Muramylpeptide shedding modulates cell sensing of Shigella flexneri

Authors

  • Giulia Nigro,

    1. Dipartimento di Biologia Cellulare e dello Sviluppo, Sapienza-Università di Roma, Via dei Sardi 70, 00185 Roma, Italy.
    Search for more papers by this author
  • Luigi Lembo Fazio,

    1. Dipartimento di Biologia Cellulare e dello Sviluppo, Sapienza-Università di Roma, Via dei Sardi 70, 00185 Roma, Italy.
    Search for more papers by this author
  • Maria Celeste Martino,

    1. Dipartimento di Biologia Cellulare e dello Sviluppo, Sapienza-Università di Roma, Via dei Sardi 70, 00185 Roma, Italy.
    Search for more papers by this author
  • Giacomo Rossi,

    1. Facoltà di Medicina Veterinaria, Università di Camerino, 62024 Matelica, MC, Italy.
    Search for more papers by this author
  • Ivan Tattoli,

    1. Dipartimento di Biologia Cellulare e dello Sviluppo, Sapienza-Università di Roma, Via dei Sardi 70, 00185 Roma, Italy.
    Search for more papers by this author
    • Present address: University of Toronto, Toronto, Canada.

  • Valeria Liparoti,

    1. Dipartimento di Chimica Organica e Biochimica, Università di Napoli Federico II, I-80126, Napoli, Italy.
    Search for more papers by this author
  • Cristina De Castro,

    1. Dipartimento di Chimica Organica e Biochimica, Università di Napoli Federico II, I-80126, Napoli, Italy.
    Search for more papers by this author
  • Antonio Molinaro,

    1. Dipartimento di Chimica Organica e Biochimica, Università di Napoli Federico II, I-80126, Napoli, Italy.
    Search for more papers by this author
  • Dana J. Philpott,

    1. Groupe d'Immunité Innée et Signalisation, Institut Pasteur, 75724 Paris Cedex 15, France.
    Search for more papers by this author
    • Present address: University of Toronto, Toronto, Canada.

  • Maria Lina Bernardini

    Corresponding author
    1. Dipartimento di Biologia Cellulare e dello Sviluppo, Sapienza-Università di Roma, Via dei Sardi 70, 00185 Roma, Italy.
    2. Istituto Pasteur-Fondazione Cenci Bolognetti, Sapienza-Università di Roma, Piazzale Aldo Moro 5, 00185 Roma, Italy.
    Search for more papers by this author

*E-mail marialina.bernardini@uniroma1.it; Tel. (+39) 6 49917850; Fax (+39) 6 49917594.

Summary

Bacterial infections trigger the activation of innate immunity through the interaction of pathogen-associated molecular patterns (PAMPs) with pattern recognition molecules (PRMs). The nucleotide-binding oligomerization domain (Nod) proteins are intracellular PRMs that recognize muramylpeptides contained in peptidoglycan (PGN) of bacteria. It is still unclear how Nod1 physically interacts with PGN, a structure internal to the Gram-negative bacterial envelope. To contribute to the understanding of this process, we demonstrate that, like Escherichia coli, Bordetella pertussis and Neisseria gonorrheae, the Gram-negative pathogen Shigella spontaneously releases PGN fragments and that this process can be increased by inactivating either ampG or mppA, genes involved in PGN recycling. Both Shigella mutants, but especially the strain carrying the mppA deletion, trigger Nod1-mediated NF-κB activation to a greater extent than the wild-type strain. Likewise, muramylpeptides spontaneously shed by Shigella are able per se to trigger a Nod1-mediated response consistent with the relative amount. Finally, we found that qualitative changes in muramylpeptide shedding can alter in vivo host responses to Shigella infection. Our findings support the idea that muramylpeptides released by pathogens during infection could modulate the immune response through Nod proteins and thereby influence the outcome of disease.

Introduction

A certain number of invariant bacterial structures, so-called pathogen-associated molecular patterns (PAMPs), are selectively recognized by the host via pattern recognition molecules (PRMs), like Toll-like receptors (TLRs) and nucleotide-binding oligomerization domain (Nod) molecules (Medzhitov, 2001; Inohara and Nunez, 2003). Upon interaction between a specific PAMP and its cognate PRM, the innate immune system is alerted essentially through NF-κB activation and subsequent cytokine production (Janssens and Beyaert, 2003).

Nod proteins are intracellular PRMs that recognize pathogens in the cytosol through sensing peptidoglycan (PGN) motifs. In particular, Nod1 recognizes the core dipeptide structure, γ-d-glutamyl-meso-diaminopimelic acid (iE-DAP), contained in the PGN of Gram-negative bacteria (Chamaillard et al., 2003; Girardin et al., 2003a). Nod2, on the other hand, recognizes muramyldipeptide (MurNAc-l-Ala-d-isoGln) present on PGN of both Gram-negative and Gram-positive bacteria (Girardin et al., 2003b; Inohara et al., 2003). Genetic variants of Nod proteins are associated with inflammatory disorders (Fritz et al., 2006), thus reinforcing the link between bacterial sensing and inflammation.

Dramatic colonic inflammation is the feature of shigellosis, a human infectious disease caused by the infection of as few as 100 bacteria of the enteroinvasive pathogen Shigella spp. (DuPont et al., 1989). Shigellae penetrate the baso-lateral pole of intestinal epithelial cells (IEC) through injection of Ipa proteins via a type III secretion system (TTSS) (Tran Van Nhieu et al., 2000). In IEC, shigellae stimulate NF-κB activation (Philpott et al., 2000) upon recognition of iE-DAP by Nod1 (Girardin et al., 2001; 2003a). Following this step, inflammation mounts via the production of pro-inflammatory cytokines, primarily IL-8 that stimulates polymorphonuclear leukocyte (PMN) recruitment (Philpott et al., 2000). Simultaneously, the TTSS secreted protein OspF acts to dephosphorylate mitogen-activated protein kinases in the host nucleus, thus preventing activation of a subset of NF-κB-responsive genes (Arbibe et al., 2007). Likewise, another TTSS effector, OspG, antagonizes degradation of the inhibitor IkBα by blocking its ubiquitinylation thus interfering with NF-κB activation (Kim et al., 2005).

Therefore, specific virulence factors and likely host cell responses mediated by PAMP–PRM interactions could account for colonic inflammatory response following infection with Shigella.

However, despite the great number of studies focused on the interaction of PRMs with PAMPs, it is still unclear how in epithelial cells, which are not normally committed to ingest and digest pathogens, Nod1 could physically interact in vivo with PGN, a structure internal to Gram-negative bacteria envelope.

