Caspase-11 Activation in Response to Bacterial Secretion Systems that Access the Host Cytosol

Inflammasome activation is important for antimicrobial defense because it induces cell death and regulates the secretion of IL-1 family cytokines, which play a critical role in inflammatory responses. The inflammasome activates caspase-1 to process and secrete IL-1β. However, the mechanisms governing IL-1α release are less clear. Recently, a non-canonical inflammasome was described that activates caspase-11 and mediates pyroptosis and release of IL-1α and IL-1β. Caspase-11 activation in response to Gram-negative bacteria requires Toll-like receptor 4 (TLR4) and TIR-domain-containing adaptor-inducing interferon-β (TRIF)-dependent interferon production. Whether additional bacterial signals trigger caspase-11 activation is unknown. Many bacterial pathogens use specialized secretion systems to translocate effector proteins into the cytosol of host cells. These secretion systems can also deliver flagellin into the cytosol, which triggers caspase-1 activation and pyroptosis. However, even in the absence of flagellin, these secretion systems induce inflammasome activation and the release of IL-1α and IL-1β, but the inflammasome pathways that mediate this response are unclear. We observe rapid IL-1α and IL-1β release and cell death in response to the type IV or type III secretion systems of Legionella pneumophila and Yersinia pseudotuberculosis. Unlike IL-1β, IL-1α secretion does not require caspase-1. Instead, caspase-11 activation is required for both IL-1α secretion and cell death in response to the activity of these secretion systems. Interestingly, whereas caspase-11 promotes IL-1β release in response to the type IV secretion system through the NLRP3/ASC inflammasome, caspase-11-dependent release of IL-1α is independent of both the NAIP5/NLRC4 and NLRP3/ASC inflammasomes as well as TRIF and type I interferon signaling. Furthermore, we find both overlapping and non-redundant roles for IL-1α and IL-1β in mediating neutrophil recruitment and bacterial clearance in response to pulmonary infection by L. pneumophila. Our findings demonstrate that virulent, but not avirulent, bacteria trigger a rapid caspase-11-dependent innate immune response important for host defense.

Caspase-11 participates in the activation of a non-canonical inflammasome that induces cell death and the secretion of IL-1a and IL-1b in response to Gram-negative pathogens, such as Escherichia coli and Vibrio cholerae, and to particular toxins, such as the cholera toxin B subunit [26][27][28][29]. This non-canonical, caspase-11-dependent response to Gram-negative bacteria is independent of virulence-associated secretion systems that deliver bacterial molecules into the host cytosol and requires LPS-induced TLR4 signaling through the adaptor TIR-domain-containing adaptorinducing interferon-b (TRIF) and TRIF-dependent type I interferon (IFN) production. Type I IFN signaling through the type I IFN receptor (IFNAR) is required for caspase-11 upregulation and activation, but how type I IFN mediates activation of caspase-11 is not well-defined [27][28][29]. Caspase-11 contributes to NLRP3-dependent activation of caspase-1 and subsequent caspase-1-dependent IL-1b secretion and cell death. Caspase-11 also facilitates an NLRP3-and caspase-1-independent pathway that results in cell death and release of IL-1a [26][27][28][29]. This caspase-11-dependent, caspase-1-independent pathway is responsible for LPS-induced septic shock in vivo [26,30]. Although caspase-11 is activated in response to signals from Gram-negative pathogens and certain pore-forming toxins, whether caspase-11 contributes to inflammasome activation in response to virulenceassociated secretion systems that deliver bacterial ligands into host cytosol is unknown.
Bacterial pathogens use evolutionarily conserved secretion systems, such as type III or type IV secretion systems (T3SS or T4SS), to translocate effector proteins into the cytosol of host cells [31,32]. In addition to bona fide virulence factors, these secretion systems also translocate bacterial molecules such as flagellin or structural components of the secretion machinery itself, which results in inflammasome activation [14,16,[33][34][35][36]. Legionella pneumophila, an opportunistic pathogen that causes a severe pneumonia known as Legionnaires' disease [37,38], utilizes its dot/icm-encoded T4SS as a virulence factor to translocate bacterial effector proteins into the host cell cytosol and establish a replicative vacuole [39][40][41][42][43][44][45][46]. L. pneumophila induces T4SS-dependent inflammasome activation through two genetically distinct pathways [47]. T4SS-mediated translocation of flagellin into the cytosol triggers caspase-1 activation and pyroptosis through the NLR NAIP5 in conjunction with another NLR, NLRC4 [16,36,[47][48][49][50]. Caspase-1 activation is also triggered independently of the NLRC4/flagellin pathway through the adaptor protein ASC, but the bacterial factor that is recognized and the upstream proteins that regulate this pathway remain unknown [47,51]. However, although ASC is necessary for robust secretion of IL-1b in response to L. pneumophila as well as a number of pathogens, such as Salmonella or Yersinia species which employ T3SSs, ASC is dispensable for induction of pyroptosis that is rapidly triggered in response to these infections. We therefore considered the possibility that in addition to its role in delayed inflammasome activation in response to Gram-negative bacteria, caspase-11 might participate in rapid cell death and release of IL-1a in response to the presence of bacterial pathogens that access the host cell cytosol by means of type IV and type III secretion systems.
Here, we demonstrate that IL-1a and IL-1b are rapidly released in response to bacterial T4SS activity independently of bacterial flagellin. In this system, we find IL-1b secretion requires caspase-1, but caspase-1 is dispensable for cell death and IL-1a release in response to a functional L. pneumophila T4SS. Instead, caspase-11 is required for both IL-1a release and cell death in response to L. pneumophila T4SS activity. Consistent with recent findings, caspase-11 contributes to optimal NLRP3-mediated caspase-1 activation and IL-1b secretion in response to L. pneumophila. However, caspase-11-dependent IL-1a release and cell death in L. pneumophila-infected cells are independent of the NAIP5/NLRC4 and NLRP3/ASC inflammasomes. In contrast to the role of TRIF and IFNAR in the response against Gram-negative bacteria, caspase-11 activation and cytokine release in response to the T4SS of L. pneumophila are independent of both TRIF and IFNAR signaling. We further demonstrate that T3SS activity of the unrelated pathogen Yersinia pseudotuberculosis induces a similarly rapid caspase-11-dependent response that also leads to cell death and release of IL-1a and IL-1b. Finally, we find that both IL-1a and IL-1b are critical in vivo for neutrophil recruitment and bacterial clearance. Overall, our data show that caspase-11 is poised to respond robustly to a conserved feature of pathogenic bacteria, bacterial access to the host cytosol through specialized secretion systems. This establishes caspase-11 as a critical regulator of immune system-mediated discrimination of pathogenic and nonpathogenic bacteria.

