also fused with endosomes (labeled with BSA-gold and indicated by large arrowheads) but lacked significant rab5 staining, even though rab5 could be detected in other vesicles near the phagosomes (panel A, arrow). Phagosomal and endosomal membranes are marked with small arrowheads.
Effect of rab5-depleted Cytosol on in Vitro Phagosome-Endosome Fusion-To further explore the requirement for rab5 and to determine whether antibody-mediated inhibition was due to membrane-bound rab5 or to cytosolic rab5, we removed rab5 from the cytosol by an immunodepletion protocol (44). Rab5 immunocomplexes were bound to protein A-Sepharose beads, as shown in Fig. 5A. Approximately 90% depletion of cytosolic rab5 was achieved by this protocol (lanes c and d of insert on Fig. 5A correspond to rab5 depleted cytosol and [rab5-anti-rab5] immunocomplexes eluted from the protein A beads, respectively). A control antibody was also included in the assay (lanes a and b, corresponding to cytosol depletion and immunocomplex elution from the protein A beads with the mAb HDP-1 antibody, respectively).
When the fusion reactions were carried out with rab5 depleted cytosol, phagosome-endosome fusion of phagosomes containing dead LM and live LM hlyϪ was substantially reduced (Fig. 5, filled circles in panels A and B, respectively). Phagosome-endosome fusion with live LM hlyϪ loaded phagosomes FIG. 4. Cryosection immunogold staining of rab5 in J774E clone cells infected with dead or live LM hly؊ . Cells were incubated for 6 min with BSA-gold (10 nm). Macrophages were then infected with dead (panels A and C) or live LM hlyϪ (panels B and D) for 10 min. Cells were fixed, and cryosections were stained for rab5 as described under "Materials and Methods." In these representative micrographs, the phagosomes containing live LM hlyϪ (B and D) stained for rab5 as shown by immunogold localization (arrows). Fusion with endosomes was detected by the presence of BSA-gold (large arrowheads). Phagosomes containing dead LM (A and C) lacked detectable amounts of rab5 immunogold. Again, fusion with endosomes was detected with BSA-gold. Rab5 staining was detected in other vesicles near the phagosomes (panel A, arrow). Phagosomal and endosomal membranes are indicated with small arrowheads. All micrographs ϫ 22,200. was more severely retarded by rab5 depletion than with dead LM containing phagosomes. Addition of exogenous prenylated GST-rab5 fusion protein restored the activity of this rab5 depleted cytosol in both assays (filled squares), suggesting that the effect of immunodepletion on fusion was totally dependent on rab5. In some experiments the restorative action of rab5 was blocked by the addition of rab5 antibody (data not shown).
GDI Extraction of Rab5 Bound to Phagosomal Membrane: Effects on Phagosome-Endosome Fusion-GDI is known to bind to GDP forms of rab proteins putatively blocking GDP/GTP exchange and causing an accumulation of rab proteins in the cytosol in an inactive form (51,52). To further analyze the role of cytosolic rab5 in mediating phagosome-endosome fusion, we added purified GDI to the assays. This protocol has been pre-viously shown (44) to inhibit transport in reactions that require recruitment of prenylated rab proteins from cytosolic GDI complexes. As shown in Fig. 5C, fusion was dramatically blocked by purified GDI (160 g/ml) in both assays, indicating that cytosolic rab5 is recruited to the membranes. When different amounts of GDI were included in the assay, concentrations as low as 16 g/ml blocked more than 50% of the fusion (Fig. 5D). The role of membrane bound rab5 on phagosome-endosome fusion was studied by treating phagosomal membranes with GDI which removes GDP-rab5 (43,44,51). One reason to extract rab5 from the membranes was to determine whether membrane bound rab5 provided some residual stimulus to fusion since phagosomes containing dead LM continue to fuse at a slow rate after depletion of cytosolic rab5 (Fig. 5A). As shown in Fig. 6A, GDI treatment extracts GDP-rab5 from the phagosomal membranes. As a control, rab5 was locked in the GTP-bound form by the addition of GTP␥S. Under these conditions, rab5 was not removed from the membranes. Fusion was readily measured after rab5 removal from the membranes in the presence of normal cytosol. As indicated above and confirmed in Fig. 6B, live LM hlyϪ phagosome-endosome fusion was completely blocked in the presence of rab5 depleted cytosol whereas dead LM phagosome-endosome fusion was still measurable. To further confirm the action of rab5 in supporting phagosome-endosome fusion, two different prenylated GST-rab5 fusion proteins were added to the in vitro phagosomeendosome fusion assays, the active mutant rab5:Q79L and the dominant negative mutant rab5:S34N. As shown in Fig. 6C, fusion was markedly activated by exogenous GST-rab5:Q79L, a mutant defective in GTP hydrolysis and inhibited by GST-rab5: S34N, a dominant negative mutant which is unable to exchange GDP for GTP. The specificity of rab5:Q79L activation of phagosome-endosome fusion was checked by including mAb 4F11 anti-rab5 in the fusion reaction. mAb 4F11 produced a dramatic reversal of the rab5 effect (from 0.88 unit of relative fusion to 0.10 unit in the presence of mAb 4F11 anti-rab5). The solubility and purification of the rab5 fusion proteins (described under "Materials and Methods") was apparently assured by the presence of REP-1, the rab escort protein that copurifies in our rab5 preparations (Fig. 6D, lane 1, corresponds to purified rab5:Q79L and lane 2 to rab5:S34N).