Several studies have detailed that following PGN remodelling (Goodell and Schwarz, 1975; 1985; Cloud-Hansen et al., 2006) around 40–50% of PGN (Park, 1993) is released during each bacterial generation and approximately 90% of this material accumulates in the periplasm, from where it is re-imported into the cytoplasm for recycling (Goodell, 1985). This process leading to the release of minimal PGN products might contribute to pathogen recognition by Nod1 during natural infection.

In order to assess the biological role of these PGN derivatives we have rationally mutagenized Shigella flexneri 5 in two genes, ampG and mppA, in the attempt to impair PGN recycling and to augment the release of muropeptides by shigellae in host cells and tissues. Our aim was to analyse whether and how these released muramylpeptides might influence host cell sensing by PRMs and ultimately Shigella virulence.

Results

Muramylpeptide identification in sterile culture supernatants of M90T, M90T ΔampG and M90T ΔmppA

In Escherichia coli, AmpG is a transmembrane protein that acts as a specific permease for intact muropeptides (tri- or tetra-) (Lindquist et al., 1993; Cheng and Park, 2002) whereas the periplasmic binding protein MppA binds murein tripeptides and utilizes general oligopeptide permease (Opp) to transfer its bound ligand into the cytoplasm where tripeptides are recycled (Park et al., 1998).

The ampG and mppA genes were deleted in S. flexneri 5 wild-type strain, M90T, and in BS176 (Sansonetti et al., 1982), a non-invasive variant of M90T, lacking the virulence plasmid pWR100 (Buchrieser et al., 2000).

The mutants M90T ΔampG, M90T ΔmppA, BS176 ΔampG and BS176 ΔmppA had the same shape and grew at the same rate as the parental strains (data not shown).

We investigated if parental and mutated strains released PGN-related compounds in growth medium. Muropeptide fractions from sterile culture supernatants of wild-type strains and mutants were purified by gel permeation chromatography and reverse-phase high-pressure liquid chromatography (HPLC) as previously reported (Garcia-Bustos and Dougherty, 1987; Jacobs et al., 1994; Folkesson et al., 2005). The amount of muramylpeptides shed by the ampG and mppA mutants was 250 μg in 2.5 l of sterile culture medium, i.e. about 10-fold higher than that obtained by M90T or BS176. The purified muropeptide fractions were analysed and identified by mass spectrometry (MS).

The accumulated products in sterile culture supernatants of M90T ΔampG and BS176 ΔampG were all identified as muropeptides containing the MurNAc (N-acetylmuramyl) moiety (Fig. 1A) while those recovered from M90T ΔmppA and BS176 ΔmppA cultures were essentially composed by peptide molecules (tri- to penta-) lacking the carbohydrate moiety. In addition, at high molecular masses, the only MurNAc-containing species was anhydro-MurNAc-Ala-Glu-meso-DAP-Ala-Ala (Fig. 1B).

Figure 1.

Mass spectrum of muramylpeptides in sterile culture supernatants of Shigella mutants. Positive ion matrix-assisted laser desorption/ ionization time-of-flight (MALDI-TOF) mass spectrum of muramylpeptides present in sterile culture supernatants of M90T ΔampG and BS176 ΔampG (A) and M90T ΔmppA and BS176 ΔmppA (B). Each muropeptide species is represented by a pair of ion peaks differing by Δm/z 22 [M+ H and M+ Na]+ except for ion at m/z 413 [M+ Na]+ that lacks the ion species [M+ H]+ at m/z 393 (see Appendix S1). Abbreviation: An, anhydro; DAP, meso-diaminopimelate; MurNAc, N-acetylmuramyl; GlcNAc, N-acetylglucosamine.

The yield of muropeptide fractions in sterile culture supernatants of M90T and BS176 growth media was very low compared with that of the ampG and mppA mutants, and contained the same compounds as the ampG mutants. In fact, the MS spectrum of the parental strains, M90T and BS176, showed the same blend of muropeptides as the ampG mutants (details on MS profiles are provided in Appendix S1).

Nod1, Nod2 and TLR2 activation by the Shigella mutants

M90T ΔampG and M90T ΔmppA were first analysed in vitro to evaluate their ability to invade, proliferate and disseminate in HeLa cell monolayers. The results demonstrated that M90T ΔampG and M90T ΔmppA do not display any defect in invasion or intracellular proliferation or intra or intercellular spreading (Fig. S1). Likewise, the absence of MppA did not prevent M90T ΔmppA from growing with haemin as the iron source as recently shown for an E. coli double mutant harbouring the mppA and dppA deletions (Létofféet al., 2006) (data not shown).

As NF-κB activation in epithelial cells infected with Shigella relies on Nod1-PGN sensing, we next evaluated the ability of M90T ΔampG or M90T ΔmppA to induce NF-κB activation in HEK293 cell monolayers expressing ectopic Nod1 (Philpott et al., 2000; Girardin et al., 2003a). HEK293 cells were invaded with Shigella at multiplicity of infection (moi) of 50 and NF-κB activity and IL-8 production were evaluated after 3 and 5 h post infection (p.i.). After 3 h p.i. with M90T, M90T ΔampG or M90T ΔmppA NF-κB was activated 4-, 6- and 16-fold over the control cells expressing Nod1 respectively (Fig. 2A). These data largely remained unaltered after 5 h p.i. IL-8 secretion followed the same trend.

Figure 2.

NF-κB activity and IL-8 production of HEK293 cells expressing either Nod1 or Nod2 or TLR2 and infected with invasive or non-invasive strains.
A. HEK293 cells transfected with a plasmid expressing hNod1 or hNod1 in association with Nod1 psiRNA were infected either with the invasive strains M90T, M90T ΔampG and M90T ΔmppA, or with the non-invasive strains BS176, BS176 ΔampG and BS176 ΔmppA, for 3 h and 5 h p.i. NF-κB activation (left) and IL-8 production (right) were evaluated. At 3 h p.i. P = 0.01 for M90T ΔampG versus M90T and P < 0.0001 for M90T ΔmppA versus M90T.
B. HEK293 cells transfected with a plasmid expressing hNod2 or hNod2 in association with Nod2 psiRNA were infected as above. At 3 h p.i. P = 0.0045 for M90T ΔampG versus M90T and P = 0.0092 for M90T ΔmppA versus M90T. At 5 h p.i. P = 0.09 for M90T ΔampG versus M90T and P = 0.35 for M90T ΔmppA versus M90T.
C. HEK293 cells stably expressing hTLR2 and transiently transfected with a plasmid encoding hTLR1 or hTLR1 in association with TLR2 psiRNA were exposed during 5 h to M90T, M90T ΔampG and M90T ΔmppA killed by gentamicin (60 μg ml−1) as detailed in Experimental procedures. NF-κB activation (left) and IL-8 production (right) were evaluated. P = 0.965 for M90T ΔampG versus M90T and P = 1 for M90T ΔmppA versus M90T.
Data are expressed as mean ± SD of three independent experiments in triplicate. NF-κB values are reported as fold induction of relative luciferase units (RLU) over unstimulated cells. Relative luciferase activity is calculated as the ratio between the value of the NF-κB-inducible Firefly luciferase and that of the constitutive Renilla luciferase reporter. The asterisk indicates a NF-κB value extremely significant, P < 0.0001. Abbreviation: TriDAP, l-Ala-d-Glu-meso-DAP (1 μg ml−1); MDP, MurNAc-l-Ala-d-isoGln (1 μg ml−1); Pam3, Pam3CysSerLys4 (500 ng ml−1).

