Intracellular Localization of Type III-delivered Pseudomonas ExoS with Endosome Vesicles*

ExoS (453 amino acids) is a bi-functional type III cytotoxin produced by Pseudomonas aeruginosa. Residues 96-219 include the Rho GTPase-activating protein (RhoGAP) domain, and residues 234-453 include the 14-3-3-dependent ADPribosyltransferase domain. Earlier studies also identified an N-terminal domain (termed the membrane localization domain) that comprises residues 51-77 and includes a novel leucine-rich motif that targets ExoS to the perinuclear region of cultured cells. There is limited information on how ExoS or other type III cytotoxins enter and target intracellular host proteins. Type III-delivered ExoS localized to both plasma membrane and perinuclear region, whereas ExoS(ΔMLD) was localized to the cytosol. Plasma membrane localization of ExoS was transient and had a half-life of ∼20 min. Type III-delivered ExoS co-immunoprecipitated 14-3-3 proteins and Rab9, Rab6, and Rab5. Immunofluorescence experiments showed that ExoS colocalized with Rab9, Rab6, and Rab5. Fluorescent energy transfer was detected between ExoS and 14-3-3 proteins but not between ExoS and Rabs proteins. Together, these results indicate that type III-delivered ExoS localizes on the host endosomes and utilizes multiple pathways to traffic from the plasma membrane to the perinuclear region of intoxicated host cells.

Intracellular localization of several type III cytotoxins has been observed previously. YopM localizes to the nucleus and stimulates the activity of PRK2 and RSK1 kinases (23,24). YopH localizes to the focal adhesion complexes and is essential for anti-phagocytosis and virulence (25). YopE localizes to cytoplasmic granules and the perinuclear region (26,27). However, how type III toxins traffic in intoxicated cells has not been established. This study shows that type III-delivered ExoS localizes temporally on the plasma membrane and associates with markers of early and late endosomes.

EXPERIMENTAL PROCEDURES
Materials-HeLa cells (CCL-2) were from the ATCC. Tissue culture media and sera were from Invitrogen. Reagents for molecular and cell biological techniques were from New England Biolabs or Invitrogen, and chemicals were from Sigma, unless noted. DNA primers were from Operon Technologies.
Cell Culture Growth-HeLa cells were cultured in complete medium (minimum essential medium ϩ 10% fetal calf serum, nonessential amino acids, sodium pyruvate, sodium bicarbonate and penicillin/streptomycin) and maintained in a 37°C humidified 5% CO 2 (v/v) incubator.
Transfection and Cellular Fractionation of HeLa Cells-HeLa cells were seeded in 85-mm dishes with 3 ϫ 10 5 cells the day before use. Cells were grown to 70% confluency and transfected with Lipofectamine and Plus reagent-mediated liposome transfer system (Invitrogen), as described by manufacturer, using 1 g of indicated DNA (31). Total DNA was normalized with pCMV10Luc (luciferase) (Kent Wilcox, Medical College of Wisconsin). After 18 -24 h, transfected cells were washed twice with PBS, suspended in 10 ml of homogenization buffer (HB1) (250 mM sucrose, 3 mM imidazole, pH 7.4), washed in 300 l of HB1, suspended in 300 l of HB2 (HB1 plus 1% mammalian protease inhibitor mixture set III (Sigma) and 0.5 mM EDTA). Cells were lysed by passage 20 times through a 25-gauge needle. The whole-cell lysate was centrifuged for 5 min at 2000 rpm in a microcentrifuge at 4°C, and the pellet (nuclei and unbroken cells) and postnuclear supernatant (PNS) were collected. The PNS was centrifuged for 30 min at 100,000 ϫ g, and the pellet (membrane) and supernatant (cytosol) were collected. Samples were normalized to volume equivalent with SDS-PAGE sample buffer, boiled, and stored at Ϫ20°C.
Type III Delivery of ExoS into HeLa Cells-HeLa cells were cultured in 6-well plates to 70% confluency and transfected with 200 ng of the indicated plasmid. After 18 -24 h, cells were infected at a multiplicity of 8:1 (bacteria:HeLa cells) of P. aeruginosa PA103 ⌬exoU, exoT::Tc (pUCP) (32) as control, and the indicated ExoS construct, including pUCPExoS (33) or pUCPExoS R146K, E379D, E381D, termed KDD. After 3.5 h, HeLa cells were washed with PBS and treated with 100 g/ml gentamicin and 200 g/ml ciprofloxacin to kill extracellular bacteria. At the indicated time, cells were washed twice with PBS, and lysates were prepared and subjected to subcellular fractionation or cells were treated with 1% paraformaldehyde and examined by direct fluorescence, subjected to immunofluorescence analysis, or phase-contrast microscopy.