Differences between Dead and Live LM hlyϪ in Phagosome-Endosome Fusion-Analysis of the energy requirements of both assays (Table II) indicate that fusion of dead LM phagosomes with endosomes was sensitive to ATP depletion, whereas fusion with the live LM hlyϪ phagosomes was not. The live LM hlyϪ fusion assay was also insensitive to deoxyglucose treatment. Another curious finding is that live LM hlyϪ phagosomeendosome fusion assay was insensitive to NEM and to anti-NSF antibodies whereas dead LM phagosome-endosome fusions were highly sensitive. Mild washing of phagosomal membranes with salt (0.5 M KCl) restored sensitivity to ATP depletion and also restored the NEM and anti-NSF (data not shown) sensitivity. Addition of cytosol in which endogenous NSF was inactivated by prior incubation at 37°C was unable to restore fusion in salt-washed vesicles. However, addition of NSF in the presence of inactivated cytosol restored fusion with both preparations. These findings indicate that, with respect to fusion, salt-washed live LM hlyϪ phagosomes behave like dead LM phagosomes and phagosomes containing S. aureus particles (39). Therefore, we looked for NSF in both dead and live LM hlyϪ phagosomes by Western blotting. As shown in Fig. 7A, NSF appeared to be enriched on phagosomal membranes containing live LM hlyϪ , whereas the levels present on dead LM phagosomal membranes was extremely low, often undectable. Mild salt treatment of these phagosomes (0.5 M KCl) removed NSF from the phagosomes containing live LM hlyϪ , but only slightly changed levels in dead LM phagosomes. Analysis of rab5 in the same salt-washed phagosome membranes indicated no effect of salt treatment on rab5 levels. Thus, recovery of almost total fusion activity (around 80%) with purified NSF (His 6 -NSFmyc) on these salt-washed phagosomes points to NSF as the NEM and ATP-sensitive factor removed by this procedure. Interestingly, when NSF was analyzed in those phagosomal membranes treated with GDI, a surprising finding was obtained. The amount of NSF on dead LM phagosomal membranes was not affected by the absence of rab5, whereas NSF associated with live LM phagosomal membranes was substantially reduced by prior GDI treatment (Fig. 7B). These data FIG. 7. Relationship between NSF and rab5 present on the membranes of dead and live Lm hly؊ phagosomes. Panel A shows the effect of KCl washing on NSF and rab5 present on the phagosomal membranes. Dead LM-or live LM hlyϪ -containing phagosomes were incubated (ϩ) or not incubated (Ϫ) with 0.5 M KCl as described in Table  II. Phagosomes were solubilized in SDS buffer, and 50 g of protein were loaded per lane. Following transfer, nitrocellulose membranes were incubated with monoclonal anti-NSF (mAb 6E6) (1:500) or with mAb 4F11 anti-rab5 (1:5,000), followed by a HRP-conjugated goat antimouse (1:10,000) antibody and detected by ECL. Panel B shows a Western blot of NSF revealing the effect of GDI treatment of phagosomal membrane associated NSF. Purified phagosomes were treated with GDI as in Fig. 6A. The phagosomes were sedimented, resuspended in SDS buffer, and following SDS-PAGE, transferred to nitrocellulose. NSF was detected with mAb 6E6 (1:500) followed by incubation with a HRP-conjugated goat anti-mouse (1:10,000) antibody using ECL.  reflect a relationship between rab5 and NSF present on phagosomal membranes. Removal of rab5 from the membranes affects the membrane-association of NSF. However, removal of NSF from the membranes does not appear to affect the membrane association of rab5. These data suggest that rab5 regulates the binding of NSF to the phagosomal membranes. DISCUSSION Upon internalization by phagocytes, L. monocytogenes phagosomes have been shown to fuse with lysosomes (10, 11); however, neither the fusion with endosomes nor its regulation have been investigated. Work from our laboratory has shown that early phagosome-endosome fusion events can be reconstituted in vitro (12,39). Here we have developed a similar assay using biotinylated-bacteria and HRP-conjugated with avidin. Binding of the endosome probe (HRP-avidin) to the biotinylated bacteria (contained within phagosomes) was used as a measure of phagosome-endosome fusion.
