Shiva Safai, how old is she

Macrophages sense and kill bacteria through carbon monoxide-dependent inflammation activation

HO-1 controls bacteria-induced IL-1β expression by macrophages. Primary macrophages exposed to bacteria or endotoxin exhibit a marked increase in expression of HO-1 and proinflammatory cytokines such as IL-1β (32, 41) To test the role of HO-1 in IL-1β expression and activation in response to bacteria challenge , we used macrophages from Hmox1–/–other LyzM-Cre Hmox1fl / fl mice and WT macrophages infected with retroviral HO-1 shRNA (Figure 1, A – C). Activation of IL-1β in response to bacteria was largely dependent on the presence and activity of HO-1, as macrophages lacking HO-1 expression showed no increase in cleaved IL-1β and expressed lower pro – IL-1β versus controls, as measured by immunoblot or ELISA (Figure 1, A – C). Of note, commercial ELISA kits detect pro-IL-1β and thus underestimate IL-1β processing by the NALP3 inflammasome pathway. The effects on IL-1β were specific and not explained by a global downregulation of macrophage activation in the absence of HO-1, as bacteria exposure induced a robust increase in TNF protein that was independent of HO-1 (Figure 1D). The effects of HO-1 on IL-1β were determined to be at the protein level (both the pro and cleaved forms of IL-1β). No statistical change in inflammasome component mRNA was observed between Hmox1+/– other Hmox1–/– macrophages in the presence or absence of bacteria (Figure 1E). Therefore, these data suggest that HO-1 is involved in part in regulation of IL-1β expression and processing via the inflammasome multiplex including NALP3 and the pro-IL-1β cleavage enzyme caspase-1.

Figure 1

Role of HO-1 on IL-1β release from bacteria-infected macrophages. (A.C.) Representative immunoblots and ELISA for IL-1β in BMDMs or PMs treated with indicated bacteria (106 CFU / ml E. faecalis or 104 CFU / ml E. coli) for 10 hours. (A.) Hmox1–/– (-) and Hmox1+/+ (+) BMDMs. (B.) Immunoblot quantitation of PMs from LyzM-Cre Hmox1fl / fl other Hmox1fl / fl injected i.p. with E. coli for 2 hours (inset; Hmox1 expression in naive PMs). (C.) Retroviral shRNA against HO-1 or latent membrane potential (LMP) control in BMDMs treated as in A.. #P. < 0.01, LyzM-Cre Hmox1fl / fl versus Hmox1fl / fl; **P. < 0.001, *P. <0.05 versus LMP. Results represent average ± SD of 2 independent experiments in triplicate. (D.) TNF ELISA in supernatants from BMDMs treated as in A.. Results represent mean ± SD of 2 independent experiments in triplicate. (E.) Real-time PCR of cDNA from cells treated as in A.. Results represent mean fold change versus control ± SD of 2 independent experiments in triplicate. *P. < 0.02. (F. other G) IL-1β ELISA and immunoblot of BMDM supernatants from Hmox1+/+ other Hmox1–/– treated ± LPS (1 ng / ml) for 5 hours followed by 100 μM, 500 μM, or 1000 μM ATP. Data represent mean ± SD of n = 4-6 from 3 independent experiments. *P. < 0.05.

To evaluate whether macrophages lacking HO-1 were deficient in their ability to activate the inflammasome, we treated macrophages with exogenous ATP, a well-characterized NALP3 stimulus. BM-derived macrophages (BMDMs) isolated from Hmox1+/+ or Hmox1–/– were activated with bacterial LPS followed by treatment with ATP. LPS treatment increased pro – IL-1β expression, which was then cleaved to IL-1β with the addition of ATP in both Hmox1–/– other Hmox1+/+ macrophages (Figure 1, F and G). Macrophages lacking HO-1 showed 25% to 30% less cleaved IL-1β compared with WT cells, which was in part related to there being less pro-IL-1β. While Hmox1–/– BMDMs respond to LPS plus ATP with a significant increase in pro-IL-1β, they have been shown to also exhibit a delay in differentiation and are therefore less responsive to LPS due to lower expression of CD14 (42). Collectively, these data suggest that ATP-dependent macrophage activation is in large part independent of HO-1, but indicate that HO-1 may play a role in IL-1β processing. The dependency of IL-1β activation on HO-1 expression in response to bacteria was corroborated pharmacologically. We saw similar effects in WT macrophages treated with tin-protoporphyrin-IX (Sn-PP-IX), which blocks HO-1 and HO-2, as well as the highly HO-1 – selective inhibitor QC-15 (ref. 43 and Figure 2A).

