Acute lung infection due to Pseudomonas aeruginosa is an increasingly serious problem that results in high mortality especially in the compromised host. In this study, we set out to ascertain what components of the TLR system are most important for innate immunity to this microorganism. We previously demonstrated that TLR2,4−/− mice were not hypersusceptible to infection by a wild-type P. aeruginosa strain. However, we now find that mice lacking both TLR2 and TLR4 (TLR2,4−/− mice) are hypersusceptible to infection following challenge with a P. aeruginosa mutant devoid of flagellin production. We demonstrate that this hypersusceptibilty is largely due to a lack of innate defense by the host that fails to control bacterial replication in the lung. Further evidence that a response to flagellin is a key factor in the failure of TLR2,4−/− mice to control the infection with the mutant strain was obtained by demonstrating that the intrapulmonary administration of flagellin over a 18 h period following infection, saved 100% of TLR2,4−/− mice from death. We conclude that the interactions of either TLR4 with LPS or TLR5 with flagellin can effectively defend the lung from P. aeruginosa infection and the absence of a response by both results in hypersusceptibility to this infection.
Infection of the lungs by Pseudomonas aeruginosa bacteria, a Gram-negative opportunistic pathogen, results in two distinct clinical syndromes. In ventilated patients, it causes an acute pneumonia with a high mortality rate, and in the genetic disorder cystic fibrosis, it is the prime cause of chronic inflammation, which is a key factor in the destructive lung disease that occurs in these patients (1). The different clinical courses of these two lung diseases likely result from the interplay of different host and bacterial factors but inflammation is a common feature of both illnesses with bacterial toxins possibly playing a significant role in acute disease (2, 3, 4). The inflammatory response in both cases is most likely due to the recognition of the lipid A portion of LPS moieties on the cell wall (5) and the flagellin of the microorganism (6, 7, 8) but there are no data that can definitively exclude other pathogen-associated molecular patterns (PAMPs)3 of P. aeruginosa.
Host recognition of PAMPs may have two entirely different consequences. An appropriate response leads to the eradication of a microorganism (9), but an exaggerated inflammatory response may lead to illnesses such as sepsis and shock (10). The corollary of the former response is that failure to recognize a microorganism results in failure to eradicate it and in susceptibility to disease. Indeed, animal models of infection have demonstrated that the susceptibility to a number of Gram-negative pathogens is linked to the lack of recognition of LPS (11, 12) in which case a defective response leads to extreme susceptibility. These experimental findings are supported by clinical observations that polymorphisms in human genes that encode pattern recognition receptors result in greater susceptibility to certain infections (13). However, in the case of P. aeruginosa, whose LPS is recognized by TLR4 (5) and possibly by TLR2 (5, 14), mice lacking both TLR2 and TLR4 (TLR2,4−/−) are not hypersusceptible to this bacterium and mount an effective innate response that clears the microorganism (15, 16). In contrast MyD88−/− mice are extremely susceptible with 100% of mice dying within 48 h with a low dose of P. aeruginosa (15, 16), indicating the potential involvement of TLR ligand interactions in host defense. In P. aeruginosa, one other possible TLR ligand is flagellin, the known TLR5 ligand, which has been implicated in a pathogenic role in acute pneumonia (6, 15) and which has been demonstrated to cause inflammation when instilled into the lungs (6, 7, 8). However, in a lung model of acute infection using a P. aeruginosa mutant devoid of flagellin production (a ΔfliC mutant), we recently demonstrated that the absence of flagellin does not significantly alter the LD50 but results in slower clearance of this microorganism from the lungs and a delay in the time to death (8). Thus, neither of the Pseudomonas interactions, LPS-TLR2,4 or flagellin-TLR5, by themselves play an essential role, although they both may participate in the innate immune response.