Results
LPS priming induces rapid IL-1a and IL-1b secretion in response to L. pneumophila T4SS activity L. pneumophila infection induces IL-1a and IL-1b secretion that requires T4SS activity [47,52]. IL-1b secretion is regulated by a flagellin-dependent NAIP5/NLRC4 inflammasome and a poorly defined ASC inflammasome that both activate caspase-1 [47,51]. The mechanisms underlying IL-1a secretion are less clear, but IL-1a secretion is still robustly induced by flagellin-deficient L. pneumophila, which do not activate the NAIP5/NLRC4 inflammasome [52]. Recent studies have described a non-canonical inflammasome triggered in response to Gram-negative bacteria. This non-canonical inflammasome requires lipopolysaccharide (LPS) for the upregulation and activation of caspase-11 and subsequent IL-1a and IL-1b release [26][27][28][29]. Whether caspase-11 is also activated in response to bacteria that use specialized secretion systems to translocate bacterial molecules into the host cytosol is unknown. We thus hypothesized that LPS priming would upregulate caspase-11, pro-IL-1a, and pro-IL-1b and allow for more robust and rapid IL-1a and IL-1b secretion in response to T4SS activity. To test this, we first compared IL-1a and IL-1b release in unprimed and LPS-primed bone marrow-derived macrophages (BMDMs). As shown previously [48,52], unprimed BMDMs secrete robust levels of IL-1a and IL-1b by 20 hours postinfection with wild-type L. pneumophila (WT Lp) ( Figure 1A). Slightly attenuated levels of secreted IL-1a and IL-1b are observed with flagellin-deficient L. pneumophila (DflaA Lp), which do not activate the NAIP5/NLRC4 inflammasome [17,18]. Secretion of both

