K + Efflux Is the Common Trigger of NLRP3 Inflammasome Activation by Bacterial Toxins and Particulate Matter

Transcription

1 Article K Efflux Is the ommon Trigger of NLRP Inflammasome Activation by acterial Toxins and Particulate Matter Raúl MuñozPlanillo, Peter Kuffa, Giovanny Martínezolón, renna L. Smith, Thekkelnaycke M. Rajendiran,, and Gabriel Núñez, Department of Pathology and omprehensive ancer enter Michigan enter for Translational Pathology University of Michigan Medical School, Ann Arbor, MI 89, USA orrespondence: gabriel.nunez@umich.edu SUMMARY The NLRP inflammasome is an important component of the innate immune system. However, its mechanism of activation remains largely unknown. We show that NLRP activators including bacterial poreforming toxins, nigericin, ATP, and particulate matter caused mitochondrial perturbation or the opening of a large membrane pore, but this was not required for NLRP activation. Furthermore, reactive oxygen species generation or a change in cell volume was not necessary for NLRP activation. Instead, the only common activity induced by all NLRP agonists was the permeation of the cell membrane to K and Na. Notably, reduction of the intracellular K concentration was sufficient to activate NLRP, whereas an increase in intracellular Na modulated but was not strictly required for inflammasome activation. These results provide a unifying model for the activation of the NLRP inflammasome in which a drop in cytosolic K is the common step that is necessary and sufficient for caspase activation. INTRODUTION A major signaling pathway of the innate immune system is the inflammasome, a multiprotein platform that activates caspase (Schroder and Tschopp, ). Once activated, caspase proteolytically processes several protein substrates including prointerleukinb (proilb) and proil8 into their biologically active forms. To date, four inflammasomes have been described, of which three inflammasomes the NLRP, NLRP, and NLR contain a PRR that belongs to the intracellular Nodlike receptor (NLR) family (Franchi et al., ). Among the NLR inflammasomes, NLRP has been under intense investigation given its link to inherited autoinflammatory syndromes (Hoffman et al., ) and to several acquired inflammatory disorders (Wen et al., ). Activation of the NLRP inflammasome is mediated by two signals. The first signal, referred as priming, is the NFkdependent transcription of NLRP and proilb, through stimulation with Tolllike receptor (TLR) agonists or certain cytokines such as tumor necrosis factora (TNFa) or ILb (auernfeind et al., 9; Franchi et al., 9). The second signal activates NLRP and is induced by nigericin, ATP, bacterial poreforming toxins (PFTs), or crystalline and particulate matter (Hornung et al., 8; Mariathasan et al., ). However, how these structurally unrelated stimuli activate NLRP remains unclear. Several events have been proposed to explain the activation of the NLRP inflammasome including the production of reactive oxygen species (ROS), mitochondrial damage, lysosomal damage, formation of large nonspecific pore in the cell membrane, and cytosolic K efflux (Franchi et al., ). The identification of the cellular event responsible for NLRP activation is complicated by the fact that NLRP activators trigger multiple cellular signals. The paradigm to explain this complexity has been ATP, which causes all the aforementioned cellular events, that is, opens a large pore permeable to monovalent cations and molecules up to 9 Da (Steinberg et al., 987), increases the production of ROS (ruz et al., 7), and damages several organelles including the mitochondria and lysosomes (Lopez astejon et al., ; Shimada et al., ). Furthermore, membrane permeation, lysosomal damage, mitochondrial damage, and ROS production are interrelated cellular events (Guicciardi et al., ), complicating even further the distinction between bystander and causative events of NLRP activation. The objective of this study was to identify the cellular signal responsible for NLRP activation in response to diverse stimuli. For that purpose, we analyzed and compared the cellular effects caused by NLRP activators, including mitochondrial perturbation, ROS generation, change in cell volume, and membrane permeability to organic molecules and ions in order to define the minimal requirement(s) to trigger NLRP. Our results suggest a unifying model for NLRP activation induced by various stimuli in which K efflux is the intracellular event that triggers NLRP activation. RESULTS Mitochondrial Perturbation Is Not Required for NLRP Activation Mitochondrial damage has been implicated in NLRP activation; therefore, we analyzed mitochondrial function in response to the Immunity 8,, June 7, ª Elsevier Inc.