Immunofluorescence Microscopy-HeLa cells seeded on coverslips in 6-well plates were infected with P. aeruginosa PA103 ⌬exoU, exoT::Tc (pUCP-ExoS-KDD) at an m.o.i. of 8:1 (bacteria:HeLa). Four hours postinfection, cells were washed three times with PBS and fixed with 1% paraformaldehyde for 1 h at 4°C. Cells were permeabilized with 0.1% Triton X-100 in 4% formaldehyde (15 min at room temperature), which was followed by three washes with PBS. Cells were blocked with 1% bovine serum albumin in PBS for 20 min at room temperature and then incubated with individual antibodies. For Rab6/ExoS double staining, cells were incubated with rabbit ␣-Rab6 IgG (1:500) and mouse ␣-HA IgG (1:1,000) for 1 h at room temperature, washed three times with PBS, incubated with goat ␣-rabbit IgG-Alexa-488 (1:500) and goat ␣-mouse IgG-Alexa-568 (1:500). For Rab5 and Rab9 staining, cells were first stained with either mouse ␣-Rab5 IgG (1:500) or mouse ␣-Rab9 IgG (1:200), followed by incubation with goat ␣-mouse IgG-Alexa-488. Next, cells were fixed with paraformaldehyde, washed with PBS, and stained with prelabeled mouse ␣-HA IgG-(Fab) 2 -Alexa-568 according to the manufacturer's protocol (Zenon prelabeling kit, Molecular Probes). Cells were then washed, mounted, and observed under a ϫ60 oil objective. Images were captured with a Ropor camera using Metamorph software and cropped in Adobe Photoshop. Confocal images were obtained on a Leica TCS SP2 laser scanning confocal microscope at the Bryant Imaging Facility, Medical College of Wisconsin. Incidence of fluorescence between two flurochromes was determined as described in ImageJ software. Briefly, ICQ numbers of the boxed regions of analyzed figures were calculated by ImageJ plugin intensity correlation analysis (National Institutes of Health, Bethesda, MD). Typical values are as follows: random (or independent) staining ICQ ϭ ϳ0; dependent staining 0 Ͻ ICQ Յ ϩ0.5; and segregated (or exclusive) staining 0 Ͼ ICQ Ն Ϫ0.5. For quantification of the percentage plasma membrane localization, random fields of HeLa cells were scored by a blinded observer. Plasma membrane association of ExoS was scored by focusing through the entire cell. Mouse ␣-hemagglutinin (HA) antibody was from Covance (catalog number MMS-101P); mouse ␣-Rab9 antibody was from Calbiochem (catalog number 552101). Rabbit ␣-Rab6 antibody was from Santa Cruz Biotechnology (C-19). Mouse ␣-Rab5 antibody was from BD Transduction Laboratories (catalog number R60020-050). Mouse ␣-Rab4 antibody was from BD Transduction Laboratories (catalog number 610888). Alex-aFluor goat ␣-mouse or ␣-rabbit IgG was from Molecular Probes. Sheep ␣-human TGN46 was from Serotec (catalog number AHP500). Rabbit ␣-calnexin was from Stress-Gen (catalog number SPA-865).
Immunoblot Analysis-The PNS was subjected to SDS-PAGE (11-12%) and transferred to a polyvinylidene fluoride membrane. Membranes were incubated with the indicated primary antibody for 1 h at room temperature, washed with PBS, and then incubated for 1 h with goat ␣-mouse-HRP IgG or goat ␣-rabbit-HRP IgG (1:40,000; Pierce). HRP was developed with SuperSignal (Pierce) and exposed to x-ray film, which was developed in a film processor or digitized with an Alpha-Inatech 8900 imaging system. Mouse ␣-H-Ras antibody was from BD Transduction Laboratories (catalog number R02120-150). Mouse ␣-GFP antibody was from Covance (catalog number MMS-118P).
Co-immunoprecipitation (Co-IP) of Host Protein Bound to Type III-delivered ExoS-HeLa cells were plated in 150-mm dishes and grown to 90% confluency. Cells were infected with P. aeruginosa expressing a vector control (pUCP), ExoS⌬MLD-KDD, or  were stained by immunofluorescence procedures as described under "Experimental Procedures." Images were generated with MetaMorph series 6.0 software exposed for the same time. B, HeLa cells were transfected with pEGFP (Clontech) and then infected with the indicated strain of P. aeruginosa. 4 h postintoxication, cells were harvested, and the postnuclear supernatants were collected. The PNS was separated to cytosol (Cyto) and membrane (Mem) fraction by ultracentrifugation and resolved on SDS-PAGE. The percentage of ExoS in the membrane fraction (right panel) was calculated based on densitometry of the Western blot using ␣-HA as the primary antibody (left panel). Fractionation of the cytosolic GFP was used as an internal control for fractionation efficiency. C, HeLa cells were transiently transfected with plasmids encoding ExoS-KDD or ExoS⌬MLD-KDD. After 18 -20 h, ExoS was immunostained with mouse ␣-HA-IgG as described in A. The amount of Alexa-546 staining was quantified as pixels by ImageJ, Research Services Branch, NIMH (Bethesda, MD). N, nucleus; P, plasma membrane; y axis ϭ gray value (arbitrary units); x ϭ axis distance (pixels).