Transfection of HEK293 cell monolayers expressing ectopic Nod1 with a plasmid encoding small interfering RNA (siRNA) for Nod1 prevented the three strains from activating NF-κB, and eliciting IL-8 production.

To verify whether the activation of Nod1 was exclusively due to the presence of the intracellular mutants, we repeated the full experimental scheme with the non-invasive variants of M90T, i.e. BS176, BS176 ΔampG and BS176 ΔmppA. HEK293 cells were exposed to non-invasive bacteria for 3 and 5 h, as above. Neither NF-κB activation nor IL-8 production was observed under these conditions (Fig. 2A), confirming that intracellular location of bacteria was a prerequisite for activation of Nod1 by M90T and its variants.

We then evaluated whether the PRM Nod2 could be activated by parental strains and the invasive and non-invasive variants. We applied the same experimental plan described for Nod1 in HEK293 cells expressing Nod2. After 3 h p.i. M90T stimulated Nod2 around threefold over the control, while both M90T ΔampG and M90T ΔmppA stimulated Nod2 around fourfold more than the uninfected cells expressing Nod2. At 5 h p.i. all three strains induced a fivefold NF-κB activation with respect to uninfected cells. IL-8 values reflected this trend. As expected, in HEK293 cells transfected with a plasmid expressing siRNA for Nod2 NF-κB activation and IL-8 release was fully abrogated. HEK293 cells expressing Nod2 were then exposed to non-invasive bacteria BS176, BS176 ΔampG and BS176 ΔmppA for 3 and 5 h, as above. Neither NF-κB activation nor IL-8 production was detected. Results are shown in Fig. 2B.

The TLR2/TLR1 heterodimer recognizes triacyl lipoproteins contained in Gram-negative bacteria. To assess whether the inactivation of ampG and mppA in Shigella could affect the recognition of this pathogen by TLR2, we analysed the ability of M90T and its mutants to activate this PRM in HEK293 cells expressing TLR2. As TLR2 is a PRM located on cell membranes and not within the cytoplasm, like Nod1 and Nod2, HEK293 cell monolayers stably transfected with TLR2 and co-transfected with a plasmid expressing TLR1 were exposed to bacteria [2 × 107 colony-forming units (cfu) of each strain] killed with gentamicin and potentially releasing lipoproteins. Luciferase activity and IL-8 production were evaluated after 5 h of contact. We found that M90T, M90T ΔampG or M90T ΔmppA stimulated these PRMs around 23-, 30- and 32-fold over the control uninfected cells expressing TLR2/TLR1 respectively. IL-8 production followed the trend observed with NF-κB. In HEK293 cells expressing TLR2/TLR1 and depleted of TLR2 by siRNA interference, the activity of NF-κB and release of IL-8 were not observed (Fig. 2C).

Activation of Nod1, Nod2 by sterile supernatants of M90T and BS176 ampG and mppA mutants

To gain further insights into the role of muramylpeptides shed by the mutants on NF-κB activation, we used the sterile purified supernatants (SPS) of M90T, M90T ΔampG or M90T ΔmppA to measure NF-κB activity and IL-8 production in HEK293 cells expressing either Nod1 or Nod2, as above. Two and a half microlitres of SPS corresponding to 1 × 107 cfu were transfected into HEK293 cells as specified in Experimental procedures. NF-κB activity and IL-8 production were evaluated 12 h later.

As shown in Fig. 3A, SPS of M90T, M90T ΔampG and M90T ΔmppA stimulated Nod1 4-, 7- and 12-fold over the control uninfected cells expressing Nod1 respectively (P < 0.0001 for both M90T ΔampG and M90T ΔmppA versus M90T). IL-8 production confirmed this difference among the strains.

Figure 3.

NF-κB activity and IL-8 production in HEK293 cells expressing either Nod1 or Nod2 following stimulation of SPS of either M90T, M90T ΔampG and M90T ΔmppA or BS176, BS176 ΔampG and BS176 ΔmppA.
A. HEK293 cells transfected with a plasmid expressing hNod1 or hNod1 in association with Nod1 psiRNA were transfected with SPS of either M90T, M90T ΔampG and M90T ΔmppA or BS176, BS176 ΔampG and BS176 ΔmppA as specified in Experimental procedures. NF-κB activation (left) and IL-8 production (right) were evaluated after 12 h of stimulation. P < 0.0001 for M90T ΔampG and M90T ΔmppA versus M90T and BS176 ΔampG and BS176 ΔmppA versus BS176.
B. HEK293 cells transfected with a plasmid expressing hNod2 or hNod2 in association with Nod2 psiRNA were transfected with SPS as above and NF-κB activation (left) and IL-8 production (right) were evaluated after 12 h of stimulation. P = 0.016 for M90T ΔampG versus M90T, P = 0.019 for M90T ΔmppA versus M90T, P = 0.024 for BS176 ΔampG versus BS176 and P = 0.007 BS176 ΔmppA versus BS176.
Data are expressed as mean ± SD of three independent experiments in triplicate. NF-κB values are reported as in Fig. 2. The asterisk indicates a NF-κB value extremely significant, P < 0.0001. Abbreviations are as in Fig. 2; Medium: sterile M9 minimal medium.

SPS of BS176, BS176 ΔampG and BS176 ΔmppA gave results similar to those of M90T and its variants (Fig. 3A), excluding any possible implication of other molecules – such as those produced by the TTSS machinery – in the Nod1-mediated NF-κB activity and IL-8 production under our experimental conditions. In HEK293 depleted of Nod1, NF-κB activity and IL-8 release were not detected.

We next analysed the contribution of Nod2 to the NF-κB induction of the SPS of all mutants in HEK293 cells expressing Nod2. With SPS of M90T, Nod2 was not activated, while with SPS of M90T ΔampG and M90T ΔmppA Nod2 was stimulated around two- and threefold over the control uninfected cells expressing Nod2. IL-8 release was low under these conditions. As above, SPS of BS176, BS176 ΔampG and BS176 ΔmppA were analysed in parallel with supernatants of M90T and M90T variants. The results obtained were consistent with those obtained with M90T and its variants (Fig. 3B). In the presence of siRNA for Nod2 NF-κB activity was abolished with SPS of all strains tested.

