Invasive Shigella flexneri Activates NF-κB Through a Lipopolysaccharide-Dependent Innate Intracellular Response and Leads to IL-8 Expression in Epithelial Cells¹

Shigella flexneri is a Gram-negative facultatively intracellular pathogen that is responsible for bacillary dysentery in humans. Infection by this pathogen leads to an intense and acute inflammatory bowel disease that is characterized by watery diarrhea with purulent discharge. Colonic biopsies from infected patients reveal massive inflammatory cell infiltration, tissue edema, and regions where the epithelium is completely destroyed (1). S. flexneri is unable to invade epithelial cells through the apical route (2). To gain entry into the colonic epithelium, S. flexneri exploits M cells, the specialized epithelial cells that overlie lymphoid follicles. M cells allow intact S. flexneri to traverse into the underlying subepithelial pocket where macrophages reside (3, 4). These macrophages engulf S. flexneri, but instead of successfully destroying the bacteria, these cells rapidly undergo apoptosis (5). Before cell death, infected macrophages release IL-1β through the direct activation of caspase-1 by S. flexneri (6). The proinflammatory nature of IL-1β results in the recruitment of polymorphonuclear cells that infiltrate the infected site and destabilize the epithelium (7). Loss of integrity of the epithelial barrier allows more bacteria to traverse into the subepithelial space and gain access to the basolateral pole of the epithelial cells (8). S. flexneri can then invade epithelial cells, spread from cell to cell, and disseminate throughout the tissue.

Although epithelial cells lining the colon are bathed in bacteria and bacterial products, they remain by and large refractory to the presence of these normally inflammatory agents. Epithelial cells are generally unresponsive to LPS (9), and therefore colonic epithelial cells do not detect the potential threat from the normal bacterial flora. Detection of the resident bacterial flora by colonic epithelial cells would have serious consequences, as the colon would be in a state of chronic inflammation. These cells, however, are not completely refractory to bacterial stimulation. Recent studies have demonstrated that epithelial cells produce proinflammatory cytokines, but this occurs only during infection by certain pathogens (10). In particular, IL-8 appears to be a major secreted product of infected epithelial cells (11). This proinflammatory chemokine is a potent chemoattractant for polymorphonuclear cells and can direct recruitment of these cells into the infected site and their infiltration of the epithelial layer (7, 12). The key role played by IL-8 during S. flexneri infection was recently demonstrated in the rabbit model of shigellosis. Using immunohistochemistry on sections of S. flexneri-infected ligated ileal loops, intense IL-8 staining was observed throughout the epithelial layer, whereas rabbits infected with a noninvasive strain showed few IL-8-positive epithelial cells. Pretreatment of rabbits with Abs to IL-8 before bacterial challenge dramatically reduced both fluid accumulation within the infected ileal loop and the level of inflammation in the infected tissue (13).

An important transcriptional regulator of IL-8 gene expression is NF-κB (14). NF-κB is an ubiquitous transcription factor involved in the inducible expression of a number of genes whose products, including many cytokines/chemokines, cell adhesion molecules, and acute phase response proteins, are involved in the inflammatory response (for a review, see Ref. 15). In most cell types NF-κB is held latent in the cytoplasm through its binding to inhibitory proteins, called IκBs,⁴ that mask the nuclear localization signal on NF-κB and thus prevent its nuclear translocation. Most of the signaling pathways stimulated by known activators of NF-κB converge on the activation of a kinase complex called the IKK signalsome. Once activated, this complex, which consists of the two kinases, IKK1 and IKK2 (16, 17, 18), and the regulatory subunit NEMO (IKKγ) (19, 20, 21), phosphorylates residues Ser³² and Ser³⁶ on the inhibitory IκBα protein (22). This phosphorylation event leads to polyubiquitination of the IκBα and subsequent degradation of the protein by the 26S proteasome (23). NF-κB is then free to translocate to the nucleus, where it binds to specific κB elements in the promoter region of responsive genes (15).

A number of studies have shown activation of NF-κB in epithelial cells by bacterial pathogens, including members of Salmonella (24), Neisseria (25), enteropathogenic Escherichia coli (EPEC) (26), and S. flexneri (27). Recently, the signaling system leading to NF-κB activation by different enteroinvasive bacteria was investigated (28); however, neither the pathway nor the bacterial products responsible for NF-κB activation have been clearly elucidated. Because of its important role in inflammation, we were interested in examining the signaling pathway leading to the induction of NF-κB in epithelial cells infected by S. flexneri and identifying the bacterial products that might activate this transcription factor. In addition, we have examined the role played by NF-κB in the induction of the proinflammatory chemokine, IL-8.

M90T is a wild-type invasive strain of S. flexneri serotype 5a. BS176 is a noninvasive variant of M90T cured of the 220kB virulence plasmid. SC301 and SC300 are derivatives of M90T and BS176, respectively, harboring the plasmid pIL22 that encodes the afimbrial adhesin from uropathogenic Escherichia coli (29). These strains have been described previously (30). The ipa mutants as well as the ipgD and icsA mutants have been described previously (31, 32, 33). The E. coli strain β2098, which is deficient in the production of N-formylated proteins, was provided by Dr. Didier Mazel (Institut Pasteur, Paris, France), and has been described previously (34). The Bacillus subtilis strain was provided by Dr. Agnes Fouet (Institut Pasteur).

Bacteria-free culture supernatants were prepared by centrifuging overnight cultures at 3000 × g for 15 min and passing the recovered supernatants through a 0.2-μm pore size filter. Phenol-water extracts of LPS from S. flexneri 1a and E. coli O111:B4 were purchased from Sigma (St. Louis, MO). LPS from S. flexneri 5a was purified by the method of Westphal (35). Rhodobacter sphaeroides LPS was purchased from List Biochemicals (Campbell, CA). LPSs were diluted in microinjection buffer (25 mM Tris (pH 7.4), 100 mM KCl, 5 mM MgCl₂, and 1 mM EGTA) and sonicated for 30 s before use.