Author Summary
The inflammasome, a multiprotein complex, is critical for host defense against bacterial infection. The inflammasome activates the host protease caspase-1 to process and secrete IL-1b. Another caspase, caspase-11, can cause cell death and IL-1a release. The bacterial signals that trigger caspase-11 activation are poorly understood. A common feature of many bacterial pathogens is the ability to inject virulence factors and other bacterial molecules into the host cell cytosol by means of a variety of virulenceassociated secretion systems. These secretion systems can introduce bacterial flagellin into the host cytosol, which leads to caspase-1 activation and cell death. However, many bacteria lack or down-regulate flagellin yet still activate the inflammasome. Here, we show that the type IV secretion system of Legionella pneumophila and the type III secretion system of Yersinia pseudotuberculosis rapidly trigger caspase-11 activation in a flagellin-independent manner. Caspase-11 activation mediates two separate inflammasome pathways: one leading to IL-1b processing and secretion, and one leading to cell death and IL-1a release. Furthermore, we find these caspase-11-regulated cytokines are critical for neutrophil recruitment to the site of infection and clearance of non-flagellated Legionella in vivo. Overall, our findings show that virulent bacteria activate a rapid caspase-11-dependent immune response that plays a critical role in host defense.
cytokines is significantly diminished during infection with L. pneumophila lacking DotA, an essential component of the T4SS (DdotA Lp), and is significantly diminished in caspase-1/caspase-11deficient (Casp1 2/2 Casp11 2/2 ) macrophages as well ( Figure 1A). The diminished IL-1 secretion induced by DdotA Lp is not due to a lack of pro-IL-1 production, as DdotA Lp and WT Lp induce robust levels of pro-IL-1b ( Figure S1A). At 4 hours postinfection, unprimed macrophages do not secrete IL-1 ( Figure 1B). However, LPS-primed cells rapidly secrete IL-1a and IL-1b, and this secretion is abrogated in Casp1 2/2 Casp11 2/2  Figure 1B). Secretion of IL-18, another IL-1 family cytokine, also requires T4SS activity and is eliminated in Casp1 2/2 Casp11 2/2 cells ( Figure S1B). Comparable levels of the caspase-1/ caspase-11-independent cytokines IL-12 and TNF-a are secreted in the absence and presence of LPS priming ( Figure S1C-D). These data suggest that LPS priming upregulates a factor required for rapid IL-1a and IL-1b release in response to L. pneumophila T4SS activity.
Caspase-1 catalytic activity is required for IL-1b but not IL-1a secretion Secretion of IL-1b in response to both canonical and noncanonical inflammasome activation requires caspase-1 [26,53,54]. In contrast, IL-1a release downstream of the non-canonical inflammasome depends on caspase-11, and does not require caspase-1 [26]. To test if the catalytic activity of caspase-1 is required for IL-1a secretion in response to L. pneumophila, we inhibited caspase-1 catalytic activity with the pharmacological inhibitor YVAD-cmk (YVAD). Consistent with previous studies [53], IL-1b secretion in response to L. pneumophila is substantially inhibited by YVAD. However, YVAD has no effect on IL-1a secretion, indicating that IL-1a release in response to L. pneumophila does not require caspase-1 catalytic activity ( Figure 1C), as has been shown for other inflammasome activators [55]. Given that IL-1a secretion occurs more rapidly upon LPS priming, is abrogated in Casp1 2/2 Casp11 2/2 macrophages, and does not require caspase-1 catalytic activity, we considered the possibility that caspase-11 might participate in inflammasome activation during L. pneumophila infection.
Caspase-11 contributes to inflammasome activation in response to flagellin-deficient L. pneumophila To test the genetic requirement for caspase-11 in the inflammasome response to L. pneumophila, we infected BMDMs from either caspase-1-deficient (Casp1 2/2 ) or caspase-11-deficient (Casp11 2/2 ) mice. In the absence of flagellin, caspase-11 is required for IL-1a secretion, whereas it is not essential for IL-1b secretion but contributes to maximal secretion ( Figure 2A). These data suggest that caspase-11 is activated in response to L. pneumophila infection independently of flagellin. Indeed, there is robust processing and secretion of caspase-11 in response to WT Cell death (% cytotoxicity) was measured by LDH release into the supernatants relative to Triton X-100-lysed cells. Graphs show the mean 6 SEM of triplicate wells. (C) Levels of processed caspase-1 (casp-1 p10) in the supernatants and full-length caspase-1 (pro-casp-1) and b-actin in the cell lysates were determined by immunoblot analysis. Data are representative of three independent experiments. *** is p,0.001 by two-way ANOVA with Bonferroni post-test, ** is p,0.01 by two-way ANOVA with Bonferroni post-test, and * is p,0.05 by unpaired t-test. NS is not significant. doi:10.1371/journal.ppat.1003400.g002 Specialized Secretion Systems Activate Caspase-11 PLOS Pathogens | www.plospathogens.org and DflaA Lp ( Figure S2). In accordance with previous findings [26,53], caspase-1 is absolutely required for IL-1b secretion. In contrast, we observe robust IL-1a release even in the absence of caspase-1. Both IL-1a and IL-1b release in response to DflaA Lp are caspase-11-dependent in both primed and unprimed macrophages (Figures 2, S3A-B), making L. pneumophila distinct from other Gram-negative bacteria that require priming to induce robust caspase-11 upregulation and activation [27]. Thus, while caspase-11 contributes to maximal caspase-1-dependent IL-1b secretion, it is both necessary and sufficient for IL-1a release in response to flagellin-deficient L. pneumophila.
Cell death in B6 BMDMs is partially flagellin-dependent, but is flagellin-independent in Casp1 2/2 BMDMs ( Figure 2B). Importantly, cell death in response to flagellin-deficient L. pneumophila requires caspase-11, thus correlating caspase-11-dependent cell death with IL-1a release from host cells. In contrast, and consistent with previous findings [26], LPS+ATP induces canonical caspase-1-dependent pyroptosis and secretion of IL-1a and IL-1b that is independent of caspase-11. Because caspase-1 must be processed to mediate IL-1b secretion [53], we examined whether caspase-1 processing is decreased in the absence of caspase-11, which could account for the decreased IL-1b secretion in response to DflaA Lp. Caspase-1 processing is slightly attenuated but not abrogated in response to DflaA Lp in Casp11 2/2 macrophages, consistent with the slight decrease in IL-1b secretion (Figures 2C, S3C). Thus, flagellin-deficient L. pneumophila trigger a canonical caspase-1-dependent inflammasome as well as a non-canonical caspase-11-dependent inflammasome.
We next sought to determine whether IL-1a is also released independently of ASC and NLRC4 during in vivo infection. Because flagellin-deficient L. pneumophila do not activate the NLRC4 inflammasome [16,17,47], infecting Asc 2/2 mice with DflaA Lp eliminates both the ASC and NLRC4 inflammasome pathways. Importantly, the level of IL-1b in the bronchoalveolar lavage fluid (BALF) 24 hours post-infection is significantly attenuated in Asc 2/2 mice infected with DflaA Lp ( Figure 3C). In contrast, the level of IL-1a in the BALF is unaffected even in the absence of both the ASC and NLRC4 inflammasomes. Both IL-1a and IL-1b release are significantly diminished in caspase-1/ caspase-11-deficient mice ( Figure S5). Collectively, our data indicate that L. pneumophila triggers caspase-11 activation and IL-1a release independently of the ASC and NLRC4 inflammasomes during both in vitro and in vivo infection.