Finally, we assessed whether the results obtained with SPS could be influenced by the presence of residual lipopolysaccharide (LPS) by evaluating NF-κB activity and IL-8 release in HEK293 cells expressing TLR4, MD2 and CD14. In this case SPS were added from ‘outside’ to HEK293 cells expressing this PRM and cofactors. NF-κB activity was evaluated after 5 h of contact. Neither activation of NF-κB nor IL-8 production was observed under these conditions. On the contrary Shigella LPS added from outside elicited both NF-κB activation and IL-8 production. Likewise, in HEK293 cells expressing TLR2/TLR1 the exposure of SPS of either of the strains elicited neither NF-κB stimulation nor IL-8 release in the presence of a positive response given by Pam3CSK4 used as a control (data not shown).

Activation of NF-κB by either M90T or BS176 mutants or SPS in HeLa cells

To verify the biological activity of these strains in conditions more close to those occurring in host tissues, we repeated the experimental scheme described above on HeLa cells naturally expressing the majority of PRMs (Uehara et al., 2007). Initially we measured NF-κB and IL-8 production upon either cell infection with M90T, M90T ΔampG and M90T ΔmppA (internal stimulation) or exposure to the non-invasive BS176, BS176 ΔampG and BS176 ΔmppA (external stimulation). After 3 h p.i. with the invasive strains NF-κB was activated around ninefold with M90T, 14-fold with M90T ΔampG and 10-fold with M90T ΔmppA with respect to uninfected cells (P < 0.0001 M90T ΔampG versus M90T). At 5 h p.i. the NF-κB values elicited by all strains were roughly double those observed at 3 h p.i. with these same strains (P < 0.0001 for M90T ΔampG versus M90T and P = 0.009 for M90T ΔmppA versus M90T). Production of IL-8 followed the same trend. On the contrary, exposure of HeLa cells to BS176, BS176 ΔampG and BS176 ΔmppA did not stimulate the activation of NF-κB nor IL-8 production. Under conditions of Nod1 depletion, NF-κB and IL-8 values were drastically reduced at 3 h p.i. and abrogated at 5 h p.i. (Fig. 4A).

Figure 4.

NF-κB activity and IL-8 production in HeLa cells upon either infection with invasive or non-invasive strains or transfection with SPS.
A. HeLa cells transfected with plasmid vector or with either hNod1 psiRNA or with hNod2 psiRNA or with hTLR2 psiRNA were infected with invasive strains M90T, M90T ΔampG and M90T ΔmppA or non-invasive strains BS176, BS176 ΔampG and BS176 ΔmppA during 3 h and 5 h p.i. NF-κB activation (left) and IL-8 production (right) were assessed. At 3 h p.i. P < 0.0001 for M90T ΔampG versus M90T and P = 0.105 for M90T ΔmppA versus M90T. At 5 h p.i. P = 0.0003 for M90T ΔampG versus M90T and P = 0.0086 for M90T ΔmppA versus M90T. TriDAP and MDP were transfected into cells as described and used as positive controls for Nod1 and Nod2 respectively. Pam3 were exposed to HeLa cells as specified and used as a control for TLR2.
B. HeLa cells transfected with plasmid vector or with either hNod1 psiRNA or hNod2 psiRNA or hTLR2 psiRNA were transfected with SPS of M90T, M90T ΔampG and M90T ΔmppA or of BS176, BS176 ΔampG and BS176 ΔmppA as detailed in Experimental procedures. NF-κB activation (left) and IL-8 production (right) were observed after 6 h of stimulation. P < 0.0001 for M90T ΔampG and M90T ΔmppA versus M90T and BS176 ΔampG and BS176 ΔmppA versus BS176. Positive controls were as above.
Data are expressed as mean ± SD of three independent experiments in triplicate. NF-κB values are reported as in Fig. 3. The asterisk indicates a NF-κB value extremely significant, P < 0.0001. Abbreviations are as in Fig. 2; Medium: sterile M9 minimal medium.

We proceeded to analyse the possible contribution of Nod2 to NF-κB activation in this cell line infected with M90T and its variants. Depletion of this PRM by Nod2 interference did not change significantly the results obtained with either of the strains in wild-type HeLa cells.

Several studies highlighted that antagonism and synergy among the different PRMs play a crucial role in determining final NF-κB activation (van Heel et al., 2005; Uehara et al., 2005). Therefore, we examined the role played by TLR2 in this cell line by depleting the cells of the expression of this PRM. At 3 h p.i. in HeLa cells infected with either of the strains NF-κB activity was unaffected. However, at 5 h p.i. under these conditions luciferase values were lower than those obtained in wild-type HeLa cells although this decrease was not statistically significant. We reasoned that external bacteria killed by gentamicin could contribute to NF-κB activation at late times of infection as shown in HEK293 cells expressing TLR2/TLR1. To analyse this hypothesis we repeated the same experimental scheme applied on HEK293 expressing TLR2/TLR1 in HeLa cells naturally expressing TLR2. Results confirmed that exposure of HeLa cells to shigellae killed by gentamicin could stimulate NF-κB activation and that TLR2 depletion significantly reduced this activity (Fig. S2).

Finally, we continued to analyse the biological activity of SPS in this cell line. We adapted to HeLa cells the procedure described for the experiments exploiting HEK293 cells. Briefly, SPS were transfected into HeLa cells and NF-κB activation and IL-8 production were measured after different times. We found that in this cell line maximal stimulation was achieved around after 6 h of stimulation whereas at 12 h no signal (including that given by PRM agonists) was observed. Therefore, we show only findings obtained after 6 h of stimulation.

M90T ΔampG SPS elicited NF-κB values ninefold higher with respect to controls while those of M90T ΔmppA and M90T resulted eight- and fourfold higher levels respectively (P < 0.0001 for both M90T ΔampG and M90T ΔmppA versus M90T and P = 0.07 M90T ΔampG versus M90T ΔmppA). IL-8 production followed the same trend. Likewise, NF-κB and IL-8 values obtained by transfection of SPS of BS176, BS176 ΔampG or BS176 ΔmppA reflected those seen with SPS of M90T and its mutants. Again, in the presence of Nod1 siRNA both NF-κB activation and IL-8 levels were significantly lower. In contrast to these results, the luciferase values remained roughly unaltered following silencing of either Nod2 or TLR2 (Fig. 4B).

In vivo virulence assessment

M90T, M90T ΔampG and M90T ΔmppA were assessed in mice by using two models of infection, the intranasal (i.n.) infection (Voino-Yasenetsky and Voino-Yasenetskaya, 1962) and the intravenous (i.v.) infection (Martino et al., 2005a). Following i.n. infection with a dose of 108 cfu (LD50) of Shigella mice develop a severe pneumonia and die. Likewise, after i.v. injection of a dose of 107 cfu (LD70) mice develop fulminant hepatitis and die. In the analysis of M90T, M90T ΔampG and M90T ΔmppA, survival of animals, bacterial counts in relevant organs, immunohistopathological features of the infected tissues and cytokine expression were examined.