HeLa cells were grown routinely in MEM (Life Technologies, Paisley, U.K.) with 10% FCS (Life Technologies) and supplemented with penicillin and streptomycin (Life Technologies). The intestinal cell line Caco-2 was grown in DMEM (Life Technologies) with 10% FCS and antibiotics. The human embryonic kidney cell line 293 was cultured in MEM with 10% FCS and antibiotics.

Antisera to NEMO and IκBα were provided by Dr. Robert Weil (Institut Pasteur), and antisera to p50 and c-Rel were supplied by Dr. Nancy Rice (Frederick, MD) and have been described previously (19, 36). Antisera to p65, IκBβ, and IκBε were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). FITC-dextran (10,000 m.w.) was obtained from Molecular Probes (Eugene, OR). Recombinant human IL-1 and TNF-α were purchased from R&D Systems Europe (Oxon, U.K.). Genistein, GF109203X, H7, wortmannin, aLLnL, and MG132 were obtained from Biomol (Plymouth Meeting, PA), and all other reagents were obtained from Sigma unless otherwise specified.

Overnight cultures of S. flexneri strains were diluted 1/100 in trypticase soy broth (Diagnostics Pasteur, Marnes la Coquette, France) and grown to midexponential phase at 37°C (37). HeLa cells that had been serum-starved overnight were washed three times, and the medium was replaced with antibiotic-free, serum-free MEM. Bacteria were then added to the cells and incubated for 10 min at room temperature to allow adherence of bacteria to cells under noninvasive conditions. Plates of infected cells were moved to 37°C and incubated for an additional 15 min. Subsequently, cells were washed five times, and the medium was replaced with MEM containing 50 μg/ml gentamicin to kill extracellular bacteria. Incubations were then conducted for the indicated periods of time. The number of intracellular bacteria was determined by lysing HeLa cells with 0.5% sodium deoxycholate in physiological saline and subsequent plating of serially diluted samples on TCA plates.

Plasmids expressing GST-IκBα1–72(1–72) wild-type and GST-IκBα1–72(1–72)S32A/S36A mutant polypeptides have been described previously (21). Purification of the GST-linked proteins was conducted using standard protocols (38).

Plasmids expressing dominant negative forms of IKK1 and IKK2 were provided by Dr. Michael Karin (University of California a San Diego, La Jolla, CA) and described previously (16). The NIK KK429–430AA expression plasmid (39) was provided by Dr. David Wallach (The Weizmann Institute, Rehovot, Israel). The NF-κB reporter Igκ-luciferase has been described previously (40). The wild-type IL-8 promoter-luciferase construct and IL-8 promoter constructs with site-directed mutations in AP-1, NF-κB, or NF-IL-6 transcription factor binding sites were provided by Drs. Andrew Keates and Ciaran Kelly (Beth Israel Deaconess Medical Center, Boston, MA).

For transfections, HeLa cells were plated in six-well plates at a density of 1 × 10⁵ cells/ml and transfected the following day using FuGene reagent (Roche, Indianapolis, IN) as recommended by the manufacturer, with 0.5 μg of Igκ-luciferase reporter and either the vector or the effector plasmid. Cotransfection with 0.5 μg of a β-galactosidase reporter plasmid was used to normalize transfection efficiencies. Cells were infected with bacteria as described above or treated with TNF-α (100 ng/ml) or IL-1 (10 ng/ml) 36 h posttransfection. Following 5 h with the indicated treatments, cells were lysed in luciferase lysis buffer (25 mM Tris (pH 7.4), 8 mM MgCl₂, 1 mM DTT, 1% (v/v) Triton X-100, and 15% (v/v) glycerol), and luciferase activity was determined. Experiments were performed in duplicate or triplicate and were repeated at least three times. Values represent the mean ± SD. Student’s t test was used to determine statistical significance between the different groups.

Following the indicated treatments, nuclear and cytoplasmic extracts were prepared as previously described (41). Cells were washed twice in ice-cold PBS and scraped from the dishes, and cell pellets were resuspended in hypotonic buffer (20 mM HEPES (pH 7.8), 10 mM KCl, 1 mM DTT, 0.5 mM EDTA, 0.5 mM EGTA, and 1 μg/ml each of leupeptin, aprotinin, and pepstatin). Following 10-min incubation on ice, Nonidet P-40 was added to a final concentration of 1%, and samples were centrifuged for 20 s in a microfuge at maximum speed. The supernatant, which represented the cytoplasmic fraction, was recovered and stored at −80°C until further use. The nuclear pellet was washed briefly with hypotonic buffer and resuspended in extraction buffer (10 mM HEPES (pH 7.8), 400 mM NaCl, 0.1 mM EDTA, 1 mM DTT, supplementary protease inhibitors, and 25% glycerol). Following a 30-min incubation at 4°C with occasional vortex mixing, samples were centrifuged at maximum speed for 5 min. The supernatant, containing the nuclear fraction, was recovered and immediately frozen at −80°C until use.

Five micrograms of nuclear extracts were combined with binding buffer (10 mM HEPES (pH 7.8), 100 mM NaCl, 1 mM EDTA, and 10% glycerol), 1 μg poly(dI-dC), and 0.5 ng of ³²P-labeled DNA probe corresponding to the κB site of the H-2KB promoter (42). Where indicated, EMSAs were also performed using a ³²P-labeled DNA probe corresponding to the region between bp −84 and −68 of the IL-8 promoter: 5′-TCGTGGAATTTCCTCTG (14). Following 30-min incubation at room temperature, samples were run on a 5% polyacrylamide gel in 1× TBE. The gel was dried, exposed to a PhosphorImager screen (Molecular Dynamics, Sunnyvale, CA), and analyzed using ImageQuant software (Molecular Dynamics).