Figure 2

Role of HO-1 and CO in IL-1β activation in macrophages. (A.) IL-1β ELISA and immunoblot in BMDMs ± selective HO-1 inhibitors Sn-PP-IX and QC-15 ± CO started 4 hours after bacteria administration (106 CFU / ml) for 6 hours. Note that CO reversed the loss in HO-1 activity found in air treatment. *P. <0.05, air versus air + Sn-PP-IX; &P. <0.01, air versus CO; #P. <0.02, CO + Sn-PP-IX versus Sn-PP-IX. (B.) Upper panel: IL-1β in BMDMs from the indicated mice ± E. coli (104 CFU / ml) ± CO as above. Note that CO rescues the IL-1β response in LyzM-Cre Hmox1fl / fl that is absent in air-treated controls (comparing lanes 2 and 5). Lower panel: immunoblot showing disappearance of HO-1 in LyzM-Cre Hmox1fl / fl differentiated over 5 days in response to MCSF (20 ng / ml). Note that as the Lyz promoter becomes active, HO-1 expression decreases. (C.) Representative IL-1β and active caspase-1 p20 immunoblots in BMDMs. CO was administered for 6 hours starting 4 hours following E. faecalis infection (106 CFU / ml). (D.) TNF ELISA in supernatants from BMDMs infected with E. faecalis as above. All blots represent 2 to 3 independent experiments expressed relative to β-actin as loading control. All ELISA data represent mean ± SD of 2 to 3 independent experiments in triplicate.

CO modulates IL-1β activation and substitutes for HO-1. HO-1 generates biologically active CO that not only imparts antiinflammatory effects, but also augments immune responses against pathogens (44). We therefore next tested whether administration of exogenous CO could rescue HO-1 deficiency and reestablish the IL-1β response. Administration of exogenous CO to macrophages treated with the selective inhibitors of HO-1 was as follows: Sn-PP-IX or QC-15 or macrophages harvested from LyzM-Cre Hmox1fl / fl restored IL-1β secretion in response to bacteria in the presence of E. coli (Figure 2, A and B). Moreover, WT cells (either BMDMs or primary peritoneal macrophages [PMs]) showed an enhanced response to CO treatment with significantly greater IL-1β expression and caspase-1 activation compared with air-treated controls, suggesting that CO targeted in part of the inflammasome machinery that generates active IL-1β (Figure 2C). We also tested an intracellular pathogen, Salmonella typhimurium, which mediates IL-1β secretion via a NALP3 inflammasome – independent pathway and observed no difference in IL-1β secretion in Hmox1+/+ other Hmox1–/– macrophages, further suggesting that CO produced by HO-1 specifically targets the NALP3 inflammasome pathway (Supplemental Figure 1A; supplemental material available online with this article; doi: 10.1172 / JCI72853DS1). Further, biliverdin, the other product of HO-1 activity, had no effect on bacteria-induced IL-1β expression (Supplemental Figure 1B). Finally, the effects of CO, like those observed with HO-1, were specific to IL-1β and not explained by global modulation of macrophage function because treatment with CO had no therapeutic effect on bacteria-induced TNF expression (Figure 2D).

HO-1 and CO are important in bacteria killing by macrophages. We next sought to study whether the effects of HO-1 and CO on IL-1β resulted in a functional effect on macrophage bacterial killing. IL-1β has been shown to influence bacterial phagocytosis (45). Bacterial killing was impaired in macrophages lacking HO-1 as expected, an effect rescued by administration of exogenous CO (Figure 3A). Initial experiments employed the well-characterized model of bacterial clearance in which bacteria growth is controlled by the addition of antibiotics. We found that addition of antibiotics 2 hours after infection to eliminate extracellular bacteria further amplified CO effects, with more rapid bacterial processing and less intracellular bacteria viability (Figure 3B). All subsequent experiments therefore were performed in the absence of antibiotics, which we considered a stronger correlate with our in vivo sepsis models where no antibiotics were used. The effects of CO on bacteria clearance were similar in mouse and primary human macrophages (Figure 3C). Strikingly, macrophages treated with CO were able to more effectively clear bacteria at concentrations that were otherwise unable to be cleared in air-treated control macrophages (Figure 3D). These effects mediated by CO were unrelated to direct bactericidal effects, as CO had no effect on bacterial growth in the absence of macrophages (Figure 3E). Macrophage viability was also not altered by CO when tested in the presence of Enterococcus faecalis, where we observed 97% ± 2% viability in air versus 93% + 4% in CO measured by trypan blue exclusion. Collectively, these data show that expression of HO-1 and its bioactive product CO are required in part for bacterial clearance by macrophages.