Our goal was to address the question of whether live microorganisms regulate or modulate phagosome-endosome fusion events. In these studies, we have taken advantage of LM hlyϪ , a mutant deficient in hemolysin which remains inside the phagolysosome. Thus, both dead and live organisms used in this study are ultimately degraded by the host cell. An extensive analysis of the parameters regulating in vitro fusion between phagosomes containing dead LM or live LM hlyϪ with endosomes revealed a dependence on cytosolic proteins, salts (i.e. KCl), and membrane-bound proteins irrespective of whether live or dead LM were used. GTP␥S stimulated fusion in both assays implicating multiple GTPases, as previously reported for endocytosis (16,31,53) and phagocytosis (39,54).
GTP-binding proteins such as rab5 and rab7 have been localized to phagosomal membranes (21). Using specific antibodies, our studies indicate that rab5 regulates early phagosomeendosome fusion events. Rab7, while present, appears not to play a role in these events. The inhibitory effect of anti-rab5 antibody was dependent on the concentration of cytosol and antibody. Phagosome-endosome fusion with dead LM-loaded phagosomes was consistently less sensitive to antibody inhibition than fusion with the live LM hlyϪ phagosomes.
To explore more fully the effects of the antibody, we prepared purified dead LM or live LM hlyϪ phagosomes and checked for the presence of rab5. Both preparations contained rab5, but rab5 was enriched on live LM hlyϪ phagosomal membranes. Interestingly, rab7 was found only on live LM hlyϪ phagosomes. However, since specific antibodies were without effect rab7 appears to play no role in the phagosome-endosome fusion. A faster fusion rate was observed with live LM hlyϪ -loaded phagosomes compared with dead LM phagosomes, although the maximum amount of fusion observed was similar. Substantial phagosome-endosome fusion occurred within 5-10 min when live bacteria loaded phagosomes were used while the phagosomes containing dead bacteria required 25-30 min to obtain a similar relative fusion. These differential rates in the fusion kinetics cannot be explained by (i) differences in uptake, (ii) by differential biotinylation of the bacterial proteins, or (iii) by different numbers of bacteria per phagosome, since these parameters were very similar. To further explore the role of rab5, we depleted rab5 from the cytosol (44). Rab5 depleted cytosol was unable to support fusion of live LM hlyϪ -loaded phagosomes. Dead LM-loaded phagosomes fused poorly after rab5 depletion. Total recovery of fusion was achieved by adding GST-rab5 to the rab5-depleted cytosol. Another approach utilized GDI. GDI binds rab proteins in the GDP state and thereby extracts them from membranes (35,44,52). Addition of GDI blocked phagosome fusion with both preparations, although more inhibition was observed with the live LM hlyϪ -loaded pha-gosomes. Consistent with earlier results, addition of rab5 in the GTP-bound form was active in promoting fusion, and GTP hydrolysis was not required. Moreover, the dominant negative mutant of rab5 (rab5:S34N) inhibited both phagosome-endosome fusion assays. These findings indicate that rab5 is required for phagosome-endosome fusion, but that the live preparation is consistently more sensitive to the absence of rab5. The remaining fusion found in dead LM-loaded phagosomes in the absence of cytosolic rab5 could be due the presence of membrane-bound rab5. To resolve this point, we removed membrane-bound rab5 from dead LM phagosomes by GDI treatment. Even in the absence of detectable membrane bound rab5 and rab5-depleted cytosol, dead LM phagosomes were still marginally active in the fusion assay while live LM hlyϪ phagosomes were completely inactive. The mechanism of rab5-independent phagosome-endosome fusion with dead LM phagosomes remains to be explored.