For detection of IκB proteins, cytoplasmic fractions equalized based on protein content were run on 12% SDS-PAGE gels and transferred to nitrocellulose, and specific Ab binding was revealed using an enhanced chemiluminescence detection system (Amersham, Aylesbury, U.K.).

For IKK kinase assays, the method described previously was followed (19). Briefly, cytoplasmic extracts were subjected to immunoprecipitation with anti-NEMO in TNT buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100, 100 μM Na₃VO₄, and supplementary protease inhibitors). Samples were mixed with protein G-Sepharose beads for 1 h at 4°C on a wheel and then washed three times with TNT and three times with kinase buffer (20 mM HEPES (pH 7.5), 50 mM NaCl, 10 mM MgCl₂, 100 μM Na₃VO₄, 20 mM β-glycerophosphate, 2 mM DTT, and 20 μM ATP). Kinase reactions were conducted at 30°C for 30 min, using 5 μCi [γ-³²P]ATP and GST-IκBα1–72(1–72) wild-type or GST-IκBα1–72(1–72)S32A/S36A mutant polypeptides as substrates. Reaction products were run on 12% SDS-PAGE gels and revealed by phosphorimaging.

To measure IL-8 secreted from Shigella-infected HeLa cells, cells were seeded into 12-well dishes and infected as described above or treated with 100 ng/ml TNF-α for various times. IL-8 protein levels were determined by ELISA using Abs and human recombinant IL-8 from R&D Systems Europe.

Fifteen microliters of bacteria-free supernatants from overnight cultures of wild-type S. flexneri serotype 5a were mixed with 5 μg of either dimeric IgA specific for S. flexneri 5a LPS or IgA specific for Salmonella typhimurium LPS for 1 h at 4°C. These Abs have been described previously (43). Protein G-Sepharose beads were then added, and the samples were left rotating overnight at 4°C. Samples were spun at 14,000 rpm in an Eppendorf microfuge, and 5 μg of additional Abs were then added to the supernatants. This process was repeated for a total of three cycles of immunodepletion; each of the subsequent incubations with the Abs was conducted for 1 h. A significant decrease (>4-fold) in the amount of LPS in the supernatants immunodepleted with the S. flexneri LPS 5a-specific Abs, but not the S. typhimurium LPS-specific Abs, was confirmed by SDS-PAGE and silver staining (44).

HeLa or Caco-2 cells were plated onto coverslips, and microinjection was performed the following day in serum-free conditions. Samples containing FITC-dextran were loaded into glass capillary micropipettes. Approximately 100 cells were microinjected within a 5-min period using a Narishige 200 microinjection system (Narishige Group, London, U.K.) and an IMT-2 inverted microscope (Olympus Optical, Tokyo, Japan). Following incubation for 30 min at 37°C in 5% CO₂, cells were fixed with 3.7% paraformaldehyde in PBS for 10 min, permeabilized with 0.5% Triton-X in PBS containing 2% BSA, and subsequently stained with Abs to the p65 subunit of NF-κB. Following labeling with Cy3-linked anti-rabbit Abs (Jackson ImmunoResearch Laboratory, West Grove, PA), coverslips were mounted and viewed with a conventional immunofluorescence microscope (BX50, Olympus Optical).

The NF-κB response in S. flexneri-infected HeLa cells was investigated by EMSA. Fig. 1 A shows that wild-type S. flexneri, M90T, expressing the afrimbrial adhesin activated NF-κB DNA binding activity in HeLa cells 45 min postinfection. This activation was specific to invasive S. flexneri and not to mere adherence to HeLa cells, because the noninvasive plasmid-cured strain also expressing the adhesin did not stimulate NF-κB DNA binding activity. In addition, these studies were performed in the absence of serum to avoid LPS activation through soluble CD14 and the LPS binding protein (9). Moreover, E. coli LPS or S. flexneri 5a LPS added extracellularly at concentrations up to 10 μg/ml did not activate NF-κB (data not shown). NF-κB DNA binding induced by S. flexneri was specific, because the addition of excess unlabeled probe competed the binding activity. In contrast, the addition of excess unlabeled DNA probe corresponding to the STAT3 binding site did not affect the NF-κB complex induced by the bacteria (data not shown).

Dose-response experiments were conducted to determine the minimum number of intracellular bacteria that were necessary for inducing NF-κB DNA binding activity. As few as 10⁴ CFU/10⁶ HeLa cells of intracellular S. flexneri (a multiplicity of infection of 0.1) were capable of stimulating NF-κB DNA binding activity; however, at 45 min postinfection, this activation was relatively weak (data not shown). An initial inoculum corresponding to an MOI of 50 resulted in 10⁷ CFU/10⁶ HeLa cells of intracellular S. flexneri and a strong activation of NF-κB following 45 min of infection that was comparable to the stimulation induced by TNF-α. Therefore, this inoculum was used in all other subsequent experiments.

To ensure that the S. flexneri-induced NF-κB response was not limited to HeLa cells, the intestinal cell line Caco-2 and the human embryonic kidney cell line 293 were infected, and NF-κB DNA binding activity was examined by EMSA. S. flexneri stimulated NF-κB DNA binding activity in CaCo-2 cells with a similar time course as that observed in HeLa cells (data not shown). 293 cells, recently shown to be unresponsive to high concentrations of LPS in the absence of serum (45), showed similar NF-κB induction in response to S. flexneri infection compared with HeLa epithelial cells (data not shown).