Figure 3

CO enhances bacterial killing by macrophages. (A.) CFU in BMDM lysates ± E. coli (104) administered for 1 hour followed by exposure to CO or air for 6 hours. #P. <0.05, CO versus air in LyzM-Cre Hmox1fl / fl other Hmox1fl / fl macrophages; *P. < 0.05, LyzM-Cre Hmox1fl / fl versus Hmox1fl / fl macrophages. (B.) Bacterial counts in BMDM lysates ± E. coli administered for 2 hours followed by the addition of penicillin / streptomycin ± CO for 6 hours. Results represent average ± SD of 2 independent experiments in triplicate. *P. <0.001, CO versus air. (C.) CFU in supernatants of mouse BMDMs or primary human peripheral blood monocytes (hMo) + E. faecalis (106) for 1 hour prior to CO (white bars) or air (black bars) administration for an additional 4 hours. *P. < 0.007; **P. <0.01, CO versus air (C). (D.) Growth kinetics of E. faecalis in BMDM supernatants ± CO measured as absorbance at 600 nm. Blue indicates 102, red indicates 106, and violet indicates 1010 CFU / ml. CO, dotted lines; air, solid lines. Data represent mean ± SD of n = 3 / group / time point. *P. < 0.005; **P. <0.05, versus air at the same bacterial concentration and time point. (E.) Growth kinetics of E. faecalis in the medium in the presence or absence of CO. Results represent mean ± SD from 3 independent experiments. Note: similar effects of CO on growth were observed with E. coli.

CO increases bacterial-induced IL-1β secretion and bacterial clearance by macrophages through a NALP3 inflammasome-dependent mechanism. The ability of macrophages to generate IL-1β is significantly impaired in the absence of functional NALP3 and caspase-1 (4). Macrophages harvested from both Nalp3–/– other Casp1–/– mice showed a significant impairment in ability to process and kill bacteria, which was most evident in caspase-1 – deficient macrophages, suggesting at least in part that IL-1β was participating in the bactericidal function of macrophages (Figure 4, A and B) . Treatment with CO was unable to augment killing in macrophages deficient in NALP3 or caspase-1 as compared with CO-treated Nalp3+/+ other Casp1+/+ macrophages (Figure 4, A and B). These findings further confirmed CO as promoting bacterial clearance by macrophages in part through a mechanism involving NALP3, caspase-1, and IL-1β activity. IL-1β is known to autoamplify macrophage activation in an autocrine or paracrine fashion (24, 46).

Figure 4

NALP3 is targeted by CO to enhance IL-1β expression and mediate macrophage bacteria killing. (A.) CFU in WT (air), Nalp3–/–, and Casp1–/– BMDM supernatants infected with E. faecalis ± CO. *P. <0.01 versus air-treated controls. (B.) Quantitation and representative images (inset) of cells positive for lysozyme (red) and E. faecalis in BMDMs from WT (air), Casp1–/–, and Nalp3–/– mice. Nuclei are stained blue. BMDMs were treated with E. faecalis for 1 hour followed by 4 hours of CO (white bars) or air (black bars). **P. <0.001, CO versus air; #P. < 0.05, Nalp3–/– versus WT; &P. < 0.05, Casp1–/– versus WT. Original magnification, × 40. (C.) Bacterial counts in BMDM treated with supernatants ± E. faecalis for 1 hour and then treated ± anti-IL-1β or IL-1β recombinant protein. CO (white bars) or air (black bars) was applied as above. *P. <0.001, CO versus air; **P. <0.01, CO versus air. Results represent mean ± SD of 2 to 4 independent experiments in triplicate. (D.) Flow cytometry in BMDMs treated with E. faecalis for 1 hour or BMDMs with antibody against ILRa-APC. IgG-APC as negative control. Mean ± SD from 3 independent experiments. *P. < 0.05, E. faecalis versus control (C) or IgG. (E.) IL-1β levels in E. faecalis–Treated BMDM supernatants as above for 4 hours followed by 1 hour treatment with cathepsin B inhibitor, cytochalasin B, or apyrase. Cells were then exposed to CO or air for 4 hours. Results are presented as percentage of control to account for vehicle effects and represent mean ± SD of 3 independent experiments in duplicate. *P. <0.05, comparing CO versus air in both the vehicle-treated control group and the cathepsin B inhibitor-treated group.