2 A ILβ (ng/ml) E 8 OR (pmolo /min) OR (pmolo /min) Nlrp / Oua OR (pmolo /min) EAR (mph/min) EAR (mph/min) Oua EAR (mph/min) F ILβ (ng/ cells) G OR (pmolo /min) Oua Nlrp / ATP (%) ' ' 8 Oua Oligo NS Oua p p D Lactate (nmoles / cells) FP 7 NS Nlrp / 9 Rot Figure. Mitochondrial Perturbation Is not Required to Activate NLRP (A) LPSprimed and Nlrp / MDMs were stimulated for min with mm nigericin () or. mm gramicidin (), and supernatants were analyzed for ILb by ELISA. () Effect of nigericin ( mm, ) and gramicidin (. mm, ) on mitochondrial function. Oxygen consumption rate (OR) and extracellular acidification rate (EAR) were measured in MDMs. lack arrows indicate the time of addition of the stimuli. ( and D) MDMs were treated for min with nigericin ( mm, ) or gramicidin (. mm, ), and the intracellular levels of ATP () and lactate (D) were determined. (E) Role of Na /K ATPase in mitochondrial perturbation by gramicidin. OR and EAR triggered by the addition of. mm gramicidin () were measured in the absence (upper panels) and presence (lower panels) of mm ouabain (Oua). lack arrows indicate the time of addition of gramicidin. (F) LPSprimed and Nlrp / MDMs were treated for min with. mm gramicidin or vehicle in the presence or absence of mm ouabain (Oua) and changed to their respective medium without gramicidin for an additional min. ILb was measured in supernatants (left panel) and caspase in cell extracts by immunoblotting (right panel). (G) MDMs were stimulated for min or min with. mm gramicidin or left untreated in the presence of mm ouabain (Oua), and the mitochondrial function was evaluated immediately after by performing a bioenergetic profile. Vertical lines indicate the injection of the specified mitochondrial inhibitors. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments. NS, not statistically significant (p R.). p <.. See also Figure S. NLRP agonists nigericin and gramicidin (Figure A; Allam et al., ; Mariathasan et al., ). We monitored mitochondrial function in realtime during stimulation with the NLRP agonists by measuring the O consumption rate (OR) and the extracellular acidification rate (EAR). To ensure that the measured changes in mitochondrial function are upstream to NLRP and are not secondary to caspase activation, we performed all the bioenergetics studies in Nlrp / macrophages unless otherwise specified. oth nigericin and gramicidin caused a rapid increase in OR and EAR in MDMs (Figure ). ericin, however, caused a much greater increase in EAR than gramicidin (Figure ). Furthermore, treatment with nigericin, but not with gramicidin, led to a significant decrease in the intracellular pool of ATP and an increase in lactate concentrations (Figures and D). These results suggest that although both nigericin and gramicidin are robust activators of NLRP, gramicidin causes much less mitochondrial perturbation than nigericin. Thus, we sought to further understand the effects of gramicidin on the mitochondria to elucidate whether mitochondrial perturbation is required for the activation of the NLRP inflammasome. micidin forms pores in lipid bilayers that are permeable to monovalent cations and H O, collapsing the transmembrane gradient of Na and K in treated cells (Andersen et al., ). Therefore, we hypothesized that the rapid rise in the OR triggered by gramicidin is secondary to an increase in energy consumption caused by the activation of the Na and K ATPase. Treatment of MDMs with ouabain, an inhibitor of the Na and K ATPase, did not alter the basal amounts of the OR and EAR (Figure E). However, ouabain abolished the increase in the OR and EAR triggered by gramicidin (Figure E) without impairing its ability to activate NLRP (see Figure SA available online). Similar to gramicidin, the NLRP agonist ATP, by acting on the Px7 receptor (Prx7), permeates the cell membrane and collapses cation gradients (Surprenant et al., 99). onsistently, ATP also elicited a rapid increase in the OR, which was inhibited by ouabain (Figure S) and absent in Prx7 / MDMs (Figure S). To exclude the possibility that the observed inhibition in mitochondrial function by ouabain is due to offtarget effects, we also inhibited the Na and K ATPase by culturing the cells in K free medium because the enzyme cannot function in the absence of extracellular K (Rose and Ransom, 997). Like ouabain, media lacking K did not change the basal OR and EAR but completely blocked the increase in the OR and EAR triggered by gramicidin (Figure SD). ollectively, these results suggest that the effect of gramicidin on mitochondrial function as observed by increases in the OR and EAR is indirect and caused by the activation of the Na and K ATPase. onsistently, inhibition of the Na and K ATPase prevented gramicidininduced changes in the OR and EAR but did not impair the ability of gramicidin to activate NLRP. micidin can also perturb the mitochondria directly by permeating the mitochondrial membranes. To explore this Immunity 8,, June 7, ª Elsevier Inc.

3 A Figure. ROS Production Is Not Required to Activate NLRP (A) LPSprimed and Nlrp / MDMs were stimulated for hr with mm rotenone, mg/ml antimycin A, mm methyladenine (MA), or mmh O, or treated min with. mm of gramidicin. NLRP activation was detected by measuring the secretion of ILb (upper panel) and caspase activation (lower panel). () MH DFDAlabeled MDMs were incubated in medium containing ROS scavengers NacetylLcysteine (NA,. mm), N(Mercaptopropionyl)glycine (MPG,. mm), and ascorbic acid (AA, mm). The oxidation of MH DFDA was measured (upper panel) and the oxidation rate was calculated (lower panel). () The effect of ROS scavengers on signal and signal of NLRP activation induced by gramicidin was evaluated. MDMs were primed with LPS for hr in the presence of ROS scavengers and D E G subsequently stimulated with. mm gramicidin (signal ) or primed for hr with LPS and stimulated with gramicidin in the presence of ROS scavengers (signal ). NLRP activation was assessed by measuring ILb release (upper panel) and caspase activation (lower panel). (D) MDMs labeled with the fluorescent ROS probe MH DFDA were stimulated with. mm gramicidin (), mm H O, or medium. The oxidation of MH DFDA was monitored (upper F panel) as in (), and the oxidation rate was calculated (lower panel). (E) LPSprimed MDMs were stimulated with nigericin ( mm, ), gramicidin (. mm, ), or mm ATP for min with the indicated amount of NA, and caspase activation was analyzed. (F and G) LPSprimed (F) and Nlrp / (G) MDMs were stimulated with nigericin ( mm, ), gramicidin (. mm, ), or ATP ( mm) for min, with Salmonella (MOI, Sal) for hr or with pdadt ( mg/ml) for hr in the presence of the indicated amounts of DPI. aspase activation (F) and the intracellular content of K (G) were measured. aspase activation was analyzed by immunoblotting and ILb by ELISA. ells were treated for min with the inhibitors or medium before adding the agonists. The intracellular levels of K were quantified by IPOES in Nlrp / cells. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments. NS, not statistically significant (p R.). p <. (stimulated versus unstimulated). possibility, we studied the effect of gramicidin on the coupling efficiency and the maximal respiratory capacity of MDMs by performing bioenergetic profiles. For this purpose, macrophages were first stimulated with gramicidin, and the OR was monitored while we sequentially administered the mitochondrial inhibitors oligomycin, FP, and rotenone. To demonstrate that the inhibitors were used at maximally effective doses, we consecutively stimulated MDMs twice with the same dose of inhibitor (Figure SE). Importantly, stimulation of MDMs with. mm gramicidin for min resulted in NLRP activation (Figure F) without affecting mitochondrial function (Figure G). Longer stimulation with gramicidin ( min), however, led to mitochondrial damage demonstrated by the uncoupling of the oxidative phosphorylation and a decrease in the maximal respiratory capacity (Figure G). ollectively, these results demonstrate that perturbation of and damage to the mitochondria that can occur with NLRP activators are not necessary events to activate the NLRP inflammasome. ROS Production Is Not Necessary for NLRP Priming or Activation ROS generation resulting from mitochondrial damage has been implicated in NLRP activation (Zhou et al., ). Unlike this study, we could not detect NLRP activation following stimulation with the mitochondrial toxicants rotenone and antimycin A, the autophagy inhibitor methyladenine, or H O (Figure A). To further clarify the role of ROS in NLRP activation, we studied the effect of the ROS scavengers NacetylLcysteine (NA), ascorbic acid (AA), and N(Mercaptopropionyl)glycine (MPG) on priming (signal ) and NLRP activation (signal ) in response to gramicidin. y using concentrations of ROS scavengers that had a profound inhibitory effect on the cellular redox state (Figure ), we found no effect of these inhibitors on either the priming or activation of the NLRP inflammasome (Figure ). In addition, treatment with gramicidin did not lead to an increase in ROS production (Figure D). High concentrations of NA have been reported to inhibit NLRP activation (ruz et al., 7). However, we did not observe this effect by using NA Immunity 8,, June 7, ª Elsevier Inc.