Supershift experiments were performed with Abs to the various NF-κB proteins to determine which components of the S. flexneri-induced NF-κB complex were binding to the consensus κB DNA element from the promoter region of the H-2KB gene. Two distinct DNA binding complexes were evident in extracts prepared from S. flexneri-infected HeLa cells (Fig. 1 B). These two complexes were resolved when the gel was run for extended time periods (i.e., 2.5 h vs 1.5 h). Addition of Abs to the p50 component of NF-κB shifted both complexes, whereas Abs to the p65 (RelA) protein shifted the top complex only. Abs to c-Rel did not lead to any appreciable change in the S. flexneri-induced NF-κB complexes (data not shown). This indicated that S. flexneri infection resulted in the formation of NF-κB complexes that were composed mainly of p50/p65 heterodimers (the complex commonly referred to as NF-κB) and homodimers of the non-trans-activating complex p50/p50.

To fully characterize the NF-κB response in HeLa epithelial cells infected with wild-type S. flexneri, time-course experiments examining NF-κB DNA binding by EMSA were performed and compared with the time course of NF-κB activation with TNF-α. As shown in Fig. 2,A, NF-κB DNA binding activity was induced after 45 min of S. flexneri infection that was sustained 4 h postinfection and was evident 6 h postinfection, which was the latest time point examined (data not shown). In contrast, NF-κB DNA binding activity was induced 15 min after addition of TNF-α and was down-regulated by 2 h posttreatment (Fig. 2 A). To rule out the possibility that S. flexneri-infected cells were producing a signal that was, in turn, responsible for NF-κB activation, time-course studies similar to those described above were conducted in the presence of 20 μg/ml cycloheximide to block host cell protein synthesis. Cycloheximide had no effect on the activation of NF-κB by S. flexneri, indicating that infection directly activates NF-κB (data not shown).

Western blot analysis was used to examine the degradation of the IκB proteins in the cytoplasm of infected HeLa cells and cells treated with TNF-α (Fig. 2,B). S. flexneri infection induced the degradation of IκBα, IκBβ, and IκBε 45 min postinfection, and this degradation was sustained 4 h postinfection, thus paralleling what was observed for NF-κB DNA binding activity induced by S. flexneri infection seen in Fig. 2 A. TNF-α induced degradation of IκBα, IκBβ, and IκBε 15 min posttreatment. Although levels of IκBβ and IκBε remained low 4 h posttreatment, levels of IκBα increased after 2-h treatment with TNF-α.

To characterize the signaling pathway induced by invasive S. flexneri leading to NF-κB activation, a number of inhibitors were used to investigate whether they could block NF-κB DNA binding activity in response to S. flexneri infection and TNF-α treatment. Neither inhibitors of protein tyrosine kinases, including genistein and herbimycin, nor inhibitors of protein kinases, including H7 and GF109203X, had any effect on NF-κB DNA binding activity induced by S. flexneri infection or TNF-α treatment. In addition, wortmannin, an inhibitor of PI3 kinase recently shown to inhibit pervanadate-induced (46) and IL-1-induced (47) NF-κB activity, did not affect either S. flexneri- or TNF-α-induced NF-κB DNA binding activity. NF-κB activation induced by both S. flexneri infection and TNF-α treatment, however, was inhibited by the anti-oxidants, N-acetyl-cysteine and pyrrolidine dithiocarbamate, as well as the proteasome inhibitors, aLLnL and MG132. These data suggested that, like TNF-α, NF-κB activation by S. flexneri infection involved the classical pathway, resulting in the antioxidant-sensitive phosphorylation of IκBs through the IKK signalsome and subsequent ubiquitination and degradation of this protein by the 26S proteasome.

To more thoroughly examine this S. flexneri-activated signaling pathway, an approach was taken in which cells were transiently transfected with vector constructs expressing dominant negative versions of signaling molecules that have been identified as key regulators of NF-κB activation in response to cytokines. Since IKK1 and IKK2 have been shown to be important kinases within the IKK signalsome, we examined the roles of these two proteins in S. flexneri-induced NF-κB activity on an NF-κB-responsive luciferase gene reporter. The IKK1(KM) and IKK2(KA) mutant proteins are catalytically inactive, since they have methionine or alanine, respectively, substituted for the lysine codon at position 44, which presumably results in defective ATP binding (16). Fig. 3,A shows the effects of overexpression of IKK1(KM) and IKK2(KA) in HeLa cells infected with S. flexneri or treated with TNF-α. Expression of IKK1(KM) or IKK2(KA) in infected cells resulted in a 76 ± 7 or 69 ± 12% decrease, respectively in luciferase activity compared with that in cells transfected with the vector alone (Fig. 3,A). As previously reported (16), TNF-α-treated cells expressing IKK1(KM) or IKK2(KA) showed a 63 ± 8 or 60 ± 20% decrease, respectively, in relative luciferase activity compared with cells transfected with the vector alone (Fig. 3 A). Transfection of both IKK1(KM) and IKK2(KA) led to similar decreases in luciferase activity in S. flexneri-infected cells (77 ± 9%) and TNF-α-treated cells (70 ± 14%), suggesting that either of these dominant negative proteins expressed alone resulted in maximum inhibition of NF-κB activity.

We next examined the role of NIK in the activity of NF-κB in S. flexneri- and TNF-α-treated cells. NIK activates the IKK signalsome and has been shown to be involved in cytokine-induced NF-κB induction (39). Fig. 3 B shows that overexpression of dominant negative NIK significantly decreased both S. flexneri- and TNF-α-induced activation of NF-κB. Dominant negative NIK inhibited S. flexneri-induced activation of the luciferase reporter by 70 ± 16% compared with infected cells transfected with the vector alone. Dominant negative NIK also efficiently blocked NF-κB activation by TNF-α as previously reported (41) (84 ± 9% decrease in reporter activation compared with cells transfected with vector alone).