Non-canonical inflammasome responses to L. pneumophila occur independently of TRIF and IFNAR
Recent data demonstrate that caspase-11 activation in response to a wide variety of Gram-negative bacteria requires TLR4 signaling through its adaptor TRIF and subsequent type I IFN production [27][28][29]. To determine if L. pneumophila engages a similar TRIF and type I IFN receptor (IFNAR)-dependent pathway for caspase-11 activation, we infected TRIF-deficient (Trif 2/2 ) and IFNAR-deficient (Ifnar 2/2 ) BMDMs. Unlike the response to E. coli, L. pneumophila infection of unprimed macrophages triggered robust cell death and secretion of IL-1a and IL-1b that was independent of IFNAR and TRIF ( Figure 5A-B). Consistently, priming with the TLR1/2 agonist Pam3CSK4, which results in TRIF-and IFNAR-dependent cytokine secretion and cell death in response to E. coli [27], still induced cell death and cytokine secretion in TRIF-and IFNAR-deficient cells in response to L. pneumophila ( Figure S8A-B). These data suggest that during L. pneumophila infection, caspase-11 is upregulated and activated independently of TRIF and IFNAR signaling. Indeed, caspase-11 is still robustly processed and secreted independently of IFNAR and TRIF ( Figures 5C, S9). Notably, substantially upregulated levels of pro-caspase-11 are not observed in the lysates of cells infected with WT or DflaA Lp because both the pro and cleaved forms of caspase-11 are rapidly secreted into the cell supernatant upon infection ( Figures 5C, S9). Accordingly, lysates from IFNAR-and TRIF-deficient macrophages infected with L. pneumophila express comparable levels of pro-caspase-11 to wild-type macrophages, whereas TRIF and IFNAR do contribute to upregulation of pro-caspase-11 in response to E. coli ( Figure S10A-C). When the macrophages are primed with LPS prior to infection, there is a moderate contribution of TRIF and IFNAR signaling to inflammasome activation, consistent with the observation that LPS stimulates the TLR4-TRIF-IFNAR axis involved in caspase-11 upregulation ( Figure  S8C-D). Because the caspase-11-dependent response to L.
pneumophila is TRIF-independent, we investigated whether the TLR signaling adaptor MyD88 contributes to caspase-11 upregulation. When immortalized macrophages deficient for both MyD88 and Trif (iMyd88 2/2 Trif 2/2 ) were infected, caspase-11 upregulation was abrogated in response to both WT and DflaA Lp ( Figure S11A-B), and we were unable to detect caspase-11 activation (data not shown). Thus, although TRIF is not required for caspase-11 activation, a TLR-dependent signal is likely required as the loss of both MyD88 and TRIF eliminates caspase-11 upregulation and activation. Caspase-11 mediates inflammasome activation in response to Yersinia pseudotuberculosis type III secretion system activity Because caspase-11 activation in response to L. pneumophila expressing a functional T4SS is so rapid and robust, we sought to test whether this robust caspase-11-dependent inflammasome activation might be a general response to the activity of specialized secretion systems that allow for bacterial access to the host cytosol. The Yersinia pseudotuberculosis type III secretion system (T3SS) induces inflammasome activation independently of bacterial flagellin and the known secreted effector proteins, and this inflammasome activation is important for bacterial clearance [57]. Since wild-type Yersinia induces cell death that is independent of both caspase-1 and -11 and requires the secreted effector YopJ [57,58], we instead infected Casp1 2/2 Casp11 2/2 , Casp1 2/2 , and Casp11 2/2 BMDMs with a strain of Y. pseudotuberculosis that expresses a T3SS but lacks the six known secreted effectors (D6 Yp). Similarly to L. pneumophila infection, both IL-1a and IL-1b release in response to D6 Yp are caspase-11-dependent ( Figure 6A). Again, caspase-1 is absolutely required for IL-1b secretion, whereas IL-1a is released independently of caspase-1. Secretion of IL-12, an inflammasomeindependent cytokine, is unaffected ( Figure S12). Cell death in response to D6 Yp is both caspase-1 and caspase-11-dependent, with a more dramatic reduction in death in Casp11 2/2 BMDMs ( Figure 6B). Furthermore, Y. pseudotuberculosis-induced release of both IL-1a and IL-1b requires the presence of a functional T3SS, as Y. pseudotuberculosis unable to form a functional T3SS pore in the host cell plasma membrane (DyopB Yp) do not induce secretion of either cytokine. These data indicate a general role for caspase-11 in the induction of rapid cell death and robust release of IL-1a and IL-1b in response to bacterial secretion systems that are capable of accessing the host cell cytosol, but may be independent of the activities of specific virulence factors per se.