4 A ILβ (ng/ml) ILβ Nlrp / 8 Nlrp / Al(OH) PPD 8 D 8 8 yt Lat ILβ (ng/ml) ILβ (ng/ml) E ILβ (ng/ml) 9 SiO 9 yt Lat 8 ILβ (ng/ml) ILβ (ng/ml) 9 LLOMe Nlrp / yt Lat ILβ (ng/ml) Figure. Phagocytosis of Particulate Matter Triggers K Efflux and Activates NLRP (A and ) LPSprimed and Nlrp / MDMs were treated for min with mm nigericin (),. mm gramicidin (), mg/ml S. aureus ahemolysin (ah), ng/ml A. hydrophila aerolysin (Aero), or mm ATP. Secreted ILb (A) and the intracellular content of K () were measured. () LPSprimed and Nlrp / MDMs were stimulated with mg/ml of Al(OH), silica (SiO ) or calcium pyrophosphate crystals (PPD), or with mm LleucylLleucine methyl ester (LLOMe), and secreted ILb and the intracellular content of K were determined at the specified time points. (D F) Effect of inhibition of phagocytosis in K efflux and NLRP activation caused by particulate matter and LLOMe. LPSprimed and Nlrp / MDMs were incubated for min with phagocytosis inhibitors cytochalasin (yt, mm) or latrunculin (Lat, nm) and subsequently treated with mg/ml of Al(OH), silica (SiO )or PPD, or with mm LLOMe. The intracellular content of K (D) and ILb release in and Nlrp / MDMs (E) were measured. K determinations were performed by IPOES in Nlrp / cells. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments. Asterisks () indicate NLRP activation (p <., versus Nlrp / ). rosses ( ) indicate a drop in intracellular content of K (p <., stimulated versus nonstimulated). See also Figure S. at neutral ph (Figure E). Similarly, the NAPDH inhibitor DPI did not impair caspase activation at mm (Figure F), a concentration that causes maximal NAPDH inhibition (Decleva et al., ). At a concentration times higher, DPI prevented caspase activation without altering the efflux of K (Figures F and G). However, mm DPI also impaired the activation of the NLR and the AIM inflammasome (Figure F). Thus, ROS do not play a role in NLRP activation. Phagocytosis of Particulate Matter Triggers K Efflux and NLRP Activation A role of K efflux in NLRP activation has been proposed because several NLRP activators can permeate the cell membrane to K (Figures A and ; Perregaux and Gabel, 99) and increasing the extracellular [K ] inhibits inflammasome activation by all tested NLRP activators (Pétrilli et al., 7). However, particulate matter has not been reported to trigger the efflux of K, and there is no evidence that reduction of cytosolic [K ] alone is sufficient to trigger NLRP activation. Therefore, we studied whether NLRP activators that have been proposed to act via lysosomal damage, i.e., particulate matter and the lysosomaldamaging dipeptide LLOMe (Hornung et al., 8), also cause efflux of K. To establish a reliable correlation between intracellular K concentrations and NLRP activation, we measured ILb release and K efflux in parallel. We determined K concentrations in Nlrp / macrophages because caspase activation can lead to pyroptosis and nonspecific membrane permeation (Figure SA). Timecourse experiments revealed that a drop in the intracellular content of K preceded the release of ILb induced by Al(OH), silica, calcium pyrophosphate crystals (PPD), and LLOMe (Figure ). Phagocytic uptake is also a requirement for NLRP activation induced by particulate matter (Hornung et al., 8; Martinon et al., ). Therefore, we investigated a role for phagocytosis in K efflux elicited by particulate NLRP activators. Pretreatment of MDMs with the phagocytosis inhibitors cytochalasin and latrunculin strongly impaired both the efflux of K (Figure D) and NLRPdependent ILb secretion triggered by particulate matter (Figure E), but not by LLOMe (Figures D and E). We did not observe any difference in K efflux caused by NLRP agonists among and Nlrp / unprimed MDMs (Figures S). However, we observed a major effect of LPS priming on K efflux caused by particulate matter and LLOMe. Specifically, LPS priming enhanced K efflux caused by SiO, Al(OH), PPD crystals, and LLOMe, but not by nigericin, gramicidin, and ATP (Figure S; data not shown). These findings are consistent with a role of phagocytosis and pinocytosis in the uptake of these stimuli (Figure D) because LPS treatment has been shown to enhance both processes (hen et al., ; Peppelenbosch et al., 999). The lysosomal inhibitors a7 Me and afilomycin A prevent NLRP activation induced by particulate matter, but not ATP (Hornung et al., 8). In accord with these results, a 7 Me and afilomycin A prevented K efflux and NLRP activation triggered by particulate matter and LLOMe (Figures SD and SE), but not by nigericin or ATP (data not shown). However, ILb release was not impaired in cathepsin deficient MDMs (Figure SF), suggesting that the inhibition by a7 Me is due Immunity 8,, June 7, ª Elsevier Inc.

