Noncanonical Inflammasome Activation by Intracellular LPS Independent of TLR4

See allHide authors and affiliations

Science  13 Sep 2013:
Vol. 341, Issue 6151, pp. 1246-1249
DOI: 10.1126/science.1240248

Move Over, TLR4

The innate immune system senses bacterial lipopolysaccharide (LPS) through Toll-like receptor 4 (TLR4) (see the Perspective by Kagan). However, Kayagaki et al. (p 1246, published online 25 July) and Hagar et al. (p. 1250) report that the hexa-acyl lipid A component of LPS from Gramnegative bacteria is able to access the cytoplasm and activate caspase-11 to signal immune responses independently of TLR4. Mice that lack caspase-11 are resistant to LPS-induced lethality, even in the presence of TLR4.


Gram-negative bacteria including Escherichia coli, Citrobacter rodentium, Salmonella typhimurium, and Shigella flexneri are sensed in an ill-defined manner by an intracellular inflammasome complex that activates caspase-11. We show that macrophages loaded with synthetic lipid A, E. coli lipopolysaccharide (LPS), or S. typhimurium LPS activate caspase-11 independently of the LPS receptor Toll-like receptor 4 (TLR4). Consistent with lipid A triggering the noncanonical inflammasome, LPS containing a divergent lipid A structure antagonized caspase-11 activation in response to E. coli LPS or Gram-negative bacteria. Moreover, LPS-mutant E. coli failed to activate caspase-11. Tlr4–/– mice primed with TLR3 agonist polyinosinic:polycytidylic acid [poly(I:C)] to induce pro-caspase-11 expression were as susceptible as wild-type mice were to sepsis induced by E. coli LPS. These data unveil a TLR4-independent mechanism for innate immune recognition of LPS.

Canonical inflammasome proteins NLRP3 and NLRC4 engage the adaptor protein ASC and activate caspase-1 to promote cleavage and secretion of interleukin-1β (IL-1β) and IL-18, as well as cell death termed “pyroptosis” (14). Gram-negative bacteria such as Escherichia coli elicit noncanonical activation of caspase-1 in that caspase-11 (5, 6) is required in addition to NLRP3 and ASC. Intriguingly, caspase-11 but neither NLRP3, ASC, nor caspase-1 is required for pyroptosis in response to Gram-negative bacteria (fig. S1A) (7). In mice, caspase-11 is critical to the immune response to Gram-negative bacteria (712), but the bacterial pathogen-associated molecular pattern (PAMP) that is responsible for triggering the noncanonical inflammasome is unknown.

Mouse bone marrow–derived macrophages (BMDMs) activate the noncanonical inflammasome in response to a combination of E. coli lipopolysaccharide (LPS) and cholera toxin B (CTB) (7). CTB is the nonenzymatic carbohydrate-binding component of the holotoxin (13). Inflammasome activity, be it canonical or noncanonical, appears to require two signals. A priming signal increases expression of inflammasome components, including pro-IL-1β, pro-caspase-11, and NLRP3, and then another signal triggers caspase activation (fig. S1, A to C) (3, 7, 14, 15). We observed that LPS from only one E. coli serotype, O111:B4, combined with CTB to elicit caspase-1 processing, the secretion of IL-1β and IL-18, and pyroptosis (Fig. 1A and fig. S2, A and B). This finding was unexpected because triggering of the canonical NLRP3 inflammasome by adenosine 5′-triphosphate (ATP) was not dependent on the LPS serotype.

Fig. 1 Triggering of the noncanonical inflammasome by cytoplasmic LPS.

(A) IL-1β released from BMDMs that were cultured for 6 hours in medium (cont), the indicated LPS serotypes, or lipid A and then supplemented with additional medium (cont) or CTB for 16 hours, or ATP for 4 hours. (B) Western blots to detect pro-caspase-11 in BMDMs cultured for 6 hours with the PAMPs indicated. (C) IL-1β secreted from wt or Casp11–/– BMDMs that were primed with Pam3CSK4 and then stimulated for 16 hours with the indicated LPS serotypes or lipid A in medium alone (cont) or in combination with CTB. (D) Binding of LPS O111:B4 to CTB by surface plasmon resonance. (E and F) Flow cytometry histograms showing labeling of BMDMs with fluorescein isothiocyanate (FITC)–conjugated LPS O111:B4 and LPS O55:B5 in the presence or absence of soluble GM1. Bar graph indicates the effect of soluble GM1 on IL-1β secretion from BMDMs stimulated for 16 hours with LPS plus CTB or 4 hours with ATP. (-) CTB indicates FITC-LPS alone. (G) IL-1β and LDH (used to measure cytotoxicity) released from primed BMDMs after culture or transfection with LPS for 16 hours. Graphs show the mean ± SD of triplicate wells and are representative of three independent experiments.