We and others have demonstrated that HO-1 deficiency results in impaired bacterial clearance (32), but provided no mechanism other then greater susceptibility to infection. The observed effects of CO on IL-1β cleavage / secretion and clearance combined with the lack of effects in Nalp3–/– other Casp1–/– macrophages prompted us to the next test whether the enhanced IL-1β cleavage / secretion was participating in the effect exerted by CO on bacterial phagocytosis and / or clearance. The notion that CO promotes bacteria clearance via a mechanism that relies on IL-1β cleavage / activation was tested and reinforced by the observation that addition of recombinant IL-1β dose dependently increased E. faecalis clearance by macrophages similar to that observed with CO (Figure 4C). Further, the effect of CO on bacterial killing was abrogated using an anti-IL-1β neutralizing antibody (Figure 4C). It is therefore likely that IL-1β acts on macrophages in an autocrine manner, as we also observed an increase in IL-1β receptor expression in E. coli-Treated macrophages (Figure 4D). Given that CO exposure was initiated hours after bacteria administration, we elected not to focus on TLR receptors as potential mechanisms of action and targets for CO. Additionally, there was a significant reduction in bacteria-elicited IL-1β expression in BMDMs from Tlr2–/– other Tlr4–/– mice likely due to deficiencies in their ability to respond appropriately to bacteria wall components, resulting in less pro – IL-1β induction (47–49).

CO requires the presence of nucleotides to increase IL-1β in macrophages. Efforts to delineate the mechanism by which CO induces IL-1β cleavage / activation and bacterial clearance directed us initially to test the effect of CO on well-established activators of the NALP3 inflammasome. CO can induce mitochondrial ROS generation in macrophages, an effect that could enable NALP3 inflammasome activation (50-52). E. faecalis induced a modest but significant ROS generation in macrophages, an effect that was not further modulated by CO (Supplemental Figure 2, A and B). This suggested to us that the mechanism by which CO modulates the NALP3 multiplex activation does not involve modulation of ROS production in bacteria-treated macrophages (52). It is likely, however, that the elevated ROS do contribute to bacterial killing in the lysosomal compartment. Inhibition of cathepsin B, a protease known to activate NALP3 inflammasome, also did not modulate CO-induced IL-1β cleavage / secretion (Figure 4E).

Somewhat surprisingly, the effects of CO were dependent on the presence of ATP as well as on a functional cytoskeletal apparatus. We augmented ATP catalysis using apyrase and used cytochalasin B to block actin microfilament formation, and both of these agents suppressed the ability of CO to induce IL-1β secretion (Figure 4E). Remarkably, neither apyrase nor the cytochalasin B inhibitor completely returned IL-1β to basal levels in air-treated controls, suggesting other mechanisms in place, including elevated expression of pro-IL-1β. These data corroborate the effects of CO on macrophage lysosomal activity, in which CO exposure increased the number of bacteria contained in lysosomes (Figure 4B).