5 to offtarget effects. Alternatively, a7 Me could be targeting another cathepsin or multiple cathepsins that are responsible for membrane permeation following the uptake of particulate matter. Extracellular a Activates NLRP through K Efflux Lee et al. have recently proposed that calcium signaling plays a crucial role in NLRP activation (Lee et al., ). The authors showed that high extracellular a activates NLRP in RPMI medium. We found that NLRP activation by a was abolished by high extracellular K (Figure SG) and correlated with a decrease in the intracellular amounts of K (Figure SH). Furthermore, we found that high extracellular a causes K efflux and NLRP activation in RPMI medium, but not in IMDM, DMEM, or HSS (Figure SI). However, gramicidin causes K efflux and NLRP activation in all tested media (Figure SI). The concentration of PO in RPMI is significantly higher than that in IMDM, DMEM, and HSS, and calcium phosphate salts are insoluble in aqueous medium and can activate NLRP (Jin et al., ). We noticed that the addition of a to RPMI leads to the formation of particulate matter (Figure SJ). On the contrary, addition of either a or PO to IMDM did not cause any particulate precipitation (Figure SJ). However, when IMDM was supplemented with both a and PO, particulate matter was formed (Figure SJ), which correlated with K efflux and NLRP activation that was inhibited by high extracellular K (Figure SK). These results suggest that high extracellular a activates NLRP acting as particulate matter, i.e., triggering K efflux. The ASR agonist R8 and the PL activator mmfs have also been reported to activate NLRP (Lee et al., ). We could detect NLRP activation by mmfs, but not R8 (Figure SL). Notably, caspase activation by mmfs was inhibited by high extracellular K (Figure SL). K Efflux Is a Specific Upstream Requirement for the Activation of the NLRP Inflammasome AIM recognizes cytosolic doublestranded DNA and activates caspase, which was found to be inhibited by concentrations of extracellular K higher than mm (FernandesAlnemri et al., ). Given the potential toxic effects of very high extracellular K concentrations, we determined the extracellular [K ] sufficient to prevent NLRP activation by nigericin, bacterial PFTs, ATP, particulate matter, and LLOMe. Increasing the extracellular [K ] from to mm had a major inhibitory effect on NLRP activation by all tested stimuli (Figures A and ). Furthermore, increasing the extracellular [K ] from to mm had a minor or no additional inhibitory effect on NLRP activation, and raising it from to mm did not have any additional inhibitory effect for any of the stimuli (Figures A and ). Thus, an extracellular [K ] of mm provides maximal inhibition of NLRP activation by all activators tested. Activation of AIM is associated with recruitment of the bipartite adaptor Asc, leading to its oligomerization and the activation of caspase (FernandesAlnemri et al., ). onsistent with this study, ILb secretion induced by the AIM activator polydadt was totally dependent on Asc, but independent of NLRP (Figure ). Unlike gramicidin, stimulation with polydadt did not elicit a decrease in the intracellular content of K (Figure D). Furthermore, ILb secretion triggered by polydadt was not inhibited by an extracellular [K ] of mm, which effectively blocked gramicidininduced ILb secretion (Figure E). ollectively, these results indicate that the efflux of K is specific to NLRP activation and does not play a role in the activation of the AIM inflammasome. NLRP also requires the adaptor Asc to activate caspase (Mariathasan et al., ). Similar to ATP (Juliana et al., ), gramicidin elicited Asc oligomerization and its migration to the detergentinsoluble protein fraction in cellular extracts (Figure F). Asc oligomerization was inhibited by mm extracellular K (Figure F), suggesting that K regulates the NLRP inflammasome by acting upstream of caspase. To better define the step in the NLRP inflammasome signaling pathway regulated by K, we analyzed the role of K in caspase activation in MDMs harboring the Nlrp R8W mutation. This mutation corresponds to the RW mutation in human NLRP, which is associated with MuckleWells syndrome. In agreement with a previous study (Meng et al., 9), treatment of Nlrp R8W MDMs with LPS alone was sufficient to activate caspase and was blocked by the caspase inhibitor YVAD (Figures G and H). However, caspase activation elicited by LPS was not inhibited by media containing mm of K and did not correlate with an efflux of K (Figure G I). In sum, these results indicate that K regulates the activation of the NLRP inflammasome upstream of Asc; i.e., K efflux acts either on NLRP or upstream of NLRP. The Formation of a Large Pore Is Not Required to Activate NLRP We showed above that particulate matter share the ability with poreforming toxins to cause K efflux. However, it is unknown whether a drop in cytosolic [K ] is sufficient to activate NLRP because ATP also induces the formation of a large unspecific pore in the cell membrane that was suggested to be involved in caspase activation (Pelegrin and Surprenant, ). Therefore, we determined whether permeation of the cell membrane to moieties larger than K is a common feature among NLRP activators. Stimulation of MDMs with nigericin and the bacterial PFTs gramicidin, ahemolysin, and aerolysin did not permit the uptake of ethidium (molecular weight ), as opposed to ATP (Figure A). Strikingly, stimulation with particulate matter and LLOMe increased ethidium uptake (Figure ). In contrast to ATP, ethidium uptake triggered by particulate matter and LL OMe was unimpaired in Prx7 / MDM (Figure ). Furthermore, ethidium uptake elicited by all stimuli was upstream to NLRP as it was undiminished in Nlrp / MDMs (Figure ). Notably, inhibition of phagocytosis with cytochalasin and latrunculin strongly reduced ethidium uptake triggered by particulate matter, but not by LLOMe (Figure D). Next, we analyzed whether NLRP activators can permeate the cell membrane to the organic osmolyte taurine (molecular mass of ). Incubation of MDMs in the presence of [ H] taurine revealed that MDMs avidly incorporate the organic osmolyte (Figure SA). acterial PFTs and ATP elicited the efflux of [ H]taurine but, remarkably, nigericin did not (Figure E). [ H] taurine efflux occurred upstream of NLRP for all stimuli and was independent of the Prx7 for bacterial PFTs, but not for ATP (Figure E). onsistent with their ability to elicit the uptake of ethidium, particulate matter also triggered the efflux of the Immunity 8,, June 7, ª Elsevier Inc.