To obtain direct evidence that S. flexneri infection activates the IKK complex in HeLa cells, immune complex kinase assays were performed. Abs to NEMO, a component of the IKK complex, were used to immunoprecipitate the complex from untreated, S. flexneri-infected, or TNF-α-treated cells, and kinase activity was then determined by analyzing the incorporation of ³²P into IκBα wild-type and IκBα mutant polypeptides. Fig. 4,A shows that immunoprecipitated IKK from cells infected with S. flexneri for 30 min specifically phosphorylated the wild-type, but not the mutant, IκBα polypeptide. TNF-α treatment for 30 min showed a similar level of IKK activation that was specific for the wild-type IκBα (Fig. 4 A).

Time-course studies were then performed to examine the kinetics of IKK activation induced by S. flexneri infection and TNF-α treatment. An increase in IKK activity induced by infection with S. flexneri was observed 15 min postinfection; maximal activity was seen at 45 min postinfection (Fig. 4,B). At 90 min postinfection, levels of IKK activity were similar to levels in untreated cells. IKK activation by TNF-α treatment demonstrated a different pattern of kinetics; maximum levels of IKK activity were observed 15 min posttreatment, and IKK activity was similar to that in untreated cells by 45 min posttreatment (Fig. 4 B).

To determine whether there was a link between NF-κB activation and IL-8 production, we first investigated whether the NF-κB complexes induced by S. flexneri infection of HeLa cells bound the IL-8 promoter. To this end, EMSAs were performed using the −84 to −68 bp region of the IL-8 promoter that encompass the κB site (14). Fig. 5 A shows a representative EMSA as well as a supershift assay used to reveal the different NF-κB complexes that bound to the κB site of the IL-8 promoter. Three different NF-κB complexes from nuclear extracts of S. flexneri-infected HeLa cells were observed to bind to this DNA probe. The lower complex was nonspecific, since it was also present in uninfected cells. Abs to the p50 subunit of NF-κB shifted the middle complex, whereas Abs to the p65 subunit shifted both the top and middle complexes. Abs to c-Rel did not appreciably alter the NF-κB complexes induced following S. flexneri infection. Therefore, it appears that p65 homodimers and p50/p65 heterodimers are the main NF-κB complexes binding the κB site of the IL-8 promoter following S. flexneri infection.

The lack of c-Rel in the NF-κB complex binding to the IL-8 promoter was surprising, since previous studies have shown that, in addition to p65 and p50, c-Rel is a component of the IL-8 promoter-specific NF-κB complex (48). However, similar to the findings of Brasier et al. (49), Western blot analysis revealed increased levels of c-Rel in nuclear extracts from S. flexneri-infected cells, while in uninfected cells, c-Rel remained in the cytoplasmic fraction (data not shown). Thus, it is possible that c-Rel was also part of the complex, but was not detected in supershift assays, perhaps due to epitope masking.

We next examined whether S. flexneri infection of HeLa cells could activate an IL-8 reporter gene and what role NF-κB played in this activation. The 5′-flanking region of the IL-8 gene contains transcription factor binding sites for AP-1, NF-IL-6, and NF-κB located ∼120 bp upstream from the TATA box. When the wild-type construct was transfected into HeLa cells, which were subsequently infected with invasive S. flexneri for 5 h, an ∼3-fold activation of luciferase activity was observed compared with that in uninfected cells (Fig. 5,B). TNF-α treatment of these cells for 5 h resulted in a 6-fold activation of luciferase activity. Infection with the noninvasive plasmid-cured S. flexneri strain did not lead to an increase in luciferase activity (Fig. 5,B). To determine which of the transcription factor binding sites was essential for responsiveness to S. flexneri infection, HeLa cells were transfected with plasmid constructs that contained point mutations in the AP-1, NF-IL-6, or NF-κB binding sites. Point mutations in the AP-1 site did not affect the ability of S. flexneri infection or TNF-α treatment to activate the IL-8 reporter gene. Mutation of the NF-IL-6 site did not affect S. flexneri-induced activation of the reporter, but decreased TNF-α-induced activity by ∼50%, similar to previous studies (14). Mutation of the NF-κB site, however, blocked both S. flexneri- and TNF-α-induced activation of the IL-8 reporter, demonstrating that this site is indispensable for activation by both these inducers of IL-8 gene expression (Fig. 5 B).

To examine whether HeLa cells secreted IL-8 in response to S. flexneri infection, IL-8 ELISAs were performed on conditioned medium from HeLa cells infected with either wild-type S. flexneri or the plasmid-cured strain or treated with 100 ng/ml of TNF-α for various time periods (Table I). Small amounts of IL-8 were detected as early as 2 h postinfection with wild-type S. flexneri, and IL-8 levels increased to 325 ± 51 pg/ml following 6 h of infection. TNF-α treatment also led to increases in IL-8 protein levels (228 ± 27 pg/ml following 6 h of treatment). In contrast, infection with the noninvasive plasmid-cured strain of S. flexneri led to relatively small amounts of IL-8 secreted from the infected cells (34 ± 19 pg/ml after 6 h of infection). Transfection of cells with a nondegradable form of IKBα (IκBα-A32/36) before infection led to a 31% decrease in IL-8 protein levels secreted by S. flexneri-infected cells compared with infected cells transfected with vector alone. These data complement the above findings and stress the importance of NF-κB as a regulator of IL-8 gene expression during S. flexneri infection.