IL-1a and IL-1b control bacterial burden and neutrophil recruitment in vivo
As caspase-11 contributes to flagellin-independent IL-1a and IL-1b release from infected macrophages in vitro and IL-1a and IL-1b secretion is flagellin-independent in vivo, we wanted to determine the contribution of IL-1a and IL-1b to host defense against L. pneumophila in vivo. IL-1a and IL-1b both bind the IL-1 receptor (IL-1R), which signals through the MyD88 adaptor protein [59][60][61]. As MyD88 is critical for control of L. pneumophila replication during in vivo infection but deletion of an individual MyD88-dependent TLR or a combination of TLRs does not recapitulate MyD88 deficiency, it is likely that other MyD88-dependent receptors, including the IL-1R, may play a role [62][63][64][65][66]. IL-1R signaling contributes to chemokine production by non-hematopoietic cells during infection with wild-type, flagellinexpressing L. pneumophila [67]. However, the role of IL-1R  signaling during infection with flagellin-deficient L. pneumophila, which do not activate the NAIP5/NLRC4 inflammasome, has not been investigated. We therefore infected B6 and IL-1R-deficient (Il1r1 2/2 ) mice intranasally with DflaA Lp and measured bacterial burden in the lung over the course of seven days. Though both B6 and Il1r1 2/2 mice received similar initial bacterial burdens, Il1r1 2/2 mice show a defect in bacterial clearance as early as 24 hours post-infection ( Figure 7A). Bacterial burden remains elevated in the absence of IL-1R signaling, with the Il1r1 2/2 mice still exhibiting a log-increase in bacterial load at 120 hours post-infection. Since IL-1R signaling is important for neutrophil recruitment [68], we examined whether Il1r1 2/2 mice have a defect in neutrophil recruitment to the pulmonary airway during L. pneumophila infection. Indeed, Il1r1 2/2 mice exhibit a significant decrease in neutrophil recruitment to the airway 24 hours postinfection, possibly contributing to their inability to efficiently clear the pathogen (Figure 7B-C).
The IL-1R signals in response to both IL-1a and IL-1b; however, these cytokines can play non-redundant roles in antibacterial defense [69]. To determine the relative contributions of IL-1a and IL-1b to neutrophil recruitment and bacterial clearance during L. pneumophila infection, we utilized neutralizing antibodies to selectively block either IL-1a or IL-1b prior to infection. Specific cytokine neutralization in the BALF could be observed 24 hours post-infection ( Figure S13). Critically, IL-1a neutralization alone significantly diminishes the percentage of neutrophils recruited to the BALF at 24 hours post-infection and results in a half-log increase in bacterial CFUs, in marked contrast to isotype control antibody or neutralization of IL-1b, which on its own did not have a significant effect ( Figure 7D-F). However, neutralization of both IL-1a and IL-1b fully recapitulates the magnitude of neutrophil reduction and defect in bacterial clearance observed in the Il1r1 2/2 mice. Collectively, these data indicate that although there are some overlapping roles for these cytokines during L. pneumophila infection, IL-1a plays a distinct role from IL-1b in driving neutrophil recruitment to the airway and mediating bacterial clearance.