6 A 7 mm K αh Aero ATP ILβ (ng/ml) D G ILβ (ng/ml) 8 mm K LPS YVAD E ILβ (ng/ml) Nlrp R8W H YVAD LPS mm K p p ILβ (ng/ml) mmk mm K Asc / Nlrp R8W F I mm K αas 8 mm K ILβ (ng/ml) Supernatant LPS % NP Asc / Nlrp / Pellet KDa Figure. ytosolic K Is a Specific Upstream Regulator of the NLRP Inflammasome (A) LPSprimed MDMs were stimulated for min with mm nigericin (),. mm gramicidin (), mg/ml S. aureus ahemolysin (ah), ng/ml A. hydrophila aerolysin (Aero), or mm ATP in medium containing the specified [K ], and secreted ILb was measured. () LPSprimed MDMs were stimulated for hr with mg/ml of Al(OH), silica (SiO ) or calcium pyrophosphate crystals (PPD), or with mm LleucylLleucine methyl ester (LLOMe) in medium containing the specified [K ], and secreted ILb was measured. ( and D) LPSprimed, Asc /, and Nlrp / MDMs were stimulated with mg/ml pdadt for hr or. mm gramicidin () for min, and the release of ILb () and the intracellular content of K (D) were measured. K determinations were performed in Asc / macrophages. (E) LPSprimed and Asc / MDMs were treated with mg/ml pdadt for hr or. mm gramicidin () for min in medium containing ormmk, and ILb was quantified in the supernatants. (F) LPSprimed MDMs were stimulated min with. mm gramicidin () in medium containing or mm K. The cells were lysed with % NP and separated by centrifugation in supernatants and pellets. Proteins in cell pellets and supernatants were crosslinked with DSS and immunoblotted with antiasc antibody. (G and H) MDM from and Nlrp R8W mice were stimulated as indicated for hr (. mg/ml LPS, mm YVAD) and released ILb (G), and caspase activation (H) were analyzed. (I) MDMs from Nlrp R8W mice were treated with. mg/ml LPS or vehicle for hr, and the intracellular content of K was determined. ILb was measured by ELISA, caspase activation by immunoblotting and intracellular K by IPOES. In experiments with high K medium, the osmolarity was maintained at mosm by isosmotic substitution of Nal with Kl. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments. smaller molecule [ H]taurine, which was detectable as early as min after stimulation (Figure S) and strongly impaired by the phagocytosis inhibitors cytochalasin and latrunculin (Figure S). These results demonstrate that membrane permeation to chemical species larger than K is commonly caused by NLRP activators because eight out of the nine agonists tested permeated the cell membrane to [ H]taurine and/or ethidium (Figure E; Figure S). However, NLRP activation is unlikely to be triggered by the formation of a large nonselective pore as previously suggested (Pelegrin and Surprenant, ), because nigericin activates NLRP without permeating the cell to the small metabolite taurine (Figure E). NLRP Activation orrelates with Efflux of K and the Influx of Na but Not with the Permeation to l To better define the minimal membrane permeation events required to activate NLRP, we investigated whether NLRP activation requires increased cell membrane permeability to Na and l ions. Treatment of MDMs with gramicidin caused an increase in the intracellular levels of both Na and l in addition to a decrease in the intracellular content of K (Figure F). ericin only caused a decrease in intracellular K and an increase in intracellular Na but did not produce a change in l concentrations (Figure F). Hence, NLRP activation correlates with the efflux of K and the influx of Na, but not with l fluxes. K Free Medium Alone Activates the NLRP Inflammasome To further evaluate the role of K in NLRP activation, we tested whether decreasing the intracellular content of K by incubating the cells in lowk media is sufficient to activate the NLRP inflammasome. We examined this hypothesis by using two different approaches. First, we incubated MDMs from wildtype () and Nlrp / mice in medium containing mm or mmk and analyzed the secretion of ILb and the intracellular content of K at different time points. When macrophages were incubated in K free medium, ILb secretion was detected as early as min in the absence of any stimuli and correlated Immunity 8,, June 7, ª Elsevier Inc. 7

7 A Ethidium fluorescence (FU) ΔF Ethidium (FU) F Intracellular (%) 8 αh Aero ATP D Nlrp / Prx7 / ΔF Ethidium (FU) K 8 Na l 7 ΔF Ethidium (FU) αh Aero ATP Ethidium fluorescence (FU) yt Lat E [ H]Taurine efflux (%) Al(OH) SiO PPD LLOMe Nlrp / Prx7 / Figure. NLRP Activation orrelates with K Efflux and Na Influx but Not with the Opening of a Nonselective Pore (A) Ethidium uptake kinetics in MDMs treated with mm nigericin (),. mm gramicidin (), mg/ml S. aureus ahemolysin (ah), ng/ml A. hydrophila aerolysin (Aero), or mm ATP. The total increase of ethidium fluorescence during the stimulation (DF ethidium) is shown in the right panel. () Ethidium uptake kinetics in MDMs treated with mg/ml of Al(OH), silica (SiO ) or calcium pyrophosphate crystals (PPD), or with mm LleucylLleucine methyl ester (LLOMe) for hr. (), Nlrp /, and Prx7 / MDMs were stimulated for hr with mg/ml of Al(OH), silica (SiO ), PPD crystals, and mm LLOMe, and the ethidium uptake was quantitated. ATP ( mm, min) was used as a control for Prx7 signaling. (D) Ethidium uptake by MDMs stimulated for hr with mg/ml Al(OH), silica (SiO ), PPD crystals, and mm LLOMe in the presence of phagocytosis inhibitors cytochalasin ( mm, yt ) and latrunculin ( nm, Lat ). (E) [ H]taurine efflux was determined in, Nlrp /, and Prx7 / MDMs treated min with mm nigericin (),. mm gramicidin (), mg/ml S. aureus ahemolysin (ah), ng/ml A. hydrophila aerolysin (Aero), or mm ATP. (F) The intracellular content of K,Na, and l was determined in MDMs treated min with mm nigericin () or. mm gramicidin (). Na and K were measured by IPOES and l by IPMS in Nlrp / cells. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments. FU, fluorescence units. See also Figure S. with an intracellular content of K % 7% ±.% (average ± SEM of three independent experiments) (Figure A). In a second approach, we incubated MDMs for hr in media containing decreasing [K ] and measured the secretion of ILb, caspase activation, and the intracellular content of K (Figure ). Under these conditions, ILb secretion in MDMs was only detected when the extracellular [K ] was %. mm and correlated with an intracellular content of K of % 77% ±.