LPS from E. coli is composed of a conserved, Toll-like receptor 4 (TLR4)–activating moiety called lipid A, a core oligosaccharide, and a hyper-divergent O-specific polysaccharide (OPS) chain (fig. S2C) (16). Lipid A failed to produce noncanonical inflammasome activity when combined with CTB (Fig. 1A and fig. S2, A and B). We considered that LPS O111:B4 might be unique in up-regulating expression of pro-caspase-11, but lipid A and the different LPS O-serotypes induced comparable expression of pro-caspase-11 (fig. S2D). These data point to pro-caspase-11 induction being a consequence of lipid A engagement of TLR4 (16). Pam3CSK4 (TLR2 agonist) or R837 (TLR7 agonist) (17) that stimulate signaling via the adaptor protein MyD88 also induced expression of pro-caspase-11 (Fig. 1B and fig. S2E) (7), yet neither combined with CTB to promote pyroptosis or IL-1β secretion (fig. S2F). Even when BMDMs were primed with Pam3CSK4 in advance, only LPS O111:B4/CTB caused caspase-11–dependent secretion of IL-1β (Fig. 1C) and pyroptosis (fig. S2G). Therefore, a distinct property of LPS O111:B4, beyond priming, enables it and CTB to trigger the noncanonical inflammasome.

CTB binds to the pentasaccharide moiety of the GM1 gangliosaccharide (13, 18). We found that LPS O111:B4, or its OPS chain alone (19), could bind to CTB, whereas LPS O55:B5 or lipid A could not (Fig. 1D and fig. S3A). Given that binding of CTB to GM1 on the cell surface promotes endocytosis of the cholera toxin complex (13, 18), we investigated whether CTB delivers LPS O111:B4 into the cell. Consistent with this hypothesis, CTB increased BMDM uptake of LPS O111:B4, but not LPS O55:B5 (Fig. 1E and fig. S3, B to D). In addition, soluble GM1 attenuated uptake of CTB (fig. S3C) and CTB/LPS O111:B4 (Fig. 1F), and this coincided with reduced IL-1β secretion in response to CTB/LPS.

Next, we explored whether CTB was dispensable for noncanonical inflammasome activation if LPS was transfected into BMDMs (fig. S4, A and B). LPS transfection stimulated caspase-11–dependent pyroptosis and caspase-11, caspase-1, NLRP3, and ASC-dependent secretion of IL-1β and IL-18, irrespective of LPS serotype or whether the cells were primed with Pam3CSK4 (Fig. 1G and fig. S4, C to F). BMDMs electroporated with LPS gave similar results (fig. S4, G and H). We conclude that E. coli LPS of any serotype can trigger the noncanonical inflammasome if it gets into the cell. Hereafter, we used LPS O111:B4 unless specified.

Because the hexa-acylated lipid A moiety within E. coli LPS stimulates signaling by TLR4 (fig. S5A) (16), we tested whether intracellular lipid A triggers the noncanonical inflammasome. BMDMs were primed with Pam3CSK4 so as to circumvent the different priming efficiencies of lipid A and LPS. Transfection (Fig. 2, A and B, and fig. S5B) or electroporation (fig. S5C) of synthetic monophosphoryl lipid A stimulated caspase-11–dependent secretion of IL-1β and IL-18, pyroptosis, and the release of cleaved caspase-1. In contrast, the tetra-acylated lipid IVa precursor of lipid A had mouse TLR4 agonist activity (fig. S5A) (20, 21) but was a poor trigger of the noncanonical inflammasome. Therefore, intracellular lipid A triggers the noncanonical inflammasome, but its sensor probably recognizes a different part of the molecule than TLR4.

Fig. 2 Lipid A triggers the noncanonical inflammasome.

(A) IL-1β and LDH released from wt and Casp11–/– BMDMs at 16 hours after transfection with LPS, synthetic monophosphoryl lipid A, or lipid IVa. (B) Western blots to detect pro-caspase 1, pro-IL-1β, cleaved IL-1β (p17), and cleaved caspase-1 (p20) from BMDMs transfected with E. coli LPS or lipid A for 16 hours. (C) IL-1β and LDH released by BMDMs at 16 hours after transfection with LPS from different bacteria. Graphs show the mean ± SD of triplicate wells and are representative of three independent experiments.

