Cdc42 and Rac1 regulate late events in Salmonella typhimurium-induced interleukin-8 secretion from polarized epithelial cells.

Salmonella typhimurium colonization of the intestinal epithelium initiates biochemical cross-talk between pathogen and host that results in the secretion of chemokines, such as interleukin (IL)-8, that direct neutrophil migration to the site of infection. In nonpolarized cells, Rac1 and Cdc42 have been shown to regulate both bacterial invasion and signaling events leading to nuclear responses and IL-8 secretion. However, because the underlying actin cytoskeleton and the associated signaling machinery are distributed much differently in polarized epithelial cells, we used polarized Madin-Darby canine kidney monolayers to investigate the role of Rac1 and Cdc42 in S. typhimurium-induced pro-inflammatory responses in the more physiologically relevant polarized state. In Madin-Darby canine kidney monolayers expressing dominant-negative Rac1 or Cdc42, both Salmonella- and tumor necrosis factor alpha-induced activation of NFkappaB and mitogen-activated protein kinase signaling cascades proceeded normally, but IL-8 secretion was inhibited. We found that Rac1 and Cdc42 were not involved in early pro-inflammatory signaling events, as in nonpolarized cells, but rather regulated the basolateral exocytosis and secretion of IL-8. In contrast, dominant-negative Rac1 inhibited apical actin pedestal formation, indicating that pedestal formation and nuclear signaling for pro-inflammatory activation are not linked. These findings indicate that there are significant differences in the requirements of pathogen-induced host cell signaling pathways in polarized and nonpolarized cells.

The intestinal epithelium provides a selectively permeable barrier that maintains internal homeostasis by separating the internal environment from the external milieu. Intestinal epithelial cells have coevolved over millions of years of close contact with normal gut flora and have developed mechanisms to discriminate between pathogenic and nonpathogenic organisms. It is this ability of the epithelium to discriminate between nonpathogenic intestinal flora and pathogenic bacteria that is central to the host mounting an effective immune response. During colonization of the gut by enteropathogenic bacteria like Salmonella typhimurium, the intestinal epithelium is the first line of defense against the invading microorganisms and reacts to the colonization with a rapid innate immune response. This highly localized response involves the polar secretion of cytokines, predominantly interleukin-8 (IL-8), 1 that sets up a subepithelial chemotactic gradient directing polymorphonuclear leukocytes (neutrophils) to the site of infection (1,2), a response that is paramount to the inflammatory diarrhea associated with S. typhimurium infection (3). The initiation of this rapid innate immune response requires two-way biochemical cross-talk between host and pathogen.
During its coevolution with the intestinal epithelium, S. typhimurium has developed mechanisms to subvert host cell functions allowing it to enter the normally nonphagocytic epithelial cells and multiply (for reviews see Refs. 4 -6). Physical contact between S. typhimurium and the apical epithelial surface, in addition to poorly understood biochemical signals, induce the activation of a specialized protein-secretion system, known as a type III secretory system (7)(8)(9). It is by way of this highly complex secretory system, comprised of multiple protein components, that Salmonella delivers bacterial effector proteins into the eukaryotic host cell cytoplasm (for review see Ref. 9). Lately, much attention has been directed at understanding the functions of the S. typhimurium type III secretory proteins on host cell cytoskeletal remodeling and nuclear signaling events. Microinjection and overexpression of recombinant type III proteins have been shown to cause the rapid reorganization of the actin cytoskeleton and membrane ruffling (10,11) that are similar to that seen during S. typhimurium invasion of the host cell. Some type III secretory proteins, like SipB, SipC, and SipD, are required for translocation other type III proteins into the cell (12), and others, like SipA, have been shown to have actin bundling abilities (10), whereas yet others, like SopE, SopB, and SptP, may act in concert to disrupt the cytoskeleton and regulate GDP-GTP exchange on the small GTPases Rac1 and Cdc42 (11,13). In nonpolarized cells these latter type III proteins seem to also regulate nuclear signal transduction pathways within the host cell in a Cdc42/Rac1-dependent manner (11). However, it is not known whether these same pathways are required for Salmonella-induced events in fully polarized epithelial monolayers where the underlying actin cytoskeleton and signaling machinery are quite different. Evidence from recent studies in polarized cells suggests that Salmonella-induced nuclear signaling events as well as basolateral IL-8 secretion requires the translocation of flagellin across the epithelium where it interacts with the basolateral surface receptor (14 -17).
