Flagellin is the structural component of flagella produced by many pathogenic bacteria and is a potent proinflammatory molecule that mediates these effects through Toll-like receptor (TLR) 5. In Listeria monocytogenes (LM), flagellin expression is regulated by temperature and has been described as being shut off at 37°C. In this study, we demonstrate that TLR5-mediated cell activation and flagellin expression is maintained at 37°C in some laboratory-adapted strains and in ≈ 20% of LM clinical isolates. To determine the role of flagellin in LM infection, a targeted mutation in the structural gene for flagellin (flaA) was generated in a parental LM strain that expressed flagellin under all conditions examined. In vitro studies demonstrated that this ΔflaA mutant was (i) non-motile; (ii) not able to activate TLR5-transfected HeLa cells; and (iii) induced tumour necrosis factor (TNF)-α production in ≈ 50% fewer CD11b+ cells in splenocytes from normal mice compared with the parental strain. However, there was no significant alteration in virulence of the ΔflaA mutant after either intravenous or oral murine infection. Similarly, there was no difference in the generation of LM-specific CD8 or CD4 T cells after intravenous or oral infection. These data indicate that flagellin is not essential for LM pathogenesis or for the induction of LM-specific adaptive immune responses in normal mice.
Flagella are complex bacterial organelles that are well conserved among diverse bacterial species. In addition to their role in allowing bacteria to migrate towards favourable and away from unfavourable environments, flagella contribute to virulence for many pathogenic bacteria (Carsiotis et al., 1984; Feldman et al., 1998). In vitro studies suggest that these effects on virulence may be attributable to either increased host cell invasion (Liu et al., 1988) or increased survival within macrophages (Weinstein et al., 1984). Bacterial flagella are composed of monomeric flagellin subunits and have potent proinflammatory activity (McDermott et al., 2000; Eaves-Pyles et al., 2001a,b; Gerwitz et al., 2001a). Purified flagellin, when injected into mice at low doses, causes systemic inflammation characterized by the production of inflammatory cytokines, chemokines and nitric oxide (Eaves-Pyles et al., 2001a). At higher doses, purified flagellin induces hypotension, reduced vascular contractility and death. As a result of these proinflammatory effects, exogenous flagellin is an effective in vivo adjuvant in the generation of pathogen-specific adaptive T-cell immune responses (McSorley et al., 2002).
All TLRs identified to date use the signalling molecule MyD88 to transduce activation signals leading to proinflammatory cytokine secretion. Although TLR3 and TLR4 can also use MyD88-independent pathways for cell activation, the response is delayed and cytokine production is less robust in the absence of MyD88 (Kawai et al., 1999; Alexopoulou et al., 2001; Kaisho et al., 2001). Recently, an essential role for MyD88 in the early phase of Listeria monocytogenes (LM) infection has been demonstrated (Edelson and Unanue, 2002; Seki et al., 2002; Way et al., 2003). The lethal dose of LM for MyD88-deficient mice is decreased at least 3 log10 (Seki et al., 2002) and, by day 3 after infection, MyD88-deficient mice had ≥3 log10 more LM colony-forming units (cfus) in their spleens and livers compared with control mice (Edelson and Unanue, 2002; Way et al., 2003). The extreme susceptibility observed for MyD88-deficient mice is not the result of defective interleukin (IL)-1 or IL-18 signalling, as caspase-1-deficient mice (that cannot produce biologically active forms of IL-1 and IL-18, which also signal through MyD88) showed only a modest increase in susceptibility to LM infection (Edelson and Unanue, 2002). These findings suggest a role for TLR recognition of LM PAMPs in innate immunity to LM infection.
LM peptidoglycan activates cells through TLR2 (Flo et al., 2000). However, in contrast to the marked susceptibility of MyD88-deficient mice to LM infection, TLR2-deficient mice do not differ from controls in susceptibility to LM (Edelson and Unanue, 2002). This suggests that TLRs other than TLR2 participate in innate immunity to LM. As LM flagellin induces inflammation through TLR5-mediated cell activation (Hayashi et al., 2001), is differentially expressed at environmental and mammalian body temperatures (Peel et al., 1988; Dons et al., 1992), and as flagella are known virulence determinants for other pathogenic bacteria (Carsiotis et al., 1984; Feldman et al., 1998), we sought to determine the role of LM flagellin in the activation of innate immunity and pathogenesis after infection.
