Inflammasome Activation by Bacterial Outer Membrane Vesicles Requires Guanylate Binding Proteins

ABSTRACT The Gram-negative bacterial cell wall component lipopolysaccharide (LPS) is recognized by the noncanonical inflammasome protein caspase-11 in the cytosol of infected host cells and thereby prompts an inflammatory immune response linked to sepsis. Host guanylate binding proteins (GBPs) promote infection-induced caspase-11 activation in tissue culture models, and yet their in vivo role in LPS-mediated sepsis has remained unexplored. LPS can be released from lysed bacteria as “free” LPS aggregates or actively secreted by live bacteria as a component of outer membrane vesicles (OMVs). Here, we report that GBPs control inflammation and sepsis in mice injected with either free LPS or purified OMVs derived from Gram-negative Escherichia coli. In agreement with our observations from in vivo experiments, we demonstrate that macrophages lacking GBP2 expression fail to induce pyroptotic cell death and proinflammatory interleukin-1β (IL-1β) and IL-18 secretion when exposed to OMVs. We propose that in order to activate caspase-11 in vivo, GBPs control the processing of bacterium-derived OMVs by macrophages as well as the processing of circulating free LPS by as-yet-undetermined cell types.

and secretion of proinflammatory cytokines interleukin-1␤ (IL-1␤) and IL-18 by caspase-1 (4,5). Therefore, compartmentalization of LPS receptors allows host cells to respond differentially to the presence of LPS at three distinct subcellular locales.
Long thought to be the sole LPS sensor in mammals, TLR4 has been studied in great detail, leading to the discovery and characterization of several auxiliary proteins required for the activation of TLR4 signaling (2). The discovery of the noncanonical inflammasome protein caspase-11 in mice and CASP-4/-5 in humans as cytosolic sensors for LPS is more recent (6), and accordingly, little is known about the existence of additional host factors controlling its activity. One notable exception is the regulation of caspase-11-dependent host responses by guanylate binding proteins (GBPs). GBPs constitute a family of proteins induced by interferon gamma (IFN-␥) as well as type I IFNs. GBPs exhibit potent antimicrobial activities both in cell culture models and in vivo (7)(8)(9). GBPs were additionally shown to promote caspase-11 activation in macrophages infected with Salmonella, Legionella, or Chlamydia (10)(11)(12), all Gramnegative bacteria that occupy pathogen-containing vacuoles (PVs) inside infected host cells (13). Mechanistic in vitro studies led to an attractive model, according to which GBPs lyse PVs and thereby release bacteria and their associated LPS into the host cell cytosol for recognition by caspase-11 (10). This model was based on results obtained from Salmonella infection studies in cultured macrophages (10) but is contradicted by our own studies using Legionella and Chlamydia infection models (11,12). Thus, while GBPs have been identified as key regulators of caspase-11 function, the mechanism by which they do so remains controversial.
Because caspase-11 is a cytosolic LPS receptor (6,14), its activation depends on the delivery of LPS across host cell membranes into the host cell cytosol. Injection of purified LPS into mice activates caspase-11 in vivo (4,5), thus arguing that mechanisms exist by which circulating LPS can traverse eukaryotic membranes and enter the host cell cytosol. However, the nature of these mechanisms is unknown, and their existence may be restricted to specific cell types. Indeed, cultured macrophages lack the ability to import extracellular LPS into their host cell cytosol and therefore do not induce caspase-11 activation when purified "free" LPS is added extracellularly (4,5). However, in contrast to free LPS, adding bacterial outer membrane vesicles (OMVs) comprised of LPS and other cell wall components to culture medium is sufficient to activate caspase-11 in macrophages, which are able to extract and import LPS from ingested OMVs into their cytosol by a poorly characterized mechanism (15). Because bacteria can produce immunomodulatory OMVs (16,17), we hypothesized that GBP-dependent caspase-11 activation during macrophage infections was mediated by OMVs. In support of this hypothesis, we observe that GBPs are essential for OMV-induced pyroptosis and IL-1␤/IL-18 secretion. Additionally, we find that GBP-deficient mice injected with OMVs or purified LPS display lower IL-1␤/IL-18 serum levels and lower mortality rates than wild-type mice. Our studies thus reveal GBPs as critical regulators of inflammation induced by circulating free LPS or OMVs in vivo.
times postinfection. Next, we investigated whether activation of the noncanonical inflammasome was altered in GBP chr3Ϫ/Ϫ BMDMs. Whereas unprimed BMDMs exposed to E. coli K-12 failed to undergo marked cell death at 8 hpi across a broad range of multiplicities of infection (MOIs), we found that IFN-␥-primed wild-type but not GBP chr3Ϫ/Ϫ or Casp11 Ϫ/Ϫ BMDMs rapidly succumbed to E. coli-triggered cell death, as measured by lactose dehydrogenase (LDH) release (Fig. 1A) or host cell nuclear incorporation of propidium iodide (Fig. 1B). These observations demonstrate that GBPs are required for the induction of caspase-11-dependent pyroptosis by nonpathogenic E. coli.