The Rho subfamily of small GTPases are known to regulate a wide variety of cellular processes including actin cytoskeletal changes, the modulation of signaling pathways, and vesicle transport (see reviews in Refs. 18 -20). Consisting of at least 14 known members, the Rho subfamily is regulated by an ever increasing number of guanine nucleotide exchange factors, GTPase-activating proteins, and GDP dissociation inhibitors, all of which not only regulate activity but also actively participate in signal transduction (20). These complex connections produce a signaling system full of promiscuity and redundancy that is both cell type-specific and dependent on the nature of the original stimulus. The role of Rho GTPases in Salmonellainduced cytoskeletal changes and invasion of nonpolarized cells has been well documented (9,21); however, recent findings in polarized cells (22) raise concern about the physiological relevance of studies done in nonpolarized or cell-free systems.
To examine the involvement of Rho GTPases in S. typhimurium-induced basolateral IL-8 secretion from polarized epithelial cells, we have adapted a Madin-Darby canine kidney (MDCK) cell system with inducible expression of Rac1 and Cdc42 dominant-negative mutants. MDCK cell lines with doxycycline-repressible expression of the dominant-negative mutants of Rac1 and Cdc42 have previously been established to examine the role of these small GTPases in the early events leading to epithelial development of polarity and tight junction formation (23,24). These same cell lines provide a unique tool to examine the role of Rho GTPases in the signal transduction pathways activated during Salmonella colonization of polarized epithelial monolayers. Using this inducible MDCK cell system has several advantages over the nonpolarized cell systems that have been previously used to investigate S. typhimurium-induced cell signaling events. MDCK cell monolayers have long been used in the study of S. typhimurium-epithelial cell interactions such as adherence and invasion (25,26), cytoskeletal rearrangements (27), bacterial translocation across the monolayer (28 -31), and polymorphonuclear transmigration in response to bacterial invasion. MDCK cell monolayers also mirror the responses of the T84 human intestinal epithelial monolayers and more closely represent the in vivo conditions during infection of intestinal enterocytes than do nonpolarized epithelial cells.
Using this more physiologically relevant system, we demonstrated a significant difference between the signaling mechanisms employed by nonpolarized cells and those used in polarized epithelial monolayers in response to S. typhimurium colonization. Overexpression of dominant-negative Rac1 and Cdc42 in nonpolarized cells has been previously shown to regulate early events in S. typhimurium-induced signaling (11,21); however, we find that this is not the case in polarized epithelia expressing these dominant-negative GTPases. In polarized epithelial monolayers expressing dominant-negative Rac1 or Cdc42, S. typhimurium-induced activation of MAPK, p38-MAPK, JNK, IB␣ kinase, translocation of NFB to the nucleus, and production of IL-8 message all proceed normally; however, basolateral IL-8 secretion is inhibited, suggesting that the small GTPases Rac1 and Cdc42 may regulate exocytosis and thus control the basolateral secretion of IL-8. This also suggests that both Rac1 and Cdc42 play a much different role during S. typhimurium infection depending on the polarity of the cell line being used in the experiments.

EXPERIMENTAL PROCEDURES
Cell Culture-MDCK epithelial cells (T23 clone) expressing dominant-negative Myc-tagged Rac1N17 or Cdc42N17 under the control of a tetracycline-repressible transactivator were grown as previously described (23). Briefly, the cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 0.1 mM nonessential amino acids (Invitrogen), 100 units/ml penicillin, 100 g/ml streptomycin, and 20 ng/ml doxycycline (Sigma) at 37°C in a humidified atmosphere containing 5% CO 2 . MDCK cells were seeded on 0.45 M pore Transwell Clear inserts (Corning-Costar, Cambridge, MA) in Dulbecco's modified Eagle's medium/fetal bovine serum medium containing doxycycline and refed 24 h later and every 2 days thereafter. Monolayers were induced to express mutant GTPases 4 -6 days after seeding by incubating monolayers in Dulbecco's modified Eagle's medium/fetal bovine serum medium lacking doxycycline and containing 2.5 mM sodium butyrate for 24 h. The parental cell line transfected with an empty plasmid and treated in an identical manner served as the control; additionally, noninduced mutant cells were also used and produced similar results. All of the antibiotics were washed out prior to incubation with bacteria and had no effect on Salmonella viability. Human T84 intestinal epithelial cells (passages 70 -95) were maintained as confluent monolayers on collagen-coated permeable supports (32,33).
Bacterial Strains and Growth Conditions-S. typhimurium 3306 is a naladixic acid-resistant (gyrA1816) strain derived from S. typhimurium strain SR-11 (34). LB was made as previously described (35). L agar is LB containing 7 g of Bacto Agar (Difco Laboratories, Detroit, MI)/liter. Bacterial growth conditions were as follows: nonagitated microaerophilic bacterial cultures were prepared by inoculating 10 ml of LB with 0.01 ml of a stationary phase culture followed by overnight incubation (ϳ18 h) at 37°C, as previously detailed (1,36).
