Francisella tularensis disrupts TLR2-MYD88-p38 signaling early during infection to delay apoptosis of macrophages and promote virulence in the host

ABSTRACT Francisella tularensis is a zoonotic pathogen and the causative agent of tularemia. F. tularensis replicates to high levels within the cytosol of macrophages and other host cells while subverting the host response to infection. Critical to the success of F. tularensis is its ability to delay macrophage apoptosis to maintain its intracellular replicative niche. However, the host-signaling pathway(s) modulated by F. tularensis to delay apoptosis are poorly characterized. The outer membrane channel protein TolC is required for F. tularensis virulence and its ability to suppress apoptosis and cytokine expression during infection of macrophages. We took advantage of the F. tularensis ∆tolC mutant phenotype to identify host pathways that are important for activating macrophage apoptosis and that are disrupted by the bacteria. Comparison of macrophages infected with wild-type or ∆tolC F. tularensis revealed that the bacteria interfere with TLR2-MYD88-p38 signaling at early times post infection to delay apoptosis, dampen innate host responses, and preserve the intracellular replicative niche. Experiments using the mouse pneumonic tularemia model confirmed the in vivo relevance of these findings, revealing contributions of TLR2 and MYD88 signaling to the protective host response to F. tularensis, which is modulated by the bacteria to promote virulence. IMPORTANCE Francisella tularensis is a Gram-negative intracellular bacterial pathogen and the causative agent of the zoonotic disease tularemia. F. tularensis, like other intracellular pathogens, modulates host-programmed cell death pathways to ensure its replication and survival. We previously identified the outer membrane channel protein TolC as required for the ability of F. tularensis to delay host cell death. However, the mechanism by which F. tularensis delays cell death pathways during intracellular replication is unclear despite being critical to pathogenesis. In the present study, we address this gap in knowledge by taking advantage of ∆tolC mutants of F. tularensis to uncover signaling pathways governing host apoptotic responses to F. tularensis and which are modulated by the bacteria during infection to promote virulence. These findings reveal mechanisms by which intracellular pathogens subvert host responses and enhance our understanding of the pathogenesis of tularemia.

more virulent in humans compared to subsp. holarctica (1). An attenuated F. tularensis live vaccine strain (LVS), derived from subsp. holarctica, causes lethal infection in mice that closely mimics the pathogenesis of human virulent strains, making it a valuable experimental tool (1). The LVS is not licensed for use as a vaccine in the United States because its mechanism of attenuation is not completely understood and it does not fully protect against infection with subsp. tularensis strains. Therefore, there is a need to better understand F. tularensis virulence mechanisms and host responses to infection, which will facilitate the development of improved vaccines and therapeutic approaches.
F. tularensis primarily replicates within macrophages early during infection (2). Upon phagocytosis, the bacteria escape the phagosome as early as 30 min post infection (p.i.) and then begin replicating in the host cell cytosol. After several rounds of intra cellular replication, host cell death is induced, contributing to bacterial dissemination and further infection (3)(4)(5)(6). Critical to the success of F. tularensis is its ability to remain immunologically silent during early replication and spread within the host. In mouse pneumonic tularemia models, strong inflammatory responses are not observed until after the first 2-3 d p.i (7)(8)(9). Similar observations have been made in the human disease (10)(11)(12). To evade detection by the host, F. tularensis actively interferes with host innate immune responses by dampening pro-inflammatory cytokine expression and delaying programmed cell death pathways (5,(13)(14)(15)(16)(17)(18)(19)(20)(21). During infection of macrophages, F. tularensis delays host cell death until late time points (~24 h p.i.), at a stage where the bacteria have replicated to high levels; the induction of apoptosis then allows bacte rial spread under non-inflammatory conditions. Despite being critical to virulence, the mechanism by which Francisella delays apoptosis during infection is poorly understood (18,19,(21)(22)(23)(24).