In addition to the induction of pyroptosis, caspase-11 activation promotes the formation of the canonical NLR family pyrin domain-containing 3 (NLRP3) inflammasome (18). Because GBPs were required for caspase-11-dependent pyroptosis ( Fig. 1A and B), we tested their role in the activation of the canonical inflammasome in E. coli-infected BMDMs. A hallmark of canonical inflammasome assembly is the formation of intracellular foci comprised of the adapter protein apoptosis-associated speck-like protein containing a CARD (ASC). We noticed that the number of cells with ASC specks was significantly reduced in GBP chr3Ϫ/Ϫ BMDMs compared to wild-type BMDMs (Fig. 1C). Similarly, processed caspase-1 and IL-1␤ were detectable by Western blotting in cell supernatants of E. coli-infected wild-type BMDMs but undetectable in cell supernatants of Casp11 Ϫ/Ϫ and GBP chr3Ϫ/Ϫ BMDMs (Fig. 1D). Third, both IL-1␤ and IL-18 secretion in response to E. coli infections was dramatically diminished in GBP chr3Ϫ/Ϫ BMDMs ( Fig. 1E and F). Because GBP chr3Ϫ/Ϫ BMDMs by and large phenocopied Casp11 Ϫ/Ϫ BMDMs in all of these functional assays ( Fig. 1), we conclude that GBPs are critical for the induction of the pyroptotic death pathway and the activation of the NLRP3 inflammasome by caspase-11 in E. coli-infected BMDMs.
Induction of pyroptosis and interleukin secretion by OMVs requires GBPs. Previous studies proposed that GBPs extract LPS from intracellular bacterial pathogens through PV lysis and bacteriolysis, processes that require direct binding of GBPs to PVs or bacteria, respectively (10,(19)(20)(21). In agreement with our recent finding that GBPs specifically detect phagosomes occupied by virulent bacteria but fail to associate with phagosomes containing avirulent bacteria (22), we observed that the association of GBP2 with phagosomes containing E. coli K-12 was infrequent (Fig. S2). This observation suggested an alternative mechanism by which GBPs promote caspase-11 activation in macrophages exposed to E. coli. Recently, it was reported that OMVs are the predominant inducer of caspase-11 activation through E. coli infections (15). Because GBPs are essential for E. coli-induced caspase-11 activation ( Fig. 1), we asked whether GBPs were also required for inflammasome activation in response to purified OMVs added to macrophages extracellularly. As shown previously (4,5), the addition of extracellular LPS, even at high concentrations (1 g/ml), had minimal impact on cell viability in both unprimed and IFN-␥-primed BMDMs ( Fig. 2A). In contrast to LPS, OMVs added to the medium of IFN-␥-primed BMDMs induced rapid, caspase-11-dependent cell death in a concentration-dependent manner ( Fig. 2A and B). Similarly to E. coli-induced pyroptosis ( Fig. 1), this OMV-induced cell death required the expression of GBPs ( Fig. 2A and B). ASC speck formation (Fig. 2C), proteolytic processing of pro-caspase-1 and pro-IL-1␤ (Fig. 2D), and IL-1␤ and IL-18 secretion ( Fig. 2E and F) triggered by the addition of OMVs were also dependent on GBPs. Together, these data show that pronounced activation of noncanonical and canonical inflammasome responses by extracellular OMVs requires GBPs.
GBP2 but not GBP5 controls caspase-11-dependent pyroptosis and interleukin secretion in cultured macrophages. The chromosomal deletion in GBP chr3Ϫ/Ϫ mice eliminates 5 Gbp genes, namely, Gbp1, Gbp2, Gbp3, Gbp5, and Gbp7 (23). Previous studies demonstrated that expression of GBP2 promotes the induction of caspase-11dependent cell death in macrophages infected with Gram-negative bacterial pathogens such as Salmonella and Legionella (10,12). In support of a possible role for GBP2 in OMV processing, we detected partial colocalization between GBP2 and LPS in OMV-treated BMDMs (Fig. S3). We therefore tested whether GBP2 was required for caspase-11 activation by nonpathogenic E. coli and OMVs. We found that GBP2 Ϫ/Ϫ BMDMs mimicked GBP chr3Ϫ/Ϫ BMDMs in their unresponsiveness to E. coli infections, as  measured by LDH release as a marker of cell death (Fig. 3A) and secretion of IL-1␤ (Fig. 3B) and IL-18 (Fig. 3C). Similarly, GBP2 was essential for the robust induction of cell death (Fig. 3D) as well as IL-1␤ (Fig. 3E) and IL-18 (Fig. 3F) secretion by OMV treatment.