% (average ± SEM of three independent experiments) (Figure ). These results demonstrate that incubating MDMs in lowk medium is sufficient to activate NLRP in the absence of an NLRP agonist. Furthermore, our results indicate that the threshold of intracellular K to engage NLRP is in the range of 7% 8%. The Na /K ATPase inhibitor ouabain enhanced NLRP activation induced by K free medium, and at higher doses it was sufficient to activate NLRP (Figure ). ompan et al. proposed that a regulatory volume decrease (RVD) following cell swelling is necessary in addition to K efflux to activate NLRP (ompan et al., ). However, the latter experiments were performed by diluting isotonic medium with distilled water and therefore did not consider the effect of reducing the extracellular [K ] in NLRP activation. Thus, we analyzed the individual contribution of lowering the osmolarity and the extracellular concentration of K to NLRP activation. Notably, only incubation of macrophages in highly hypotonic medium (9 mosm) activated the NLRP inflammasome (Figure SA), which correlated with nonspecific membrane permeation and significant cellular toxicity as evidenced by increased LDH release and a drop in ATP levels (Figure S). 9 mosm medium produced robust NLRP activation only when the extracellular [K ] was below mm (Figure SA). Increasing the extracellular [K ] prevented NLRP activation, but did not decrease cytotoxicity, further suggesting that K efflux and not cytotoxicity resulted in NLRP activation (Figure S). Furthermore, incubation of macrophages in mosm medium induced significant cell swelling (Figure SD) and a RVD response (Figures S) but did not activate NLRP (Figure SA). Notably, K free medium activates NLRP (Figures A and ) without causing cell swelling (Figure SE) and a RVD (Figure S). ollectively, these results suggest that K free medium activates NLRP without swelling the cells. Furthermore, cell swelling or RVD does not activate NLRP, and NLRP activation by severe hypoosmolarity is largely due to the efflux of K. In another study, Lee et al. reported that a decrease in camp leads to NLRP activation (Lee et al., ). However, we could not detect a decrease in the camp levels upon stimulation with K free medium (Figure SF). Na Influx Is Not an Absolute Requirement for NLRP Activation We next evaluated the role of Na influx in NLRP activation by isoosmotically substituting extracellular Na by the cation 8 Immunity 8,, June 7, ª Elsevier Inc.

8 A ILβ (ng/ml) ILβ (ng/ml) μm Ouabain ILβ ILβ Nlrp / 7 9 ILβ ILβ Nlrp / choline. Remarkably, reducing the extracellular [Na ] had a strong dosedependent inhibitory effect on NLRP activation induced by K free medium, gramicidin, or nigericin (Figures 7A and 7; Figure SA). These stimuli required an extracellular [Na ] R,, and mm, respectively, to activate NLRP (Figures 7A and 7; Figure SA). Furthermore, lowering the extracellular [Na ] also reduced the drop in the intracellular K caused by all three stimuli (Figures 7A and ; Figure SA). None of these stimuli activated NLRP in medium containing mm Na despite the fact that all of them induced a reduction in the intracellular content of K below % (Figures 7A and 7; Figure SA). Indeed, lowering the extracellular [Na ] decreased the K threshold for NLRP activation induced by lowk medium from 7% 8% (Figures A and ) to % % (Figures 7A and 7; Figure SA). These results suggest that Na influx can modulate NLRP activation independently of K efflux. However, substitution of extracellular Na with choline did not impair K efflux or NLRP activation elicited by ATP (Figure 7), aerolysin, Al(OH) and silica (Figure 7D; Figures S and S). Thus, our results demonstrate that Na influx can modulate NLRP activation by certain agonists, but it is not a strict requirement for NLRP activation. To further assess a role for Na in NLRP activation, we tested whether the influx of Na is sufficient to activate NLRP. Treatment of MDMs with doses of the Na ionophore monensin previously shown to cause significant Na influx (Gurcel et al., ) did not cause NLRP activation (Figure S). Next, we addressed a possible role of membrane depolarization in NLRP activation. micidin forms pores in the cell membrane that allow K efflux and Na influx leading to membrane depolarization in highna medium (hifflet et al., ). How IMDM mmk mm K ILβ (ng/ml) p p. ILβ ILβ Nlrp / Nlrp / Figure. Incubation in LowK Medium Is Sufficient to Activate NLRP (A) LPSprimed and Nlrp / MDMs were incubated in K free medium, and the release of ILb (bars) and intracellular content of K (solid squares) were measured at the specified time points. () LPSprimed and Nlrp / MDMs were incubated for hr in medium containing the specified [K ], and the release of ILb (bars) and intracellular content of K (solid squares) were measured. () LPSprimed and Nlrp / MDMs were treated with ouabain for hr in IMDM or K free medium, and the release of ILb and the intracellular content of K were determined. ILb secretion was analyzed by ELISA. K determinations were performed in Nlrp / cells by IPOES. Values represent mean ± SD (n = )., statistically significant (p <., versus Nlrp / ). Results are representative of at least three separate experiments. See also Figure S. ever, K efflux with decreased Na influx leads to hyperpolarization (laustein and Goldring, 97; Langheinrich and Daut, 997). To test whether membrane depolarization is required for NLRP activation, we treated the cells with gramicidin in K free medium in which Na was substituted by the membrane impermeable cation choline (Figure SD). Under these hyperpolarizing conditions (laustein and Goldring, 97; Langheinrich and Daut, 997), gramicidin caused robust caspase activation (Figure SD). These results indicate that membrane depolarization is not required for NLRP activation and further support K efflux as the trigger of the NLRP inflammasome. DISUSSION The fact that NLRP is activated by an array of chemically and structurally unrelated stimuli has led to