Requirement for N-Ethylmaleimide-sensitive Factor Activity at Different Stages of Bacterial Invasion and Phagocytosis*

Bacterial invasion, like the process of phagocytosis, involves extensive and localized protrusion of the host cell plasma membrane. To examine the molecular mechanisms of the membrane remodeling that accompanies bacterial invasion,soluble NSF attachment proteinreceptor (SNARE)-mediated membrane traffic was studied in cultured cells during infection by Salmonella typhimurium. A green fluorescent protein-tagged chimera of VAMP3, a SNARE characteristic of recycling endosomes, was found to accumulate at sites of Salmonella invasion. To analyze the possible role of SNARE-mediated membrane traffic in bacterial infection, invasion was measured in cells expressing a dominant-negative form of N -ethylmaleimide-sensitivefactor (NSF), an essential regulator of membrane fusion. Inhibition of NSF activity did not affect cellular invasion by S. typhimurium nor the associated membrane remodeling. By contrast, Fcγ receptor-mediated phagocytosis was greatly reduced in the presence of the mutant NSF. Most important, dominant-negative NSF significantly impaired the fusion of Salmonella-containing vacuoles with endomembranes. These observations indicate that the membrane protrusions elicited by Salmonella invasion, unlike those involved in phagocytosis, occur via an NSF-independent mechanism, whereas maturation of Salmonella-containing vacuoles is NSF-dependent.

Invasive bacterial species gain access to intracellular compartments, where they survive and replicate by subverting host cell function. Although it is not yet clear how a variety of bacterial species induce the formation of the vacuoles that engulf them, it has been established that the process of inva-sion has much in common with that of phagocytosis (1)(2)(3). Both processes typically involve membrane remodeling events, namely pseudopodial extension during phagocytosis and ruffling during bacterial invasion, and both phenomena are associated with changes in the actin cytoskeleton of the host cell (1). As well, like phagosomes, bacteria-containing vacuoles can undergo a maturation process (2,4). However, whereas phagocytosis and invasion are superficially similar, it is not clear whether the molecular mechanisms of internalization are shared.
Current models of phagocytosis suggest that phagosomes are generated by receptor-mediated "zippering" of the host cell plasma membrane against the surface of opsonized particles, accompanied by the active extension of pseudopodia (5)(6)(7)(8). A source of intracellular membrane is required for the membrane remodeling that allows for pseudopod formation, and recent evidence suggests that specific endosomal compartments are involved in particle engulfment (9 -11). It is now apparent that the soluble NSF attachment protein receptor (SNARE) 1 protein VAMP3, an integral component of the recycling endosome, plays an important role in phagocytosis, possibly participating in localized fusion events at the site of phagosome formation and thus driving pseudopodial extension (12).
It is not known if endosomal compartments have a central role in the process of bacterial invasion, but the extensive membrane remodeling that accompanies cellular infection by Salmonella typhimurium suggests that invasion by this bacterial pathogen may mechanistically parallel phagocytosis. In the current study, we examined the role of SNARE-mediated membrane fusion in S. typhimurium invasion. Because of the established similarities with phagocytosis, the role of VAMP3 in invasion was analyzed first. However, because the exact intracellular compartments and the repertoire of SNARE proteins that may be involved in bacterial invasion are not known, the role of membrane fusion was also investigated through the general inhibition of SNARE-mediated membrane traffic, specifically by blocking the activity of N-ethylmaleimide-sensitive factor (NSF). NSF is an ATPase/chaperone that is responsible for the disassembly of SNARE complexes and is therefore an essential regulator of intracellular membrane fusion events (13). We generated a mutant form of NSF (E329Q-NSF) that is deficient in ATP hydrolysis and therefore inhibits the function of endogenous NSF, resulting in a dominant-negative effect (14,15). We found that acute expression of E329Q-NSF inhibits phagocytosis in a macrophage cell line. By contrast, invasion of S. typhimurium into epithelial cells was not impaired by the dominant-negative NSF; however, maturation of Salmonellacontaining vacuoles (SCVs) was inhibited.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Sheep red blood cells (SRBC), goat anti-SRBC IgG, and N-ethylmaleimide (NEM) were purchased from ICN. Human IgG was purchased from Baxter Healthcare Corp. Dulbecco's modified Eagle's medium (DMEM) and fetal bovine serum (FBS) were obtained from Wisent (Montreal, Quebec, Canada). Brefeldin A and G418 were obtained from Calbiochem and Fugene 6 from Roche Molecular Biochemicals. Tetramethylrhodamine isothiocyanate-transferrin (Tfn) and tetramethylrhodamine-phalloidin were from Molecular Probes. Restriction enzymes were obtained from New England Biolabs. Latex beads (3 m diameter), ATP, and p-nitrophenyl phosphate (pNPP) and all other reagents were purchased from Sigma.