We previously identified the F. tularensis TolC protein as a virulence factor of both the LVS and human virulent Schu S4 strain that is critical for the ability to delay apoptosis and dampen pro-inflammatory cytokine expression during infection (21,23,41). TolC belongs to a class of Gram-negative bacterial outer membrane channel proteins involved in the type I protein secretion pathway and efflux of small molecules (42). F. tularensis TolC deletion mutants elicit increased release of pro-inflammatory cytokines and increased cytotoxicity during infection of macrophages, corresponding to prema ture activation of the apoptotic pathway and loss of the intracellular replicative niche (21,23,41,43). Both the LVS and Schu S4 ∆tolC mutants are highly attenuated in mouse tularemia infection models (21,23,41). While the ∆tolC mutants are competent for phagosomal escape, intracellular replication, and dissemination to distal organs during in vivo infection, compared to the wild-type (WT) bacteria, the mutants activate increased caspase-3 cleavage, replicate to lower numbers, and are eventually cleared by the host (21,23,41,43).
In the current study, we utilized ∆tolC mutants in both the LVS and Schu S4 strain to uncover signaling pathways by which the host recognizes F. tularensis to control infection. Comparison of primary murine macrophages infected with WT or ∆tolC LVS revealed that F. tularensis disrupts MAPK signaling at early times p.i. to reduce cytotox icity and preserve its intracellular replicative niche. Specifically, we present evidence that the detection of F. tularensis and subsequent phosphorylation of the p38 MAPK is mediated by TLR2-MYD88 signaling, which activates the host apoptotic response. Furthermore, we found that the human virulent Schu S4 strain similarly modulates TLR2-and MAPK-dependent signaling early during infection to dampen macrophage cell death and pro-inflammatory cytokine responses. Finally, experiments using the mouse pneumonic tularemia model revealed that both TLR2 and MYD88 signaling contribute to the early host response to LVS infection in vivo and to the protection against lethal infection. Notably, the LVS ∆tolC mutant regained full virulence in MYD88 −/− mice, identifying a critical protective role for MYD88-dependent signaling in the pathogenesis of tularemia.

F. tularensis dampens MAPK signaling in macrophages to reduce cytotoxicity during intracellular infection
F. tularensis delays the apoptotic response of infected macrophages to preserve its intracellular replicative niche (5,19,(21)(22)(23)(24). We previously demonstrated that this delay in apoptosis is dependent on the TolC protein (21,23). We, therefore, took advantage of the LVS ∆tolC mutant to probe host cell signaling pathways modulated during infection. Given the role of MAPKs in the host response to F. tularensis (33)(34)(35)(36)(37), we examined if infection of murine bone marrow-derived macrophages (BMMs) with the LVS ΔtolC mutant leads to increased p38, SAPK/JNK, or ERK1/2 activation in comparison to infection with the WT LVS. BMMs were infected with either the WT or ΔtolC LVS, or left uninfected, and MAPK activation (i.e., phosphorylation) was measured via immuno blot analysis of cell lysates. As shown in Fig. 1A through D, infection with the WT LVS did not cause significant increases in phosphorylation of the MAPKs compared to the uninfected controls. In contrast, infection with the ∆tolC LVS triggered increased levels of phosphorylated p38 and SAPK/JNK at 30 min p.i. compared to both the WT LVS-infected macrophages and uninfected controls ( Fig. 1A through D). These differences were no longer observed by 60 and 120 min p.i. and were also not observed at later time points (12 h p.i.; data not shown). In contrast to p38 and SAPK/JNK, we did not observe differences in levels of phospho-ERK1/2 between the infected BMMs and uninfected controls ( Fig. 1A and D). These results demonstrate that both the p38 and SAPK/JNK MAPK pathways are dampened by F. tularensis, via a TolC-dependent mechanism, during the early macrophage response to infection.