CO induces active release and production of ATP from bacteria. The requirement for ATP generation in the CO response directed us to first study whether the macrophages were the source for increased ATP. We observed that CO had no effect on intracellular ATP production by macrophages (0.82 ± 0.07 μM versus 0.96 ± 0.15 μM air versus CO, respectively). CO also had no effect on ATP hydrolysis by ectonucleotidase activity (air plus E. faecalis: 213.5 ± 45.41 nmols / min / mg protein versus CO plus E. faecalis: 226.61 ± 33.3 nmols / min / mg protein). This implied that ATP was actively produced and released and / or prevented from being metabolized by bacteria. ATP as a DAMP is known to influence NALP3 inflammatory activation. CO exposure of heat-inactivated bacteria showed no increase in ATP levels or IL-1β secretion (data not shown), which suggested to us active release of ATP from bacteria in response to CO. While CO had no effect on bacterial growth (Figure 3E), it did induce a dose-dependent and rapid increase in bacteria-derived ATP versus air-treated controls, as assessed by time-lapse imaging and colorimetric analyzes (Figure 5, A and B).The amount of ATP generated by the bacteria in response to CO was similar in magnitude to levels of ATP released from dying cancer cells, which contribute critically to NALP3 inflammasome activation via a mechanism involving the ATP P2 receptor (53). Important here is that the ATP generated by bacteria was present at concentrations ranging from 3 to 5 μM in CO versus 1 to 2 μM basally and remained elevated and continuously generated by the bacteria when CO was present. This increase in ATP in response to CO likely underestimates the local concentrations at the macrophage cell surface, where ATP is released by the bacteria and likely continues inside the phagosome. We next asked whether bacteria-derived ATP contributed to modulation of macrophage activation by CO. We measured bacterial ATP release and binding to macrophages in response to CO by loading E. coli with radiolabeled phosphate (32P) and measuring the amount of 32ATP bound to P2X7 receptors immunoprecipitated from macrophages. A significant increase in 32ATP binding to P2X7 was observed in CO versus air-treated macrophages (Figure 5C). There was no binding of ATP in P2x7–/– macrophages (Figure 5D), suggesting that this is indeed a receptor for ATP originating in the bacteria and then binding to macrophages. To test whether endogenous CO arising from HO-1 exerted the same effects on bacteria, HO-1 was maximally induced in macrophages in response to LPS, and the conditioned medium, when rapidly transferred to 32P-loaded bacteria cultures, showed a significant increase in 32ATP formation (Figure 5E). 32ATP was captured with a bacterial ATP-binding protein, which was subsequently immunoprecipitated and subjected to thin-layer chromatography; the radiolabel was densitometrically quantitated. Importantly, blockade of HO-1 with QC-15 in this scenario completely abrogated the 32ATP increase (Figure 5E). Finally, to demonstrate a functional impact of the elevated ATP, macrophages were treated with exogenous ATP, which recapitulated that observed with CO in enhancing bacterial clearance (Figure 5F). Collectively, these data support a mechanism by which pathogens are recognized and destroyed by the host that involves activation of macrophages by ATP (54, 55).

Figure 5

Effects of CO on bacteria-derived ATP generation and macrophage-killing response. (A.) ATP generation by E. faecalis (106) ± CO for 1 hour measured colorimetrically (inset) or fluorometrically. (B.) ATP fluorescence in supernatants from E. coli (104) treated for 6 hours ± CO. *P. < 0.05; **P. < 0.001. (C. other D.) Interaction between E. coli–Derived 32ATP and immunoprecipitated BMDM P2X7 receptor from WT and P2rx7–/– mice (D.) treated with supernatant from CO-exposed 32ATP-producing bacteria. Neg, IgG control; A, ampicillin control. *P. < 0.03; #P. <0.05, CO versus air. (E.) 32ATP-producing bacteria supernatant exposed to BMDMs where HO-1 was induced by LPS to increase endogenous CO by HO-1 ± QC-15 to block HO-1. *P. <0.01, LPS versus control (C); #P. <0.05, LPS versus LPS + QC. (F.) Bacterial counts in BMDM supernatants infected with E. faecalis for 1 hour, then ± ATP (50 μM) for an additional 6 hours. *P. < 0.05. (G) ATP in supernatants of WT E. coli MG1665 or ΔatpA E. coli (104) treated 30 minutes with CO or air expressed as fold change to account for proliferation rate differences. Data represent mean ± SD of 3 experiments in triplicate. *P. <0.02, CO + M1655 versus air + M1655. (H) Immunoblot with antibodies against IL-1β lysates of WT or mutant E. coli–Treated BMDMs ± CO as above. (I.) Bacterial counts expressed as fold change due to differences in proliferative rates. BMDMs were infected for 1 hour, then treated ± CO for an additional 6 hours. **P. <0.01, CO versus WT E. coli (102). *P. <0.05, CO + ΔcyoB versus air + ΔcyoB; &P. <0.04, CO + Δcyd versus air + Δcyd. (J) WT or mutant E. coli growth ± CO. Results represent mean ± SD from 3 independent experiments.