ExoS-KDD at an m.o.i. of 8:1 (bacteria:HeLa). Note, both ExoS⌬MLD-KDD and ExoS-KDD used in this co-IP had a 3xFLAG (Sigma) epitope engineered onto the C terminus ( Fig.  1). After 4 -5 h of infection, HeLa cells were harvested, and 600 l of PNS were collected as described in the cell fractionation protocol. NaCl was added to a final concentration of 150 mM. One hundred and fifty l of 3x-FLAG antibody-conjugated agarose beads (50% slurry, Sigma) were aliquoted into a microcentrifuge tube. The beads were washed twice with HB2 buffer to remove glycerol. PNS (500 l) were added to the beads, and the tubes were incubated for 2 h at 4°C on a wheel rotator. The beads were pelleted with a pulse spin; the supernatant was discarded, and the beads were washed three times with TBS buffer (50 mM Tris-HCl, 150 mM NaCl, 1% mammalian protease inhibitor mixture set III (Sigma), and 0.5 mM EDTA) for 10 min each. The beads were incubated overnight in 150 g of 3xFLAG peptide in 100 l of TBS buffer. The suspension was pulse-spun, and the supernatant was analyzed for proteins that were co-immunoprecipitated with ExoS by two-dimensional IEF SDS-PAGE as described previously (34). After extraction, beads were boiled to determine ExoS that remained bound to beads following extraction with the 3xFLAG peptide. Mouse ␣-FLAG M2 affinity gel was from Sigma (catalog number A2220). 3xFLAG peptide was from Sigma (catalog number F4799).
MALDI-TOF Analysis-Bands from five gels were excised and subjected to trypsin digestion (1 g; Promega, Madison, WI) in 50 l of 100 mM NH 4 HCO 3 , pH 8, at 37°C for 24 h. After digestion, gel slices were sonicated twice in 200 l of 80% acetonitrile and 1% formic acid (in H 2 O) for 10 min. Eluted material was combined and evaporated, and the pellet was dissolved in 15 l of 0.1% trifluoroacetic acid (in H 2 O). Peptide solutions were desalted with C18 Zip Tips (Millipore, Bedford, MA) that had been equilibrated successively in 15 l of acetonitrile, 15 l of 50% acetonitrile (H 2 O), and 15 l of 0.1% trifluoroacetic acid (H 2 O). Resin was washed twice with 0.1% trifluoroacetic acid (H 2 O). Peptides were eluted in 2 l of 60% acetonitrile and 0.1% trifluoroacetic acid (H 2 O saturated with ␣-cyano-4-hydroxycinnamic acid) and applied to a sample plate to air-dry. Samples were ionized by an N2 UV laser using a PE-pro mass spectrometer (Applied Biosystems). Two hundred laser shots were conducted at an accelerating voltage of 25,000 and laser intensity of 2075 (repetition rate 3 Hz). Scans were processed using Biosystems Voyager 6004 software. Peptide fingerprinting was used to identify proteins present in the band, using Protein Prospector (University of California, San Francisco).
Fluorescence Resonance Energy Transfer (FRET) Determinations-DNA encoding 14-3-3-␥, 14-3-3-⑀, or 14-3-3-␤ was subcloned into pYFP-C1 (Clontech), and DNA encoding g/ml of ciprofloxacin and 100 g/ml of gentamicin and fixed at 15-min intervals. ExoS and ExoS⌬MLD were detected by immunofluorescence staining for the HA epitope as described under "Experimental Procedures." HA staining (upper) and phase contrast images (lower) for the time course for type III-delivered ExoS and the 4.00-h time point for type III-delivered ExoS-⌬MLD are shown. There results shown are from independent representative experiments repeated several times. B, random fields were scored blinded for the % of cells that stained positive for ExoS on the plasma membrane. This is determined by visual inspection of cells that stained for ExoS on the plasma membrane/total number of cells that stained ExoS ϫ100 and is reported as the average of three independent experiments. Dashed line (total ExoS level) represents the amount of cell-associated ExoS determined by measuring the total amount of HA reactivity in each well as arbitrary units of fluorescence. The fluorescence intensity was calculated by Metamorph software.
ExoS-KDD, ExoS-⌬MLD-KDD, or ExoS-(1-416)-KDD was subcloned into pCFP-N1 (Clontech). HeLa cells cultured on coverslips were co-transfected with 150 ng of pYFP-14-3-3-␥ and 150 ng of either p-ExoS-KDD-CFP or p-ExoS-(1-416)-KDD-CFP. Control experiments showed that this ratio of the acceptor to donor yielded optimal FRET measurements (data not shown). 18 h posttransfection, cells were fixed and mounted onto microscope slides. Acceptor photobleach experiments were carried out with a Leica TCS SP2 laser scanning confocal microscope according to manufacturer's instructions. Briefly, the YFP signal in a selected region of interest was bleached with a 514 nm laser until fluorescence was Ͻ20% of the original intensity; Fig. 7A shows the bleached area as a box. Average fluorescence intensity of donor (D, CFP) and acceptor (A, YFP) before (pre) and after photobleach (post) was measured. FRET efficiency was calculated with Equation 1, Control experiments were performed with one frame of photobleach with 0% laser power (35).