Salmonella-induced IL-8 Secretion-Confluent monolayers of T84 cells and MDCK cells, grown on 6.5-mm-diameter collagen-coated Transwell inserts, were washed three times with HBSS and placed into 300 l of HBSS and incubated for 30 min at 37°C. The monolayers were placed in dry wells, and 25 l of Salmonella-containing HBSS (1.6 ϫ 10 10 bacteria/ml) was placed on the apical surface of each monolayer. This inoculum has been previously shown to correspond to 30 associated bacteria/cell (1). 1 h later, the monolayers were returned to the wells containing 300 l of HBSS. 5 h after adding the bacteria, basolateral cell supernatants were removed and assayed for IL-8. Human IL-8 from T84 cells were measured by ELISA as previously described (1). Canine IL-8 from MDCK cells was measured by ELISA in 96-well plates (Linbro/Titertek; ICN Biochemicals, Costa Mesa, CA) coated overnight with 1 g/ml anti-rabbit IL-8 monoclonal antibody and detected with rabbit anti-canine IL-8 polyclonal antibody from Dr. Randal Mrsny (Genentech Inc., South San Francisco, CA). Canine fibronectin from MDCK cells was measured by ELISA in 96-well plates (Linbro/ Titertek; ICN Biochemicals) coated overnight with basolateral supernatants and detected with a monoclonal anti-fibronectin antibody (Transduction Laboratories).
Cell Extraction and Immunoblotting-Filter-grown cells were rinsed twice in ice-cold HBSS, lysed in protein loading buffer (50 mM Tris, pH 6.8, 100 mM dithiothreitol, 2% SDS, 0.1% bromphenol blue, 10% glycerol), and sonicated using a Branson Sonifier 450 (Branson, Danbury, CT). Equal amounts of total cell protein were separated by SDS-polyacrylamide gel electrophoresis (37), transferred to nitrocellulose according to standard procedure (38), and processed for immunoblotting with specific antibodies. Immune complexes were visualized with the appropriate peroxidase-conjugated antibodies and developed using an ECL kit (Amersham Biosciences). Chemiluminescent signals were collected and scanned from ECL Hyperfilm (Amersham Biosciences) with a Fluor-S MultiImager (Bio-Rad) or a UMAX Astra 1200S flatbed scanner (UMAX Technologies, Inc., Fremont, CA). For figures, the contrast of images was adjusted, arranged, and labeled in Adobe Photoshop and Adobe Illustrator (Adobe Systems Incorporated, San Jose, CA). The digital images are representative of the original data.
GTPase Activation Assay-GTP-bound Rac1 and Cdc42 were isolated from MDCK cell lysates using a Rac/Cdc42 activation assay kit (Upstate Biotechnology, Inc., Lake Placid, NY) according to the manufacturer's protocol. Filter-grown MDCK cell monolayers were washed in HBSS ϩ at 4°C and lysed at 4°C in cell lysis buffer (25 mM HEPES, pH 7.5, 10 mM MgCl 2 , 150 mM NaCl, 1% Nonidet P-40, 10% glycerol) supplemented with a protease inhibitor mixture (Roche Molecular Biochemicals). Cell debris was removed by centrifugation at 4,000 rpm for 5 min at 4°C, and a fraction (1/100) of the total cell lysate was saved to be probed for total Rac1 or Cdc42. The remaining supernatant was incubated with gentle mixing for 1 h at 4°C with glutathione S-transferase-PAK1 beads (Upstate Biotechnology, Inc.). The beads were washed extensively, and proteins bound to the beads were separated on 12% acrylamide gels and transferred to nitrocellulose. The membranes were incubated with either an anti-Rac1 (BD Transduction Laboratories) or anti-Cdc42 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) antibody, followed by a peroxidase-conjugated sheep anti-mouse or donkey anti-rabbit antibody (Amersham Biosciences). Immune complexes were visualized by ECL as described above.
Northern Blots-MDCK cell monolayers were induced to express the mutant GTPases and treated as indicated. Two hours after treatment, the cells were lysed with TRIzol reagent (Invitrogen) for 5 min at room temperature and then extracted with chloroform. RNA was precipitated with isopropanol, washed with 75% ethanol, and resuspended in distilled H 2 O treated with 0.01% diethylpyrocarbonate. 30 g of total RNA was denatured, fractionated by electrophoresis in a 1% agarose gel, and transferred to BrightStar-Plus membrane (Ambion, Inc., Austin, TX) using NorthernMax-Gly Northern blotting kit procedures (Ambion, Inc.). Canine IL-8 cDNA and glyceraldehyde-3-phosphate dehydrogenase probes were labeled with BrightStar psoralen-biotin nonisotopic labeling kit (Ambion, Inc.). The blots were prehybridized for 60 min at 42°C in ULTRAhyb solution (Ambion, Inc.). The blots were hybridized with 1.0 pM psoralen-biotin-labeled, heat-denatured cDNA probe in the same solution for 24 h at 42°C. The blots were washed one time in low stringency wash at room temperature and two times in high stringency wash at 42°C followed by detection using a BioDetect kit (Ambion, Inc.) according to the manufacturer's instructions and exposed to film. The blots were stripped for10 min in 1% SDS-SSC at 100°C and reprobed as above with a biotinylated DECAtemplate™-glyceraldehyde-3-phosphate dehydrogenase mouse (905 bp) probe (Ambion Inc.).