Possible bacterial factors capable of inducing NF-κB activity were analyzed by using mutant S. flexneri strains deficient in certain virulence factors, including IpaA, IpaB, IpaC, IpaD, IpgD, IcsA, and a S. flexneri strain cured of the virulence plasmid. Fig. 6,A shows that mutant strains of S. flexneri that are noninvasive are unable to stimulate NF-κB DNA binding activity. IpaA strains, which exhibit an entry efficiency ∼10% that of wild-type S. flexneri (50), still activated NF-κB. Likewise, invasive strains carrying a mutation in ipgD or icsA activated NF-κB DNA binding activity to similar extents as the wild-type organism (data not shown). The role of bacterial entry in the activation of NF-κB was examined more thoroughly using cytochalasin B. This inhibitor of S. flexneri entry caps actin filaments and prevents the dynamic cytoskeletal rearrangements that are necessary for S. flexneri invasion of epithelial cells (30). Cytochalasin B (1 μg/ml) added to the epithelial monolayer at the same time as S. flexneri infection prevented NF-κB activation (Fig. 6 B). Gentamicin protection assays revealed ∼200 CFU/monolayer treated with cytochalasin B, whereas without this inhibitor, 5 × 10⁶ CFU/monolayer was present. This same concentration of cytochalasin B had only a small effect on NF-κB activation by TNF-α. Taken together, these studies show a role for bacterial entry in NF-κB activation of epithelial cells by S. flexneri.

Since lack of invasiveness of IpaB, IpaC, and IpaD mutants precluded the ability to test these proteins for their ability to activate NF-κB in HeLa cells, microinjection of bacterial supernatants containing these secreted virulence proteins was performed. Bacteria-free supernatants from overnight cultures of wild-type M90T or the plasmid-cured strain BS176 were microinjected into individual HeLa cells and compared with cells microinjected with broth medium alone. Following subsequent staining with Abs to the p65 subunit of NF-κB, immunofluorescence revealed activated NF-κB in the nucleus of cells microinjected with supernatants from M90T (Fig. 7,A) and BS176 (data not shown). Cells microinjected with medium alone showed only cytoplasmic staining of NF-κB (Fig. 7 B). NF-κB-inducing activity in the supernatants of wild-type and plasmid-cured S. flexneri strains appeared to be common to Gram-negative organisms, because the microinjection of supernatants from the E. coli strain DH5α into HeLa cells resulted in localization of NF-κB in the nucleus of microinjected cells (data not shown). Supernatants from the Gram-positive organism B. subtilis were also tested in this system, but nuclear staining of NF-κB was only observed when these supernatants were concentrated ∼100-fold (data not shown). Addition of supernatants from S. flexneri, E. coli, or B. subtilis to the cell medium (1:2 ratio) did not activate NF-κB. To investigate whether the chemotactic bacterial peptide, fMLP, was responsible for this activation, supernatants from an E. coli strain deficient in the production of N-formylated proteins (34) were microinjected into HeLa cells. Nuclear staining of NF-κB was detected in cells microinjected with these supernatants, ruling out this possibility (data not shown).

Further characterization of the NF-κB-inducing factor in supernatants from M90T was conducted. Diluting the supernatants 1/4 in PBS significantly reduced NF-κB-inducing activity, and the activity was undetectable at a dilution of 1/16. The activity of supernatants from E. coli DH5α was higher than that in wild-type S. flexneri supernatants, in that loss of NF-κB-inducing activity was observed only after diluting the supernatants 1/256 with PBS. All subsequent characterizations of the NF-κB-inducing activity of wild-type S. flexneri supernatants were performed on samples diluted 1/4. The NF-κB-inducing activity of M90T supernatants was not affected by 1) boiling for 60 min, 2) treatment with proteinase K (150 μg/ml for 1 h at 60°C, overnight at 37°C, and subsequently boiled), or 3) DNase (1 μg/ml for 15 min at 37°C and subsequently boiled; data not shown). In addition, the NF-κB-inducing activity of the bacterial supernatants was stable for extended periods at 4°C. These observations suggested that LPS might be responsible for activating NF-κB. Immunodepletion of LPS from the bacteria-free supernatants with dimeric IgA Abs specific for S. flexneri 5a LPS resulted in a partial loss of NF-κB-inducing activity with <50% of microinjected cells showing localization of NF-κB in the nucleus (Fig. 7,C). M90T supernatants immunodepleted of LPS with IgA Abs specific for Salmonella LPS showed localization of NF-κB in the nucleus of microinjected cells (data not shown). Microinjection of purified LPS from different bacterial species was then conducted. Purified S. flexneri LPS from serotypes 5a (Fig. 7,D) and 1a, R. sphaeroides LPS (data not shown), and E. coli LPS from strain O111:B4 (Fig. 7 E) all activated NF-κB following microinjection. The NF-κB-inducing activity of S. flexneri LPS from both 5a and 1a as well as LPS derived from R. sphaeroides was weak, in that ∼50% of the microinjected cells showed nuclear localization of NF-κB. This was in contrast to cells microinjected with LPS from E. coli strain O111:B4, which consistently showed NF-κB activation (100% of cells). The insolubility of purified lipid A prevented us from microinjecting this product. However, microinjection of detoxified (i.e., delipidated) LPS derived from E. coli O111:B4 did not lead to activated NF-κB in the nucleus, suggesting that the lipid portion of LPS was likely to be responsible for this activation (data not shown).

To ensure that this response was not limited to HeLa epithelial cells, Caco-2 cells were grown on coverslips and microinjected with bacteria-free supernatants or purified LPS. Similar to what was observed in HeLa cells, the microinjection of bacteria-free supernatants from overnight cultures of M90T (data not shown) or the microinjection of purified LPS from E. coli O111:B4 (Fig. 7 F) led to the translocation of NF-κB into the nucleus. Cells microinjected with either medium or buffer alone did not show activated NF-κB in the nucleus (data not shown).