Together, these data demonstrate that GBP2 is critical for OMV-mediated and caspase-11-dependent pyroptosis and canonical inflammasome activation. The specific activities of individual GBPs are poorly characterized, but a number of recent studies indicated that individual GBPs fulfill specialized, nonredundant functions in cell-autonomous immunity (24)(25)(26)(27)(28)(29). One study proposed that GBP5 specifically assists assembly of the NLRP3 inflammasome and enhances responsiveness to a subset of NLRP3 priming agents (27). Based on these previous findings, we asked whether GBP5 modulated inflammasome activation in response to E. coli infections or OMV treatment in macrophages. We found that GBP5 was dispensable for E. coli-and OMV-induced pyroptosis as well as IL-1␤ and IL-18 secretion in BMDMs (Fig. 3). These data indicate that GBP5 is nonessential for canonical inflammasome responses that occur downstream from caspase-11 activation in cultured macrophages.

GBPs control OMV-and LPS-induced inflammation in vivo.
To assess the role of GBPs in caspase-11-dependent inflammation, we first injected mice with the TLR3 agonist poly(I·C), which prompts elevated expression of GBPs in vivo (Fig. S4) and skews LPS-triggered inflammation toward caspase-11-rather than TLR4-dominated responses (4,5). According to previously established protocols (15), we first administered poly(I·C) for 6 h and then injected mice with OMVs and monitored IL-1␤ and IL-18 serum levels in wild-type and GBP chr3Ϫ/Ϫ mice at 6 h post-OMV injection. We found that both IL-1␤ and IL-18 serum levels were significantly reduced in GBP chr3Ϫ/Ϫ mice, indicating that GBPs promote OMV-induced inflammasome activation in vivo ( Fig. 4A and B). We then assessed whether GBPs are also required for the activation of caspase-11 in response to in vivo administration of purified LPS. We found that GBP chr3Ϫ/Ϫ (Fig. 4C and D) as well as GBP2 Ϫ/Ϫ (Fig. 4E and F) mice displayed lower IL-1␤ and IL-18 serum levels at 4 h postinjection. To monitor the role of GBPs in caspase-11-biased sepsis, poly(I·C)-treated mice received an injection of purified LPS at a concentration of 20 mg/kg body weight. We observed higher survival rates among GBP chr3Ϫ/Ϫ and GBP2 Ϫ/Ϫ mice than among wild-type mice (Fig. 4G and S5). Together, these data demonstrate that GBPs are in vivo regulators of OMV-and LPS-induced inflammation and sepsis.   ϭ 7), and GBP2 Ϫ/Ϫ (n ϭ 9) mice were i.p. injected with 2 mg/kg body weight of poly(I·C) and then 6 h later i.p. injected with LPS (20 mg/kg body weight). Morbidity and mortality were observed for 42 h at 3-h intervals. For panels A to F, mean Ϯ standard deviation is shown. Each symbol represents an individual mouse. Significance was defined as follows: ***, P Ͻ 0.001; **, P Ͻ 0.01; *, P Ͻ 0.05; N.S, not statistically significant. Significance was measured by two-way ANOVA with Sidak's multiple-comparison test (A to D), unpaired t test (E and F), or log rank test (G).

DISCUSSION
Caspase-11 directly detects LPS within the host cell cytosol (6). To explain how LPS gains access to the host cell cytosol, four distinct LPS delivery pathways were proposed: (i) some intracellular Gram-negative bacteria escape vacuoles to enter the host cell cytosol, where they release LPS (30); (ii) host GBPs execute membranolytic activities to extrude intracellular Gram-negative bacteria from PVs and extract LPS through bacteriolysis (10,(19)(20)(21); (iii) endocytosed bacterial OMVs release LPS into the host cell cytosol potentially through fusion with or transport across endosomal membranes (15,31); and (iv) circulating free LPS (in the form of aggregates or bound to LPS-binding proteins) is consumed in vivo by an undefined cell population able to present LPS for caspase-11-mediated recognition (4,5). Here, we present evidence that GBPs play previously unknown roles in the latter two pathways.
GBPs assist caspase-11 activation in response to infections with Gram-negative bacteria (10)(11)(12). It was proposed that GBPs lyse vacuoles containing Gram-negative bacteria and thereby release LPS into the host cell cytosol (10). However, we observed that GBPs were able to promote caspase-11 activation in response to Chlamydia muridarum infections without any detectable recruitment of GBPs to Chlamydia-containing vacuoles, thus arguing against the vacuolar lysis model (11). In addition, we observed that GBPs accelerated caspase-11 activation in cells transfected with LPS aggregates (12), demonstrating that GBPs can impact the kinetics of caspase-11 activation independently of an infection. Because of the recent discovery that Gram-negative OMVs serve as vehicles for the delivery of LPS into the host cell cytosol (15), we hypothesized and here demonstrate a central role for GBPs in the activation of caspase-11 by OMVs.