Confocal Laser Scanning Microscopy-Monolayers of MDCK cells were rinsed three times in PBS, fixed for 10 min in 3.7% paraformaldehyde, and then rinsed three times in PBS. The cells were then permeabilized for 10 min with 0.2% Triton X-100 and rinsed three times with PBS containing 10% bovine serum albumin. Permeabilized cells were then incubated with either goat anti-Salmonella antibody (Kirkegaard & Perry Laboratories) for 20 min at room temperature or mouse anti-NFB p65 (Transduction Laboratories) or rabbit anti-c-Myc (Santa Cruz Biotechnology, Inc.) antibodies for 1 h at 37°C. After staining with primary antibodies, the cells were rinsed three times with PBS and incubated with the appropriate FITC-conjugated secondary antibodies for 20 min to 1 h at 37°C. In some experiments, actin filaments were stained with rhodamine-conjugated phalloidin for 45 min at 37°C and rinsed three times in PBS. Monolayers attached to membranes were excised from the inserts and mounted cell side up on a glass slide. The membrane was covered with 10 l of SlowFade reagent (Molecular Probes, Eugene, OR) followed by a coverslip, and the edges were sealed to prevent drying. Specimens were examined with a Zeiss LSM410 scanning laser confocal microscope using the 488/568 nm wavelength lines of an argon-krypton laser. The cell monolayer was optically section every 0.5 m. Image resolution using a Zeiss 100ϫ Neofluor objective. Zeiss LSM software was 512 ϫ 512 pixels.

Apical Colonization by S. typhimurium Induces Basolateral IL-8 Secretion from MDCK Cell Monolayers-Recent studies
have relied on nonpolarized cells and in vitro assays to investigate the molecular mechanisms by which Salmonella stimulates IL-8 secretion from host cells. Here, we have instead chosen the well characterized MDCK polarized epithelial cell model system to investigate the interactions between host and pathogen. We began by comparing the Salmonella-induced IL-8 secretion in both human T84 and canine MDCK polarized monolayers using human-and canine-specific ELISAs. Both T84 and MDCK cells form polarized monolayers with physically and biochemically distinct apical and basolateral membranes when grown on porous Transwell cell culture inserts (39 -41). When filter-grown monolayers were infected with wild-type S. typhimurium (3306) added to the apical surface, significant IL-8 secretion was detected in the basolateral supernatant at levels comparable with that stimulated by the addition of basolateral TNF␣ (Fig. 1). Untreated monolayers did not secrete IL-8 during the same time period (Fig. 1). These results confirm that, despite differences in origin, MDCK cell monolayers respond to Salmonella colonization in a manner identical to T84 cells and thus provide an excellent model system for the study of pathogen-epithelial cell interactions.

Expression of Dominant-negative Rac1 Inhibits Salmonellainduced Apical Pedestal Formation, whereas Both Dominantnegative Cdc42 and Rac1 Inhibit Basolateral IL-8 Secretion-
Recent studies in nonpolarized cells have indicated that Cdc42 and to a lesser extent Rac1 were directly involved in the Salmonella-induced membrane ruffling, signaling cascades leading to nuclear responses and ultimately IL-8 secretion (11). As a result, our first objective was to determine whether expression of the dominant-negative Cdc42 or Rac1 molecules were capable of inhibiting Salmonella-induced membrane ruffling and IL-8 secretion in our polarized epithelial cell system. We began by examining the expression of the dominant-negative Cdc42 or Rac1 in polarized monolayers both by immunoblot and confocal laser scanning microscopy. When monolayers were incubated in the presence of doxycycline, only the lower molecular weight endogenous GTPases were observed in immunoblots of total cellular protein ( Fig. 2A, MDCK). Filtergrown monolayers expressing the dominant-negative Cdc42 or Rac1 had higher molecular weight bands corresponding to the Myc-tagged mutant proteins, with expression levels equal to or slightly greater than that of the endogenous GTPase ( Fig. 2A,  arrow). We also examined the expression in filter-grown monolayers stained with an anti-Myc antibody that recognizes only the mutant GTPases, followed by a secondary FITC-conjugated antibody. Once again we found that monolayers incubated in the presence of doxycycline expressed only the endogenous GTPases and lacked expression of the Myc-tagged dominantnegative mutants (Fig. 2B, MDCK). Monolayers expressing the dominant-negative Cdc42N17 or Rac1N17 exhibited cytoplas- mic staining that was consistent with observations made previously by others (Fig. 2B) (22).