Polyclonal antibody to ␣-mannosidase II was a generous gift of Dr. Marilyn Farquhar (University of California, San Diego). The phycoerythrin (PE)-labeled antibody to mouse Fc␥II/III receptors (CD16/ CD32) was purchased from PharMingen. Rabbit polyclonal antiserum to S. typhimurium lipopolysaccharide was purchased from Difco. Monoclonal antibody to human LAMP-1 was obtained from the Developmental Studies Hybridoma Bank. Fluorescein isothiocyanate (FITC)-and tetramethylrhodamine isothiocyanate (TRITC)-conjugated donkey antimouse IgG and donkey anti-rabbit IgG were purchased from Jackson ImmunoResearch Laboratories. Polyclonal antibody to GFP, Alexa 350-, and Alexa 594-conjugated donkey anti-rabbit antibodies were purchased from Molecular Probes. Horseradish peroxidase-conjugated goat anti-rabbit antibody was obtained from Pierce.
To produce an antiserum to NSF, His 6 -tagged recombinant NSF protein was expressed in bacteria and purified according to Whiteheart et al. (15). The protein was then mixed with Freund's adjuvant and used to immunize rabbits by following standard procedures. The NSF antibody was affinity-purified from rabbit serum by binding to and eluting from an NSF-coupled Affi-Gel 10 column (Bio-Rad).
cDNA Constructs and Mutagenesis-pEGFP-N1 was purchased from CLONTECH Laboratories Inc. (Palo Alto, CA). The acylation motif of Lyn kinase fused to GFP (PM-GFP) was a kind gift of Dr. Tobias Meyer (Stanford University, Palo Alto, CA). The VAMP3-GFP chimera is described elsewhere (12). Wild-type (WT) NSF in pQE-9 (pQE-9-NSF; Whiteheart et al. (15)) was a generous gift of Dr. S. Whiteheart (University of Kentucky). WT NSF was subcloned into pcDNA3.1 (Invitrogen, Carlsbad, CA) using a multistep strategy. First, the 3Ј-coding region was excised from pQE-9-NSF using HindIII and ligated into pcDNA3.1 (pcDNA3.1-3ЈNSF). Second, a fragment encompassing most of the coding region of NSF, including the ATG start codon, was excised from pQE-9-NSF using BamHI and NheI and ligated independently into pcDNA3.1 (pcDNA3.1-5ЈNSF). Third, pcDNA3.1-5ЈNSF was digested with XbaI and NheI, and this fragment was then ligated into a pcDNA3.1-3ЈNSF backbone that had been digested with NheI, yielding a complete and contiguous coding sequence of NSF within pcDNA3.1 (pcDNA3.1-NSF). The fidelity of this, and all NSF constructs employed in this study, was confirmed by DNA sequence analysis.
Generation of an NSF gene bearing a point mutation in the ATPase domain of the D1 region of NSF was accomplished by polymerase chain reaction-mediated, plasmid-based, site-specific mutagenesis. To replace the glutamate at position 329 of NSF with glutamine (Gln substituted at Glu 329 ), pQE-9-NSF was used as the target vector. Two complementary polymerase chain reaction primers containing the desired point mutation in NSF were made and used along with the QuikChange Site-directed Mutagenesis Kit from Stratagene Cloning Systems. Transformed bacterial clones were screened by DNA sequencing to identify clones containing the mutated form of the NSF gene in pQE9. This clone of mutant NSF cDNA was designated E329Q NSF. To introduce the E329Q mutation into NSF in pcDNA3.1, BspEI/BstEII digestion of E329Q NSF in pQE-9 was used to generate a cDNA cassette, containing the mutated region of NSF, that was subcloned into like digested WT NSF in pcDNA3.1.
ATPase Assay-NSF constructs in pQE9 were expressed in Escherichia coli, and the His 6 -tagged proteins were purified as described above. Purified wild-type and E329Q NSF proteins were used for in vitro ATPase assays using pNPP as a chromogenic ATP analogue, based on the method of Mignaco et al. (16). Briefly, 5 mM ATP and 6 mM pNPP (final concentrations) were freshly added to a buffer of 25 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.5 mM dithiothreitol, 0.65 mM ␤-mercaptoethanol, 1 mM MgCl 2 , and 10% glycerol. With this was mixed 20 -25 g/ml purified NSF protein, and the samples were incubated at 37°C for 1-2 h. Reactions were quenched with NaOH, to a final concentration of 60 mM, and absorbance at 425 nm was read on an Hitachi U-2000 spectrophotometer. Activities were calculated by determining quantities of pNPP produced based on a molar extinction coefficient of 1.1 ϫ 10 4 M Ϫ1 cm Ϫ1 (16). For controls with NEM, protein samples were incubated in the presence of 2.5 mM NEM for 15 min on ice prior to addition to the assay.