While M90T ΔampG displayed virulence levels similar to those of M90T, M90T ΔmppA was markedly attenuated in both models. Following i.n. inoculation of 108 cfu (LD50) of M90T ΔmppA only 5% of the animals died. Likewise, only 8% of mice died after i.v. injection of 107 cfu (LD70) of this same strain. In accordance with mortality data, the number of cfu in the lung and liver, following i.n. and i.v. infection, respectively, was significantly reduced. As already described (Cersini et al., 2003), after 72 h in lungs of mice infected i.n. with the wild-type strain we found approximately 106 cfu; a similar value was calculated with M90T ΔampG while only 104 cfu were recovered in lungs infected with M90T ΔmppA.

At 48 h after i.v. infection with either M90T or M90T ΔampG the livers of infected animals contained approximately 106 cfu of Shigella (Martino et al., 2005a); this value was only 102 cfu in animals infected i.v. with M90T ΔmppA.

Following i.v. infection LD50 of M90T ΔmppA was 1 × 108 cfu (versus 5 × 106 with M90T) while when administered i.n. LD50 was 8 × 108 cfu. M90T ΔmppA harbouring the wild-type copy of mppA carried by pSTBlue-1-mppA was restored to full virulence in both infection models. Data are summarized in Table 1.

Table 1.  Assessment of virulence of M90T, M90T ΔampG and M90T ΔmppA in the pulmonary (i.n.) and systemic (i.v.) models of shigellosis.
StrainMurine pulmonary modelMurine systemic model
Mortality (%)acfu per lung (×104)bMortality (%)ccfu per liver (×104)dcfu per spleen (×104)
  • a. 

    n = 24 per strain. The animals were infected intranasally (i.n.) with a dose of 1 × 108 cfu and the LD50 was obtained at 72 h p.i.

  • b. 

    Surviving animals were sacrificed at 72 h p.i. and lungs were removed and tested for bacterial counts.

  • c. 

    n = 18 per strain. The animals were infected intravenously (i.v.) with a dose of 1 × 107 cfu and the LD70 was obtained at 48 h p.i.

  • d. 

    Surviving animals were sacrificed at 48 h p.i. and liver and spleen were removed and tested for bacterial counts.

  • e. 

    Colony-forming units (cfu) in either the lung or liver of animals infected with M90T versus cfu in the lung or liver of mice infected with M90T ΔmppA, P < 0.0001.

M90T50465.42 ± 3470873.15 ± 74.319.88 ± 0.34
M90T ΔampG47234.95 ± 1167970.02 ± 85.712.66 ± 0.63
M90T ΔmppA55.65 ± 0.3e80.02 ± 0.01e0.004 ± 0.002
M90T ΔmppA pSTblue-1-mppA56165.00 ± 72751247 ± 5435.76 ± 0.59

Histopathology, immunohistochemistry of and cytokine production in lungs of mice infected i.n.

Histopathological analysis was performed on tissue sections of lungs removed 72 h after i.n. infection. Alterations in pulmonary tissues were quantified and recorded as the mean of different scores related to inflammation as previously reported (Cersini et al., 2003; Martino et al., 2005b) and are summarized in Appendix S2.

Haematoxylin-eosin (HE)-stained lung sections were examined to assess the degree of: (i) thickening of alveolar septum, infiltration and diffusion of neutrophils in alveolar spaces, (ii) bronchiolar inflammation and epithelial damage and (iii) bronchiolar-associated lymphoid tissue (BALT) activation. Lung sections were also treated with monoclonal antibodies (mAbs) to evaluate the expression and distribution of IL-6 and Shigella LPS.

Wild-type strain and M90T ΔampG caused maximal lesions due to large areas of severe suppurative bronchopneumonia in which bronchioles were filled with neutrophils and sometimes with a mixture of cell debris, mucus and macrophages (Fig. 5A). This severe inflammation was accompanied with a strong expression of IL-6 and with intense LPS staining. In M90T-infected tissues, LPS was associated with inflammatory cells and free in the mucus while in lungs infected with M90T ΔampG, LPS was strongly stained in mononuclear cells (MNs) within BALT (Fig. 5A and Appendix S2). In lungs infected with M90T ΔmppA the tissue architecture was not significantly altered and septa and airways remained substantially unaffected. A strong reaction of BALT was observed accompanied with the development of characteristic perivascular cuffings. IL-6 expression was high and LPS staining moderate, especially located inside MNs associated with BALT (Fig. 5A and Appendix S2). In infected lungs, the production of IL-6 assessed at 12, 24 and 72 h p.i. was similar with all strains. In contrast, IFN-γ yield was significantly higher with M90T ΔmppA at all time points (Fig. 5B).

Figure 5.

Lung and liver reactions following intranasal infection or intravenous infection with M90T, M90T ΔampG or M90T ΔmppA.
A. Histochemical and immunohistochemical (IH) analysis of lungs infected with M90T, M90T ΔampG or M90T ΔmppA. The animals were sacrificed at 72 h p.i. From top to bottom: HE staining of infected lung sections and avidin-biotin immunoperoxidase labelling of IL-6, and S. flexneri 5 LPS.
HE: In uninfected control the absence of inflammatory infiltrates and BALT activation are evident; samples infected with either M90T or M90T ΔampG show a suppurative bronchopneumonia and a poor BALT activation; with M90T ΔmppA alveolar spaces and airways are free from inflammatory cells while areas of BALT activation are associated with mild phlogosis. Original magnification ×200, bars 50 μm.
IL-6: In sections of lungs infected with either M90T or M90T ΔampG the amount and distribution of cells expressing IL-6 are similar. With M90T ΔmppA IL-6 production is mainly localized in MNs of BALT. Arrows point to areas showing IL-6 staining. Original magnification ×200, bars 50 μm.
Shigella flexneri 5 LPS: With M90T LPS is mainly present extracellularly in bronchial lumen (arrows); with M90T ΔampG LPS is found in some MNs (arrowheads) and free in bronchial lumen (arrows); with M90T ΔmppA LPS immunostaining is weak and restricted to MNs (arrowheads). Original magnification ×1000, bars 250 μm.
B. Production of IL-6 and IFN-γ in lungs of animals infected with M90T, M90T ΔampG and M90T ΔmppA (n = 3 per group). Uninfected controls represent four mice. Lungs were removed at 12, 24 and 48 h p.i., homogenized in PBS and supernatants served as samples for IL-6 ELISA. The asterisk indicates a statistically significant difference in value compared with wild-type strain (P < 0.005).
C. Histochemical and IH analysis of livers infected with M90T, M90T ΔampG or M90T ΔmppA. The mice infected i.v. were sacrificed 48 h p.i. From top to bottom: HE staining, avidin-biotin immunoperoxidase labelling of S. flexneri 5 LPS, colorimetric TUNEL analysis of infected liver section.
HE: With M90T a diffuse hepatocyte swelling with necrotic areas is accompanied with the presence of a microgranuloma (dashed line); with M90T ΔampG a microgranuloma is observable (dashed line) surrounded by a degenerate parenchyma; in the M90T ΔmppA-infected sample an extended microgranuloma (dashed line) displays a large central area of necrosis associated with the absence of hepatocyte degeneration.
Shigella flexneri 5 LPS immunostaining: With either M90T or M90T ΔampG, S. flexneri LPS is located extracellularly in a large area in the central portion of a microgranuloma (arrowheads) or in a few MNs outside this area (arrows); with M90T ΔmppA LPS is exclusively localized in the central area of a microgranuloma (arrowheads), associated with necrotic material or expressed inside a certain number of MNs.
TUNEL colorimetric assay: Some TUNEL-positive cells are present in sections of the liver of the uninfected control mouse; in samples infected with either M90T or M90T ΔampG, a great number of TUNEL-positive nuclei of hepatocytes and lymphoid cells within (arrowheads) and outside (arrows) the microgranulomatous areas confirms a high apoptotic index; with M90T ΔmppA a small number of TUNEL-positive cells are present inside (arrowheads) as well as outside (arrows) the microgranulomatous area.
Original magnification of all panels ×400, bars 100 μm.