Results and Discussion
TLR5-mediated cell activation by LM
Previous studies have demonstrated that the expression of flagellin in LM is temperature dependent. Flagella are present, as detected by electron microscopy, and the structural gene for flagellin, flaA, is transcribed maximally after growth at 22°C or below, but expression is repressed at 37°C (Peel et al., 1988; Dons et al., 1992). Downregulation of flagellin by LM at 37°C may represent a mechanism by which the bacterium attempts to evade TLR5-mediated host immunity. To test this hypothesis, we examined the temperature-dependent regulation of flagellin expression in a larger panel of LM wild-type strains. Six well-characterized laboratory-adapted strains of LM were tested for the ability to activate TLR5-transfected HeLa cells after growth at 22°C or 37°C (Table 1, Fig. 1A). Although all LM strains tested could activate cells through TLR5 after growth at 22°C, surprisingly, one LM strain (10403s) retained the ability to activate cells through TLR5 after growth at 37°C.
Table 1. . TLR5-mediated cell activation for LM laboratory-adapted strains and clinical isolates after growth at 22°C or 37°C.
Fold induction of NF-κB through TLR5 after growth at
. Ratios > 2.5.
Cell activation (fold induction of NF-κB) through TLR5 is expressed as the ratio of relative light units from TLR5-transfected cells/control plasmid-transfected cells after stimulation with the indicated LM strain.
To determine whether TLR5-mediated cell activation by LM 10403s grown at 37°C is unique to this laboratory-adapted strain, we examined TLR5-mediated cell activation in a panel of 25 LM clinical isolates obtained from sterile site patient cultures (Table 1, Fig. 1A). Similar to the laboratory-adapted strains, all LM clinical isolates activated cells through TLR5 after growth at 22°C. However, at 37°C, only five out of 25 clinical isolates (20%) retained the ability to activate cells through TLR5. Furthermore, all LM strains that activated cells through TLR5 were also motile at 37°C in soft agar. These results indicate that ≈ 20% of naturally occurring pathogenic LM strains retained flagella expression and the ability to activate cells through TLR5 after growth at 37°C in vitro.
Flagellin expression by LM
In LM, there is a direct correlation between transcription of the flagellin structural gene, flaA, and the presence of flagellin protein (Peel et al., 1988; Dons et al., 1992). Thus, to validate the fact that TLR5-mediated cell activation resulted from flagellin expression, the presence of flaA mRNA was examined for LM. Two clinical isolates and one laboratory-adapted LM strain that activated cells through TLR5 after growth at 37°C, and four clinical isolates and two laboratory-adapted LM strains that did not, were examined (Table 2). Although all LM strains expressed the flaA gene after growth at 22°C, only strains that could mediate cell activation through TLR5 after growth at 37°C expressed the flaA gene. In contrast, for all LM strains that did not activate cells through TLR5 after growth at 37°C, flaA transcript could not be detected.
Table 2. . RNA expression of the LM flagellin structural gene, flaA, in LM strains after growth in vitro at either 22°C or 37°C as determined by RT-PCR.
flaA gene expression at
Activates cells through
TLR5 after growth at 37°C
Does not activate cells through
TLR5 after growth at 37°C
As our primary interest was to examine the contribution of LM flagellin expression during infection, we picked two well-characterized laboratory-adapted LM strains [one that expressed flaA and retained the ability to activate cells through TLR5 at 37°C (10403s) and one that did not (43251)] to examine whether the distinct differences in flagellin expression and TLR5-mediated cell activation for these strains would extend to similar differences during in vivo infection. This was performed by examining the expression of flaA by reverse transcription polymerase chain reaction (RT-PCR) in LM-infected spleens directly ex vivo from mice infected with LM strain 10403s or 43251. Although flaA mRNA was consistently expressed in LM recovered from mice infected with 10403s, it could not be detected in LM recovered from mice infected with 43251 (Fig. 1B). As a positive control for LM RNA isolation from infected tissue, hly mRNA, the expression of which is upregulated by LM during infection, was uniformly detected in tissue from mice infected with either 10403s or 43251. Thus, the expression of flaA during in vivo infection by LM strains 10403s and 43251 is identical to the pattern of flaA expression and TLR5-mediated cell activation by these strains when grown at 37°C in vitro. These data demonstrate that the difference in flagellin expression at 37°C in vitro between LM strains 10403s and 43251 extends directly to differences during in vivo infection.