Cell Culture and Transfections-RAW 264.7 and COS-1 cells were obtained from the American Tissue Culture Collection, and Fc␥RIIAtransfected COS-1 cells (COS-2A) have been described previously (8,(17)(18)(19). All cells were cultured in DMEM supplemented with 10% FBS and incubated at 37°C and 5% CO 2 . COS-2A were periodically maintained in the presence of 200 g/ml G418. For cell-based assays, cells were grown on 25-mm glass coverslips to 50 -75% confluence. Transfections of exponentially growing populations of cells were done with the Fugene 6 reagent according to the manufacturer's instructions. Cells were transfected with either WT NSF or E329Q NSF alone or cotransfected with EGFP and then incubated, depending upon the assay, for periods of 5, 8, or 16 h prior to analysis. For cotransfections, a mass ratio of EGFP:NSF cDNA of 10:1 was used, and immunofluorescent staining of the cells was performed to confirm that greater than 95% of EGFP-positive cells were ectopically expressing NSF.
Transferrin Uptake and Indirect Immunofluorescence-To study transferrin endocytosis, cells were serum-starved for 1 h in DMEM and subsequently incubated with 50 g/ml labeled Tfn for 30 -45 min, after which excess Tfn was washed away. For indirect immunofluorescence, cells were fixed with 4% paraformaldehyde/PBS at room temperature for 20 -30 min. If permeabilization was necessary for staining of intracellular markers, fixed cells were bathed in 0.1% Triton X-100/PBS for 10 min. Cells were then blocked with 5% skim milk powder/PBS for at least 1 h. For staining, cells were incubated in the presence of primary antibody, at 37°C, for 1 1 ⁄2 h, washed, incubated with secondary antibody for 1 h, washed, and mounted using DAKO mounting medium (DAKO Corp.).
Western Blotting-For analysis of VAMP3-GFP expression and cleavage by TeTx, transfected cells were lysed in PBS, pH 7.4, containing 1% Triton X-100, 0.1% SDS, 0.5% deoxycholate, and protease inhibitors. Lysates were cleared by centrifugation at 10,000 ϫ g for 10 min, and the protein concentrations of the extracts were determined using the Bio-Rad protein assay. Samples were boiled in sample buffer, and the proteins were separated by SDS-polyacrylamide (10%) gel electrophoresis under nonreducing conditions. Proteins were electrophoretically transferred to polyvinylidene difluoride membranes (Immobilon P, Millipore), and the membranes were then blocked with 5% milk powder in TBS-T (20 mM Tris-HCl, 137 mM NaCl, 0.1% Tween 20, pH 7.6). The membranes were probed with a polyclonal antibody to GFP, washed, and incubated with a horseradish peroxidase-conjugated goat antirabbit secondary antibody. After washing, the labeled proteins were detected using an enhanced chemiluminescence (Amersham Pharmacia Biotech).
Flow Cytometry and Quantitative Image Analysis-To analyze cell surface Fc receptor expression, transfected RAW cells were incubated with PE-labeled anti-Fc receptor antibody for 1 h on ice. The cells were then fixed in 4% paraformaldehyde and analyzed using a Becton Dickinson FACStar Plus flow cytometer and CELLQUEST software (Becton Dickinson). The filter settings were 530/30 and 575/26 nm for GFP and PE, respectively.
Quantitation of NSF expression and transferrin labeling in RAW cells and of surface Fc receptor expression in COS-2A cells was done by examining labeled cells with a Leica DM IRB inverted microscope. Cell Images were captured using WinView 32 software (Princeton Instruments) and subsequently analyzed using NIH Image software. Individual cell boundaries were defined manually, and fluorescence intensity was integrated over calculated cell areas.
Phagocytosis and Particle-binding Assays-For phagocytosis in RAW 264.7 cells, 3-m latex beads were opsonized with human IgG, 1 h room temperature, and washed with PBS to remove unbound IgG. Opsonized beads were then added to RAW cells plated in DMEM containing 10% FBS and incubated for 30 min at 37°C. Unbound beads were then washed away, and internalized beads were quantified under Nomarski optics. For phagocytosis in COS-2A cells, SRBCs were opsonized with goat anti-SRBC IgG for 1 h at 37°C. After washing with PBS, opsonized SRBCs were added to COS-2A cells plated in complete medium and incubated for 30 min or 1 h at 37°C. Extracellular SRBCs were removed by hypotonic lysis (water, 30 s), and internalized SRBCs were quantified as above. When assessing phagocytosis in cells transfected with EGFP, it was possible to quantify internalized particles using fluorescence microscopy due to the exclusion of GFP from the interior of the phagosomes. In this way, cells could be fixed after phagocytosis and counter-stained, for example with NSF antibody in experiments where cells had been cotransfected with NSF cDNA.
To assess particle binding efficiency, transfected RAW 264.7 cells were incubated with opsonized SRBCs for 30 min at 4°C. Nonadherent SRBCs were washed away with PBS, and the number of bound SRBCs/ cell was quantified under light microscopy.