Lipid A is conserved among different serotypes of E. coli, but several pathogenic Gram-negative bacteria make a modified lipid A and evade TLR4-mediated immune surveillance. For example, Heliobacter pylori makes a tetra-acyl lipid A moiety with long carbon fatty acid chains that stimulates TLR4 poorly (20, 22, 23). LPS from H. pylori or Rhizobium galegae (24) elicited little TLR4 signaling in BMDMs (fig. S5A). Their divergent lipid A moieties also failed to trigger the noncanonical inflammasome after transfection (Fig. 2, B and C, and fig. S5, D and E). In contrast, LPS from Salmonella typhimurium, which has a similar lipid A structure to that of E. coli, stimulated robust TLR4 signaling (fig. S5A) (16) and noncanonical inflammasome activity (Fig. 2C and fig. S5, D and E). These data suggest that lipid A modifications in pathogenic bacteria combat caspase-11–mediated surveillance mechanisms.

TLR4 plays a central role in detecting LPS (25, 26), but our experiments with lipid IVa implied that TLR4 does not trigger the noncanonical inflammasome. We sought genetic proof with Tlr4–/– BMDMs that were primed with either Pam3CSK4 or poly(I:C). These cells responded just like wild-type BMDMs to intracellular LPS or lipid A in that they released cleaved caspase-1, IL-1β, IL-18, and lactate dehydrogenase (LDH) (Fig. 3, A and B, and fig. S6, A to D). MD-2 and MD-1 are the lipid A-binding proteins that complex with TLR4 and its homolog RP105, respectively (27). BMDMs lacking these proteins also exhibited normal noncanonical inflammasome activity, as did BMDMs lacking RP105 or the LPS coreceptor CD14 (fig. S6E). Therefore, TLR4, RP105, CD14, MD-1, and MD-2 are all dispensable for intracellular LPS to trigger the noncanonical inflammasome. The existence of multiple sensors that recognize the same PAMP is not without precedence: Both TLR5 and NLRC4 sense bacterial flagellin (2830).

Fig. 3 TLR4 is dispensable for triggering of the noncanonical inflammasome by cytoplasmic LPS.

(A) Western blots showing IL-1β and caspase-1 that is released from BMDMs at 16 hours after E. coli LPS or lipid A is delivered by transfection or CTB. Control cells were transfected with medium alone (cont). (B) IL-1β and LDH released by wt, Tlr4–/–, and Casp11–/– BMDMs at 16 hours after delivery of E. coli LPS or lipid A. (C) IL-1β and IL-6 secreted from BMDMs infected with E. coli W3110 (parental) or CMR300 (LPSmt lacking mature LPS) for 16 hours. The multiplicity of infection is indicated on the x axis or was 5. (D) IL-1β secreted by Tlr4–/– BMDMs transfected for 2 hours with dimethyl sulfoxide (cont), lipid IVa, or LPS (from H. pylori or R. galegae) and then transfected with LPS from E. coli for 16 hours or cultured with ATP for 4 hours. (E) IL-1β secreted by BMDMs transfected for 2 hours as in (D) and then infected with the bacteria indicated for 16 hours. Graphs show the mean ± SD of triplicate wells and are representative of three independent experiments.

TLR4 signaling via the adaptor protein TRIF produces type I interferons (IFN-α and IFN-β) that induce pro-caspase-11 expression (8, 9, 12). Whether these IFNs play a role beyond priming BMDMs is unknown. We primed BMDMs lacking either the type I IFN receptor or TRIF with Pam3CSK4, which signals via MyD88 rather than TRIF (fig. S2E) (17). Subsequent transfection or electroporation with LPS caused Ifnar–/– and Trif–/– BMDMs to secrete IL-1β and undergo pyroptosis to the same extent as wild type or Tlr4–/– BMDMs (fig. S6F). Therefore, the entire TLR4-TRIF–type I IFN pathway is dispensable for intracellular LPS to trigger the noncanonical inflammasome. Previously, induction of pro-caspase-11 protein and its autoproteolysis was proposed to be sufficient to engage the noncanonical inflammasome (9). We believe that this scenario is unlikely because LPS transfection triggered pyroptosis in primed Tlr4–/– BMDMs despite unchanging pro-caspase-11 levels (fig. S6G).