To further strengthen the concept that bacteria-derived ATP was the DAMP responsible for macrophage activation, we next utilized an E. coli strain lacking the ATP synthase subunit A (ΔatpA). This strain survives on alternative, weaker ATP-generating systems. This mutant allowed us to study the molecular target underlying the effect of CO on bacteria-derived ATP. Macrophages were infected with the ΔatpAE. coli or with the E. coli MG1655 control strain and treated with or without CO. As shown in Figure 5G, unlike what occurred in the MG1655 control strain, CO was unable to increase ATP levels over baseline in the ΔatpA mutant, demonstrating that CO promotes production of ATP via a mechanism that in part requires a fully functional ATP synthase. Moreover, macrophages infected with ΔatpA in the presence of CO generated very little to no cleaved IL-1β (Figure 5H). We used 2 additional E. coli mutants, ΔcyoB other ΔcydB. These mutants lack the indicated cytochrome oxidase in the bacteria, and it is these heme-containing oxidases that are known to be targets for CO (64). When macrophages were treated with these 2 strains in the presence of CO, we observed an effective increase in IL-1β in the ΔcyoB strain that correlated with augmented clearance, but no effect of CO on either IL-1β cleavage or bacteria clearance in the ΔcydB strain (Figure 5, H-J). Importantly, both pro and cleaved IL-1β were markedly reduced in ΔcydB. From these data, we conclude that CO targets and perhaps binds to heme in the oxidase of ΔcydB to compel ATP generation much like that observed in the ATP synthase mutant. Absence of either protein resulted in a loss in CO effects on IL-1β and bacteria killing. While each of the strains exhibited slightly different growth rates, these were not affected by exposure to CO (Figure 5J). Collectively, these data strongly support the notion that bacteria-generated ATP can bind and activate macrophage P2 receptors triggering NALP3 / IL-1β / inflammasome-dependent bacterial clearance (57).

CO increases bacteria-derived ATP generation in vivo. The effects of CO on bacterial ATP production were recapitulated in vivo in mice infected i.p. with WT E. coli. We took advantage of the fact that luciferase requires ATP to metabolize luciferin and used the luciferase signal intensity as an indirect measure of ATP. Mice exposed to CO showed a significantly enhanced luciferin signal versus air-treated controls, demonstrating the presence of ATP required for luciferase activity (Figure 6, A and B). The luciferin signal was localized to the site of bacteria inoculation in the peritoneal cavity and measured 1 hour after CO, when the amount of bacteria would still be similar between air and CO. These results suggested that the highly diffusible CO accessed the peritoneal cavity and interacted with the bacteria to compel the 2- to 3-fold increase in ATP levels in a manner similar to that generated by bacteria in response to CO in vitro.

Figure 6

CO increases ATP levels in bacteria-infected mice. (A. other B.). ATP measured by IVIS bioluminescence or fluorescence in mouse peritoneal lavages. Mice (n = 4 / group) were infected with E. coli (109) for 1 hour followed by CO for 1 hour. ATP levels were imaged and quantitated 30 minutes after injection of luciferase. *P. <0.05, CO versus air. The stronger signal intensity reflects increased ATP levels evidenced by greater luciferase activity.

CO promotes IL-1β secretion via K+ channel-dependent activation in macrophages. Given that CO is reported to modulate K.+ channel activity via ATP purinergic receptor signaling (58, 59), we hypothesized that CO modulated NALP3 inflammasome activation through modulation of an ATP-dependent K+ channel that likely involves one or more P2 receptors (7, 20). First, we measured intracellular K+ concentrations in macrophages in response to ATP and, as expected, observed a marked decrease in intracellular K+ levels. Macrophages treated with either bacteria or CO alone had little effect on intracellular K+ concentrations (Figure 7A). In contrast, macrophages treated with E. coli (WT or MG1655) plus CO showed a decrease in intracellular K+ equivalent to that observed with ATP alone (Figure 7A). Employing the selective K.+ channel inhibitor tetraethylammonium (TEA), we tested the relative contributions of ATP and K+ channel activity on NALP3 inflammasome activation in response to CO. Mice infected with E. faecalis were treated with or without CO and with or without the addition of TEA. PMs were harvested and assessed for caspase-1 cleavage and IL-1β expression. TEA suppressed the ability of CO to increase caspase-1 activation and IL-1β secretion compared with controls linking K+ channel to inflammatory activity. (Figure 7B). The inability of TEA to block IL-1β expression to levels below that observed with E. faecalis alone suggests other ATP-dependent, non-K+ channel-mediated mechanisms involved in the activation of IL-1β, such as P2Y receptors (57). Collectively, these findings speak to the selectivity of CO in modulating IL-1β via P2 receptor signaling, perhaps suggesting differences in ATP affinity for each purinergic receptor.