Bacterial Infection-The S. typhimurium strain 1344 was used and has been described previously (20). Overnight bacterial cultures were diluted 1:30 into fresh Luria-Burtani broth and incubated at 37°C, shaking for 3 h. Bacteria were pelleted by centrifugation at 10,000 rpm for 2 min and then resuspended in PBS. The bacteria were diluted in Earle's buffered salt solution (EBSS), pH 7.4, and added to COS-1 cells that had been plated on coverslips at a density of ϳ50 bacteria per cell. Infection was carried out at 37°C for 15 min. Excess bacteria were washed away with PBS, and the cell monolayers were then incubated in growth medium containing 100 g/ml gentamicin for 15 min at 37°C. The cells were processed for immunofluorescence by fixing in 2.5% paraformaldehyde/PBS for 10 min, washing, and blocking with 1% BSA/2% donkey serum/PBS. Samples were stained for external bacteria by incubating with rabbit anti-S. typhimurium LPS antibody followed by Alexa 594-conjugated donkey anti-rabbit antibody. Subsequent to this, the samples were again incubated in blocking buffer, this time containing 0.1% saponin. The samples were then stained for total bacteria (external and internalized) with the same primary antibody followed by Alexa 350-conjugated donkey anti-rabbit antibody. For examination of vacuole maturation, fixed and permeabilized samples were stained for total bacteria with the anti-LPS antibody and Alexa 594-conjugated donkey anti-rabbit antibody and counterstained for LAMP-1 with the monoclonal anti-human LAMP-1 antibody and FITCconjugated goat anti-mouse IgG.

VAMP3-GFP Localizes to Sites of Salmonella Invasion-Pre-
vious studies from our laboratory demonstrated the focal exocytosis of VAMP3-containing vesicles at sites of particle internalization during phagocytosis (12). Here, we examined the possibility that VAMP3-containing vesicles may play a role in the membrane remodeling that accompanies cellular invasion by S. typhimurium. Cells expressing a chimeric VAMP3-GFP construct (12) were subjected to invasion by S. typhimurium as described under "Experimental Procedures." Shortly after infection (6 -8 min), the distribution of VAMP3-GFP was monitored by fluorescence microscopy, whereas the localization of bacteria was detected by indirect immunofluorescence using an antibody to S. typhimurium lipopolysaccharide (anti-LPS). In resting COS-1 cells VAMP3-GFP was present in vesicles that were observable throughout the cytoplasm but accumulated particularly in a juxta-nuclear cluster, typical of recycling endosomes ( Fig. 1a), consistent with previously published observations (21). Comparatively little VAMP3 was present on the cell surface in otherwise untreated cells. In cells undergoing invasion by S. typhimurium, in contrast, VAMP3-GFP was seen to accumulate at or near the plasmalemma at sites of bacterial attachment (Fig. 1, b and c).
Accumulation of VAMP3-GFP occurred both at sites where Salmonella bacteria were external and at sites where the bacteria had been internalized. However, although VAMP3-GFP accumulated at these sites, it was not clear if this represented only intracellular movement of the fluorescent protein or mobilization and exocytosis of this SNARE at sites of invasion. To examine this possibility, we tested for the appearance of the COOH terminus of VAMP3-GFP (bearing the GFP moiety) on the cell surface by staining the cells with an antibody against GFP. Following invasion, COS-1 cells transfected with VAMP3-GFP were gently fixed (using 2.5% paraformaldehyde) but not permeabilized. Staining of the cells then revealed external GFP (Fig. 1d) at sites of VAMP3-GFP accumulation (Fig. 1e) that corresponded with sites of attachment of external bacteria (Fig.  1f). These observations indicate that accumulation of VAMP3-GFP can occur prior to internalization of S. typhimurium and that the accumulation of this chimera is accompanied by its insertion into the plasma membrane.
Tetanus Toxin Blocks Recruitment of VAMP3-GFP to Sites of Salmonella Internalization but Does Not Inhibit Invasion-The localization and plasmalemmal insertion of VAMP3-GFP at sites of bacterial invasion suggest that membrane fusion events mediated by this SNARE may contribute to the membrane remodeling that is involved in bacterial invasion. To test this possibility, a cDNA encoding the light chain of the tetanus toxin (TeTx) was cotransfected into COS-1 cells along with VAMP3-GFP. To confirm that cotransfection of these two vectors resulted in significant cleavage of VAMP3-GFP, transfected cell populations were harvested and examined by Western blotting with an anti-GFP antibody. To study the effect of TeTx on the mobilization of VAMP3-GFP that accompanies Salmonella invasion, COS-1 cells were transfected with both VAMP3-GFP and TeTx prior to invasion assays with S. typhimurium. As depicted in Fig. 3b, the expression of TeTx altered the distribution of VAMP3-GFP within COS-1 cells. Tetanus toxin is known to cleave the cytoplasmic portion of VAMP3, and the observed redistribution of this SNARE is indicative of the accumulation of the transmembrane/lumenal domain-GFP moiety of the chimera in the Golgi apparatus (22). Most important, in COS cells coexpressing VAMP3-GFP and TeTx, accumulation of VAMP3-GFP at sites of S. typhimurium invasion was markedly reduced (Fig. 3, b  and c).