the hypothesis that NLRP does not directly detect these stimuli but instead senses a commonly induced intracellular signal. In line with this notion, several intracellular events have been proposed as the common signal upstream to NLRP, including a change in the intracellular concentration of K, the formation of a large pore in the cell membrane, lysosomal destabilization, mitochondrial damage, the production of ROS, changes in cell volume, and a signaling. The elucidation of the mechanism of NLRP activation is further complicated by the pleiotropic action of NLRP agonists. For example, ATP permeates the cell membrane to molecules up to 9 Da, damages lysosomes and the mitochondria, and increases the production of ROS. In this manuscript, we have analyzed the cellular events that have been proposed to serve as the common conduit to activate the NLRP inflammasome by using a panel of stimuli. Our work indicates that intracellular K depletion alone acting on or upstream of NLRP is the minimal common cellular event that is necessary and sufficient to activate the NLRP inflammasome. Immunity 8,, June 7, ª Elsevier Inc. 9

9 A ILβ ILβ Nlrp / mm Na ILβ (ng/ml) ILβ (ng/ml) 8 ILβ ILβ Nlrp / Nlrp / mm Na ATP Mitochondrial damage has been suggested to be the upstream signal responsible for NLRP activation (Zhou et al., ), and cells treated with nigericin and ATP were found to have decreased OR after min of stimulation (Shimada et al., ). Unlike this study, we found that nigericin, gramicidin, and ATP caused a rapid increase in the OR. Robust NLRP activation by nigericin, gramicidin, and ATP occurs within min of stimulation. Therefore, it is possible that the mitochondrial damage in response to nigericin and ATP observed by Shimada et al. () at later time points is due to cytotoxic effects of these stimuli and not involved in triggering NLRP activation. We found that the rapid increase in OR and EAR elicited by gramicidin was mediated by activation of the Na /K ATPase and was unrelated to the activation of the NLRP inflammasome. Although gramicidin can damage the mitochondria after prolonged stimulation, we found conditions in which robust NLRP activation was observed in the absence of mitochondrial 8 p p p p ILβ (ng/ml) ILβ ILβ Nlrp / 8 8 mm Na 7 p D Aero Al(OH) SiO mm Na p p Immunity 8,, June 7, ª Elsevier Inc. p Figure 7. Na Influx an Modulate NLRP but Is Not a Strict Requirement for Inflammasome Activation (A) LPSprimed and Nlrp / MDMs were incubated for hr in K free medium containing the specified [Na ], and the release of ILb (bars) and the intracellular content of K (solid squares) were measured. ( and ) LPSprimed and Nlrp / MDMs were stimulated for min with. mm gramicidin () () or mm ATP () in media containing mm K and the specified [Na ]. The release of ILb (bars), the intracellular content of K (solid squares), and caspase activation were analyzed. (D) LPSprimed MDMs were stimulated for min with. mm gramicidin () or ng/ml aerolysin (Aero) or for hr with mg/ml Al(OH) or silica (SiO ) in medium containing mm K and either or mm Na, and caspase activation was analyzed. In low Na medium, Nal was isosmotically substituted with choline chloride to maintain a final osmolarity of mosm. ILb was measured by ELISA and caspase activation by immunoblotting. K determinations were performed in Nlrp / cells by IPOES. Values represent mean ± SD (n = ). Results are representative of at least three separate experiments., statistically significant (p <., versus Nlrp / ). See also Figure S. perturbation. ollectively, these results indicate that although gramicidin can alter the mitochondrial function through different mechanisms, these effects are not required for NLRP activation. ROS generation secondary to mitochondrial damage has also been implicated in NLRP activation because manipulation of the respiratory chain with chemical inhibitors was reported to trigger NLRP activation (Zhou et al., ). However, unlike this study, we could not detect NLRP activation in cells treated with the mitochondrial toxicants rotenone and antimycin A at maximally effective doses or with the autophagy inhibitor MA or H O. We do not have an explanation to account for the difference in results. High concentrations of ROS inhibitors block NFkmediated priming of the NLRP inflammasome, but not NLRP activation induced by nigericin and silica (auernfeind et al., ). y using lower concentrations of ROS scavengers that had a strong inhibitory effect on the cellular redox state, we could not see an inhibitory effect in either priming or NLRP activation. Thus, our results suggest that ROS generation does not play a crucial role in NLRP activation. Pioneering studies proposed that the efflux of K is responsible for the maturation of proilb because ATP and nigericin permeate the cell membrane to K and high extracellular K prevents proilb processing (Perregaux and Gabel, 99). Subsequent studies further supported the hypothesis that NLRP activation by all tested stimuli is inhibited by high extracellular [K ](Pétrilli et al., 7). However, the role of K in NLRP

10 activation by particulate matter has been questioned because, to date, there is no experimental evidence that K efflux occurs during stimulation with particulate matter (Tschopp and Schroder, ). Furthermore, a recent report suggested that particulate matter and soluble agonists activate NLRP through two distinct mechanisms (Shenoy et al., ). Notably, we found that K efflux precedes the activation of NLRP induced by particulate matter and the lysosomal damaging dipeptide LLOMe. In addition, we found that NLRP activation by high extracellular a is due to K efflux and could be explained by the formation of particulate matter. Therefore, our work supports a unifying model for NLRP activation by membrane permeating molecules and particulate matter in which a decrease in the cytosolic concentration of K engages the NLRP inflammasome. These results suggest that the internalization of particulate matter via phagocytosis induces lysosomal membrane damage, which triggers the opening of one or more membrane pores permeable to K. It has also