Bacteria constitutively produce OMVs. Bacterial OMV production is further increased by physiological stressors such as changes in the redox state, nutrient availability, or pH that bacteria experience during cell entry, and especially phagocytosis by macrophages (16,(32)(33)(34). We therefore propose that caspase-11 activation during the infection of macrophages with pathogens such as Salmonella, Legionella, or Chlamydia is induced at least in part by the bacterial secretion of OMVs. OMVs released during infection are then processed in a GBP-dependent manner to trigger caspase-11-dependent pyroptosis and canonical inflammasome activation. More detailed studies are required to delineate the mechanism(s) by which GBPs enable caspase-11 activation following OMV treatment. Considering that GBPs are part of the membrane-remodeling dynamin protein superfamily (35), GBPs could potentially play a role in controlling membrane dynamics at the OMVendosomal interface and thereby expose the lipid A moiety of LPS toward the cytosolic face of endosomes. Extensive cell biological and biochemical studies will be required to test this and alternative hypotheses regarding the molecular mechanism by which GBPs promote caspase-11 activation in OMV-exposed macrophages.
Whereas the addition of purified LPS to culture medium is insufficient to trigger robust caspase-11 activation in wild-type BMDMs, injection of LPS into mice induces caspase-11-dependent sepsis (4,5). Therefore, circulating LPS in vivo is most likely ingested and presented to caspase-11 by designated cell types. Hepatic macrophages (Kupffer cells), sinusoidal endothelial cells, and hepatocytes are credited to be mainly responsible for the removal and detoxification of LPS from the bloodstream (36)(37)(38) and therefore are also candidates to mediate in vivo caspase-11 activation in response to circulating LPS. Our study demonstrates that GBPs play a critical role in the host response to circulating LPS. Future studies will need to define how the host promotes caspase-11 activation in response to free LPS in vivo and the specific roles that GBPs play in this process.
GBPs were shown to act as positive regulators of infection-induced caspase-11 activation (10,12), and yet their precise functional role in this process has remained poorly defined (7)(8)(9). Our study delineates a unique role for GBPs in controlling caspase-11 activation in response to the sterile delivery of OMVs or free LPS in vivo. Our report therefore sets a novel framework within which the cellular and molecular activities of GBPs as regulators of inflammasome activation can be explored. removed, and pellets were rinsed with acetone twice. Pellets were then air dried and suspended in 8 M urea. Whole-cell extracts were prepared by lysing cells in RIPA lysis buffer (Sigma-Aldrich). Samples were loaded on a 4 to 20% gradient SDS-PAGE gel and transferred to polyvinylidene difluoride (PVDF). Membranes were blocked in Tris-buffered saline-0.1% Tween 20 (TBST) with 5% BSA or 5% nonfat dry milk. Membranes were incubated with primary antibodies overnight at 4°C, and secondary antibody incubations were performed for 1 h at RT. The following primary antibodies and dilutions were used: rabbit anti-caspase-1 (AG-20B-0042; Adipogen; 1:1,000), goat anti-IL-1␤ (AF-401-NA; R&D Systems; 1:1,000), and mouse anti-␤-actin (A2228; Sigma-Aldrich; 1:1,000).
In vivo challenges. All mice used for in vivo challenges were at the age of 8 to 12 weeks. For OMV challenge experiments, mice were injected intraperitoneally (i.p.) with poly(I·C) at a dose of 2 mg/kg body weight or PBS control and then 6 h later injected i.p. with 4 g purified OMVs per mouse or an equal volume of PBS control. Serum was obtained 6 h after OMV injection. For LPS challenge experiments, mice were injected i.p. with LPS (E. coli O111:B4 LPS; L3024; Sigma) at a dose of 8 mg/kg body weight or PBS control. Serum was obtained 4 h postinjection. Serum IL-1␤ and IL-18 concentrations were measured by ELISA. For the study of lethal endotoxemia, mice were first challenged i.p. with poly(I·C) (2 mg/kg) followed 6 h later by i.p. injection of LPS (20 mg/kg). Mice were monitored every 3 h for 48 h following initial injection. Mice were considered moribund and euthanized if they dropped below 80% starting weight or if they exhibited severe ataxia, indicated by lack of righting response.
Statistical analyses. Data analysis was performed using GraphPad Prism 6.0 software. Data shown are means Ϯ standard errors of the means (SEM) unless otherwise indicated. Statistical significance was calculated using the two-way analysis of variance (ANOVA) (with Tukey's or Sidak's multiple-comparison test), unless otherwise noted in figure legends.