To quantify better the relative ability of LM to activate cells through TLR5 after growth at 37°C compared with 22°C, the degree of activation for LM 10403s grown at 37°C was compared with serial twofold dilutions of LM 10403s grown at 22°C (Fig. 1C). The induction of NF-κB through TLR5 for LM grown at 22°C was consistently higher than for LM grown at 37°C in undiluted samples, and became comparable only after five to six serial twofold dilutions [relative amount: 1/64–1/32 (2–3%)]. In contrast, the induction of NF-κB through TLR2 for LM grown at 37°C was similar to that for LM grown at 22°C. These data demonstrate that, although some LM strains retained the ability to activate cells through TLR5 at 37°C, the relative ability to activate cells through TLR5 after in vitro growth at 37°C is considerably less than that at 22°C.
Pathogenesis of flagellin-deficient LM
As only ≈ 20% of LM strains causing invasive human infection retained the ability to activate cells through TLR5 after growth at 37°C, this suggests that flagellin expression in vivo and at 37°C does not play a critical role in LM pathogenesis. To address this prediction, we generated a flagellin-deficient LM strain (LM ΔflaA) by targeted disruption of the flaA gene (Shetron-Rama et al., 2003) and compared the virulence after intravenous inoculation of mice with flagellin-deficient and wild-type LM. This mutant was constructed in LM strain 10403s, which expressed flagellin and activated cells through TLR5 under all conditions tested. As predicted, this mutant was unable to activate cells via TLR5 at either 37°C or 22°C, although the defect in TLR5-mediated cell activation for this mutant could be fully restored by complementation in trans with a construct containing the full-length flaA gene (Fig. 1D).
As TLR5 is one of at least three known TLRs (TLR2 for LM peptidoglycan and TLR9 for bacterial CpG DNA) capable of recognizing PAMPs present in LM, we examined the relative contribution of LM flagellin to proinflammatory cytokine production in primary splenocytes. In splenocytes from normal mice, an ≈ 50% reduction in tumour necrosis factor (TNF)-α-producing CD11b+ splenocyte cells was observed after stimulation with 10403s ΔflaA compared with wild-type 10403s (Fig. 2). TNF-α production in splenocytes is dependent on the TLR/MyD88 signalling pathway, as the response to both 10403s and 10403s ΔflaA was substantially reduced in splenocytes from MyD88-deficient mice. These data demonstrate that LM flagellin contributes to, but is not the only signal inducing, proinflammatory cytokine production in primary cells containing multiple TLRs capable of LM recognition.
After intravenous inoculation of C57 H-2bxd mice with 1.0 × 105 cfus of either 10403s ΔflaA or wild-type 10403s (n = 6 per group), 100% of the mice became moribund and were euthanized or died by days 3–5 after infection. After infection with an 80% lower inoculum, 2.0 × 104 cfus of 10403s ΔflaA or wild-type 10403s, no mortality was observed (n > 20 per group). These data are consistent with the previously defined LD50 (104.7−105 cfus) after intravenous inoculation for LM strain 10403s in C57Bl/6 mice (White et al., 2000; Shedlock et al., 2003). Although no mortality was observed after intravenous infection with 2.0 × 104 cfus of either LM strain 10403s ΔflaA or wild-type 10403s, bacterial replication occurred in vivo as determined by the numbers of LM recovered from the spleen and liver beginning at day 3 after infection compared with the initial inoculum (Fig. 3). For mice infected with either wild-type 10403s or 10403s ΔflaA, reduced numbers of LM were present by day 7 after infection, indicating that bacterial clearance had occurred. However, at each of these time points, there were no differences in the number of LM recovered from mice infected with 10403s ΔflaA compared with wild-type 10403s (Fig. 3). Thus, LM flagellin does not affect bacterial replication or clearance after intravenous inoculation.
While these findings indicate that flagellin does not affect pathogenesis after intravenous challenge, it is possible that flagellin may do so in the context of infection acquired orally. In polarized intestinal epithelial cells, TLR5 is expressed only on the basolateral surface and is excluded from the luminal surface (Gerwitz et al., 2001a,b). This pattern of TLR5 expression in intestinal epithelial cells is hypothesized to limit inflammation to endogenous gut flora, allowing inflammation and recruitment of professional APCs only in response to invasive flagellated bacteria able to transverse the intestinal epithelial cell barrier (Sierro et al., 2001). Thus, the contribution of TLR5 in immunity to LM infection may vary with the route of experimental infection.