Epithelial cells are by and large refractory to the proinflammatory nature of bacteria and bacterial products, yet these cells can be provoked to respond to certain bacterial pathogens. One cellular response to bacterial infection is activation of the transcription factor, NF-κB. Previous studies have shown NF-κB activation in epithelial cells infected with S. flexneri (27) as well as other Gram-negative pathogens, including Salmonella (24), Neisseria (25), and EPEC (26). However, neither the pathway leading to NF-κB activation by these organisms nor the bacterial products involved have been clearly defined. In this study we characterized the NF-κB response in epithelial cells infected with S. flexneri and compared its activation with that induced by TNF-α. TNF-α treatment induced rapid and transient NF-κB DNA binding activity. Down-regulation of NF-κB following TNF-α treatment probably occurs because of three events. First, we found that the kinase activity of the IKK complex was rapidly down-regulated by 30 min post-TNF-α treatment, thus halting the activation cascade. Negative regulation of IKK activity was recently shown to be due to the phosphorylation of C-terminal serines on the IKK2 subunit (51). Second, because NF-κB induces the expression of IκBα (52), newly synthesized IκBα translocates to the nucleus, where it sequesters NF-κB and shuts down the activation of NF-κB-regulated gene expression (53). Indeed, we showed that IκBα reappears in the cytoplasm of TNF-α-treated cells by 2 h posttreatment and is thus available to down-regulate activation of NF-κB. Finally, down-regulation of NF-κB activation following TNF-α treatment is also attributed to endocytosis of ligand-bound receptors, thus preventing reactivation (54, 55).

In S. flexneri-infected cells, however, this negative regulatory loop was apparently missing. We found that S. flexneri infection resulted in sustained activity of the IKK complex, similar to what has been previously observed in human monocytic cells treated with LPS (56). Cytoplasmic levels of IκBα remained undetectable for the duration of the time course studied, and accordingly, NF-κB DNA binding activity was also sustained. The level at which S. flexneri infection disrupts the negative regulation of this cascade is not yet known. One possibility is that the NF-κB-inducing factor from S. flexneri is constantly presented to the cell, and this continually drives the activation of the cascade. Alternatively, S. flexneri infection could lead to inefficient phosphorylation of the negative regulatory domain of IKK2, resulting in its sustained activation. This inability of infected cells to down-regulate NF-κB activation may have important implications during infection in vivo; sustained NF-κB activity could cause inflammation to go unchecked, resulting in the severe damage to the epithelial layer seen during shigellosis.

NF-κB is an important transcriptional regulator of IL-8 (14), a proinflammatory chemokine produced by infected epithelial cells. The production of this chemokine appears to be part of an innate program initiated by bacterial infection (10, 11, 28, 57). The importance of IL-8 during S. flexneri infection was recently demonstrated in the rabbit ileal loop model of shigellosis (13). Induction of IL-8 by infected epithelial cells was shown to drive the massive inflammatory response and tissue destruction following S. flexneri infection. This response, although destructive to the host, is necessary to clear the organism from the infected tissue and prevent bacteriemia. In the present study S. flexneri infection was shown to activate an IL-8 promoter-driven reporter gene, and IL-8 protein levels were also shown to increase following infection. In addition, S. flexneri-induced IL-8 promoter activity relied on the NF-κB binding site within the promoter region. These findings stress the importance of NF-κB in regulating IL-8 induction during S. flexneri infection and implicate this transcription factor as a key regulator of the inflammatory process during infection in vivo.

The transcription factor NF-κB has been known for more than a decade; however, the complex signaling pathway leading to its activation is just beginning to be elucidated. The signaling pathways leading from TNF-α and IL-1 receptor activation to NF-κB have been best characterized. The recently described human Toll-like receptors (TLR) are members of the IL-1 receptor family of proteins, and signaling from these receptors appears to be identical with IL-1 signaling (58). Although the upstream components of activation are distinct, signaling to NF-κB by TNF-α and Toll/IL-1 involves activation of NIK and the IKK complex (19, 39). By using transient transfection to overexpress dominant negative versions of these proteins in HeLa cells we were able to examine whether S. flexneri infection could activate a NF-κB-dependent reporter. Like TNF-α and Toll/IL-1 signaling to NF-κB, activation of NF-κB by invasive S. flexneri required IKK1, IKK2, and NIK. However, our current studies suggest that the S. flexneri-induced signaling pathway upstream of NIK is potentially novel; TNF-α- or IL-1/TLR-specific signaling molecules, such as TRAF2 or MyD88 and IRAK1, do not play a role in NF-κB activation by S. flexneri (D. J. Philpott, X. Li, M. Mavris, G. Stark, and P. J. Sansonetti, manuscript in preparation).

Innate immune recognition of bacterial products is an ancient system of host defense that shares striking similarities in organisms as diverse as humans, fruit flies, and, to a certain extent, plants (reviewed in Ref. 59). It is not surprising, therefore, that the bacterial products that are recognized are invariant molecules including structural components such as LPS of Gram-negative organisms and peptidoglycan from the cell walls of Gram-positive organisms (60). More recently, bacterial lipoproteins have also been shown to trigger host defense mechanisms (61, 62). In both Drosophila and humans, a family of proteins termed Toll or TLRs is central to innate immunity. In mammals, innate recognition of microbial pathogens probably proceeds through a system by which CD14, one example of a pattern recognition receptor (PRR), detects LPS and by an unknown mechanism activates TLR4, thereby transducing the signal. A number of human TLRs have been identified in the database, leading to speculation that different TLRs may be involved in detecting different bacterial products to bring about distinct cellular responses.