Two recent studies have shed more light on the role played by these interactions in defense of the lungs but the reported data are not entirely consistent on whether these TLRs are critical for the control of a Pseudomonas lung infection and survival of infected mice. Feuillet et al. (17) demonstrated that TLR4,5−/− mice are hypersusceptible to infection by wild-type P. aeruginosa strain PAK but the innate immune response to the strain and bacterial clearance were not measured. A second study by Skerrett et al. (16) using TLR2−/−, TLR4−/−, or TLR2,4−/− mice and a P. aeruginosa strain PAK lacking flagellin, in effect abrogating the TLR5 response, concluded that none of the mutant mouse strains were hypersusceptible to this strain that lacked flagellin and that bacterial clearance was not affected in any of these mutant mice but was defective in MyD88−/− mice. Thus, the results of these studies have not been consistent in regard to control of bacterial replication and mortality. This inconsistency may be due to the fact that the outcome of infection by P. aeruginosa is much more complex and may involve inflammatory injury in response to the microorganism, the initiation of an effective innate immune response to clear the organism, as well as the pathogenic effects of secreted and injected toxins. As evidence for this complexity, strains of P. aeruginosa defective in the production of various exoenzymes have been shown to be less virulent in the mouse model of pneumonia (2, 3, 4). With respect to the role of inflammation, the inflammatory response to the inhalation of Pseudomonas LPS leads to lung injury (18) and the inflammatory response to strains of P. aeruginosa that produce an excess of flagellin results in rapid death when compared with a wild-type strain (8).
To ascertain what factors are involved in mortality and control of bacterial replication, we examined different combinations of wild-type, TLR2,4−/−, TLR2−/−, and TLR4−/− mice infected by P. aeruginosa or its ΔfliC mutant. We demonstrate that TLR4 and TLR5 play major but redundant roles in controlling bacterial replication and in host defense against P. aeruginosa pneumonia.
Materials and Methods
Males of several mouse strains were used for the experiments. TLR2−/− and TLR4−/− mice were obtained from S. Akira (Osaka University, Osaka, Japan) and were backcrossed eight times with C57BL/6 to ensure similar genetic backgrounds. TLR2,4−/− mice were generated by breeding TLR2−/− mice and TLR4−/− mice. C57BL/6 mice used as control mice were supplied by the Centre d’Elevage R. Janvier and used at ∼8 wk of age. Mice were fed normal mouse chow and water ad libitum and were reared and housed under standard conditions with air filtration. Mice were cared for in accordance with Pasteur Institute guidelines in compliance with the European Animal Welfare regulations.
The wild-type strain PAK, a commonly studied P. aeruginosa strain, was obtained from S. Lory (Harvard Medical School, Boston, MA), as originally isolated by D. Bradley (Memorial University of Newfoundland, St. John’s, Canada). This strain of P. aeruginosa is known to contain and express a full complement of virulence factors, including pili, flagella, the type II secreted enzymes, exotoxin A, elastases and phospholipases and the type III secreted exoenzymes S, T, and Y. PAK ΔfliC and PAK-L88 are mutants of the parent strain in which the fliC gene has been deleted (19), and in which there is an L88A substitution in the TLR5 binding site of Pseudomonas flagellin (20), respectively. PAO1 ΔfliC is a fliC mutant of strain PAO1 (21), another well-studied strain of P. aeruginosa. Luminescent strains of PAK and its ΔfliC mutant, were constructed by inserting luxAB into the neutral att site of the chromosome of strain PAK and its derivative PAK-C using a mini-Tn7-lux plasmid provided by Microbiotix, in which luxAB is driven by the lac promoter (22).
Bacteria were grown overnight in Luria-Bertani broth then transferred to fresh medium and grown for 4–5 h to mid-log phase. The cultures were centrifuged at 4000 × g for 15 min and the cell pellets washed twice with PBS. The bacterial pellet was diluted in its original volume and the OD adjusted to give the approximate desired inocula. The inocula were verified by serial 10-fold dilutions of the bacterial suspensions and plating on Luria-Bertani agar.
Mice were anesthetized by i.p. injection of a mixture of ketamine-xylazine and were placed supine. A plastic catheter (diameter of 0.86 mm) was inserted into the trachea via the oropharynx. The proper insertion was verified by checking the formation of mist due to expiration on a mirror placed in front of the external end. A 50-μl bacterial suspension was laid down at the internal end of the catheter with a micropipette using a sterile disposable tip for gel loading, introduced into the catheter. Mice were then immediately held upright to facilitate bacterial inhalation and until normal breathing resumed. A series of experiments were done to ascertain the effects of the TLR mutations in bacterial clearance of the fliC mutant of P. aeruginosa. Groups of mice were infected by the intratracheal route using inocula of bacteria that were known to be less than LD50 for wild-type mice from previous experiments (8, 15). In one series of experiments, recombinant P. aeruginosa flagellin prepared as described (20) was mixed with the bacterial inoculum before intratracheal instillation and at several times after infection.