To address the role of LPS in triggering the noncanonical inflammasome during a Gram-negative bacterial infection, we primed wild type and Tlr4–/– BMDMs with Pam3CSK4 and then infected them with either E. coli “wild-type” strain W3110 (parental) or the LPS-mutant derivative strain CMR300 (LPSmt), the latter lacking mature hexa-acyl lipid A and accumulating the tetra-acyl lipid IVa precursor because of a defect in LPS biosynthesis (31). W3110 bacteria triggered caspase-11–dependent IL-1β secretion (Fig. 3C) and pyroptosis (fig. S6H) in both wild type and Tlr4–/– BMDMs, whereas the LPS-mutant CMR300 strain did not. The two strains of E. coli did, however, stimulate comparable IL-6 production, which is consistent with lipid IVa being a mouse TLR4 agonist (fig. S5A) (20, 21). In addition, Tlr4–/– BMDMs preloaded with LPS from H. pylori or R. galegae exhibited less caspase-11–dependent secretion of IL-1β when they were subsequently transfected with LPS from E. coli (Fig. 3D) or infected with Gram-negative bacteria (Fig. 3E). This antagonistic effect of LPS from H. pylori or R. galegae was specific to the noncanonical inflammasome because IL-1β secretion in response to canonical stimuli ATP or Pseudomonas aeruginosa (3, 7, 32) was not impaired. Collectively, these data indicate that macrophages use the noncanonical inflammasome to sense intracellular LPS during a Gram-negative bacterial infection, and they explain why noncanonical inflammasome activity is not observed with Gram-positive bacteria (712). Although it is unclear exactly where in the cell the noncanonical inflammasome is activated during bacterial infection, a cytosolic location has been proposed (10, 11).

We investigated the relevance of our in vitro findings with a mouse model of acute septic shock. Wild type, Casp11–/–, and Tlr4–/– mice were challenged with a lethal dose of LPS, and consistent with previous reports (7, 25, 26, 33), caspase-11 or TLR4 deficiency conferred resistance, with 9 out of 10 Casp11–/– mice and 10 out of 10 Tlr4–/– mice surviving (Fig. 4). The caveat of this study is that LPS-challenged Tlr4–/– mice probably fail to induce pro-caspase-11. We circumvented this priming issue by pretreating the mice with a nonlethal dose of the TLR3 agonist poly(I:C) (Fig. 4 and fig. S7A). Subsequent LPS challenge caused morbidity in all Tlr4–/– and wild-type mice within 60 hours, whereas the onset of morbidity in the Casp11–/– mice was delayed significantly (Fig. 4 and table S1). Therefore, both TLR4-dependent and -independent mechanisms for sensing LPS contribute to this model of lethal sepsis (fig. S7B). Lethal sepsis is probably driven by caspase-11–dependent pyroptosis rather than caspase-1–dependent secretion of IL-1β and IL-18 because Casp1–/– mice as susceptible to LPS as are wild-type mice are (7). We postulate that high-dose LPS is sufficient to trigger the noncanonical inflammasome in vivo because BMDMs cultured with large quantities of LPS display spontaneous LPS internalization (fig. S7C) (34, 35). It is unclear, however, which cell types trigger the noncanonical inflammasome to drive lethal sepsis in vivo.

Fig. 4 Lethal sepsis mediated by TLR4-independent sensing of LPS.

Kaplan-Meier survival plots for mice challenged with 4 mg/kg poly(I:C), 54 mg/kg LPS, or poly(I:C) followed by LPS. Experiments used 6 to 10 mice of each genotype. Data are representative of four independent experiments. Adjusted P values are supplied in table S1.

We show that the hexa-acyl lipid A moiety of LPS from E. coli or S. typhimurium is sensed inside the macrophage, albeit in an ill-defined manner, and this triggers caspase-11 activation by the noncanonical inflammasome. The resistance of mice lacking caspase-11 to LPS-induced lethality suggests that antagonists of the activation of human caspases 4 and 5 (the orthologs of mouse caspase-11) may benefit patients suffering from sepsis.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Table S1


References and Notes

  1. Acknowledgments: We thank S. Louie, J. Dong, J. Starks, K. O'Rourke, E. Suto, B. L. Lee, and P. Martinez for technical support; M. van Lookeren Campagne, G.-J. Boons, M. Roose-Girma, E. Naik, and K. Newton for discussions and manuscript editing; S. Akira, O. Takeuchi, and W. W. Leitner for mice; and T. M. Stephen for CMR300. Structural work on LPS was supported, in part, by U.S. Department of Energy grant DE-FG02-09ER20097 to the Complex Carbohydrate Research Center. Generation of Md1–/–Md2–/– and Rp105–/– mice was supported by grants from the Japan Society for the Promotion of Science (KAKENHI 23590564) and Ministry of Education, Culture, Sports, Science and Technology (KAKENHI 21117002).
View Abstract

Stay Connected to Science

Navigate This Article