Both the p38 and SAPK/JNK MAPK pathways are important for regulating apoptosis (32,44,45). To determine if activation of p38 and SAPK/JNK during F. tularensis infection impacts the macrophage apoptotic response, BMMs were treated with the inhibitors SB203580 or SP600125, which target p38 or SAPK/JNK, respectively (36,(46)(47)(48)(49)(50). BMMs were treated with inhibitor, or DMSO as a control, and at 24 h p.i., release of lactate dehydrogenase (LDH) was measured as a marker for host cell death and CFUs were determined to measure intracellular bacterial replication. In the control BMMs, infection with the ΔtolC LVS resulted in an approximately threefold increase in host cell death and~1 log decrease in bacterial replication compared to infection with the WT LVS ( Fig.  1E and F). Thus, as previously observed (21,41), the F. tularensis ∆tolC mutant is defective in dampening host cell death and, as a result, exhibits decreased intracellular replication. Treatment of BMMs with the SAPK/JNK inhibitor SP600125 did not alter these ∆tolC mutant phenotypes ( Fig. 1E and F). In contrast, for BMMs treated with the p38 inhibitor SB230580, infection with the ∆tolC mutant resulted in decreased host cell death, and the intracellular replication defect of the mutant was diminished ( Fig. 1E and F). Together, these results implicate p38 in the macrophage apoptotic response to F. tularensis infection and suggest that the bacteria dampen this response early during infection to preserve the intracellular niche. or DMSO as a vehicle control, and then infected at an MOI of 50 with the WT or ΔtolC LVS. At 24 h p.i., supernatant fractions were collected and assayed for levels of LDH release (E) or cell lysates were plated for colony-forming units (CFUs) (F). Data in (B-F) represent means ± SEM of three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant; calculated by two-way ANOVA test with Tukey's multiple-comparison posttest.

P38 phosphorylation and the downstream apoptotic response occur via a TLR2-MYD88 signaling pathway
Activation of p38 in response to microbial infection occurs downstream of TLR-signaling pathways (45). To determine if TLR2 is required for activation of p38 during F. tularensis infection, we compared BMMs isolated from WT or TLR2 −/− mice. BMMs were infected with the WT or ∆tolC LVS, or left uninfected, and then probed for phosho-p38. In agreement with the preceding results, increased p38 phosphorylation was detected for the WT BMMs infected with the ΔtolC mutant at early times p.i. ( Fig. 2A and B). However, no increased p38 phosphorylation was observed for the TLR2 −/− BMMs ( Fig. 2A and B).
To test if MYD88 is required together with TLR2 for MAPK activation in response to F. tularensis infection, we examined p38 phosphorylation in BMMs isolated from WT or MYD88 −/− mice. Similar to the TLR2 −/− macrophages, increased p38 phosphorylation was no longer observed for the MYD88 −/− BMMs infected with the ΔtolC LVS ( Fig. 2C and D). F. tularensis can also activate p38 through redox signaling via reactive oxygen species (ROS) produced by the NADPH oxidase (NOX2) (13). In macrophages, the production of ROS by NOX2 occurs downstream of TLR activation and requires the p47 phox subunit (51,52). However, in contrast to infection of the TLR2 −/− and MYD88 −/− macrophages, infection of p47 phox−/− BMMs with the ∆tolC LVS still resulted in increased p38 phosphorylation at 30 and 60 min p.i. (Fig. 2E and F). Together, these results confirm that macrophages sense F. tularensis infection via TLR2 and demonstrate that the bacteria interfere with the TLR2-MYD88 pathway, independent of the host NADPH oxidase, to prevent p38 signaling. We next examined the connection of TLR2-MYD88 signaling to the downstream macrophage cell death response to F. tularensis infection. BMMs isolated from WT, TLR2 −/− or MYD88 −/− mice were infected with WT or ΔtolC LVS, and host cell death was measured by LDH release at 24 h p.i. Whereas infection of WT BMMs with the ∆tolC mutant led to an approximately twofold increase in cell death compared to infection with the WT LVS, this was no longer observed for the TLR2 −/− and MYD88 −/− BMMs ( Fig. 3A and C). In agreement with these results, intracellular replication of the ∆tolC mutant was similar to the WT LVS in both the TLR2 −/− and MYD88 −/− BMMs ( Fig. 3B and D). Thus, the intracellular replicative niche for the ∆tolC mutant is preserved in the absence of host cell signaling through TLR2 and MYD88. No changes in the host cell death or intracellular replication phenotypes of the LVS ∆tolC mutant were observed during infection of p47 phox−/− BMMs (Fig. S1). The LVS ∆tolC mutant triggers activation of the apoptotic pathway as early as 6 h p.i., whereas WT bacteria actively suppress and delay this apoptotic response (21). To determine if TLR2 signaling contributes to the downstream apoptotic response during F. tularensis infection, we compared levels of cleaved caspase-3 in WT or TLR2 −/− BMMs infected with the WT or ΔtolC LVS. In WT BMMs, increased caspase-3 cleavage was detected at 18 h p.i. in response to the ΔtolC mutant when compared to infection with the WT bacteria or uninfected control ( Fig. 3E and F). In contrast, no difference in caspase-3 cleavage was observed for infection of the TLR2 −/− macrophages ( Fig. 3E and F). Thus, F. tularensis delays macrophage apoptosis by dampening TLR2 signaling via a TolC-dependent mechanism.