Tetanolysin Experiments-HeLa cells were permeabilized with tetanolysin (List Biologicals, Campbell, CA) using a procedure adapted from Ahnert-Hilger et al. (36). At the first indication of cell rounding, P. aeruginosa-infected HeLa cells (6-well plate) were washed with PBS at room temperature and incubated in 6 ml total of ice-cold HG1 buffer (20 mM PIPES, 2 mM Na ϩ -ATP, 4.8 mM Mg(CH 3 COO) 2 , 150 mM potassium glutamate, 2 mM EGTA, 1 mM dithiothreitol, and KOH to obtain pH 7.0) with 2.4 g of tetanolysin for 15 min and then washed with ice-cold HG1 buffer. Next, 6 ml of total of HG1 buffer containing 20 nM [ 32 P]adenylate phosphate-NAD (10 Ci) were added, and cells were incubated for 40 min at 37°C in 5% CO 2 . Cells were washed in PBS and lysed with the addition of 100 l of sample buffer and subjected to SDS-PAGE.
ADP-ribosylation of Ras by Type III-delivered ExoS-pEGFP-His-Ras (C181S,C184S) and pEGFP-His-Ras(WT) were generated, and 600 ng of plasmid encoding GFP-Ras(C181S,C184S) or GFP-RasWT were transfected into HeLa cells. After ϳ18 h, cells were intoxicated with P. aeruginosa strain PA103(⌬exoU, exoT::Tc) (pUCPExoS) for 3 h when cells were washed, collected, lysed in sample buffer, and subjected to 11% SDS-PAGE. Ras proteins were detected by ECL, using mouse ␣-Ras IgG as the primary antibody (BD Transduction Laboratories). Ratio of ADP-ribosylation was calculated as electrophoretically shifted Ras(C181S,C184S) divided by unshifted Ras(C181S,C184S) and analyzed by densitometry. siRNA Knockdown Assay-Protocol is based on Ambion siRNA protocol. Briefly, siPORT TM Amine and NeoFX transfection reagents were mixed with siRNA oligonucleotides and incubated with HeLa cells for 48 h in complete media. Rab9 siRNA was designed by Ambion, and Rab6A/AЈ siRNA was designed as described previously (37). Knockdown efficiency of the respective Rab protein was calculated by Western blot as described above for immunoblot analysis. Primers are as follows: Rab9 siRNAϩ, 5Ј-GCAGUGUAUCAUCUACUA-Att-3Ј, and Ϫ, 5Ј-UUAGUAGAU-

Type III-delivered ExoS Associates with the Plasma Membrane and
Perinuclear Region-How type III cytotoxins localize and traffic within mammalian cells is not clear. To limit toxic effects on the host cell, several noncatalytic forms (ExoS⌬MLD- . Co-immunoprecipitation of host proteins with type III-delivered ExoS. A, HeLa cells (150-mm dishes) were grown to 90% confluency and intoxicated with P. aeruginosa ⌬exoU, exoT::Tc (pUCP), pUCP-ExoS-⌬MLD-KDD-3xFLAG, ⌬MLD, or pUCP-ExoS-KDD-3xFLAG, KDD at an m.o.i. of 8:1. After 4 h, cells were lysed, and cell lysate (PNS) was prepared. Co-immunoprecipitation of host protein with ExoS was performed as described under "Experimental Procedures." Immunoprecipitated material was acetone precipitated and subjected to 11% SDS-PAGE. Gels were silver-stained. The amount of ExoS precipitated was determined by Western blot (WB) (top). Arrows show the migration of ExoS and 14-3-3 proteins (factor-activating exoenzyme S (FAS)) as determined by Western blotting. B, immunoprecipitated material released with the 3xFLAG peptide was also acetone-precipitated and subjected to isoelectric focusing, using pH 4 -7 strips with two-dimensional standards. The IEF strips were subjected to 11% SDS-PAGE and silver-stained. The two-dimensional gel electrophoresis of the IP material was performed five independent times. In each of the five independent experiments, bands 1-2 and 1-6 were observed in the co-immunoprecipitates with ExoS⌬MLD and ExoS, respectively. Bands 1 and 2 were subjected to MALDI-TOF (Table 1), whereas bands 3-6 were subjected to MALDI-MS/post-source decay ( Table 2). ExoS⌬MLD and ExoS were detected by Western blotting of an identical sample and circled (ExoS⌬MLD) or ExoS(KDD).