Discussion
Inflammasomes respond robustly to conserved features of pathogenic microbes, such as pore-forming toxins or specialized secretion systems that access the host cytosol. Inflammasomes therefore play a central role in enabling the immune system to discriminate between virulent and avirulent bacteria [70]. Recent reports show a role for caspase-11 in regulating the activation of a non-canonical inflammasome that promotes cell death as well as IL-1a and IL-1b secretion. This non-canonical inflammasome Figure 8. Caspase-11 controls multiple pathways of inflammasome activation in response to bacterial secretion systems that access the host cytosol. Three distinct inflammasome pathways are induced upon interaction of virulent bacteria with host cells. Translocation of flagellin into the host cytosol by specialized secretion systems triggers a NAIP5/NLRC4/caspase-1 inflammasome that leads to cell death, IL-1a, and IL-1b release. Virulent bacteria induce two separate pathways of caspase-11-dependent inflammasome activation through a two-signal model. First, TLR stimulation by PAMPs (signal one) leads to upregulation of pro-IL-1a, pro-IL-1b, NLRP3, and pro-caspase-11. Next, cytosolic detection of virulence activity, namely type III or type IV secretion (signal two), leads to caspase-11 processing and activation. Active caspase-11 contributes to NLRP3mediated inflammasome activation and caspase-1-dependent IL-1b secretion. Caspase-11 also mediates caspase-1-independent cell death and IL-1a release through a pathway that is independent of the NLRP3/ASC and NAIP5/NLRC4 inflammasomes and involves an unknown host sensor. doi:10.1371/journal.ppat.1003400.g008 responds to both pathogenic and non-pathogenic Gram-negative bacteria independently of specialized secretion systems that translocate bacterial molecules into the host cytosol [26][27][28][29]. This pathway involves the TRIF-and IFNAR-dependent upregulation and activation of caspase-11 and occurs with relatively delayed kinetics in comparison to the response to pathogenic bacteria. Intriguingly, we find that the activity of the L. pneumophila Dot/Icm T4SS leads to rapid and robust caspase-11 activation independently of the TRIF-IFNAR axis, and this activation triggers rapid cell death and release of both IL-1a and IL-1b (Figure 8). We extend these results to show that the evolutionarily distinct T3SS of another pathogen, Y. pseudotuberculosis, also rapidly triggers caspase-11-dependent responses. Collectively, our findings demonstrate that caspase-11 is critical for inflammasome activation in response to the secretion systems of virulent bacteria that enable bacterial molecules to access the host cell cytosol and demonstrate that IL-1a and IL-1b together play a crucial protective role during acute infection in vivo.
We demonstrate that in response to the activity of bacterial secretion systems that enable cytosolic access, caspase-11 contributes to NLRP3-mediated inflammasome activation and caspase-1dependent IL-1b secretion and to a second ASC and NLRC4independent pathway that does not require caspase-1 and leads to cell death as well as robust IL-1a release. These L. pneumophilainduced pathways are similar to recent findings with a number of Gram-negative bacterial pathogens, including C. rodentium, E. coli, and S. typhimurium [26][27][28][29]. However, we observe rapid and robust T4SS-dependent activation of these two caspase-11-mediated pathways by L. pneumophila, whereas the response to Gram-negative bacteria lacking specialized secretion systems occurs less robustly and with much slower kinetics. Intriguingly, we observe a similarly rapid caspase-11-dependent induction of cell death and IL-1 release in response to the structurally and evolutionarily unrelated T3SS of Y. pseudotuberculosis. Importantly, this pathway is independent of host sensing of flagellin, as it is triggered by flagellin-deficient L. pneumophila, and Y. pseudotuberculosis downregulates flagellin expression when the T3SS is expressed [71]. Thus, our data suggest that the caspase-11 inflammasome is poised to respond robustly and rapidly to the activity of bacterial secretion systems that are capable of delivering microbial products to the host cell cytosol and may enable the host to respond to pathogens that evade flagellindependent responses. This could have significance for understanding the role of caspase-11 activation at mucosal sites colonized by large numbers of commensal bacteria. At mucosal barriers, it would be expected that the non-canonical inflammasome pathway would not be robustly activated by commensal bacteria but could respond rapidly to the presence of bacterial secretion systems that enable pathogen access to the host cytosol.
Our findings are consistent with recent observations that the L. pneumophila Dot/Icm T4SS triggers the caspase-11-dependent noncanonical inflammasome [72], as well as the finding that bacteria that enter the cytosol either due to failure to maintain integrity of their replicative vacuoles or natural entry into the cytoplasm also trigger rapid caspase-11 activation [73]. Thus the pathway that leads to caspase-11 activation appears to be particularly sensitive to pathogens that 'violate the sanctity of the cytosol' [74], either through the activity of specialized secretion systems that translocate bacterial molecules into the cytosol or through their direct entry into the host cell cytosol. Whether other pathogens that replicate within the cytosol, such as Listeria or Shigella, or cytosolic viruses possess mechanisms to evade this pathway remains to be determined.
L. pneumophila T4SS-mediated activation of caspase-11 differs from the other pathways of non-canonical inflammasome activa-tion in several ways. First, L. pneumophila-mediated activation of caspase-11 does not require TRIF or IFNAR signaling. We observe a moderate dependence on TRIF and IFNAR signaling when macrophages are primed with LPS prior to infection, consistent with LPS-dependent upregulation of caspase-11 expression through the TLR4-TRIF-IFNAR axis [27][28][29]. However, in the absence of LPS priming, TRIF and IFNAR signaling are dispensable for L. pneumophila-dependent caspase-11 activation. In this context, it is likely that MyD88 compensates for the absence of TRIF, as cells deficient for both MyD88 and TRIF failed to activate caspase-11 in response to L. pneumophila. Thus, although the TLR4-TRIF-IFNAR axis is required for caspase-11 activation in response to Gram-negative bacteria, a MyD88-dependent signal is sufficient for caspase-11 activation in response to pathogens that utilize virulence-associated secretion systems to translocate bacterial molecules into the host cytosol. It is possible that different signals are capable of activating caspase-11 through distinct pathways, but these pathways occur with distinct kinetics because they may indicate distinct levels of pathogenicity. Thus, while caspase-11 is robustly upregulated by LPS priming, this upregulation alone is insufficient for rapid activation in response to bacteria that lack specialized secretion systems, as DdotA or DyopB bacteria do not induce rapid cell death even in primed cells. Collectively, these data indicate a two-signal model for rapid caspase-11 activation during infection with virulent bacteria, where bacterial PAMPs induce caspase-11 upregulation, but rapid caspase-11 activation requires a second, secretion system-dependent signal ( Figure 8).
The specific secretion system-dependent signals responsible for caspase-11 activation are currently unknown. While rapid activation of caspase-11 requires the presence of a functional type III or type IV secretion system or cytosolic access of the bacteria, whether the signal is an as-yet-undefined translocated bacterial molecule or a cellular response to the pore forming activity of these systems remains to be determined. The delayed NLRP3-and caspase-11-dependent response to Gram-negative bacteria suggests that in addition to LPS-induced upregulation of inflammasome components, bacterial mRNA provides an additional signal for activating the NLRP3 inflammasome [18,75], although the role of caspase-11 in this response has not been formally demonstrated. Activity of the type III or IV secretion systems may bypass the need for bacterial mRNA. Alternatively, these secretion systems may translocate bacterial RNA [70,76,77], and the rapid caspase-11-dependent response they induce could be due to more rapid delivery of bacterial mRNA into the host cell cytosol.
Furthermore, the host factors required for activation of the NLRP3-independent caspase-11-dependent inflammasome also remain to be identified. As this pathway is independent of flagellin sensing, NLRP3, ASC, and NLRC4, an unknown upstream sensor and/or adaptor may be involved in caspase-11 activation in response to a translocated bacterial substrate or an endogenous signal induced by infection. This sensor may also be upregulated by type I IFN signaling itself [27][28][29].
Our data show that IL-1a release during L. pneumophila infection is controlled by two independent pathways, one involving the flagellin-dependent NAIP5/NLRC4 and caspase-1-dependent inflammasome and a second pathway involving the NLRP3independent caspase-11-dependent inflammasome (Figure 8). Though we demonstrate that IL-1a release has an important biological consequence in vivo for neutrophil recruitment and bacterial clearance, it is unclear if IL-1a release is regulated by unconventional secretion, as is the case for IL-1b [78]. As both pathways that control IL-1a release also lead to cell death, our data are consistent with a model in which IL-1a is an endogenous alarmin that is released during cell death [79].
Interestingly, caspase-11 also contributes to control of flagellinexpressing L. pneumophila by serving as a component of an NLRC4-dependent inflammasome that promotes trafficking of the L. pneumophila-containing vacuole to lysosomes [80]. Thus, caspase-11 may function in multiple ways to control L. pneumophila infection. Importantly, we find that IL-1a, IL-1b, and IL-1R signaling play an important role in the control of L. pneumophila infection through efficient neutrophil recruitment to the airway. IL-1a and IL-1b play both distinct and overlapping roles in mediating neutrophil recruitment and controlling bacterial replication, as depletion of IL-1a alone showed a more pronounced defect in neutrophil recruitment and bacterial clearance than depletion of IL-1b alone, but loss of both cytokines resulted in a further reduction of neutrophil recruitment and an increased defect in bacterial clearance. Further analysis is required to define the relative contributions of the various caspase-11-mediated effector functions to the control of L. pneumophila replication in vivo.
In conclusion, these studies demonstrate that T3SS and T4SS activities trigger rapid and robust activation of caspase-11. This activation contributes to maximal NLRP3-dependent IL-1b secretion as well as to NLRP3-independent IL-1a release and host cell death. The downstream effector functions of these pathways are important for host defense against L. pneumophila in vivo, as IL-1a and IL-1b promote neutrophil recruitment to L. pneumophila-infected lungs and control pulmonary bacterial replication. Our results highlight the contribution of caspase-11 to rapid inflammasome activation and discrimination between pathogenic and nonpathogenic bacteria.