The observation that TeTx blocked the accumulation of VAMP3 near sites of Salmonella invasion prompted us to test whether TeTx would also block Salmonella invasion, in a manner analogous to its known effect on phagocytosis (10). To quantitate the effect of TeTx on invasion, transfected cells were exposed to S. typhimurium, and the fraction of cells harboring intracellular bacteria was estimated as described under "Experimental Procedures" (see also Fig. 3, d-f). Surprisingly, whereas expression of TeTx altered the basal distribution of VAMP3 and prevented its accumulation at sites of bacterial internalization, it had no discernible effect on the number of cells invaded by Salmonella (Fig. 4). As well, the number of bacteria that invaded individual COS-1 cells was not affected by the expression of tetanus toxin (data not shown). Furthermore, expressing EGFP, alone, or EGFP and TeTx in COS-1 cells had no effect on Salmonella invasion (not shown). These results suggest that, although VAMP3 accumulates at sites of invasion, accumulation and function of this protein are not required for S. typhimurium entry into the cells.
Characterization and Expression of Dominant-negative NSF-The finding that TeTx had no effect on Salmonella invasion contrasts with previous studies that demonstrated a role of VAMP proteins in phagocytosis (10). These results are not necessarily incompatible because, although VAMP3 may be required for phagocytosis, other SNARE proteins, possibly including TeTx-insensitive isoforms of VAMP, may be essential for S. typhimurium invasion. It is impractical and in some cases impossible to use specific toxins to test the function of all the SNARE proteins expressed in COS-1 cells that may be involved in Salmonella invasion. Instead, a general inhibitor of SNARE-mediated membrane traffic was sought. A logical target for this purpose was NSF, an ATPase/chaperone protein responsible for the disassembly of SNARE complexes and an essential regulator of membrane fusion events. To inhibit NSF activity, we introduced a point mutation in the Chinese hamster NSF gene which resulted in the substitution of glutamine for glutamic acid at residue 329 (E329Q) within the catalytic site of the D1 domain of NSF. E329Q-NSF has been described elsewhere and shown to form hexamers that bind ATP but do not hydrolyze it (14,15). When mixed with wild-type protein, E329Q-NSF forms mixed hexamers that also lack activity, leading to a dominant-negative effect. To verify the effectiveness of the mutation, purified, recombinant wild-type, and mutant NSFs were assayed for nucleotidase activity in vitro, as described under "Experimental Procedures." Wild-type NSF had an activity of 0.382 Ϯ 0.019 mol pNPP/mg/h (means Ϯ S.E.). Wild-type NSF, incubated with 2.5 mM NEM, had an activity of 0.05 Ϯ 0.023 mol of pNPP/mg/h, and the activity of E329Q-NSF was 0.142 Ϯ 0.011 mol of pNPP/mg/h. Consistent with the findings of others (15), the intrinsic activity of E329Q NSF was ϳ30% that of wild-type NSF. The remaining nucleotidase activity of E329Q-NSF is attributable to the second catalytic site in the D2 domain of the protein (15).
To characterize the function of E329Q-NSF in intact cells, we chose to express the mutant protein in a macrophage cell line (RAW 264.7). This provided us with the opportunity to compare NSF function in phagocytosis, a receptor-mediated event initiated by the macrophage, and in invasion, an opsonin receptorindependent process triggered by Salmonella. Wild-type and mutant NSF genes in pcDNA 3.1 were expressed in mouse macrophages by transient transfection along with EGFP. As illustrated in Fig. 5, b and d, the endogenous levels of NSF (see EGFP-negative cells) were barely detectable by immunofluorescence, but the ectopic expression of either wild-type (WT) or mutant NSF was clearly observable. Quantitation of the expression of E329Q-NSF (Fig. 6) revealed that 5 h after transfection the amount of the dominant-negative NSF was slightly increased over that of the endogenous protein. By 8 h posttransfection, the level of ectopic expression of NSF was ϳ3 times the endogenous NSF (Fig. 6).