F. tularensis Schu S4 modulates MAPK signaling and TLR2-dependent innate immune responses during macrophage infection
While the LVS provides an excellent model for studying F. tularensis pathogenesis, findings obtained with attenuated strains do not always recapitulate in human virulent strains (53)(54)(55). Therefore, we assayed whether the human virulent Schu S4 strain similarly dampens MAPK activation and TLR2 signaling via TolC during infection of BMMs, as observed for the LVS. To examine MAPK responses, BMMs were infected with WT or ∆tolC Schu S4, or left uninfected, and phosphorylated, and total levels of p38, SAPK/JNK, and ERK1/2 were determined at 30 and 60 min p.i. Infection with the Schu S4 ∆tolC mutant triggered increased levels of phospho-p38 at both time points when compared to macrophages infected WT Schu S4 or uninfected controls ( Fig. 4; Fig. S2). This agrees with the results obtained with the LVS. Increased SAPK/JNK phosphorylation was also observed for BMMs infected with the Schu S4 ∆tolC mutant ( Fig. 4; Fig. S2). In contrast to the LVS, increased ERK1/2 phosphorylation was additionally observed at 30 min p.i. for BMMs infected with the Schu S4 ∆tolC mutant ( Fig. 4; Fig. S2). These results show that MAPK signaling is broadly activated early during infection of macrophages by shown. (B, D, F) Ratio of phosphorylated to unphosphorylated p38, relative to the uninfected control, determined from densitometric analysis of corresponding blots from three independent experiments. Data represent means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; P < 0.0001; ns, not significant; calculated by two-way ANOVA test with Tukey's multiple-comparison posttest.
Research Article mBio the human virulent F. tularensis Schu S4 strain, and the bacteria dampen this host response in a TolC-dependent manner. We next tested whether TLR2 signaling is required for the downstream cell death response of macrophages to Schu S4 infection, as found for the LVS. WT or TLR2 −/− BMMs were infected with WT or ∆tolC Schu S4, and cell death was measured by LDH release at 24 h p.i. Infection of WT BMMs with the Schu S4 ∆tolC mutant led to an approximately sixfold increase in cell death compared to infection with WT bacteria (Fig. 4B). However, the increased cytotoxicity of the ∆tolC mutant was no longer observed in the TLR2 −/− BMMs (Fig. 4B). In agreement with this, intracellular replication of the Schu S4 ∆tolC Research Article mBio mutant was rescued in the TLR2 −/− vs WT BMMs (Fig. 4C). Thus, the F. tularensis LVS and Schu S4 strain both delay macrophage cell death by interfering with TLR2-dependent signaling to maintain the intracellular niche and maximize bacterial replication. In addition to its ability to delay programmed cell death pathways during infec tion, the Schu S4 strain dampens pro-inflammatory cytokine responses of BMMs in a TolC-dependent manner (23). We, therefore, examined the role of TLR2 signaling in the downstream cytokine response elicited by the SchuS4 ∆tolC mutant during BMMs infection. WT or TLR2 −/− BMMs were infected with WT or ∆tolC Schu S4, or left uninfec ted, and release of the pro-inflammatory cytokine IL-6 into the culture medium was determined by ELISA at 24 h p.i. As previously observed (23), the Schu S4 ∆tolC mutant elicited elevated release of IL-6 ( Fig. 4D) compared to infection with WT Schu S4 or uninfected controls. However, no increase in IL-6 levels was observed for infection of the TLR2 −/− BMMs (Fig. 4D). Thus, TLR2 is integral to the macrophage pro-inflammatory response to infection by the human virulent F. tularensis Schu S4 and the bacteria disrupt TLR2 signaling via TolC to suppress cytokine responses of the host.