The energy requirements for the two phagosome-endosome fusion assays are quite different. Similar to earlier results using S. aureus (39,54), phagosome-endosome fusion with the dead LM preparation was sensitive to ATP depletion. However, fusion of endosomes with phagosomes loaded with live LM hlyϪ was insensitive to ATP depletion. It is possible that the source of ATP in the two fusion assays might be different, e.g. the live organism may generate ATP. Alternatively, ATP-sensitive factors indispensable for docking and fusion may have already carried out their function or the need for certain components (e.g. NSF) may have been bypassed. We favor the latter possibility supported by the following findings: (i) a lack of sensitivity to NEM and anti-NSF antibodies, (ii) presence of NSF on phagosomes containing live LM hlyϪ and its virtual absence on dead LM phagosomes, (iii) a mild salt treatment of live LM hlyϪ phagosomes removes NSF and renders the fusion events sensitive to ATP and to NEM treatment, and (iv) fusion with salt-washed vesicles was restored by the addition of recombinant NSF to the assay in the presence of heat-inactivated cytosol. This is consistent with the findings of Rodriguez et al. (55), who found that fusion in salt-washed endosomes was fully recovered by NSF. When the relationship between membranebound NSF and rab5 was examined, differences between the live and dead LM preparations emerged. NSF was removed from both preparations with a mild salt wash, a procedure that did not affect rab5 associated with the membranes. However, following removal of rab5 with GDI, binding of NSF to live LM phagosomal membranes was lost, while the low levels of NSF associated with dead LM preparations were unaffected. These data suggest that the enhanced binding of NSF to live LM phagosomal membranes is coupled to rab5 or some other unknown rab. Why does NSF accumulate on the membranes of LM hlyϪ phagosomes? It is possible that the point in the temporal sequence of events where NSF acts has already been passed, and NSF is no longer needed. Similar conclusions have been drawn from experiments with Golgi membranes using an in vitro transport assay (56). It is also possible that other fusion factors are active here as suggested by Simons and colleagues (57), and the accumulation of NSF is secondary and perhaps, as indicated below, simply a consequence of rab5 accumulation on the LM hlyϪ membranes. However, at this point it is not possible to delineate the role played by NSF in live LM phagosomeendosome fusion.
What is the mechanism by which rab5 is recruited to the phagosomal membrane and why are phagosomes containing live LM hlyϪ so much more active in rab5 and NSF recruitment? The mechanism may be related to the guanine nucleotide exchange rate and the nucleotide status of rab5. The nucleotide status of rab5 would affect its membrane accumulation, since hydrolysis of GTP triggers release of rab5 to the cytosol and subsequent binding to GDI. Accordingly, the live organism may inhibit a rab5 GAP activity, resulting in an accumulation of rab5:GTP on the membrane. Alternatively, rab5 may hydrolyze GTP and remain on the membrane as a complex with other downstream proteins which in turn may lock NSF on the membranes. What could be the reason why LM drives binding of rab5 and NSF onto the membranes? First, it is clear from our results that cytosolic rab5 is required for fusion of live and dead phagosomes. Thus, rab5, and possibly NSF, accumulated on the membranes may not be functional. A more reasonable possibility is that the presence of rab5 and NSF on the membranes prevents maturation of the phagosome. Perhaps, the sequential binding and dissociation of rab5 precedes in an obligatory way, the binding of other factors that are necessary for fusion with elements of the lysosomal compartment. Thus, the live organism may extend its lifetime in the endosomal compartment from whence it can translocate into the cytosolic compartment. It might be speculated that the microorganism has a requirement to access the endocytic compartment of the host cell, thereby providing maximal surface area available in a relatively nonlethal intracellular compartment from which translocation to the cytosol can occur. LM phagosomes containing live bacteria may be more immature than their counterparts containing dead bacteria. Thus fusion would be up-regulated because maturation was delayed in the phagosomes containing live Lm hlyϪ . That would be in agreement with the faster kinetics of their fusion with endosomes and with the more avid binding of rab5. Phagosomes containing dead LM may have matured faster; they have less avidity for rab5 and fuse more slowly with endosomes. In summary, our results with rab5 represent the first documentation of an active participation of a live microorganism on membrane trafficking along the phagocytic pathway. The bacteria appear to modulate the function of rab5 putatively by affecting nucleotide exchange. Clearly more work is needed to delineate the biochemical mechanisms involved.