been proposed that the opening of an unspecific pore formed by the hemichannel pannexin after Prx7 stimulation by ATP or a RVD response is necessary for NLRP activation in addition to K efflux (ompan et al., ; Pelegrin and Surprenant, ). However, recent studies using MDMs deficient in pannexin showed that this hemichannel is not the molecular substrate of the large pore opened by ATP (Qu et al., ). To clarify the role of a large pore in NLRP activation, we studied membrane permeation caused by NLRP activators to a set of molecular markers of decreasing size: ethidium, [ H] taurine, l, and Na. ATP, particulate matter, and LLOMe caused membrane permeation to K and the larger chemical species [ H]taurine and ethidium. Membrane permeation to [ H]taurine and ethidium by particulate matter was not mediated by the Prx7 and was secondary to phagocytosis because it was strongly inhibited by cytochalasin and latrunculin. In contrast, the bacterial PFTs gramicidin, ahemolysin, and aerolysin did not permeabilize the cell membrane to ethidium but caused the efflux of the smaller marker [ H] taurine. However, the H /K ionophore nigericin caused the efflux of K and the influx of Na but did not permeabilize the cell membrane to l, [ H]taurine, or ethidium. Therefore, the minimal membrane permeabilization events associated with NLRP activation are the efflux of K and the influx of Na. Depletion of intracellular K by incubating the cells in low K medium was sufficient to activate NLRP, which occurred when the intracellular content of K dropped below 8%. In addition, K free medium activated NLRP without causing a change in cell volume or a RVD response. It was suggested that K efflux might not be sufficient to activate NLRP because decreasing extracellular Na prevents NLRP activation (Perregaux and Gabel, 998). Accordingly, substituting of extracellular Na by the cation choline had a strong inhibitory effect in NLRP activation by K free medium, nigericin, and gramicidin. However, NLRP activation by ATP, aerolysin, Al(OH), and silica did not require extracellular Na. Therefore, although all NLRP activators tested permeabilized the cell membrane to Na, the influx of Na or membrane depolarization was not an absolute requirement for NLRP activation. Thus, our results demonstrate that a drop in the cytosolic content of K acts as the common signal triggered by bacterial PFTs and particulate matter, which is sufficient to engage the NLRP inflammasome. EXPERIMENTAL PROEDURES Elemental Analysis Intracellular K and Na measurements were performed by inductively coupled plasma optical emission spectrometry (IPOES) with a PerkinElmer Optima DV spectrometer using yttrium as internal standard. l was measured by inductively coupled plasma mass spectrometry (IPMS) at the W. M. Keck Elemental Geochemistry laboratory (University of Michigan). The culture media were thoroughly aspirated, and cells were extracted after min in % ultrapure HNO.K determinations were done in 9well plates. When K,Na, and l were simultaneously analyzed, well plates were used. For accurate measurement of the intracellular ions, a control was performed in every experiment to determine the extracellular amount of the investigated ion remaining after aspiration, and this value was subtracted from every measurement. Also for accurate elemental determinations, analyses were performed in inflammasomedeficient cells to avoid ion fluxes due to unspecific membrane permeation secondary to pyroptosis. In experiments using low extracellular K and/or Na, cells were washed with medium low in the respective ion prior to stimulation to avoid ion carryover. Metabolic Analysis Oxygen consumption rate and extracellular acidification rate were measured by using an XF Extracellular Flux Analyzer (Seahorse ioscience). MDMs were seeded in well plates (. /well). The following day, cells were changed to bicarbonatefree DMEM and incubated in a nono incubator for hr before the experiment. Mix, wait, and measure times were,, and min, respectively. Values were obtained from the average of four wells. Lactate was measured on the Siemen s Advia autoanalyzer by using a lactate oxidase colorimetric method. Intracellular ATP was measured by using the elltitergloò Luminescent ell Viability Assay (Promega). [ H]Taurine Release MDMs were seeded in 9well plates, and hours later the medium was replaced by ml of medium containing. mi/ml of [, H]taurine. The following day, the cells were washed three times with PS and stimulated. The stimulations were terminated by rapid aspiration of the medium, and cells were lysed with % trichloroacetic acid. Taurine efflux was calculated as a fractional release; i.e., the radioactivity released into the extracellular medium as a percentage of the total radioactivity present initially in the cells. The latter was calculated as the sum of radioactivity recovered in the supernatant and that remaining in the cells at the end of the assay. ell volume measurements were performed with a oulter ounter (eckman). SUPPLEMENTAL INFORMATION Supplemental Information includes five figures and Supplemental Experimental Procedures and can be found with this article online at org/./j.immuni... AKNOWLEDGMENTS The studies were supported by National Institutes of Health (NIH) grants AI, DK99, and AR988 (to G.N.). ioenergetics studies were done at the Molecular Phenotyping ore, supported by the Michigan Nutrition Obesity Research enter (NIH grant DK89). G.M.. was supported by the postbaccalaureate Research Education Program grant R GM8 from the NIH. We thank Millenium Pharmaceuticals, George Dubyak, Eicke Latz, and Warren Strober for generously supplying mutant mice, Yuumi Nakamura for maintaining the Nlrp R8W knockin colony, Stephen Fisher for help with [ H]taurine experiments, ce hen and Luigi Franchi for critical review of this manuscript, Joel Whitfield for ELISA measurements, Sherry Koonse for animal husbandry, Sydney ridges for performing Seahorse experiments, Ted Huston for IPMS measurements, Keith Matz for lactate measurements, and James Windak for IPOES technical support. Immunity 8,, June 7, ª Elsevier Inc.