We next assessed the biological effects of the dominantnegative NSF in vivo by monitoring transferrin uptake. Transferrin is internalized by a receptor-mediated, clathrin-depend-  ent process and cycles through endosomal compartments before returning to the cell surface. By incubating cells in the presence of rhodamine-labeled transferrin, we were able to determine that RAW cells overexpressing WT NSF internalized transferrin in a manner that was indistinguishable from untransfected cells (Fig. 5, e and f). In contrast, transferrin uptake was strikingly decreased in cells expressing E329Q-NSF (Fig. 5, g and h). A detailed quantitation of this effect is presented in Fig. 6, as a function of time after transfection. The level of transferrin uptake (as a percentage of control cells) was reduced by ϳ60% 5 h after transfection and by ϳ95% after 8 h. Expression of E329Q-NSF also inhibited the lysosomal accumulation of material normally internalized by fluid phase and altered the morphology of the Golgi complex in a manner resembling the effects of brefeldin A (not illustrated). Comparable levels of expression of WT-NSF were without effect on the uptake of transferrin (Fig. 6) or fluid phase markers or on Golgi morphology. These experiments indicate that ectopic expression of NSF to levels that can cause altered SNARE function can be attained shortly after transfection (within 8 h). Importantly, such brief periods of expression preserved cellular via-bility and had no discernible effect on several aspects of cell function. Specifically, the acidity of lysosomes, assessed by partition of fluorescent weak bases, was preserved, as was the electrical potential across the mitochondrial inner membrane, estimated using rhodamine 143 (not shown). Unless indicated otherwise, an 8-h transfection protocol was used thereafter.
Dominant-negative NSF Inhibits Phagocytosis in Macrophages-The phagocytic ability of RAW cells expressing either WT or E329Q-NSF, along with EGFP, was examined using IgG-opsonized latex beads. As seen in Fig. 7, a and b, RAW cells transfected with WT-NSF were capable of phagocytosis as efficiently as untransfected neighboring cells. Phagocytosis by cells transfected with WT-NSF was also similar to that in cells transfected with EGFP and pcDNA 3.1 vector alone (see quantitation in Fig. 7b). In contrast, RAW cells transfected with E329Q-NSF ingested significantly fewer beads (Fig. 7, a and b). The phagocytic indices for WT-NSF and E329Q-NSF transfectants were 205 Ϯ 26 and 66 Ϯ 11, respectively (Fig. 7b), equivalent to 68% inhibition. Similar results were obtained using IgG-opsonized red blood cells.
It was conceivable that inhibition of phagocytosis by E329Q-NSF resulted from impairment of constitutive secretion, causing depletion of proteins that turn over rapidly. This possibility was tested by exposing the cells to brefeldin A, a potent inhibitor of ARF GTPase activity and COP I function (23), for 3 h prior to phagocytosis assays. This time is comparable to the period of expression of inhibitory concentrations of E329Q-NSF (see Fig. 6). Incubation with brefeldin A for 3 h had no detectable effects on phagocytosis (Fig. 7b), despite inducing very rapid dispersal of the Golgi complex. Thus, the impairment of phagocytosis in RAW cells caused by dominant-negative NSF is not due to the inhibition of constitutive secretion.

Dominant-negative NSF Does Not Block Fc␥ Receptor Function-
To define further the nature of the phagocytic defect in RAW cells expressing mutant NSF, we examined the level of Fc receptor expression on the surface of transfected RAW cells. Flow cytometry of RAW cells expressing WT and E329Q-NSF revealed that transfected cells had levels of Fc␥II/III receptors on their surface that were similar to those of untransfected cells (Fig. 8a). The presence of receptors and their ability to engage opsonized particles was also assessed by particle binding assays using both latex beads and red blood cells. As seen in Fig. 7a and quantified in Fig. 8b, RAW cells expressing WT or E329Q-NSF bound opsonized particles with equal efficiency.
Actin cup formation also proceeded normally in cells transfected with E329Q-NSF. Actin remodeling is a well characterized post-receptor event that occurs early and transiently in the process of phagocytosis (24 -26). Staining of F-actin in RAW cells exposed to opsonized particles for 5 min showed that the reorganization of actin fibers in nascent phagosomal cups was indistinguishable in WT or mutant NSF transfectants (Fig. 8c). Collectively, these results indicate that impairment of phagocytosis by the dominant-negative NSF occurred at a step distal to Fc receptor activation and is independent of actin reorganization.

Dominant-negative NSF Inhibits VAMP3-GFP Recruitment to Sites of Salmonella Internalization but Does Not Block Salmonella-induced Membrane
Ruffling-Based on the inhibitory effect of dominant-negative NSF on phagocytosis reported above, we anticipated that inhibition of NSF activity would also have adverse effects on Salmonella invasion. To ensure that the dominant-negative NSF construct was having inhibitory effects in the cells used for invasion assays, COS-1 cells were cotransfected with EGFP and either WT-NSF or E329Q-NSF for 8 h, and transferrin uptake was determined. As in RAW cells, E329Q-NSF impaired the intracellular accumulation of transferrin in COS-1 cells (data not shown). Also, immunofluorescent staining of ␣-mannosidase II in cells expressing E329Q-NSF indicated that Golgi morphology had been altered by expression of the dominant-negative mutant (not illustrated). Ectopic expression of WT-NSF had no apparent effect in either of these assays.