Histopathology and immunohistochemistry of livers of mice infected i.v.

Histopathological analysis of tissue sections of livers was performed on mice sacrificed 48 h after i.v. infection as already reported (Martino et al., 2005a). Livers infected i.v. with M90T showed severe degeneration and cloudy swelling of hepatocytes (Fig. 5C). Necrotic areas were associated with the presence of microgranulomas. These aggregates, consisting of focal leukocyte reaction to hepatic cell death, showed a central area of mild necrosis. Immunohistochemical analysis revealed that LPS of M90T and M90T ΔampG seemed to be associated with some MNs within microgranulomas as well as present in the extracellular areas (Fig. 5C). In contrast with these results, the liver of M90T ΔmppA-infected animals did not display degenerative alterations; nevertheless, large and well-defined microgranulomatous lesions were present with an enlarged necrotic central area, which were strongly stained by the anti-Shigella LPS mAb (Fig. 5C). Apoptotic cells are a hallmark of Shigella infection in the liver and apoptosis plays a crucial role in worsening the lesions in the hepatic tissue (Martino et al., 2005a,b). With M90T, a great number of apoptotic cells were observed within microgranulomas and throughout the hepatic parenchyma, while with M90T ΔmppA the number of apoptotic cells was low within microgranulomas and in the surrounding areas (Fig. 5C and Appendix S2).

Tissue alterations were quantified as previously reported (Cersini et al., 2003; Martino et al., 2005a) and summarized in Appendix S2.

Discussion

In this study we analysed the contribution of PGN sensing to the overall cellular response to Shigella infection. First, we investigated whether Shigella releases PGN fragments during PGN remodelling as originally shown for E. coli (Goodell and Schwarz, 1985) and for two human pathogens, Bordetella pertussis (Goldman et al., 1982; Cookson et al., 1989) and Neisseria gonorrhoeae (Sinha and Rosenthal, 1980). Then, we evaluated whether and how the PGN derivatives shed by Shigella could stimulate the innate immune response. Finally, we analysed whether Shigella virulence could be modulated by altering the release of these PGN fragments.

Our results demonstrate that Shigella spontaneously releases PGN fragments and that this process can be greatly increased by inactivating either ampG or mppA genes, both involved in PGN recycling, as first shown by Jacobs through the deletion of ampG in E. coli (Jacobs et al., 1994; Cloud-Hansen et al., 2006). We found that the increase of muramylpeptide shedding corresponds to significant Nod1 activation in epithelial cells. In fact, the Shigella ampG and mppA mutants trigger Nod1-mediated NF-κB activation to a greater extent than the wild-type strain. Likewise, SPS added from ‘inside’ to host cells are able to stimulate Nod1 to varying degrees, depending on their relative amount and composition. In particular, muramylpeptides shed by M90T ΔmppA whose composition differs significantly from that of M90T and M90T ΔampG were especially efficient in mediating Nod1 stimulation in HEK293 cells expressing ectopic Nod1.

In contrast, Nod2 is not activated by the PGN mixture released by wild-type bacteria and is only weakly activated by muramylpeptides released by either of the mutants.

Biological activity of muramylpeptides has been already reported in the context of N. gonorrhoeae (Sinha and Rosenthal, 1980) and B. pertussis pathogenesis (Goldman et al., 1982) as the PGN fragments released by these microorganisms cause damage to ciliated cells similar to the effects provoked by these pathogens during natural infection (Melly et al., 1984; Flak and Goldman, 1999). Furthermore, in a recent study Hasegawa et al. (2006) demonstrated that a higher level of Nod1-stimulatory activity was found in culture supernatants of various bacteria and not in extracts from whole bacterial cells. Our findings together with these observations support the idea that muramylpeptides released by pathogens during natural infection could modulate the immune response through qualitative and quantitative differences in Nod protein activation.

In HeLa cells infected with Shigella, Nod1 drives bacterial recognition (Girardin et al., 2001). Our data confirm this as, in this cell line, the absence of Nod1 dramatically decreases the activation of NF-κB, especially at late times of infection. In contrast, depletion of Nod2 does not alter the ability of Shigella to activate NF-κB, while depletion of TLR2 slightly reduces the NF-κB activation, suggesting a potential involvement of this PRM to the response of host tissues to Shigella.

In contrast with results obtained in HEK293 cultures, in HeLa cells infected with either M90T ΔampG and M90T ΔmppA, NF-κB was roughly equally stimulated. These differences between the two cell lines might be explained by the fact that NF-κB activation in treated HEK293 cells depends on Nod1 overexpression while in HeLa cells, it is the endogenous Nod1 that triggers the response. Ectopic expression of Nod1 could emphasize differences in Nod1 sensing between the two mutants that would be otherwise not detectable. Furthermore, differences in PRM expression between the two cell lines could influence the signal transduction as reported for TLR (Kurt-Jones et al., 2004). Finally, in HeLa cells, living shigellae might activate other not yet identified PRMs thus affecting the immune response initially triggered by Nod1.