R146K,E379D,E381D) (ExoS⌬MLD-KDD) and ExoS(R146K, E379D, E381D) (ExoS-KDD) were used to monitor the intracellular localization of type III-delivered ExoS (Fig. 1). In Fig. 2, the infection was stopped when wild type ExoS (ExoS-WT) fully round the cells. At this time, intoxication with P. aeruginosa expressing pUCP, ExoS⌬MLD-KDD, and ExoS-KDD had minimal effect on cell morphology ( Fig. 2A). ExoS-WT expression was detectable but less than the noncatalytic ExoS-KDD or Exos⌬MLD-KDD due presumably to the toxic nature of the protein. Low expression of the catalytically active ADP-ribosyltransferase domain in transiently transfected mammalian cells has also been observed (38). Type III-delivered ExoS⌬MLD-KDD localized throughout HeLa cells, whereas ExoS-KDD localized on both the plasma membrane and the perinuclear region ( Fig. 2A). Because both ExoS⌬MLD-KDD and ExoS-KDD have an HA epitope, the HA epitope did not appear responsible for the observed intracellular localization. HeLa cells infected with ExoS⌬MLD-KDD or ExoS-KDD were separated to soluble and membrane fractions by ultracentrifugation. Western blot analysis showed that ExoS⌬MLD-KDD was present in the cytosol, whereas ExoS-KDD was distributed in both the membranes and cytosol (Fig. 2B). Cells were pretransfected with pGFP, and GFP was used as an internal control for efficiency of cell fractionation. To determine whether the plasma membrane localization was an intrinsic property of type III-delivered ExoS, localization of transfected ExoS⌬MLD-KDD and ExoS-KDD was examined. ExoS⌬MLD-KDD localized throughout the cells, whereas ExoS-KDD accumulated near the perinuclear region without apparent plasma membrane association (Fig. 2C); quantification of the staining is also shown. This indicates that the MLDdependent association of ExoS with the plasma membrane occurs during type III delivery. Plasma membrane association of type III-delivered ExoS may require a P. aeruginosa factor(s) or may represent a difference in the experimental protocol used for expression of ExoS in a bacterial intoxication versus expression during a transfection. This is currently under investigation.
Type III-delivered ExoS Is Temporally Localized on the Plasma Membrane-To investigate the stability of plasma membrane-associated ExoS, a pulse-chase type experiment was performed to measure the stability of type III-delivered ExoS on the host cell membrane. Between 3 and 4 h postinfection, type IIIdelivered ExoS-KDD was detected on both the plasma membrane and in the perinuclear region (Fig. 3A). At 4 h postinfection, the infection was stopped by washing cells and adding gentamicin and ciprofloxacin to the culture media. During the chase, plasma membrane association of ExoS-KDD decreased by 50% in ϳ20 min, whereas the overall amount of cell associated-ExoS-KDD remained constant with the majority of ExoS-KDD now located in the perinuclear region (Fig. 3, A and B). Plasma membrane-associated ExoS-KDD was observed at the earliest time points, when total ExoS-KDD within the cell was low. Type III-delivered ExoS⌬MLD-KDD did not associate with membranes in a similar time course experiment; the 4-h time point of HeLa cells infected with P. aeruginosa expressing ExoS⌬MLD-KDD and stained for ExoS is shown in Fig. 3A  (4.00, ⌬MLD). These data suggest a sequential movement of type III-delivered ExoS from the plasma membrane to the perinuclear region.
Intracellular Type III-delivered ExoS Is Associated with 14-3-3 Proteins, Tip47, and Rab GTPases-Co-IP was used to gain insight into the cellular basis for intracellular localization of ExoS. Cell lysates were prepared from HeLa cells intoxicated for 4 h with P. aeruginosa possessing a vector control (pUCP), ExoS⌬MLD-KDD, or ExoS-KDD. To facilitate the IP of ExoS from the cell lysate, a 3xFLAG epitope was FIGURE 5. Rab9, Rab6, and Rab5 co-immunoprecipitate with type III-delivered ExoS but not ExoS⌬MLD. Proteins that co-immunoprecipitated with a vector control (pUCP), ExoS⌬MLD-KDD-3FLAG (⌬MLD), or ExoS-KDD-3FLAG (KDD) were subjected to 12% SDS-PAGE in a nonreducing gel. Proteins were transferred to a polyvinylidene fluoride membrane and probed sequentially with mouse ␣-Rab9 IgG, rabbit ␣-Rab6 IgG, mouse ␣-Rab5 IgG, and rabbit ␣-Rab4. Between probes the blot was stripped and tested for stripping efficiency using x-ray film. (Note, the migration of Rab9, Rab6, Rab5, and Rab4 are unique and do not overlap in this SDS-PAGE system.) Secondary antibodies used were goat-␣-mouse or rabbit-HRP IgG and membranes were developed with SuperSignal (Pierce). The reactive bands on the right are probed for total Rab in the indicated lysates prior to immunoprecipitation (input). These experiments were performed two times independently with similar results.