We next cotransfected COS-1 cells with VAMP3-GFP and either WT-NSF or E329Q-NSF and performed Salmonella invasion assays as described above. As seen in Fig. 9b, coexpression of E329Q-NSF influenced the intracellular distribution of VAMP3-GFP (compare Fig. 9b with Fig. 1a) and prevented the localization of VAMP3-GFP to sites of Salmonella attachment (Fig. 9, b and c). Coexpression of WT-NSF did not affect the expression pattern of VAMP3-GFP nor did it influence its redistribution to Salmonella-associated ruffles (not shown).
To determine whether membrane ruffling had been blocked by expression of E329Q-NSF, we next expressed either the WT or mutant NSF together with a cDNA encoding a chimera of GFP and the acylation motif of the membrane-associated kinase Lyn (PM-GFP). This construct has been described elsewhere and has been demonstrated to label preferentially the plasma membrane, making it a useful marker of plasmalemmal redistribution (27). The basal pattern of PM-GFP is shown in Fig. 9d. This relatively homogeneous plasmalemmal distribution was not altered by coexpression of WT-NSF. PM-GFP was found to accumulate at sites of Salmonella invasion (not illustrated), in accordance with the reported formation of membrane ruffles. Coexpression with E329Q-NSF had subtle effects on the resting distribution of PM-GFP in COS-1 cells but, importantly, did not alter the redistribution of PM-GFP to sites of Salmonella invasion (Fig. 9, e and f). Together, the results presented in Fig. 9 indicate that dominant-negative NSF blocked the participation of SNARE-mediated membrane traffic in the membrane redistribution that accompanies Salmonella invasion but did not eliminate membrane ruffling.
NSF Activity Is Not Required for Invasion of COS-1 cells by S. typhimurium but Is Necessary for Particle Ingestion by Phagocytically Competent COS-1 Cells-The combined observations that dominant-negative NSF inhibited the phagocytosis of opsonized particles but did not block Salmonella-induced membrane ruffling prompted us to quantify Salmonella invasion in COS-1 cells expressing E329Q-NSF. COS-1 cells were cotransfected with EGFP and either WT or E329Q-NSF prior to assaying Salmonella invasion as described in Fig. 1. As seen in Fig. 10a, expression of either WT or dominant-negative NSF had no effect on the invasion of S. typhimurium into COS-1 cells. The percentage of cells invaded was not affected by expression of mutant NSF (Fig. 10a) nor was the number of Salmonella bacteria per invaded cell altered (not shown). Similar results were obtained when HeLa cells were used, when NSF constructs were coexpressed with VAMP3-GFP or PM-GFP, and when transient expression of E329Q-NSF was extended to 16-h periods.
The differential effects of E329Q-NSF on phagocytosis and bacterial invasion may reflect differences in the cell types used, rather than in the mechanisms underlying the two processes. To address this possibility, we conferred phagocytic capability to COS-1 cells by stable transfection of Fc␥ receptor IIA. These cells, designated COS-2A, have been characterized previously and shown to support phagocytosis in a manner that closely resembles the behavior of "professional" phagocytes such as macrophages (8,(17)(18)(19). COS-2A cells were transiently cotransfected with EGFP and NSF constructs for use in phagocytosis assays as described under "Experimental Procedures." Cells transfected with WT-NSF ingested opsonized latex beads efficiently, whereas their E329Q-transfected counterparts showed significantly lower phagocytic indices (Fig. 10b). Cells transfected with WT-NSF ingested particles as efficiently as untransfected or mock-transfected COS-2A (data not shown). As was the case for RAW cells, inhibition of phagocytosis was not attributable to changes in receptor expression. Quantification of surface Fc␥ receptor IIA expression on transfected COS-2A cells, using indirect immunofluorescence and NIH Image software, revealed no differences between WT and E329Q-NSF-transfected samples (Fig. 10c).
Dominant-negative NSF Impairs SCV Maturation-SCVs are known to undergo a maturation process resulting in the insertion of specific endomembrane components into the SCV membrane along with changes in vacuolar contents (2,28). The appearance of lysosomal-associated membrane protein 1 (LAMP-1) in the vacuolar membrane is a marker of such maturation. To examine the possible role of NSF in vacuolar maturation, we examined the distribution of LAMP-1 in infected cells using indirect immunofluorescence microscopy. Fig. 11 shows COS-1 cells cotransfected with EGFP and WT or E329Q-NSF that have been infected with S. typhimurium. The percentage of S. typhimurium-containing vacuoles that were positively stained for LAMP-1 was significantly reduced in cells expressing dominant-negative NSF (Fig. 11e). The expression of E329Q NSF had no effect on the level of expression of LAMP-1, as determined by Western blot, or on the basal distribution of this lysosomal marker (data not shown). It is im- portant to note that the inhibition of LAMP-1 acquisition is likely to be underestimated in these experiments because scoring was done on an all-or-none basis, not taking relative degrees of staining into account. Thus, while internalization of S. typhimurium into COS-1 cells was not impeded by inhibition of NSF function, expression of dominant-negative NSF in these cells did perturb the maturation of the resultant SCVs. DISCUSSION Our findings indicate that NSF-regulated membrane fusion is required to support Fc receptor-mediated phagocytosis but not bacterial invasion. Specifically, whereas expression of dominant-negative NSF inhibited particle engulfment both in macrophages and phagocytically competent fibroblastic cells, invasive bacterial internalization was not significantly altered. This does not rule out the possibility that other AAA proteins related to NSF, such as p97 (29), are active in response to Salmonella invasion, and this alternative is currently being investigated. Importantly, however, inactive NSF did impair the maturation of Salmonella-containing vacuoles at a stage after bacterial internalization was complete. These results suggest that membrane traffic has different roles during these two processes.