Research Article mBio
The LVS ∆tolC mutant regains full virulence in MYD88 −/− mice Given our findings that F. tularensis targets TLR2-and MYD88-dependent signaling during macrophage infection, we next sought to determine the impact of this bacteriahost interplay in the mouse pneumonic tularemia model. WT and TLR2 −/− C57BL/6J mice were challenged by the intranasal route with 5 × 10 5 CFU of the WT or ΔtolC LVS, and survival and weight loss were monitored for 14 d p.i. As expected (21,41), WT mice infected with the WT LVS had a mean survival time of 7 d, whereas all WT mice infected with the ΔtolC mutant survived for the duration of the experiment (Fig. 5A and C). In comparison to the WT mice, TLR2 −/− mice infected with the WT LVS exhibited a faster time to death, with a mean survival time of 6 d (Fig. 5A). In TLR2 −/− mice infected with the ΔtolC LVS, one mouse succumbed to infection on day 9 p.i., while all other littermates survived (Fig. 5A). Thus, the ∆tolC mutant remains attenuated in vivo in the absence of TLR2.
Weight loss serves as an indicator of disease severity in mouse infection models (56). Consistent with the ability of F. tularensis to suppress host responses during the first 2-3 d p.i (17), WT mice infected with the WT LVS did not begin to exhibit weight loss until day 3 p.i., which was then followed by a steep decline in weight prior to succumbing to infection ( Fig. 5B and D). In contrast, WT mice infected with the ∆tolC mutant exhibited weight loss starting on day 1 p.i. (Fig. 5B and D). WT mice infected with the ΔtolC mutant recovered from this initial decline and started to regain weight between days 4-5 p.i. Similar to WT mice, TLR2 −/− mice infected with the WT LVS did not begin to lose weight until day 3 p.i. (Fig. 5B). Of note, the TLR2 −/− mice infected with the LVS ∆tolC mutant no longer exhibited early weight loss on days 1-2 p.i. Instead, the TLR2 −/− mice exhibited an initial delay in weight loss, followed by a steep decline on days 3-4, mirroring the pattern observed for mice infected with the WT bacteria (Fig. 5B). The ∆tolC-infected TLR2 −/− mice rebounded starting on day 5 p.i., with the mice eventually regaining full body weight. Collectively, the survival and weight loss results indicate a role for TLR2 in the early response to F. tularensis infection; however, even in the absence of TLR2, the host is able to control infection with the LVS ∆tolC mutant.
We next compared infection of WT and MYD88 −/− mice with the WT or ∆tolC LVS. The infection phenotype of the MYD88 −/− mice was distinct from both the WT and TLR2 −/− mice. MYD88 −/− mice infected with the WT LVS exhibited a faster time to death, with a mean survival time of 5 d (Fig. 5C). Notably, the virulence of the LVS ∆tolC mutant was fully restored in the MYD88 −/− mice, with all mice succumbing to infection with kinetics similar to mice infected with the WT bacteria (Fig. 5C). The weight loss patterns for the MYD88 −/− mice were also distinct compared to the WT and TLR2 −/− mice. MYD88 −/− mice infected with both the WT and ΔtolC LVS exhibited an extended delay in weight loss, with the mice maintaining full weight until day 4 p.i. Additionally, by the time the WT and ΔtolC infected MYD88 −/− mice started to lose weight, both groups quickly succumbed to infection. Together, these data demonstrate a critical role for MYD88-dependent pathways in host protection against F. tularensis infection and suggest that, in the absence of MYD88, TolC-mediated suppression of host inflammatory responses is no longer required for bacterial virulence.