Our findings show that epithelial cells do not respond to extracellular LPS, either purified or in the context of noninvasive S. flexneri strains. The Toll status of three intestinal epithelial cells lines was recently investigated, and although these cells express TLR4 and TLR2 in the case of T84 and Caco-2 cells, these cell lines are unresponsive to LPS at least in the absence of serum (63). On the other hand, our data suggest that LPS presented inside epithelial cells, either through the release of LPS by intracellular S. flexneri or by microinjection of the purified product, activates NF-κB. This response was not only observed in HeLa cells, but was also reproduced in the intestinal cell line Caco-2. It seems quite plausible, therefore, that a cytoplasmic PRR exists to detect intracellular LPS and that activation of this protein initiates a cascade leading to NF-κB activation.

During infection, S. flexneri enters the cell by a process resembling macropinocytosis, and once within the cell, lyses the membrane-bound vacuole. Free LPS is probably shed from the bacterium once inside the host cell (our unpublished observation) as is the case during S. typhimurium infection of HeLa cells (64). Because of its amphiphilic nature, free LPS is probably associated with membrane vesicles (64). It is possible that LPS reacts with the putative PRR within these vesicles or, alternatively, these LPS-containing vesicles could be trafficked to a specific compartment where this recognition takes place. The role of lipid trafficking in LPS activation of monocytes was recently examined. It was demonstrated that R. sphaeroides LPS, which is normally an LPS antagonist, becomes an LPS agonist upon treatment of cells with cationic membrane-active compounds (65). These compounds alter the packing geometry of R. sphaeroides LPS in the membrane and, in turn, its intracellular trafficking, which then correlates with the ability of this LPS to activate human monocytes. It was speculated that sorting of LPS and subsequent involvement in the activation of signaling cascades may be mediated by microbial pattern recognition receptors (65). Interestingly, we also found that R. sphaeroides LPS acts as an LPS agonist if microinjected into HeLa cells. Microinjection of R. sphaeroides LPS may allow it to be sorted in such a way that this LPS is then recognized and available to activate these cells.

Although we cannot rule out the possibility that other factors present in bacteria-free supernatants or associated with invasive S. flexneri activate NF-κB, several properties of the NF-κB-inducing factor strongly suggest that it is LPS. Firstly, microinjection of supernatants from both invasive and the plasmid-cured strains of S. flexneri activated NF-κB, ruling out the possibility that a plasmid-encoded virulence factor was responsible for this activation. Moreover, microinjection of supernatants from nonpathogenic E. coli were also able to induce NF-κB activation. Secondly, the NF-κB-inducing activity was insensitive to boiling and proteinase K and DNase treatment, suggesting that the factor is a lipid. The only treatment that significantly affected the NF-κB-inducing activity of the bacteria-free supernatants was immunodepletion of LPS. Finally, activation of NF-κB is a common response of epithelial cells to infection with a number of Gram-negative bacteria (24, 25, 26), including S. flexneri (27), implying that the responsible factor is probably a conserved component of this diverse set of pathogens. Bacterial lipoproteins also share these properties with LPS; however, its insolubility prevented us from testing this compound in microinjection experiments. We also found that microinjection of supernatants from the Gram-positive organism, B. subtilis, into HeLa cells activated NF-κB. However, positive activation was observed only after concentrating the supernatants ∼100-fold. Nevertheless, this finding raises the possibility that epithelial cells also possess a detection system for Gram-positive associated factors. Alternatively, the same receptor that detects intracellular LPS could bind Gram-positive peptidoglycan, as is the case for CD14 (66).

The different bacterial products tested by microinjection varied in their potency to induce NF-κB. Microinjection of bacteria-free supernatants invariably led to the translocation of NF-κB to the nucleus, whereas purified LPSs from different bacterial species varied in their potency to stimulate NF-κB. In terms of the bacterial supernatants tested, E. coli DH5α supernatants had the highest NF-κB-inducing activity, which correlated with the fact that the amount of released LPS in these supernatants was greater than that in S. flexneri supernatants as determined by silver-stained gels (data not shown). As for the purified LPSs tested, LPS from E. coli strain O111:B4 had the highest activity, which correlates with its high critical aggregation concentration and a tendency to form nonlamellar structures in solution (67). This property has been correlated with increased biological potency of LPS (68). What was surprising, however, was the fact that the amount of LPS in the purified samples was much greater than that observed in the bacterial supernatants, yet the supernatants exhibited higher NF-κB-inducing activity. This discrepancy in activity vs the concentration of LPS is possibly due to differences in the presentation of LPS in its purified form vs that released from bacteria. We speculate that LPS in bacteria-free supernatants may be in a form that is more available to interact with potential receptors because of its association with bacterial membrane or other bacterial surface components. The association of LPS with these factors is likely to inhibit lameller formation, and this may enhance the ability of LPS to interact with an intracellular receptor. In support of this idea, it was shown that LPS in bacterial supernatants is more biologically potent in stimulating monocytes than the same amount of purified LPS (69). Therefore, the biological context of LPS is extremely relevant and impacts on the potency of LPS released by S. flexneri during infection in vivo.

In conclusion, our studies of the mechanism of NF-κB activation by the invasive pathogen, S. flexneri, have suggested the presence of an intracellular pattern recognition receptor that detects LPS in the cytoplasm of infected epithelial cells. In colonic epithelial cells, such an intracellular receptor seems logical. It would be detrimental to the organism if colonic epithelial cells were capable of responding to extracellular LPS in the context of the normal resident flora. However, by triggering a response once a pathogen invades the cell, an intracellular PRR could initiate an inflammatory cascade in an attempt to control the threat of infection. These findings open up the possibility for a novel system of host defense against invasive enteric pathogens and may also impact on the potential of modulating the immune response to intestinal infections through this common mode of bacterial detection.

We thank Dr. Armelle Phalipon for reagents and advice, Drs. Michelle Rathman and Claude Parsot for reading the manuscript and for helpful discussions, and Dr. Richard Ferrero and Djilali Belaid for assistance with the IL-8 assays. We are also grateful to those individuals who generously provided bacterial strains, Abs, or plasmids.