The functional ability of the dominant-negative Cdc42 and Rac1 to inhibit endogenous GTPase activity and affect Salmonella-induced actin cytoskeleton reorganization was examined in MDCK monolayers. Previous reports in nonpolarized cells indicate that predominantly Cdc42 and to a lesser extent Rac1 regulate the rapid actin cytoskeletal rearrangements needed for bacterial invasion (11,21); however, more recent studies in polarized MDCK monolayers have found that only Rac1 is involved in this process at the apical surface (22). Our results concur with the latter. Although both dominant-negative MDCK cell lines express levels of mutant GTPase roughly equivalent to the endogenous proteins, only Rac1N17 was able to inhibit the local Salmonella-induced actin cytoskeletal reorganization at the apical membrane (Fig. 2C). These results are consistent with the distinct phenotypic differences seen in polarized monolayers expressing the dominant-negative Cdc42 and Rac1 GTPase (22).
We next compared the Salmonella-induced basolateral IL-8 secretion from filter-grown monolayers expressing dominantnegative Rac1N17 and Cdc42N17 with monolayers expressing only the endogenous GTPases. In all cell lines, the cells incubated with buffer alone did not secrete IL-8; however, cells expressing only endogenous GTPases secreted significant amounts of IL-8 when incubated with apical S. typhimurium or basolateral TNF␣ (Fig. 2D). Cells expressing dominant-negative Cdc42 or Rac1 showed approximately a 70% reduction in the wild-type Salmonella-induced IL-8 secretion and a 75% to almost complete reduction in TNF␣-induced IL-8 secretion (Fig. 2D).
To confirm that only Rac1 was activated during apical invasion by Salmonella, we next examined the time course of GTPase activity for Rac1 and Cdc42 in infected monolayers. Using an affinity precipitation assay that relies on the specific interaction of activated, GTP bond Rac1 or Cdc42 with the protein-binding domain (PBD) of P21-activated kinase (PAK-1) linked to agarose beads to precipitate the activated GTPases, we compared the amount GTP-bound Rac1 or Cdc42 isolated from infected monolayers with that from uninfected controls. Prior to measuring the GTPase activity induced by apical Salmonella colonization, we established the capacity of our assay by precipitating Rac1 and Cdc42 that were activated in vitro by incubation with GTP␥S. We found that the PAK-1 PBD beads had the capacity to bind activated Rac1 or Cdc42 at least 50-fold greater than basal levels (data not shown) and thus were more than sufficient to precipitate the GTPases activated by Salmonella. In monolayers incubated for various times with apical Salmonella, we saw a rapid, transient, activation of Rac1 that was initiated as early as 10 min after the addition of bacteria, peaked between 20 and 30 min, and returned to basal levels between 30 and 60 min of incubation (Fig. 3). During the remainder of the incubation (60 -240 min), Rac1 activity remained constant at basal levels (Fig. 3). We observed similar results in four separate experiments. In concurrence with a report from another lab (22), we found that apical colonization did not activate Cdc42 at any time during the incubation (data not shown), which is in complete agreement with the inability of Cdc42N17 to inhibit apical membrane ruffling (Fig. 2C).

FIG. 2. Expression of dominant-negative Rac1 and Cdc42 inhibit Salmonella and TNF␣-induced IL-8 secretion.
A, filtergrown MDCK cell monolayers induced to express transgenes were lysed, and equal amounts of total cellular protein were separated by SDS-PAGE and transferred to nitrocellulose. The membranes were probed with anti-Rac1 or Cdc42 monoclonal antibodies followed by horseradish peroxidase-conjugated anti-mouse antibody and bands were visualized with ECL. Mutant, Myc-tagged proteins are indicated by the arrow. B, monolayers grown as above were fixed, permeabilized, and labeled with anti-c-Myc followed by FITC-conjugated secondary antibody and visualized by confocal microscopy. C, monolayers were incubated with apically applied wild-type S. typhimurium (3306) (ϩ), or HBSS ϩ alone (Ϫ), for 30min at 37°C, fixed, permeabilized, and labeled with goat anti-Salmonella antibody followed by FITC-conjugated anti-goat antibody and rhodamine phalloidin to visualize actin filaments. D, MDCK monolayers expressing the transgenes were incu-bated at 37°C for 5 h in either HBSS ϩ alone (No treatment), TNF␣ (100 ng/ml) added basolaterally (TNF␣), or wild-type S. typhimurium added apically (WT (3306)). After 5 h, basolateral supernatants were collected, and IL-8 was quantified using a canine specific ELISA. The data represent the means of triplicate experiments Ϯ S.D. The data labeled MDCK represent results seen in noninduced mutant cells and cells transfected with an empty plasmid.
These results demonstrate that only Rac1 was activated by apical colonization and that this activity was transient, occurring for only 30 -60 min during the early membrane ruffling phase.