Ethics statement
This study was carried out in strict accordance as defined in the federal regulations set forth in the Animal Welfare Act (AWA), the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, and the guidelines of the University of Pennsylvania Institutional Animal Use and Care Committee. The protocols were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania (protocols #803465 and #803459).

Bacterial strains
Legionella pneumophila serogroup 1 strains were used in all experiments. Macrophages were infected with Lp02 (thyA), a thymidine auxotroph derived from strain Lp01 [40], or DdotA [81] and DflaA [16] isogenic mutant strains. For in vivo studies, mice were infected with the Lp02 DflaA or the JR32 [82] DflaA isogenic mutant strain where indicated. For in vitro and in vivo studies, L. pneumophila were cultured on charcoal yeast extract agar for 48 hours at 37uC prior to infection. Escherichia coli BL21 strain was cultured in LB broth for 16 hours at 37uC prior to infection. The Yersinia pseudotuberculosis strains used were IP2666 DyopHOJEMK (D6) [58] and DyopB [83]. Yersinia were grown overnight with aeration in 26YT broth at 26uC. The bacteria were diluted into fresh 26YT containing 20 mM sodium oxalate and 20 mM MgCl 2 . Bacteria were grown with aeration for 1 hour at 26uC followed by 2 hours at 37uC prior to infection.
In vivo infection studies [8][9][10][11][12] week-old mice were anesthetized by intraperitoneal injection of a ketamine/xylazine/PBS solution at a dose of 100 mg/kg ketamine and 10 mg/kg xylazine. Mice were infected intranasally with 40 ml of a bacterial suspension containing 1610 6 CFU L. pneumophila or PBS vehicle control. For antibody neutralization experiments, mice were injected intraperitoneally with 100 mg anti-IL-1a antibody (clone ALF-161), 100 mg anti-IL-1b antibody (clone B122), 100 mg of each anti-IL-1a and anti-IL-1b antibody, or 100 mg Armenian hamster IgG 1 isotype control antibody (eBioscience) 16 hours prior to intranasal infection. At the indicated timepoints after infection, mice were sacrificed, and the bronchoalveolar lavage fluid (BALF) and lungs were harvested. To determine bacterial load, the lungs were mechanically homogenized in sterile distilled H 2 O and a portion of the lysate was spread onto CYE plates. Animal experiments were performed in accordance with approved University of Pennsylvania Institutional Animal Care and Use Committee protocols and procedures.