DISCUSSION
F. tularensis delays both programmed cell death and pro-inflammatory cytokine respon ses during intracellular infection to ensure its replication and survival. TLR2-and MYD88dependent signaling pathways are central regulators of the pro-inflammatory response to F. tularensis (26)(27)(28)(29)(30)(38)(39)(40). However, mechanisms by which F. tularensis modulates programmed cell death responses during infection are still not understood despite being fundamental to virulence. We previously established that the outer membrane channel protein TolC is important for the ability of both the LVS and human virulent Schu S4 strain to delay apoptosis and pro-inflammatory responses during macrophage infection (21,23,41,43). In this study, we took advantage of the ΔtolC mutant phenotype to uncover host-signaling pathways that respond to F. tularensis and are disrupted by the bacteria during infection. Our results reveal that F. tularensis interferes with TLR2-MYD88-p38 signaling in a TolC-dependent manner to block macrophage apoptotic responses early during infection. Further, we show that this strategy is conserved between the attenuated LVS and human virulent Schu S4 strain. Comparison of WT and ∆tolC LVS in the mouse pneumonic tularemia model revealed distinct roles for TLR2 and MYD88 in early responses to infection and host survival. MYD88-deficient mice were unable to maintain protection against infection with the LVS ΔtolC mutant, identifying MYD88 as critical for host defense and a key target for F. tularensis subversion.
F. tularensis induces early MAPK activation in both murine and human macrophages that is dampened by ~2 h p.i. although conflicting reports exist on the extent of MAPK activation during Francisella infection (13, 16, 33-37, 57-59). Our results indicate that  Our results suggest that TolC-dependent activity counteracts early TLR2-MYD88 signaling, while the bacteria are still within the phagosome. Subsequent downmodu lation of MAPK activity then occurs independent of TolC due to rapid exit from the phagosome together with increased MKP-1 expression. The activation of p38 and SAPK/JNK is commonly associated with the induction of apoptosis (32,44). Many intracellular pathogens have evolved to block p38 signaling to prevent apoptosis, and the use of p38 inhibitors during infection of monocyte-derived cells by the LVS was shown to reduce host cell death (36,45). In line with these observa tions, we found that inhibiting p38 activity early during infection reduced cytotoxicity of the LVS ΔtolC mutant in macrophages. In contrast, treatment with the SAPK/JNK inhibitor did not alter the ΔtolC mutant phenotype. The increased SAPK/JNK phosphorylation we observed at 30 min p.i. suggests that SAPK/JNK is likely to perform other functions in the early host response to F. tularensis, such as the expression of cytokines, and the individual role of each MAPK warrants further investigation.
During F. tularensis infection of macrophages, both TLR2-and NADPH oxidasedependent redox signaling lead to p38 activation (13,33,34,61). Our data demonstrate that F. tularensis suppresses TLR2-and MYD88-dependent activation of p38 early during infection in a TolC-dependent manner. Moreover, our results show that evasion of TLR2-mediated signaling allows the bacteria to delay host apoptotic responses and thereby maximize replication in the protected intracellular niche. In contrast, we did not observe a role for the NADPH oxidase in MAPK activation or host cell death for either the WT or ΔtolC LVS. This is consistent with previous observations that Francisella inhibits the oxidative burst immediately upon infection via the activity of alkaline phosphatases and antioxidants, and that secretion of the major alkaline phosphatase AcpA occurs independently of TolC (13,14,62).