TABLE 1 14-3-3 proteins interact with both ExoS⌬MLD-KDD and ExoS-KDD
The species used in these experiments was Homo sapiens. a Co-immunoprecipitated proteins from a cell lysate of infected HeLa cells immunoprecipitated with ExoS⌬MLD-KDD (⌬MLD) or ExoS-KDD (KDD) were subjected to two-dimensional electrophoresis (see Fig. 4). Spots 1 and 2 were excised from the gel and subjected to tryptic digestion followed by MALDI-TOF analysis. b MALDI-TOF analysis was performed by Protein-Nucleic Acid Shared Facility at the Medical College of Wisconsin. MOWSE score was calculated by Protein Prospector (University of California, San Francisco) and represents the fidelity of the identification of the peptides with a data base of peptides for individual proteins. In this analysis, scores of Ͼ74 indicate identity or homology at p Ͻ 0.05. c Values were determined by comparing with protein standards of known molecular masses and pI values. engineered at the C terminus of ExoS⌬MLD-KDD or ExoS-KDD (Fig. 1). Control experiments showed that fusion of the 3xFLAG epitope to the C terminus of ExoS⌬MLD-KDD or ExoS-KDD did not interfere with toxin secretion and translocation (data not shown). To facilitate the isolation of proteins associated with ExoS, the 3xFLAG peptide was added to the IP to release antibody-bound ExoS and ExoS-associated proteins (Fig. 4, A and B). The immunoprecipitation efficiency was ϳ60% of the total ExoS in the cell lysate (data not shown). The two prominent bands in the IP were ExoS⌬MLD-KDD and ExoS-KDD (Fig. 4A). A 28-and 30-kDa protein co-immunoprecipitated with ExoS(⌬MLD)-KDD and ExoS-KDD and was identified by MALDI-TOF as various isoforms of the 14-3-3 proteins ( Fig.  4A and Table 1). Differences in the intracellular localization may be responsible for the association of ExoS and ExoS⌬MLD with different 14-3-3 isotypes. To enhance the resolution of ExoS-associated proteins, proteins in the co-IP were subjected to two-dimensional gel electrophoresis. IEF-SDS-PAGE analysis detected the two bands representing the 14-3-3 proteins that co-immunoprecitated with ExoS⌬MLD-KDD and ExoS-KDD and four proteins that co-immunoprecipitated specifically with ExoS-KDD (Fig. 5). Mass spectroscopy identified two of the four proteins that coimmunoprecipitated with ExoS-KDD as Tip47 and TPD52 (Table 2). Subsequent Western blotting identified Rab9, Rab6, and Rab5 as proteins that co-immunoprecipitated with ExoS-KDD (Fig. 5). Rab9 and Rab6 were chosen for the analysis because they are associated with late endosomes where Tip47 is found. The co-immunoprecipitation of Rabs with ExoS, but not with ExoS⌬MLD, supports the presence of ExoS on intracellular vesicles and that ExoS is associated with several endosomal compartments. The inability to detect Rab4 in the IP may reflect the limited sensitivity of the antibody, the relative abundance of Rab4, or that ExoS is not associated with recycling endosomes. Tip47/Rab9-associated vesicles have been implicated as a novel, directional vesicle transport system from the late endosome to the Golgi/ER (39,40). Together, this indicated the intracellular localization of ExoS with early and late endosomes.  -568 (1:500). Rab9 was detected with mouse ␣-Rab9 IgG, followed by goat ␣-mouse IgG-Alexa-488. Rab6 was detected with rabbit ␣-Rab6 IgG followed by goat ␣-rabbit IgG-Alexa-488, whereas Rab5 is detected with mouse ␣-Rab5 IgG followed by goat ␣-mouse IgG-Alexa-488. Fluorescent images were captured with MetaMorph series 6.0 software. B, HeLa cells were infected with P. aeruginosa PA103 ⌬exoU, exoT::Tc (pUCP-ExoS-KDD) or pUCP-ExoS-⌬MLD-KDD at an m.o.i. of 8:1 (bacterial:HeLa) and double-stained for Rab9 and HA as described in A. Images were obtained with a Leica confocal scanning microscope. Color correlation (indicated by ICQ) of boxed regions was calculated using ImageJ intensity correlation analysis (National Institutes of Health). Arrows denote examples of vesicles that co-stained positive for ExoS and Rab9. a Co-immunoprecipitated proteins from a cell lysate of infected HeLa cells immunoprecipitated with ExoS⌬MLD-KDD (⌬MLD) or ExoS-KDD (KDD) were subjected to two-dimensional electrophoresis (see Fig. 4). Protein spots 3 and 4 were excised from the gel and subjected to tryptic digestion followed by MALDI-tandem mass spectrometry analysis. Peptide(s) from each protein spot were subjected to post-source decay analysis. Protein spots 5 and 6 were also analyzed but did not yield peptides. b Values were determined by comparing to protein standards of known molecular masses and pI values. c MALDI-mass spectrometry /post-source decay was performed by John Leszyk (University of Massachusetts Medical School, Laboratory for Proteomic Mass Spectrometry). d In this analysis, MOWSE scores of Ͼ41 indicate identity or homology at p Ͻ 0.05.