Bronchoalveolar lavage (BAL) was performed on some mice at 6 h after infection following pentobarbital euthanasia. The BAL fluids (4 ml) were diluted and plated on Luria-Bertani agar plates to obtain viable bacterial counts in the BAL fluid. Total cell counts were measured in the BAL fluids with a Coulter Counter (Coulter Electronics), and cell differential counts were determined after cytospin centrifugation and staining with Diff-Quik products. Murine cytokine concentrations in BAL fluid were determined using DuoSet ELISA kits obtained from R&D Systems.
Photon emission of the luminescent bacteria in the lungs of infected mice was measured using the IVIS system (Xenogen Biosciences). In preliminary experiments, we ascertained that the system could detect 106 CFU of luminescent P. aeruginosa given intratracheally. To achieve detection in black C57BL/6 mice, it was necessary to remove hair from their chests with a depilatory agent used for this purpose in humans. After infection, analysis of photons was done in mice under isoflurane inhalation anesthesia in an IVIS charge-coupled device camera coupled to the LivingImage software package (Xenogen). A digital false-color photon emission image of the mouse was generated, and photons were counted within a constant defined area corresponding to the surface of the chest encompassing the whole lung region. Photon emission was measured as photons per second emitted.
Cytokine levels, myeloperoxidase concentrations, polymorphonuclear neutrophil (PMN) counts, and pathogen counts were expressed as mean ± SEM. Differences between groups were assessed for statistical significance using the ANOVA test followed by Fischer’s test. A value of p < 0.05 was considered statistically significant.
Effect of TLR2,4 mutations on survival of mice following infection by a ΔfliC mutant of P. aeruginosa
Survival of C57BL/6 mice and TLR2,4−/− was examined using wild-type strain PAK and its flagellin-deficient mutant PAKΔfliC. The dose of bacteria (5 × 106 CFU/mouse) was chosen after preliminary experiments showed that the mutant mice were dying rapidly when given half the LD50 of the ΔfliC mutant for wild-type mice as measured in a previous study (8). The dose chosen was expected to cause little or no mortality among wild-type mice, but should uncover any hypersusceptibility among knockout mice.
As previously demonstrated with wild-type strain PAK (15), there were no differences in the survival of the control and the TLR2,4−/− mice (Fig. 1). In contrast, the survival studies with the ΔfliC mutant demonstrated extreme susceptibility of the TLR2,4−/− mice, with all mice dying within 2 days of challenge with a dose of this strain that caused no mortality among the control mice. These results suggested that the extreme susceptibility was likely due to the absence of a response to both LPS and flagellin consequent to the mutations in TLR2 and TLR4 and the concomitant absence of flagellin from the bacterium.
Effect of the TLR2 and TLR4 mutations on bacterial proliferation in the lungs
We have previously shown that proliferation of wild-type strain PAK was controlled in TLR2,4−/− mice (15). Skerrett et al. (16) recently demonstrated that proliferation of a ΔfliC mutant of this same strain of P. aeruginosa is also controlled in TLR2,4−/− mice even though host innate immune responses are severely blunted. These recently published data did not, however, explain the hypersusceptibility that we observed (Fig. 1). We therefore examined bacterial proliferation to ascertain whether hypersusceptibility was accompanied by failure to control bacterial proliferation as occurs in MyD88-deficient mice (15, 16). When examined at 6 h postinfection (Fig. 2,A), TLR2,4−/− mice failed to control replication of the ΔfliC mutant, with bacterial counts reaching 10 times the challenge dose in these mice, whereas it was down to <20% of the inoculum in the control mice. We also measured proliferation in vivo at 6 and 20 h by detecting the growth of luminescent bacteria directly in the lungs of living anesthetized mice. Bacterial luminescence images are shown in Fig. 2,B. Measured in photons per second, luminescence increased at least 10-fold in a 24-h period in the TLR2,4−/− mice infected with the ΔfliC mutant, whereas luminescence decreased in wild-type mice (Fig. 2 C). In agreement with our previous report (15), luminescence decreased as a function of time in wild-type and TLR2,4−/− mice challenged with the wild-type strain PAK (data not shown).