Like the LVS, we found that the human virulent Schu S4 strain suppresses activation of p38 and SAPK/JNK in a TolC-dependent manner. However, different from the LVS, we also observed increased ERK1/2 phosphorylation with the Schu S4 ΔtolC mutant. Therefore, the human virulent Schu S4 strain may more broadly activate MAPK signaling in host cells compared to the attenuated LVS. As found with the LVS, TLR2 signaling was required for the increased cytotoxicity and decreased intracellular replication observed during infection with the Schu S4 ∆tolC mutant. Expression of the pro-inflammatory cytokine IL-6 is also known to be induced downstream of TLR2 and p38 (26,45,63). In agreement with this, we found that TLR2 signaling was required for increased release of IL-6 in response to infection with the Schu S4 ∆tolC mutant. Together, our results identify a key role for the TLR2-p38 axis in macrophage recognition of both attenuated and human virulent strains of F. tularensis. Future studies will be needed to define a possible role for ERK1/2 during Schu S4 infection of macrophages and to determine if F. tularensis similarly activates MAPKs in human macrophages.
Despite similar requirements for TLR2 and MYD88 in controlling intracellular replication of Francisella in macrophages, we found that MYD88 plays a more significant role in protection than TLR2 in the mouse pneumonic tularemia model. While the TLR2 −/ − mice exhibited a faster time to death upon infection with the WT LVS, the TLR2 −/− mice were still able to control and survive infection with the ∆tolC mutant. In contrast, not only did the MYD88 −/− mice succumb to infection faster with the WT LVS, but also the virulence of the ∆tolC mutant was fully restored. Our data agree with previous reports that MYD88 serves a broader role than TLR2 in mediating host protection against Francisella (26,(38)(39)(40)64). MYD88 is required for the secretion of IFN-γ in response to F. tularensis infection, and IFN-γ is capable of restricting the growth of virulent F. tularensis both in vitro and in vivo (65)(66)(67). MYD88-dependent IFN-γ production is independent of TLR2 and, instead, is driven by IL-18 signaling via the IL-1R (39,40). Thus, F. tularensis may antagonize MYD88 to block both TLR and IL-1R signaling during infection. In the absence of MYD88, both signaling networks are disrupted, and TolC-mediated suppression of host inflammatory responses is no longer required for bacterial virulence.
Our results support a central role for TolC in the ability of F. tularensis to subvert TLR2-MYD88-MAPK signaling during infection. However, the mechanism by which TolC acts on these host pathways remains to be determined. We previously demonstrated that the F. tularensis ΔtolC mutant has no observable growth defect or compromised membrane integrity and that the mutant replicates intracellularly without activating caspase-1 (21,23,41,43). This suggests that bacterial lysis or other structural defect is not responsible for the increased host cell death and pro-inflammatory responses triggered by infection with the ∆tolC mutant. Our finding that the ∆tolC mutant regains full virulence in mice lacking MYD88 further argues against a general role for TolC in bacterial integrity or other function. One possibility is that TolC (which is not a lipopro tein itself ) may act, either directly or indirectly, to limit exposure of lipoproteins or other molecules that are recognized by TLR2 during infection. A second, non-exclusive, possibility is that TolC may disrupt host cell signaling through the secretion of effector proteins as part of a type I secretion system. Further work is needed to identify potential effectors secreted via TolC or if TolC might function through a different mechanism.
In summary, we present evidence that F. tularensis delays apoptosis during the infection of macrophages by limiting TLR2-MYD88-p38 signaling at the early stages of infection. This TolC-dependent activity allows F. tularensis to preserve its intracellular niche and prolong time for replication in the protected host cell environment. Early recognition of F. tularensis by TLR2 and MYD88 contributes to host protection during both macrophage and in vivo infection, with MYD88 playing a more critical role in vivo. F. tularensis subversion of TLR2-MYD88-p38 signaling is likely central to its ability to remain immunologically silent during the early stages of infection and key to its extreme virulence. Further understanding of the mechanism by which F. tularensis disrupts host cell signaling via TolC will enhance our understanding of the pathogenesis of tularemia and provide targets for therapeutic intervention.