The finding that NSF function is required for optimal phagocytosis is consistent with an emerging body of evidence that suggests that regulated fusion of endomembranes with the plasmalemma is essential for phagocytosis (10, 11, 30 -32). In COS-1 cells transfected with EGFP alone or with EGFP together with WT or dominant-negative NSF constructs, the percentage of SCVs that were labeled with LAMP-1 antibody was determined. Means Ϯ S.E. from three independent experiments representing Ͼ100 cells for each condition are shown (*, p Ͻ 0.0025).
Recent work from our laboratory indicates that VAMP3-containing vesicles participate in the focal exocytosis that occurs at sites of particle engulfment during phagocytosis (12), and together with the results presented here, these data are compatible with a model wherein NSF-regulated, SNARE-mediated membrane fusion drives exocytic events that are essential for the completion of particle engulfment. This requirement is independent of particle size, as phagocytosis of IgG-opsinized, invasion-null Salmonella bacteria (ϳ1 m long) was also blocked by expression of dominant-negative NSF. 2 Fusion of intracellular vesicles with the plasma membrane may be required primarily to provide additional bilayer area required for pseudopod extension. Alternatively, it is possible that fusion of endomembranes is needed to deliver components of the molecular machinery or signaling complexes necessary for completion of the particle engulfment process. The examination of phagocytosis-associated downstream signaling events in the presence of dominant-negative NSF will be informative in this regard.
Whereas mechanisms of bacterial invasion are currently the subject of intensive study, it is not yet clear how a variety of bacterial species induce the formation of the vacuoles that engulf them. In the case of Salmonella species, it is known to involve the type III secretion system encoded for by the genes of pathogenicity island-1. However, the functionality of many of the gene products in this system is still poorly understood (2,33,34). It has been established that Salmonella invasion can be accompanied by localized membrane ruffling and membrane protrusions that resemble the pseudopods observed during phagocytosis. The current study suggests that, although VAMP3 does localize to Salmonella-induced ruffles, this localization is not required for the formation of ruffles (see results with PM-GFP, Fig. 8). Furthermore, internalization of bacteria during the invasion of cells by S. typhimurium was not impaired by TeTx or by inhibition of NSF, revealing an important contrast between phagocytosis and Salmonella invasion.
Collectively, these findings suggest that the ruffling noted during invasion may be largely driven by rearrangement of the underlying cytoskeleton and does not require membrane fusion. This model is in agreement with previous observations that indicate that invasion by S. typhimurium occurs independently of the phosphatidylinositol 3-kinase activity (PI3K) of the host (1,35). This enzyme is an acknowledged mediator of membrane traffic in animal cells. Remarkably, inhibition of PI3K effectively precludes phagocytosis (31,36), highlighting the differences between the latter process and bacterial invasion. It is also possible that membrane fusion events are involved in vacuole formation but that these are triggered by bacterial proteins that are independent of host NSF and PI3K and therefore insensitive to E329Q-NSF and to wortmannin, respectively. Although it has been shown that bacterial proteins are required for the fusion of Chlamydia trachomatiscontaining inclusions (37), no fusogenic bacterial products have been described in Salmonella species.
After internalization of some invasive bacterial species, the vacuoles formed can undergo a series of fission and fusion events involving endomembrane compartments such as endosomes and lysosomes (2). S. typhimurium is known to reside initially in vacuoles that transiently bear the markers EEA1 and transferrin receptor and subsequently become enriched in markers of later compartments such as the vacuolar H ϩ -ATPase and LAMP-1. We found that the percentage of S. typhimurium-containing vacuoles that were positive for LAMP-1 was significantly reduced in cells expressing dominant-negative NSF. These findings are consistent with recent studies that indicate roles for both NSF and Rab proteins in the fusion of SCVs and early endosomes in vitro (38). We are currently working to elucidate the roles of specific SNAREs involved in SCV maturation; however, the present study reveals a clear mechanistic difference between the NSF-dependent maturation of vacuoles and the NSF-independent internalization of the bacteria. Therefore, our results are consistent with a model of bacterial invasion whereby S. typhimurium autonomously invades target cells and subsequently commandeers intracellular machinery to manipulate its own microenvironment.