Early Salmonella-induced Nuclear Signaling Cascades Are Unaffected by Expression of Dominant-negative Cdc42 or Rac1-To determine whether the effects of dominant-negative Rac1 and Cdc42 were inhibiting early nuclear signaling events as well as localized cortical actin rearrangements at contact sites, we examined several mediators known to be involved in stimulating IL-8 secretion, including the IB kinase (42), p44 -42 MAPK, p38, and JNK (43). Filter-grown monolayers of Rac1N17 and Cdc42N17 cells were grown in medium supplemented with or without doxycycline to prevent or induce dominant-negative GTPase expression respectively. Monolayers incubated with doxycycline expressed only the endogenous GTPases, whereas both Rac1N17 and Cdc42N17 monolayers incubated without doxycycline had high levels of expression of the dominant-negative GTPases (Fig. 4). Degradation of IB␣ in proteosomes is an early step, mediated by IB kinase, known to be required for nuclear signaling events leading to activation of the transcription factor NFB and stimulation of IL-8 synthesis (42). When monolayers were incubated with buffer alone, IB␣ remained intact in all of the cell lines (Fig. 4, No  treatment). However, when monolayers were incubated with apical S. typhimurium (3306) or basolateral TNF␣, IB␣ levels were significantly reduced compared with monolayers that received no treatment (Fig. 4). These results are identical to those obtained from T84 monolayers treated in the same manner (data not shown). We next determined the effect of the dominant-negative GTPases on S. typhimurium activation of MAPK, p38 MAPK, and JNK by examining the phosphorylated, active forms of these kinases. Prior to experimentation, filter-grown monolayers were serum-starved for 48 h by incubation in medium containing 0.5% fetal bovine serum to reduce the background level of phosphorylated kinases because of growth factors in the serum. In both mutant expressing and nonexpressing monolayers, the level of phosphorylated (activated) MAPK, p38, and JNK was very low (Fig. 4). However, after incubation with apical S. typhimurium (3306) or basolateral TNF␣, each of the monolayers responded in a similar manner with increased activation of MAPK, p38, and JNK (Fig.  4). Interestingly, both S. typhimurium and TNF␣ seem to be activating the same early pathways even though the initial signaling event is Ca 2ϩ -dependent for Salmonella, but Ca 2ϩindependent for TNF␣ (44). These results support the idea that neither Rac1 nor Cdc42 are involved in the early Salmonellainduced nuclear signaling events in polarized epithelial mono-layers and are in direct contrast to results previously obtained in nonpolarized cells.

Salmonella-induced NFB Nuclear Translocation and IL-8 Message Production Are Not Inhibited by Dominant-negative
Cdc42 or Rac1-The next critical step central to Salmonellainduced innate immune response and IL-8 secretion is the translocation of the transcription factor NFB to the cell nucleus (45). We therefore used confocal microscopy to examine NFB nuclear translocation in monolayers expressing dominant-negative Rac1 and Cdc42. Filter-grown monolayers were incubated in the presence or absence of apical S. typhimurium (3306) followed by an additional 1-h incubation at which time the monolayers were fixed, permeabilized, and stained with anti-p65 (NFB) antibody and a secondary FITC-conjugated antibody and mounted on slides. In monolayers that received no treatment, NFB was localized in the cytoplasm in monolayers expressing only endogenous GTPases as well as those expressing dominant-negative Rac1 and Cdc42 (Fig. 5, top  row). This cytoplasmic localization of inactive NFB is identical to staining seen in T84 monolayers (not shown). When monolayers were instead incubated with apical S. typhimurium (3306), NFB was released from its inactive, IB␣-bound state and translocated from the cytoplasm to the nucleus. This occurred in each of the cell lines, regardless of whether the dominant-negative GTPases were being expressed (Fig. 5, bottom row).
We next focused our attention on the S. typhimurium-induced transcription of the IL-8 message to determine whether Rac1N17 and Cdc42N17 may be having an effect at the level of transcription. Northern blot analysis revealed that monolayers incubated with either apical S. typhimurium or basolateral TNF␣ were capable of transcribing IL-8 mRNA (Fig. 6, top   FIG. 3. Apical Salmonella colonization transiently activates endogenous Rac1. Filter-grown MDCK cell monolayers were incubated at 37°C with apically applied wild-type 3306, or HBSS ϩ alone (No treatment) for the times indicated. Activated GTP-bound Rac1 was isolated from cell lysates by affinity precipitation with PBD-PAK1 beads (see "Experimental Procedures"). GTPase pull-downs (upper panel), as well as one-tenth of the total cellular lysates (lower panel) were separated by SDS-PAGE and transferred to nitrocellulose. The membranes were probed with anti-Rac1 monoclonal antibody followed by a horseradish peroxidase-conjugated antibody, and the bands were visualized with ECL.