Macrophage experiments
Bone marrow was collected from the femurs and tibiae of mice. Bone marrow cells were differentiated into macrophages by culturing the cells in RPMI containing 30% L929 cell supernatant and 20% FBS at 37uC in a humidified incubator. The macrophages were replated one day prior to infection in RPMI containing 15% L929 cell supernatant and 10% FBS. For experiments involving LPS-primed macrophages, macrophages in 48-well plates (2.0610 5 cells/well) were pretreated with 0.5 mg/ mL LPS for 2.5 hours and either mock-infected with PBS, infected with L. pneumophila at an MOI = 10 for 4 hours, or treated with 2.5 mM ATP for 1 or 4 hours. For experiments performed in the absence of LPS priming, macrophages in 48-well plates (2.0610 5 cells/well) were either mock-infected with PBS, infected with L. pneumophila at an MOI = 10 for 16 or 20 hours, or infected with E. coli at an MOI = 25 for 1 hour followed by gentamycin treatment for 15 hours. To assess the involvement of caspase-1 catalytic activity, macrophages were treated with 20 mM or 40 mM of the caspase-1 inhibitor YVAD-cmk (Bachem) or an equivalent volume of dimethyl sulfoxide (vehicle control) 0.5 hours prior to infection. For L. pneumophila and E. coli infections, bacteria were centrifuged down onto the macrophages at 1200 RPM for ten minutes prior to incubation. For Y. pseudotuberculosis infection, bacteria were washed three times with pre-warmed DMEM, added to the cells at an MOI = 20, and centrifuged down onto the macrophages at 1000 rpm for 5 min. Cells were incubated at 37uC for 1 hour post-infection followed by addition of 100 mg/mL gentamicin. Supernatants were harvested 4 hours post infection for ELISA and LDH analysis.

Cytotoxicity assays
Cells were infected or treated as described above, and supernatants were harvested at the indicated times post-infection. Lactate dehydrogenase (LDH) release was quantified using the LDH Cytotoxicity Assay Kit (Clontech) according to the manufacturer's instructions.

Immunoblotting
Supernatants from infected cells were mixed 1:1 with 2 X SDS-PAGE sample buffer or infected BMDMs were directly lysed in 1 X SDS-PAGE sample buffer. Samples were boiled, separated by SDS-PAGE, and transferred to Immobilon P membranes (Millipore). Primary antibodies against caspase-1 p10 (Santa Cruz Biotechnology), caspase-11 (Sigma, clone 17D9), IL-1b (R&D systems), and b-actin (Sigma) were used. Detection was performed with HRP-conjugated anti-rabbit IgG (Cell Signaling Technology) or anti-rat IgG (Santa Cruz Biotechnology or Jackson Immuno).

ELISA
Harvested supernatants from infected macrophages or the BALF from infected mice were assayed using capture and detection antibodies specific for IL-18 (MBL), IL-1a, IL-1b, and IL-12p40 (BD Biosciences).

Flow cytometry
To determine neutrophil recruitment to the airway, BALF cells were stained with Live/Dead Fixable Dead Cell Stain (Invitrogen), and antibodies specific for CD45, Gr-1 (eBioscience), and Ly6G (Biolegend). Data were collected with an LSRII flow cytometer (BD Biosciences) and post-collection data was analyzed using FlowJo (Treestar). Cells were gated on singlets and live cells. Neutrophils were identified as being CD45 + , Gr-1 + , and Ly6G + .

Statistical analysis
Plotting of data and statistical analysis were performed using Graphpad Prism software, and statistical significance was determined by the unpaired two-tailed Student's t test, one-way ANOVA with Tukey post-test, or two-way ANOVA with Bonferroni post-test. Differences were considered statistically significant if the P value was ,0.05.  Figure S12 Caspase-11-deficient cells secrete comparable amounts of IL-12 in response to Y. pseudotuberculosis. B6, Casp1 2/2 Casp11 2/2 , Casp1 2/2 , or Casp11 2/2 mice were primed with 0.05 mg/mL LPS for 2.5 hours and infected with type III secretion system-deficient Y. pseudotuberculosis (DyopB Yp), effectorless Y. pseudotuberculosis DHOJMEK (D6 Yp), or PBS (mock infection) or treated with 2.5 mm ATP for 4 hours. The level of IL-12 p40 in the supernatants was measured by ELISA.