In vivo, the virulence capacity of ShigellaΔampG was comparable to that of the wild-type strain, which is in accordance with the absence of virulence attenuation for a Salmonella enterica serovar Typhimurium ΔampG strain (Folkesson et al., 2005). In contrast to this result, ShigellaΔmppA was strongly attenuated. In lungs of mice infected with M90T ΔmppA, the IL-6 production was comparable to that induced by wild-type bacteria while the level of IFN-γ was higher. IFN-γ could contribute to pathogen clearance as shown during natural and experimental shigellosis (Way et al., 1998). Similarly, the liver of animals infected i.v. with M90T ΔmppA contained larger microgranulomas (Martino et al., 2005a,b), than those induced by M90T, in which a significant presence of IFN-γ expressing cells might facilitate Shigella clearance. We are currently investigating the molecular mechanisms underlying the attenuation of M90T ΔmppA and how the reduction of virulence could rely on the peculiar ability of this strain and its supernatant to activate Nod1. An attractive hypothesis could be that M90T ΔmppA is particularly able to elicit the Nod1-dependent activation of antibacterial factors like antimicrobial peptides (AMPs) in the liver and in the respiratory tract in early time of infection. Several bacterial factors including muramyl dipeptide have been demonstrated to stimulate AMPs expression and secretion (Ayabe et al., 2000).

Recent studies have stressed the role of bacterial PGN composition in immune evasion of various pathogens (Chaput et al., 2006; 2007; Boneca et al., 2007) and have unveiled bacterial strategies to modulate the PGN structure or to inject muramylpeptides into host cells (Bartoleschi et al., 2002; Viala et al., 2004). Here we show that the immunopotential of Shigella is significantly influenced by quantitative and qualitative alterations in muramylpeptide shedding, thus suggesting that the regulation of PGN release could represent a further sophisticated bacterial strategy to survive in host tissues and to evade immune defences.

Experimental procedures

Bacterial strains, plasmids and growth conditions

The S. flexneri 5 strains used are M90T streptomycin-resistant (SmR) and its non-invasive derivative BS176 (Martino et al., 2005b). Bacteria were grown in Trypticase soy broth (TSB) (BBL, Becton Dickinson and Co., Cockeysville, MD) or agar (TSA). Minimal medium was prepared with M9 salts, glucose (0.2%) and nicotinic acid (10 μg ml−1). When necessary, kanamycin, ampicillin, streptomycin or chloramphenicol was added to cultures at 50, 100, 100 or 20 μg ml−1 respectively.

Bacterial strains' construction

DNA manipulations were carried out by standard methods following manufacturer's suggestions. To obtain M90T ΔampG two DNA fragments from M90T ampG were amplified by PCR with the following primers: ampGF1 (5′-GCTCTAGAGTTCAGCCATATTGCTGATCCTGG-3′) and ampGR1 (5′-TGCACTGCAGTTCCAGCGTTTTGGGCACAGG-3′) for fragment 1 (575 bp); ampGF2 (5′-TGCACTGCAGTCGGGTTTGATGCGGGTGAAGTA-3′) and ampGR2 (5′-CGGAATTCATCCAGCAAACCACCAAGCACGAC-3′) for fragment 2 (636 bp). The two fragments were ligated with a 829 bp PstI cassette encoding chloramphenicol resistance from pGEM-CAT (a gift from A. Covacci, Chiron Vaccines, Siena, Italy), cloned into pGP704 and introduced into M90T and BS176 to replace the wild-type ampG through allelic exchange. Likewise, to yield M90T ΔmppA and BS176 ΔmppA two DNA fragments, of 684 bp and 498 bp from M90T mppA, were amplified with the primers: mppAF1 (5′-GTCCTAGAGGTGCGCCACATAAAGATGAG CC-3′), mppAR1 (5′-TGCACTGCAGAATATCCTTCAACAGCTTCTG-3′), mppAF2 (5′-TGCACTGCAGTTACGCCGGAACCTTCGCCGTTTG-3′) and mppAR2 (5′-CGGAATTCCGCCACATCTTCAGGATTATTAATG-3′), ligated with a 1.2 kb PstI cassette encoding kanamycin resistance from pUK4K and then used to replace the wild-type mppA in M90T and BS176 (SF1841 in S. flexneri 2a 301 genome corresponds to E. coli mppA). The entire M90T ampG gene was amplified by using the following primers: ampGCF (5′-ATG TCCAGTCAATATTTACG-3′) and ampGCR (5′-TACGTCAGATGCGTTTTTCG-3′). The 1475 bp fragment was then cloned into pSTBlue-1 (Perfectly BluntTM Cloning kit, Novagen).

Likewise, a sequence (2034 bp) containing the entire M90T mppA gene was cloned into pSTBlue-1 by using the following primers: mppACF (5′-ACTGGTTTGAACGTGCGAAG-3′) and mppACR (5′-TACCAAAACCACGGTATGGG- 3′).

Analysis of muramylpeptides in sterile bacterial culture supernatants

In order to analyse muramylpeptides present in sterile culture supernatants M90T, BS176, M90T ΔampG, M90T ΔmppA, BS176 ΔampG and BS176 ΔmppA were cultured overnight in M9 medium up to OD600 = 1.5. Sterile culture medium (2.5 l) of M90T, M90T ΔampG and M90T ΔmppA was then lyophilized, redissolved in water, extensively desalted by successive gel filtration chromatography, reduced by sodium borohydride, purified by HPLC and analysed by MS (Garcia-Bustos and Dougherty, 1987; Jacobs et al., 1994; Folkesson et al., 2005). All supernatants of bacterial cultures (20 ml), lyophilized and used in in vitro assays were formerly analysed for contamination by either lipoprotein or LPS by Gas-Chromatography/Mass Spectrometry of fatty acids (Wollenweber and Rietschel, 1990).

Cell cultures and infections

HeLa and HEK293 cells were maintained in D-MEM (Cambrex Bio Science, Walkersville) supplemented with 10% FBS (Cambrex Bio Science, Walkersville). HEK293 huTLR2 (gDhTLR2), kind gift of Arturo Zychlinsky, were grown in complete medium as above supplemented with geneticin (Gibco BRL) 500 μg ml−1. Stable transfected cell line 293-hTLR4/MD2-CD14 (InvivoGen) were cultured in DMEM with 10% FBS and 10 μg ml−1 Blasticidin-S (InvivoGen) and 50 μg ml−1 HygroGold® (InvivoGen).

HEK293 and HeLa cell infections with invasive and non-invasive strains were performed as previously reported (Philpott et al., 2000; Cersini et al., 2003). Incubation of the infected cells in the presence of gentamicin was prolonged to 3 or 5 h. To evaluate TLR2 activation bacterial cultures (2 × 107 cfu) were added with gentamicin (60 μg ml−1) and then exposed to HEK293 cells expressing TLR2/TLR1 as specified above.