FIG. 4. Salmonella-induced signaling cascades are not inhibited by dominant-negative Rac1 or Cdc42.
Filter-grown MDCK cell monolayers induced to express transgenes were incubated at 37°C for 1 h in either HBSS ϩ alone (No treatment) or with wild-type S. typhimurium added apically (WT (3306)) or for 30 min with TNF␣ (100 ng/ml) added basolaterally (TNF␣). The cells were lysed, and equal amounts of total cellular protein were separated by SDS-PAGE and transferred to nitrocellulose. The membranes were probed with anti-Rac1, Cdc42, IB␣, phospho-p44/42 MAPK, phospho-p38 MAPK, or phospho-JNK monoclonal antibodies followed by horseradish peroxidase-conjugated anti-mouse antibody, and the bands were visualized with ECL. The data labeled MDCK represent results seen in noninduced mutant cells and cells transfected with an empty plasmid. The data are representative of multiple experiments. row), with no discernable difference between control monolayers and monolayers expressing Rac1N17 or Cdc42N17. Hybridization with a glyceraldehyde-3-phosphate dehydrogenase probe was equivalent under all experimental conditions, indicating that each lane was loaded with the same amount of total RNA (Fig. 6, bottom row). Our results thus far indicate that S. typhimurium-induced signaling events proceeding through IL-8 transcription were not affected by expression of dominantnegative Rac1 or Cdc42, forcing us to examine later events in IL-8 secretion.
Expression of Dominant-negative Cdc42 and Rac1 Reduce Basolateral Secretion of the Marker Protein, Fibronectin-Recent studies have indicated a potential role for Cdc42 and Rac1 in the biosynthetic, endocytic, and secretory pathways (46 -50).
To determine whether dominant-negative Cdc42 and Rac1 may have compromised the secretion of S. typhimurium-induced IL-8, we examined the constitutive basolateral secretion of fibronectin under the same conditions. Using an ELISA to detect basolateral fibronectin secretion, we compared monolayers expressing Rac1N17 and Cdc42N17 with nonexpressing monolayers. We found that during the same time course in which IL-8 was collected, constitutive basolateral fibronectin secretion was also reduced. In RacN17 monolayers we found that fibronectin secretion was reduced to ϳ65% of control monolayers, and in Cdc42N17 it was reduced even further to ϳ40% of control monolayers (Fig. 7). These results provide strong evidence that Cdc42 and Rac1 are involved in the basolateral secretion both of constitutively secreted proteins like fibronectin and of transiently expressed proteins like IL-8.

DISCUSSION
Interactions between S. typhimurium and epithelia stimulate the production and secretion of bacterial effector molecules that allow the bacterium to enter the normally nonphagocytic host cells, a process that is essential for bacterial colonization. As a result, the host cell mounts a rapid innate immune response via the polar secretion of pro-inflammatory cytokines, ultimately recruiting neutrophils (polymorphonuclear leukocytes) to the site of infection. In nonpolarized cells, the bacterial effector protein, SopE, has been shown to catalyze the activation of Cdc42 and Rac1, mediating both the actin cytoskeletal rearrangements necessary for bacterial entry as well as nuclear signaling events leading to pro-inflammatory cytokine release (11,21); however, in polarized cells the host cell requirements for membrane ruffling and bacterial invasion are quite different (22). Therefore, in the present study, we tested the hypothesis that polarized epithelia would have different molecular requirements for Salmonella-induced nuclear signaling events necessary for IL-8 secretion than those reported in nonpolarized cells. Interestingly, we have found that neither Cdc42 nor Rac1 are necessary for Salmonella-induced nuclear signaling events in polarized epithelia but rather are required for basolateral secretion of IL-8. These results are particularly relevant to future studies of bacterial-host cell interactions and demonstrate the vital importance of host cell polarity on these interactions. Understanding the process by which the host epithelium recognizes the invading pathogen and mounts a pro-inflammatory response is paramount to understanding how they cause disease. MDCK monolayers respond to apical Salmonella colonization in a manner that is both morphologically and biochemically identical to intestinally derived T84 monolayers. At the site of bacterial attachment, there is a rapid actin cytoskeletal reorganization that results in the formation of membrane ruffles that are also seen in nonpolarized cells infected with Salmonella (5). This highly localized cytoskeletal rearrangement and subsequent bacterial invasion require the injection of several bacterial Type III effector proteins that can activate host cell Rho GTPases and bundle actin filaments (for review see Ref. 51). It is at this point when Salmonella-host cell interactions in nonpolarized and polarized cells begin to diverge. In nonpolarized cells the Salmonella effector protein SopE has been shown to activate the GTPase activity of Cdc42 and Rac1, inducing both membrane ruffling and nuclear signaling events leading to pro-inflammatory responses (11,21). It is important to remember that nonpolarized cells lack the segregation of membrane receptors and signaling molecules as well as the highly organized actin cytoskeleton common to polarized cells. A recent study of the requirements for invasion in polarized epithelia eloquently demonstrates these differences. In contrast to previous reports, the authors found that Cdc42 is not necessary for apical invasion and is not even activated by Salmonella colonization (22). Instead, they show that SopE activation of Rac1 is required for membrane ruffle formation and bacterial entry at the apical surface of polarized epithelia, which is similar to our results in Figs. 2 and 3. These important differences are explained by the likelihood that SopE only activates GTPases in the immediate area surrounding the site of bacterial-epithelial contact where Rac1 is localized. Cdc42, on the other hand, is far removed from the site of bacterial invasion and seems to be functioning at the trans-Golgi network and basolateral membrane in polarized epithelia (46 -48, 52, 53). Interestingly, both Cdc42 and Rac1 are activated by Salmonella invasion of the basolateral membrane, but neither is required for invasion to occur at that surface (22).