Co-localization of Type III-delivered ExoS and Rab Proteins on Intracellular
Vesicles-Co-localization of ExoS with Rab proteins in cultured cells was next measured to support or refute the observed association of ExoS with Rab proteins. HeLa cells were infected with P. aeruginosa expressing ExoS-KDD and subjected to immunofluorescence. ExoS was stained with Alexa-594 and endogenous Rab5, Rab6, and Rab9 were stained with Alexa-488, respectively (Fig. 6). Endogenous Rab5, Rab6, and Rab9 localized in unique vesicular patterns. Rab5 was located in the periphery and the perinuclear region; Rab6 was located in the perinuclear region and was condensed within Golgi stacks (41), and Rab9 was located in the peri-nuclear region with an extension toward the periphery. ExoS partially co-localized with each of the Rab proteins. Partial co-localization of Rab5 and ExoS was observed in the periphery. Rab6 and ExoS co-localized on the crescent where Rab6 was enriched, presumably in the TGN region, and co-localization was not observed in the periphery (Fig. 6A, merge). Confocal microscopy also showed co-localization of Rab9 and ExoS and vesicles that contained either Rab9 or ExoS, which was consistent with the partial co-localization of ExoS on Rab9-associated vesicles (Fig. 6B). Color correlation analysis of the staining pattern indi-cated partial coincidence of staining between ExoS-KDD and Rab9 but not between ExoS⌬MLD-KDD and Rab9.
Type III-delivered ExoS ADP-ribosylates Golgi/ER-localized Ras-To determine whether type III-delivered ExoS can ADPribosylate a perinuclear localized host protein, HeLa cells were transfected with GFP-H-Ras(C181S,C184S), which is retained in Golgi/ER, in contrast to GFP-RasWT, which is localized at the plasma membrane (Fig. 8A). Partial co-localization of Ras-(C181S,C184S) with ER and Golgi markers was observed. The Golgi marker, TGN46, showed tighter localization than Ras-(C181S,C184S), whereas the ER marker, calnexin, expression extended beyond Ras(C181S,C184S) (Fig. 8B) (43,44). Next, HeLa cells were transfected with either RasWT or Ras(C181S,C184S) and intoxicated with P. aeruginosa expressing ExoS. Using a tetanolysin-based assay (29), where 32 P-NAD was incorporated into host cells, type III-delivered ExoS ADP-ribosylated both RasWT and Ras(C181,184S) with similar efficiency (Fig. 8C). Because RasWT and Ras(C181,184S) have preferred localizations (Fig.  8A) and Ras(C181S,C184S) has been shown previously not to traffic to the plasma membrane (45), one can conclude that Ras(C181S,C184S) is localized in the Golgi/ER and is ADP-ribosylated by ExoS. In contrast, ExoS⌬MLD did not ADP-ribosylate Ras or Ras(C181,184S), consistent with previous observations (11,46), and was used as a control for specificity. Tetanolysin did not appear to influence ExoS localization because ExoS has similar localization in tetanolysin-treated and nontreated cells (data not shown). This supported the functional trafficking of type III-delivered ExoS to the Golgi/ER. Partial Inhibition of the ADP-ribosylation of Golgi/ER-localized Ras by Dominant Negative Rab9-A gel shift assay was used to measure the influence of dominant negative Rabs and siRNA of the Rabs on the ADP-ribosylation of Ras(C181S,C184S) by type III-delivered ExoS. Rab6A/AЈ siRNA knocked down ϳ70% of total endogenous Rab6A/AЈ, whereas Rab9 siRNA knocked down Ͼ90% of total endogenous Rab9. However, knockdown of either Rab6 or Rab9 had a minimal effect on the ability of type III-delivered ExoS to ADP-ribosylate Golgi/ER-localized Ras (data not shown). Rab9DN had a partial inhibition (ϳ20%) on the ability of type IIIdelivered ExoS to ADP-ribosylate Golgi/ER-localized Ras, whereas Rab6DN did not inhibit (Ͻ10%) the ADP-ribosylation of Golgi/ER-localized Ras (data not shown).

DISCUSSION
The data in this study support a model where type IIIdelivered ExoS initially localizes on the plasma membrane and traffics to the Golgi/ER through a vesicle transport system (Fig. 9). Plasma membrane association of ExoS requires the MLD and type III delivery. The ADP-ribosylation of Ras-WT supports the association of ExoS on the cytoplasmic leaflet of the plasma membrane ( Fig. 9, step 1). The ability to chase ExoS from the plasma membrane suggests a temporal status for plasma membrane association. However, this does not rule out the possibility that ExoS may enter by two pathways, one locating ExoS directly on plasma membrane and another delivering ExoS directly to the perinuclear region. Current understanding of Tip47-vesicle transport does not resolve how ExoS associates with the plasma membrane for transport to late endosome, but ExoS may travel through Rab5-dependent early endosome vesicles (Fig. 9, step 2). Association with Tip47/Rab9 vesicles provides a directional trafficking of ExoS from the late endosome to the Golgi (Fig. 9, step 4), whereas association with Rab6AЈ/6A provides a mechanism for direct trafficking from early endosome to the TGN (step 3). These pathways may overlap and may compensate each others' function.