Transfections

HEK293 and HeLa cells were transiently transfected with PolyFect Transfection Reagent (Qiagen) according to the manufacturer's instructions. Briefly, the cells were transfected overnight with a reaction mix composed by 5 μl of Polyfect Trasfection Reagent, 25 ng of Firefly luciferase reporter constructs, pGL3.ELAM.tk, and 2.5 ng of Renilla luciferase reporter plasmid, pRLTK. Additionally, HEK293 cells were transfected with 250 ng of hNod1 (pcDNA3-Nod1-FLAG) or hNod2 (pUNO-hNOD2a – InvivoGen) while HEK293 huTLR2 were transfected overnight with 250 ng of hTLR1 (pUNO-hTLR1 – InvivoGen). Vector plasmids (pcDNA3) were used as controls in all transfection experiments.

Stimulation with sterile purified supernatants

The day before transfection, the cells were seeded into 24-well plates to obtain a semi-confluent monolayer. At the same time of the transfection 2.5 μl of SPS corresponding to 1 × 107 cfu were added to the cell culture medium. NF-κB-dependent luciferase activation and IL-8 production were measured the following day for HEK293 cells and 6 h after transfection for HeLa cells.

Luciferase assays

The cells were transiently transfected in 24-well plates. Luciferase assays were carried out using the Dual Luciferase kit (Promega). Firefly luciferase reporter constructs, containing five NF-κB elements at the promoter region, were transfected together with the Renilla luciferase reporter plasmid as an internal control. Cells were lysed in 100 μl of passive lysis buffer, and 20 μl of lysate was assayed for Firefly and Renilla activity according to the manufacturer's instructions. The data shown represent the mean values ±SD of three separate experiments performed in triplicate. Results are reported as fold induction of relative luciferase units (RLU) over unstimulated cells. Relative luciferase activity is calculated as the ratio between the value of the NF-κB-inducible Firefly luciferase and that of the constitutive Renilla luciferase reporter.

γTriDAP (l-Ala-d-Glu-meso-DAP), chemically synthesized by AnaSpec (San Jose, CA), MDP (MurNAc-l-Ala-d-isoGln) and Pam3CSK4 (N-Palmitoyl-S-[2,3- bis(palmitoyloxy)-(2RS)-propyl]-[R]-cysteinyl-[S]-seryl-[S]-lysyl-[S]-lysyl-[S]-lysyl-[S]-lysine ×3 HCl), purchased from InvivoGen, were used as positive controls for Nod1, Nod2 and TLR2/TLR1 activation respectively.

siRNA assays

siRNA inhibition of Nod1, Nod2 or TLR2 was carried out by using, respectively, psiRNAh-Nod1, psiRNAh-Nod2 and psiRNAh-TLR2 constructs, all from InvivoGen.

For HEK293 cells, 25 ng of plasmid expressing psiRNA or vector control (containing a scramble sequence) was transfected together the plasmid expressing the corresponding PRM.

Nod1, Nod2 and TLR2 depletion was verified by Western blot analysis with mouse mAbs anti-DYKDDDK (Clone 2EL-1B11, Chemicon International), anti-Nod2 (2D9, Cayman Chemical Company, Ann Arbor, MI) and anti-hTLR2 (eBiosciences, cloneTL2.1) respectively. A monoclonal anti-β-tubulin (Clone TUB 2.1, Sigma) antibody was used as a control.

In HeLa cells siRNA inhibition of Nod1, Nod2 and TLR2 was performed as follows: cells were first transfected with the siRNA expressing plasmid using the procedure described above. Seventy-two hours after the first transfection the cells were re-transfected with pGL3.ELAM.tk and pRLTK to evaluate NF-κB activation and IL-8 production as detailed above.

In HeLa cells Nod1 inhibition was evaluated through reverse transcription polymerase chain reaction (RT-PCR) and Real-Time as previously described (Welter-Stahl et al., 2006). Nod2 and TLR2 depletion was analysed as described for HEK293 cells.

Intranasal infection of mice

Five-week-old Balb/C female mice (Charles River, Italy) were infected i.n. with a dose of 108 cfu. After 72 h mice were sacrificed and lungs were removed and processed for histopathological studies, bacterial counts and RT-PCR analysis as reported (Cersini et al., 2003). Alternatively, the mice were kept under survey during 7 days after infection.

For bacterial counts and enzyme-linked immunosorbent assay (ELISA) studies the samples were treated as described (n = 8 for each condition). Each experiment was repeated at least three times.

Intravenous infection of mice

Mice were challenged by injection of the caudal vein with 200 μl of sterile saline solution (SSS) of bacterial suspension, containing 107 cfu (LD70) as already described (Martino et al., 2005a). After 48 h the surviving animals were sacrificed and liver and spleen removed and prepared for histopathology and counts of viable bacteria, as reported. Alternatively, the mice otherwise infected were kept under survey during 7 days after infection.

Uninfected mice having received 200 μl of saline solution through the caudal vein were used as controls in each experiment (n = 8 per experimental group). Each experiment was repeated at least three times.

Histopathology

For morphological and immunohistochemical (IH) examination, tissue samples were fixed in 10% buffered formalin for 48 h, processed and embedded in paraffin. Sections of 3 μm thick were stained with HE or processed for IH. For IH, tissue's sections were treated as described (Martino et al., 2005a,b). Primary antibodies are: anti-S. flexneri 5a LPS (dimeric IgA, 6 mg ml−1) (Martino et al., 2005a) and anti-mouse mAb IL-6 (Serotec, Oxford, UK).

A microgranuloma was defined as a well-circumscribed cell aggregation composed of five or more mononuclear phagocytes. Pro-apoptotic effect of S. flexneri was highlighted through a TUNEL colorimetric staining (DeadEnd, Promega) according to the manufacturer's instructions.

ELISA

Homogenized lungs were microfuged at top speed for 5 min and the supernatants were used for cytokine analysis. Murine IL-6 and IFN-γ levels in tissues were assayed by using ELISA kits (R&D Systems, Minneapolis, USA) according to the manufacturer's instructions. Human IL-8 levels in cell culture supernatants were measured using ELISA kits (R&D Systems, Minneapolis, USA).

Statistical analysis

Two-tailed t-test was used to compare cytokine titres and luciferase data. Mortality experiments were assessed using Fisher's exact test.

Data were presented as mean ± SD, and the numbers of independent experiments were indicated in each legend of the figures. A P-value < 0.05 was considered statistically significant. A P-value < 0.0001 was considered extremely significant.

Acknowledgements

We thank M. Rescigno for helpful discussion and technical support, C. Tang for critical reading of the manuscript, I. Martini, G. Levi and M. Esposito for technical help, G. Bertini, P. Piccoli and G. Cortese for care of the mice. This work was supported by grants from the Italian Ministero dell'Istruzione, Università e Ricerca (PRIN 2006) and WHO (V27/181/184).

Ancillary