Evaluation of TNF-α, IL-6, KC, and G-CSF production in response to infection of TLR2,4−/− mice by a ΔfliC mutant of P. aeruginosa
We have previously noted that when TLR2,4−/− mice are infected with the wild-type P. aeruginosa strain, the TNF-α response is severely attenuated but the IL-6 response is intact (15), suggesting that a PAMP other than LPS may also stimulate the IL-6 response. We thus examined the production of these two and other inflammatory mediators under the different experimental combinations used in the present study. Studies were conducted at 6 h postinfection, as mutant mice were often dead at 18–24 h following challenge with the ΔfliC mutant. We confirmed the earlier data that the TNF-α response is lost in the TLR2,4−/− mice (Fig. 3,A) and that the IL-6 (Fig. 3,B) response is preserved. The most striking observation was that infection of TLR2,4−/− mice with the ΔfliC mutant led to a complete suppression of host response in terms of both TNF-α and IL-6 formation. Given the complete loss of TNF-α synthesis in the TLR2,4−/− mice infected with the flagellated strain and the additional loss of IL-6 in these mice only when flagellin is absent, it is possible that the TNF-α response is mainly due to a PAMP recognized by TLR2 and TLR4, and that the IL-6 response is a sum of the effects of a PAMP recognized by TLR2/4 and the activity of flagellin. The KC and G-CSF responses (Fig. 3, C and D) to the absence of flagellin were similar to those of IL-6 in these mice. Thus absence of IL-6, G-CSF, and KC responses was correlated with hypersusceptibility but the absence of a TNF-α response was not.
Evaluation of PMN recruitment and activation
In agreement with the data presented, PMN recruitment in the airspaces and their activation as measured by the release of free myeloperoxidase were similar under the different experimental combinations, except in the case of TLR2,4−/− mice infected with the ΔfliC mutant (Fig. 4). Under the latter, the values for both parameters were not different from those of uninfected control mice suggesting a failure of recruitment and activation by PMN in these mice.
Role of flagellin in the innate immune response
To examine whether an innate immune response to flagellin as opposed to LPS (TLR2 or TLR4) was critical to the survival of these mice, one group of TLR2,4 mutant mice was given 0.1 μg of flagellin mixed with the challenge bacteria, followed by another dose of flagellin 6 h later by the intranasal route. A second group was given the same dose at the time of infection then at 6, 12, and 18 h postinfection by the intranasal route. A control group of TLR2,4−/− mice received PBS instead of flagellin but did not survive beyond 24 h. Survival of the mice that received two doses of flagellin was prolonged by about a day but all were dead by 48 h post bacterial challenge, whereas the administration of four doses of flagellin resulted in the survival of 100% of the mice even as long as 2 wk bacterial postchallenge (Fig. 5).
To further implicate the requirement for TLR5 recognition of flagellin in TLR2,4−/− mice, we challenged these mice with P. aeruginosa strain PAK-L88 (8), which is a PAK mutant with an L88A amino acid change in the putative TLR5 binding site of Pseudomonas flagellin (20). This single amino acid change results in a significant reduction in the inflammatory response of the mouse lung when recombinant L88 flagellin is instilled into the lungs (8). Using different concentrations of the L88 mutant (between 1 and 2 × 107 CFU), we observed greater mortality than that seen with the use of the wild-type strain but not as great as that seen with the ΔfliC mutant. The differences in survival observed within the first 2 days of infection of the TLR2,4−/− mice challenged with wild-type strain PAK or PAK-L88 were statistically significant (p < 0.05) (Fig. 6). Data are expressed as mean ± SEM obtained from three distinct experiments performed with 7–13 mice each. Mean survivals at 48 h of 78% with the wild-type strain and 35% with the L88 mutant. Under the same experimental conditions, the survival of the mice was 0% for the ΔfliC mutant (Fig. 6). Some possible reasons for the lack of total susceptibility of TLR2,4−/− mice to infection with the PAK-L88 mutant are discussed below.