The determination that type III-delivered ExoS is temporally associated with the plasma membrane and associates with Tip47 implicates a role for vesicle trafficking in the movement of ExoS to the Golgi/ER. The biological significance for this trafficking pathway follows the observation that ExoS⌬MLD does not associate with Tip47/Rab9 or ADP-ribosylate the Ras . After 2 h, cells were permeabilized with tetanolysin and incubated with 32 P-NAD. Cell lysates were subjected to SDS-PAGE and exposed for autoradiography (upper panel, Auto-Rad). Afterward, membrane was subjected to Western blotting for GFP-reactive proteins using mouse ␣-GFP-IgG as primary antibody and goat ␣-mouse IgG-HRP as the secondary antibody and development by ECL (lower panel, Western). Panels shown are representative independent experiments from at least two experiments. FIGURE 9. Intracellular localization and trafficking of type III-delivered ExoS. Type III-delivered ExoS is injected into cells and initially localizes to the plasma membrane in a process that requires Pseudomonas type III secretion. Membrane localization is mediated by the MLD. Plasma membrane-bound ExoS ADP-ribosylates plasma membrane-bound Ras and other host proteins. ExoS dissociates from the plasma membrane and localizes on vesicles with multiple endosome marker proteins. Trafficking of ExoS from the plasma membrane to the Golgi/ER may occur either in series or in parallel pathways. Intracellular ExoS ADP-ribosylates Golgi/ER-bound Ras. EE, early endosome; LE, late endosome; Lys, lysosome; 5, Rab5; 6, Rab6; 9, Rab9; 47, Tip47.
GTPases. Tip47 defines a novel class of vesicles that function independent of the TGN38/shiga toxin pathway (39). Although other type III cytotoxins localize within host cells, the transient plasma membrane localization of ExoS appears to be novel. Pseudomonas ExoU (47,48) and YopO/YpkA (49) localize stably to the plasma membrane, and Yersinia YopH (50) localizes to focal adhesion complexes, and Yersinia YopM (23, 51) localizes to the nucleus via a vesicle-associated pathway. Toxin localization may correlate to efficient targeting of host proteins within the cell.
Another protein present in the co-immunoprecipitate with ExoS was TPD52. TPD52-like proteins are coiled-coil motifbearing proteins first identified through expression in human breast carcinoma, which have been proposed to represent signaling intermediates and regulators of vesicle trafficking (52). TPD52 displays granular cytoplasmic distribution in breast carcinoma cells (53) and is localized on the ER of human PLC hepatoma cells (54). TPD52 may function as an adaptor protein that works in parallel or in series with Tip47 to target the trafficking of ExoS on intracellular vesicles. TPD52 has limited homology (32% similarity and 25% identity within ϳ200 amino acids) with the region of Tip47 that interacts with the cytoplasmic tail of the mannose 6-phosphate receptor. This region of homology may represent a common binding domain between the two proteins. Future studies will address the function of TPD52/Tip47 family proteins as protein adaptors for intracellular trafficking of ExoS.
There was a preferred association of ExoS with Rabs, supporting the association of ExoS with intracellular vesicles. Rab9 was initially chosen for the analysis because they are associated with late endosomes where Tip47 is found. The data indicate that ExoS is associated with several endosomal compartments. The inability to detect Rab4 in the IP may reflect the limited sensitivity of the antibody, low abundance of the protein, or that ExoS is not associated with recycling endosomes. Rab proteins play an essential role in protein trafficking pathways, regulating vesicle budding, movement, and fusion. Approximately 70 Rabs are found in the human genome (55), and at least 12 Rabs localize to the endocytic pathway of mammalian cells (56). Rab5 is important for sequestering ligands into clathrin-coated pits and subsequent fusion of vesicles with early endosomes (57). Molecules exit early endosomes along several different pathways. A direct pathway for recycling receptors to the plasma membrane depends on Rab4 (58), which recycles transferrin receptors and membrane lipids back to plasma membrane (59). Molecules transported to the trans-Golgi network from endosomes follow multiple routes. One pathway that has been defined by internalized TGN38 and several bacterial toxins involves the transport from early or recycling endosomes to the trans-Golgi network (17,60). This transport is controlled by specific members of protein families as follows: the t-SNARE proteins, syntaxin 6, syntaxin 16, Vti1a; the early endosomal v-SNARE proteins, VAMP4 and VAMP3/cellubrevin; and Rab6AЈ (18). Although Rab6A and Rab6AЈ play nonoverlapping roles in membrane trafficking (37), knockdown of Rab6A/A provided only a partial inhibition of trafficking. Recently shiga toxin (37), exotoxin A (22), and ricin toxin (61) were observed to utilize this pathway. Another pathway from early or recycling endosomes to the trans-Golgi network is used by the cation-independent mannose 6-phosphate receptor and furin, which occurs via Rab9late endosomes (62)(63)(64). Association with multiple endosome vesicles may explain the limited ability of DN-Rab9 and DN-Rab6 and si-Rab9 and si-Rab6 to inhibit the ADP-ribosylation of a Golgi/ER substrate by ExoS. Multiple endocytosis pathways allow cholera toxin access to the Golgi and ER (19) where inhibition of endocytosis by clathrin-, caveolin-, or ARF6-dependent mechanisms does not block cholera toxin movement into the cell or attenuate toxicity (20,21).
How ExoS moves within the host cell via the MLD is an intriguing question. The MLD is composed of a redundant multiple leucine motif (12) where leucines are clustered on one side of the helical face and charge residues clustered on the opposite helical face of a helical wheel model. The physical basis for the association of ExoS with endosomes is not known, but it is a topic of future investigations.