LM inoculated by orogastric gavage are invasive and cause bacteraemia (Huleatt et al., 2001; Pope et al., 2001; Kursar et al., 2002). The kinetics of bacterial replication and clearance, and the generation of LM-specific T cells after intravenous and oral infection are similar, although the oral dose required to elicit an intravenous equivalent response is significantly greater (oral 109 versus intravenous 104 cfus) (Huleatt et al., 2001; Pope et al., 2001). After orogastric infection, 33% of mice infected with 10403s ΔflaA (n = 12) and 11% of mice infected with wild-type 10403s (n = 9) became moribund and were sacrificed or died. This difference in survival did not reach statistical significance (P = 0.29). In the remaining mice, a similar ability to translocate across the gastrointestinal epithelium and disseminate to the spleen, mesenteric lymph nodes and liver was observed for 10403s ΔflaA and wild-type 10403s, as determined by the number of LM cfus by day 3 after infection (Fig. 4). Furthermore, both 10403s ΔflaA and wild-type 10403s were substantially cleared from these same tissues by day 7.
Adaptive immunity triggered by flagellin-deficient LM
As exogenously added flagellin is able to enhance the T-cell response to other bacterial pathogens (McSorley et al., 2002), the contribution of flagellin to the generation of adaptive immune responses to LM was examined (Fig. 5). After intravenous or oral infection with either 10403s ΔflaA or wild-type 10403s, no significant differences were observed in either the percentage or the total number of CD8 or CD4 T cells specific for the LM major histocompatibility (MHC) class I (LLO 91–99) or class II (LLO 189–201) immunodominant peptides. These data demonstrate that, during infection, LM flagellin does not substantially alter the generation of adaptive immune responses, and are consistent with the similar kinetics of bacterial clearance observed between days 3 and 7 after infection. These results are consistent with the generation of normal numbers of LM-specific CD8 T cells and only a modest decline in LM-specific CD4 T cells in MyD88-deficient mice (which lack functional TLR5 and TLR2) (Way et al., 2003). These findings also suggest that LM flagellin does not alter the generation of LM-specific T cells through non-TLR-mediated pathways.
In summary, the data presented in this study demonstrate that, although all LM strains are able to activate cells through TLR5 when grown at 22°C, only ≈ 20% of LM strains retain flagellin expression at 37°C in vitro and perhaps in vivo. Although less flagellin is produced by these LM strains after growth at 37°C compared with 22°C, these strains retained the ability to activate cells though TLR5. When an LM mutant with targeted flagellin deficiency was generated and compared with a parental LM strain expressing flagellin, no significant differences were observed in bacterial replication, bacterial clearance or induction of LM-specific T cells after either intravenous or oral infection. These results indicate that flagellin is neither an essential trigger for innate or adaptive immunity to LM, nor does it exact a physiologically important metabolic penalty, the relief from which (following downregulation at temperatures present in mammals in vivo) allows LM to invade or to replicate more efficiently in vivo. In the context of the extreme susceptibility of MyD88-deficient mice, and lack of susceptibility of TLR2-deficient mice, to primary LM infection (Edelson and Unanue, 2002; Seki et al., 2002), these findings suggest that functional redundancy exists between known TLRs (pattern recognition receptors) in innate immunity to LM infection. Alternatively, the lack of a significant phenotype for flagellin-deficient compared with flagellin-expressing wild-type LM may reflect the relatively low amount of flagellin produced by LM at 37°C, compared with that produced at 22°C. This alternative would support the hypothesis that LM downregulates flagellin as a means of evading host immunity through TLR5.
Laboratory-adapted wild-type LM strains 10403s, ScottA, EGD, LO28 and SLCC5764 (Mackaness) were provided by Dr Daniel Portnoy (University of California at Berkeley). Wild-type LM strain 43251 was obtained from the American Type Culture Collection (ATCC). 10403s ΔflaA was generated by targeted deletion of the flaA locus as described previously (Shetron-Rama et al., 2003). Clinical isolates of LM were provided by the microbiology laboratory at Harborview Medical Center (strains U5–U17) and the Washington State Department of Health (strains DOH102–DOH4104). LM strains were routinely stored as glycerol stocks at −80°C, streaked on to brain–heart infusion (BHI) plates and grown in BHI broth. LM electrocompetent cells were prepared, and transformations were performed as described previously (Park and Stewart, 1990).