Recent studies have begun to examine the potential regulatory role of both Rho GTPases and the actin cytoskeleton in the endocytic, biosynthetic, and secretory processes. The actinbinding proteins ankyrin, spectrin, and myosin II have all been found to be closely associated with the Golgi and the trans-Golgi network and have been implicated in vesicle budding and fusion (54) that may be regulated by the polymerization state of actin. Because both Cdc42 and Rac1 can direct the reorganization of actin filaments, it is likely that the dominant-negative mutants of these GTPases may affect the transient actin filaments that have recently been shown to be associated with vesicle budding at the Golgi and the trans-Golgi network (54,55). It is still unclear whether Cdc42 and Rac1 directly regulate vesicle budding or simply modulate pathways required for the maintenance of epithelial polarity (Fig. 8) (46 -48, 53).
Early studies of Salmonella-host cell interactions in nonpolarized cells have prompted the formation of a paradigm of Salmonella-induced inflammation in which the type III effector protein SopE is solely responsible for the activation of signaling cascades that result in nuclear responses and ultimately IL-8 secretion (11,21). This early archetype appears to be an oversimplification of the process that occurs in polarized epithelia. Recent studies have demonstrated that nonflagellated Salmonella, capable of invading host cells and possessing intact type III apparatus, lacks the ability to induce nuclear signaling and IL-8 secretion (17). Still other studies have demonstrated that flagellin alone can activate nuclear signaling in polarized cells, but only from the basolateral surface (14), by exclusively activating the basolaterally localized Toll-like receptor 5 (TLR5) (Fig. 8) (16,56). The mechanism by which flagellin activates TLR5 and pro-inflammatory signaling cascades is still unclear, but the importance of cell polarity in Salmonella-induced inflammation has become increasingly evident. These results fit very well with our own, because we find that Salmonella induces IL-8 secretion independent of SopE activation of Cdc42 or Rac1. Additionally, we have shown that Salmonellae focally infect monolayers, and yet all cells in the monolayer respond with translocation of NFB to the nucleus. This is completely consistent with the idea that soluble transcytosed flagellin interacts with the basolateral TLR5 on all cells, initiating the pro-inflammatory response (Fig. 8). Although bacterial effectors are necessary for the early events of membrane ruffling and invasion at the apical membrane, it is now clear that Salmonella-induced nuclear signaling events require the activation of basolateral TLR5 by flagellin in polarized epithelial cells (Fig. 8).
In summary, the work presented here demonstrates a clear role for Cdc42 and Rac1 in the Salmonella-induced transit of IL-8 from the trans-Golgi network to the basolateral membrane. These findings are in complete accordance with other recent work in which Cdc42 was found to regulate both biosynthetic and endocytic protein traffic to the basolateral membrane (46,47). Additionally, our results fit nicely with recent studies that demonstrate significant differences between polarized and nonpolarized cells in their requirements for Salmonella internalization (22). However, our results contrast previous work in nonpolarized cells (11,21) and suggest that the mediators of Salmonella-induced nuclear signaling are dependent on the degree of cell polarity. Our results indicate that Salmonella-induced alteration of the actin cytoskeleton and During apical colonization of intestinal epithelial cells, Salmonella type III effector molecules reorganize the underlying actin cytoskeleton and regulate bacterial invasion in a Rac1-dependent manner. Salmonella flagellin is transcytosed across the monolayer to interact with basolateral TLR5 receptors in both infected and noninfected cells, activating MAPK, p38, JNK, and IB kinase. Subsequently, IB␣ is degraded in proteosomes, releasing NFB to translocate to the nucleus and stimulate IL-8 transcription. Cdc42 and Rac1 regulate basolateral exocytosis of IL-8. induction of nuclear signaling events have very different requirements in polarized epithelia and nonpolarized cells. The precise mechanisms by which Cdc42 and Rac1 regulate the basolateral secretion of IL-8 have yet to be characterized.