Bacteria and growth conditions
The F. tularensis LVS and SCHU S4 strain (BEI Research Resources Repository, Manassas, VA, USA) were grown on chocolate II agar plates (BD Biosciences, Franklin Lakes, NJ, USA) or in modified Mueller-Hinton broth (MHB; Mueller-Hinton broth BD Biosciences containing 1% glucose, 0.025% ferric pyrophosphate, and 0.05% L-cysteine hydrochlor ide). All growth and manipulations of the Schu S4 strain were performed under biosafety level 3 (BSL3) containment conditions.
All protocols involving animals were approved by the Institutional Animal Care and Use Committee of Stony Brook University.

Mouse infections
Inoculums for the mouse infections were prepared by growing bacteria for 60 h on plates. Bacteria were then scraped, washed with phosphate-buffered saline (PBS), and resuspended in MHB supplemented with 10% sucrose (MHB-Sucrose). Inoculums were then serially diluted in MHB-sucrose and frozen at −80°C until use. Intranasal mouse infections were performed by administering 5.0 × 10 5 CFU in a 10-µL inoculum into each narris (20 µL total). Actual infectious doses were determined by retrospective CFU counts. All mice were then monitored for 14 d.

BMM infections
For all BMM infections using the LVS, cells were seeded as described above and infected with the WT or ΔtolC LVS at an MOI of 50. Plates were then centrifuged at 100 × g for 5 min to synchronize infection. For all Schu S4 infections, BMMs were infected at an MOI of 500. Plates were not centrifuged following the addition of bacteria to prevent the generation of aerosols, necessitating the higher MOI. For intracellular replication, cytotoxicity, ELISA, and caspase-3 measurements, BMMs were infected for 2 h, washed with PBS, and incubated for an additional 1 h with 10 µg/mL of gentamicin to remove extracellular bacteria. After gentamicin treatment, cells were washed with PBS and then incubated with fresh BMIM (lacking gentamicin) until the desired time points. To measure MAPK activation, BMMs were infected for the indicated times without the addition of gentamicin. For experiments using MAPK inhibitors, BMMs were pretreated for 30 min prior to infection with 10 µM of the p38 inhibitor SB203580, 15 µM of the SAPK/JNK inhibitor SP62001, or DMSO as a vehicle control. Inhibitor concentrations were chosen based on the published literature (36,(46)(47)(48)(49)(50). We also confirmed that the addition of either inhibitor at these concentrations does not interfere with bacterial replication in broth culture. The inhibitors were removed prior to gentamicin treatment.

Cytotoxicity assays
At 24 h p.i., supernatants from infected BMMs were collected and analyzed for LDH release using the CytoTox 96 nonradioactive cytotoxicity assay (Promega) following the manufacturer's protocol. Supernatant fractions from uninfected BMM controls were used to determine background LDH release, and maximum LDH release was determined from BMM that were lysed via a single −80°C/37°C freeze-thaw cycle. Percent LDH release was calculated by subtracting the background LDH release from all samples, dividing the resulting sample values by the value for the maximum LDH release, and multiplying by 100.

Intracellular replication assays
At 24 h p.i., BMMs were washed twice with PBS and then lysed in DMEM with 0.1% deoxycholate for 10 min at room temperature. Serial dilutions were then performed in PBS with the cell lysates, plated, and incubated for 3 d prior to enumeration of CFUs.

Cytokine analysis
At 24 h p.i., culture supernatants from BMMs infected with the Schu S4 strains were collected following approved BSL3 guidelines. Levels of IL-6 were determined using a mouse IL-6 ELISA kit (Biolegend) according to the manufacturer's protocol. Values were recorded as picograms per milliliter.

Statistical analysis
For the LDH release, intracellular replication, and MAPK experiments, one-tailed P values were calculated by two-way ANOVA test with Tukey's multiple-comparison posttest. The mouse survival studies were analyzed by the log-rank test. Statistical calculations were performed using GraphPad Prism.