HeLa cells (4.0 × 104) were seeded into 96-well plates and transfected using Polyfect reagent (Qiagen) with the following amounts of DNA per well: ELAM-Luc, 0.03 µg; β-actin Renilla-Luc, 0.009 µg; and haemagglutinin (HA)-tagged full-length murine TLR5 or TLR2 (minus the signal sequence) cloned in frame with the signal sequence in pDisplay (Invitrogen), 0.03 µg. The total amount of DNA per well was adjusted to 0.30 µg with empty pDisplay vector. Twenty-four hours after transfection, cells were stimulated (4 h) with either LM (1.0 × 107 cfus or serial dilutions of this inoculum per well), human IL-1β (10 ng ml−1) or flagellin (200 ng ml−1) (purified from Salmonella typhimurium and generously provided by Dr Kelly Smith, University of Washington). For stimulations, LM was diluted in Dulbecco's modified Eagle medium (DMEM) and incubated at 37°C, 5% CO2 for 4 h. Transfected HeLa cells were then lysed, and the amounts of Firefly and Renilla luciferase light units in the lysates were quantified with the Dual-Luciferase reporter assay system (Promega). Expression of NF-κB (relative light units) represents the ratio of Firefly (ELAM-Luc) to Renilla (β-actin Renilla-Luc) luciferase values per well.
The motility of LM strains was evaluated after stabbing into soft agar (0.3% agar) as described previously (Shetron-Rama et al., 2003) and incubation for 24 h at the indicated temperature.
RNA isolation and RT-PCR
TRIzol (Gibco BRL) reagent was used to isolate RNA from LM-infected tissues according to the manufacturer's instructions. Purified RNA (0.50 µg) was reverse transcribed to cDNA using random primers and Superscript II (Invitrogen). Primers used for PCR were as follows: flaA, sense 5′-AAGTAAATACTAATATCA TTAGC-3′, antisense 5′-GCTAATTGACGCATACGTTGC-3′; hly, sense 5′-TAGTTAGTCTACCAATTGCGC-3′, antisense 5′-TTG TAACCTTTTCTTGGCGGC-3′.
Intracellular cytokine staining
Intracellular cytokine staining was performed as described previously (Way et al., 2003). Briefly, red blood cell (RBC)-lysed single-cell splenocyte suspensions were incubated in the presence of the indicated LM-specific peptide (10−6 M) or LM (1.0 × 107 cfus) and GolgiStop (Pharmingen) for 5 h. Cells were then washed with fluorescence-activated cell sorting (FACS) buffer (PBS, 0.1% BSA, 0.09% sodium azide), blocked with anti-CD16/CD32 and then stained with anti-CD8, anti-CD4 or anti-CD11b antibodies. The cells were then permeabilized with Cytoperm solution (Pharmingen) and stained with anti-interferon (IFN)-γ or anti-TNF-α antibody. Stained cells were analysed on a FACScan flow cytometer (Becton Dickinson) using cellquest software (BD Biosciences).
Infections with LM
As endogenous LM class I immunodominant peptides have been described only in the murine H-2d haplotype and class II immunodominant peptides only in the H-2b haplotype, mice used were F1 (H-2b × d) derived from matings between C57Bl/6 (H-2b) and C57B10.D2 (H-2d) mice. For infection, bacteria were grown in BHI at 37°C, washed and resuspended in saline (200 µl), and inoculated via the lateral tail vein or oral gastric gavage. For each infection, the LM inoculum was confirmed by quantitative culture. The numbers of LM in lysates of infected tissues were determined as described previously (Way et al., 2003). After infection, mice were evaluated twice daily and euthanized when moribund. All experiments were performed under IACUC-approved protocols.
The difference in geometric mean cfus between groups of mice, and the percentages and numbers of activated splenocytes, were evaluated using Mann–Whitney statistical analysis (graphpad, Prism Software). The difference in survival between groups of mice after oral infection was evaluated using Kaplan–Meier survival analysis (graphpad, Prism Software). In all analyses, P < 0.05 was taken as statistically significant.
The authors would like to thank Dr Kelly Smith for the gift of purified S. typhimurium flagellin, Dr Ferric Fang (Harborview Medical Center) and Mr Mike McDowell (Washington State Department of Health) for supplying LM clinical isolates. This work was supported by NIH grants HL69503 (to A.M.H.) and HD18184 (to C.B.W.). S.S.W. is an NICHD Fellow of the Pediatric Scientist Development Program (NICHD grant award K12-HD00850). T.R.K. is a Pfizer Postdoctoral Fellow in Infectious Diseases.