Role of TLR2 vs TLR4 in innate immunity to a ΔfliC mutant of P. aeruginosa
Because TLR2,4−/− were hypersusceptible to a ΔfliC mutant of strain PAK, we sought to examine which of these two TLRs was instrumental in defense against this organism in the absence of a response to flagellin. TLR2−/− and TLR4−/− mice were infected with the ΔfliC mutant of strain PAK at a dose of 5 × 106 CFU, which does not kill wild-type C57BL/6 mice. All TLR4−/− mice died within 48 h of infection, whereas all TLR2−/− mice survived (Fig. 7). To confirm that this hypersusceptibility was more general to P. aeruginosa strains, a ΔfliC mutant of strain PAO1 was used to infect wild-type and TLR4−/− mice. In preliminary experiments, an inoculum of 106 CFU failed to kill wild-type mice. However all TLR4−/− rapidly succumbed to infection by this mutant (Fig. 7).
The outcome of an acute pulmonary infection due to P. aeruginosa is a balance between an appropriate innate immune response and bacterial virulence. The pathway to an appropriate response is through the recognition of PAMPs by TLRs and possibly other non-TLR systems such as the NOD-like receptor system and the inflammasome (23, 24). However, although it has been shown that P. aeruginosa triggers the NOD system (25) and its flagellin is recognized by the inflammasome (26), a role for host defense by these systems in vivo has not been demonstrated to date. Data from this study and of Feuilllet et al. (17) suggests that the most important systems may be the TLR system and that TLR4 and TLR5 are redundant for the recognition of P. aeruginosa and possibly other flagellated bacteria. However, there have been reported instances when defense against Gram-negative bacteria required only TLR4 (11, 12, 27, 28). On closer examination of those studies, some of the microorganisms studied do not possess a flagellum or do not express flagellin in vivo. For example Klebsiella pneumoniae and Haemophilus influenzae are not flagellated bacteria and Bordetella down-regulates flagellin production in vivo (29). One study of P. aeruginosa that showed TLR4 alone to be essential used strain PA103 (30), which is known to be a nonflagellated strain (31). Other studies have alternatively shown TLR4 mutant mice not to be hypersusceptible to Escherichia coli strains (32, 33, 34). However, whether these strains or clones of a given strain expressed flagellin and made a flagellum is unknown. It is possible that the lack of flagellin expression in the strains or clones of a given strain used in the studies that showed hypersusceptibility to E. coli was responsible for this observation, as TLR4 mutant mice would not be able to mount an effective response if flagellin were absent.
A role for TLR2 in defense against P. aeruginosa infections is more difficult to substantiate, although interactions of TLR2 with both LPS and flagellin have been described (5, 14, 35). It has been reported that Pseudomonas LPS may be recognized by either TLR2 or TLR4 depending on its structure (5), thus one cannot entirely rule out a role for TLR2. However, Feuillet et al. (17) as well as the current study, could not, however, measure a contribution of TLR2 in lung defense against P. aeruginosa. Because TLR2 does recognize P. aeruginosa, the suggestion by Skerrett et al. (16) that it may play a counter-regulatory role seems feasible but it does not appear to have a direct role in defense. The response of strains of P. aeruginosa that have been adapted to the airways as well as most common laboratory strains examined appear to be TLR4-dependent rather than to TLR2 (5). Strain PAK used by Feuillet et al. (17), Skerrett et al. (16), as well as this study fall into this group. Moreover, we see the same TLR4 dependency in PAO1, another P. aeruginosa strain.
Other features of this study that are worthy of commentary are the differences in deaths and bacterial clearance that were noted between our work and Skerrett et al. (16) for TLR4−/− and TLR2,4−/− mice, when a ΔfliC mutant of strain PAK was used. Skerrett et al. (16) reported one death in TLR2,4 mutant mice and none in TLR2 or TLR4 mutant mice, whereas we noted hypersusceptibility. Bacterial clearance in TLR2,4−/− mice was observed by Skerrett et al. (16), but was markedly reduced in our studies with actual proliferation noted at both 6 and 24 h following lung challenge. The use of different doses of bacteria could be an explanation because mortality appears to be dose-dependent (17). Indeed, it can be deduced from the data reported by Skerrett et al. (16) that the bacterial load of their mice fluctuates around 105 CFU/mouse, whereas we challenged mice with 5 × 106 CFU. In agreement, Feuillet et al. (17) observed a dramatic death rate of the mice with a challenge of 6 × 106 CFU/mouse.
Due to the unavailability of TLR2,4,5−/− mice, we used a ΔfliC mutant of P. aeruginosa in most of these studies. However, having a mutant in the flagellin binding site for TLR5 (strain PAK-L88) allowed us to examine the role of TLR5 in our studies. TLR2,4−/− mice proved not to be as hypersusceptible to this strain as the ΔfliC mutant, in which we saw 100% mortality after 2 days. However, infection of these mice with PAK-L88 did result in higher mortality than infection with the wild-type strain.
We offer several explanations for this reduced hypersusceptibility. The L88 flagellin on the bacterium is glycosylated and the Pseudomonas flagellin glycosyl moiety stimulates an inflammatory response (20). The ΔfliC mutant has no glycosyl groups, as it is a flagellin deletion mutant. Pseudomonas flagellin activates caspase-1 through a mechanism that is TLR5-independent (26), thus the L88 mutation in the TLR5 binding site still stimulates IL-1β secretion, which may offer some protection (26). Our own previous experiments (20) also demonstrate a low level of IL-8 release (10% of wild-type flagellin), which may be enough to prevent the hypersusceptibility seen with a ΔfliC mutant. Thus there may be three possible explanations for the lack of extreme hypersusceptibility as seen with the ΔfliC mutant.
The current study also points to important roles for flagellin in the generation of IL-6 and G-CSF because loss of these responses occurred only in TLR2,4−/− mice infected with the flagellin mutant. We have previously reported the loss of the IL-6 response in MyD88-deficient mice and its preservation in TLR2,4−/− mutant mice, when these mice are infected with a wild-type P. aeruginosa strain (15). Their loss when a flagellin mutant was used suggests the existence of a specific pathway that is flagellin-dependent. Similar observations, that an IL-6 response was lost in response to an intranasal challenge of TLR5−/− mice with flagellin protein were made by Feuillet et al. (17), suggesting that the IL-6 response may be TLR5 and flagellin dependent. Our study expands this by showing the lack of response is solely due to the loss of flagellin. How this response to flagellin is mediated remains to be determined.
Our observations on TNF-α are also of interest. We have noted in the past (15) and find once more that animals incapable of mounting a TNF-α response are not susceptible to Pseudomonas lung infection and do not demonstrate a defect in Pseudomonas clearance. However, most studies, including one that used a TNF-α knockout mouse (36, 37) show a defect in protection against such an infection. Thus, the role of TNF-α is not as clear as described because it has also been shown that mice deficient in TNF-α receptors do not have diminished resistance to a Pseudomonas lung infection (18), a result which is consistent with our findings.
There also exists the possibility that other systems may play a role in defense against this organism but we have not been able to substantiate this. It has recently been demonstrated that flagellin from two intracellular bacterial species Salmonella and Legionella (38, 39) is recognized by Ipaf, which activates caspase-1 (40) leading to the formation of IL-1β and IL-18, which are proinflammatory cytokines. This system also recognizes Pseudomonas flagellin (26). However, its role in infection by an extracellular pathogen such as P. aeruginosa may not be as critical and be supplemental. Alternatively, this system may be functional but is overwhelmed by high inocula of bacteria such as those used in experimental in vivo studies. Thus, we conclude that the activation of inflammatory pathways through TLR4 or TLR5 is sufficient to control an acute P. aeruginosa lung infection. Although the studies of TLR-Pseudomonas interactions have been limited to acute infections, one may conclude that these are equally important in chronic infection in cystic fibrosis. However, in this case these interactions may fail to control the infection because of host and microbial factors such as microcolony formation (41) and defective phagocytosis in the presence of mucins, (42) and instead result in cyclical bouts of uncontrolled inflammation.
We thank Herbert Schweizer (Colorado State University, Fort Collins, CO) for advice on use of the lux reporter system in P. aeruginosa, and Marie-Anne Nicola (Plate-Forme d’Imagerie Dynamique, Imagopole, Institut Pasteur, Paris) for advice on the use of the IVIS system.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by Grant AI 45014 from National Institutes of Health (to R.R.).
Abbreviations used in this paper: PAMP, pathogen-associated molecular pattern; BAL, bronchoalveolar lavage; PMN, polymorphonuclear neutrophil.