Figure 7

CO modulates potassium flux in vitro and in vivo in macrophages. (A.) Intracellular potassium (K+) levels measured fluorometrically in BMDMs infected with E. coli (104) ± CO. ATP (200 μM) was used as a positive control. Note that both ATP and CO + E. coli decreased intracellular K+ fluorescence, suggesting a K+ efflux. *P. <0.0001, ATP versus WT; #P. <0.001, M1655 + CO versus WT; &P. <0.001, M1655 or CO versus WT. White bars, air; black bars, CO. (B.) IL-1β ELISA of peritoneal lavage fluid from mice treated ± E. faecalis (106) ± K+ channel inhibition with TEA ± CO. CO was started 1 hour after infection and continued for 1 hour. *P. <0.01 CO versus control (C). All results represent mean ± SD of 2 to 4 independent experiments with n = 3-6 / group.

CO protects against sepsis-induced multiorgan failure in mice. To test our in vitro observations in a clinically relevant experimental model of bacterial infection in vivo, we inoculated mice i.p. with a lethal dose of E. faecalis prior to therapeutic administration of inhaled CO. Caspase-1 cleavage and IL-1β secretion were measured in peritoneal exudates. We observed a significant time-dependent induction of IL-1β cleavage / secretion in PMs in response to bacteria, an effect enhanced by inhaled CO (Figure 8, A – C). Of note, IL-1β cleavage / secretion was measured in total peritoneal exudates, which likely involves cell types other than macrophages. While the effects of CO on IL-1β expression and corresponding caspase-1 cleavage were evident in peritoneal leukocytes, this was not associated with a systemic increase in IL-1β, suggesting that CO acts at the primary site of infection targeting live bacteria, compelling their release of ATP, which in turn drives increased IL-1β secretion by activated macrophages. Moreover, CO also promoted more rapid resolution of inflammation, as suggested by the more rapid decrease in IL-1β expression, which occurred as early as 4 hours after infection compared with that of controls treated with air at the same time point. The more rapid resolution of the infection correlated with a greater than 50% increase in bacterial clearance over controls as well as improved survival and inhibition of multiorgan failure (Figure 8). CO protected against sepsis-induced gut epithelial cell sloughing in the colon and acute liver injury as assessed by reduced levels of serum transaminases (ALT) (Figure 8, D and E). Acute kidney injury was also abrogated. Elevated serum blood urea nitrogen (BUN) levels as a measure of kidney damage were lower in CO-treated mice compared with air-treated controls (97.3 ± 16.3 mg / dl in air versus 32.0 ± 19.4 in CO-treated mice, P. <0.02; normal BUN = 18-29 mg / dl). Real-time IVIS imaging using chemiluminescent E. coli (Xen 14) corroborated the ability of CO to augment bacterial clearance (Figure 8F). Leukocyte infiltration into the peritoneal cavity in response to bacterial administration showed an increase in neutrophils (PMN) at 4 hours after infection (10.4 ± 1.1 × 106 versus 1.3 ± -0.3 x 106 in untreated controls P. <0.02). PMN cell infiltration was modestly reduced in CO-treated mice (6.1 ± 1.9 × 106) over the same time interval, correlating with the decreased bacterial counts and modulation of the inflammatory response (Figure 8, A and F).

Figure 8

CO acts therapeutically to inhibit lethal sepsis via the inflammation. (A.) IL-1β in peritoneum from mice infected with E. faecalis (109 CFU) for 1 hour ± CO. Data represent mean ± SD of n = 6 / group in triplicate. *P. < 0.01; #P. <0.05, CO versus air. (B.) Immunoblot of cleaved caspase-1 and IL-1β in PM lysates from mice infected with E. faecalis for 1 hour prior to 1 hour treatment with air or CO. Representative blot from 3 independent experiments. (C.) Immunoblots of macrophages lysates from mice treated as in B.. (D.) ALT in infected mice treated as above. Results represent mean ± SD (n = 4-6 / group) from 3 experiments. *P. <0.001, CO versus air. (E.) H&E staining 8 hours after E. faecalis ± CO; 6 to 10 images per animal. Arrows indicate lesions. Original magnification, × 20. (F.) Bacterial counts in peritoneal lavagates. Inset: IVIS images. Results represent mean ± SD of 6 to 8 / group. *P. < 0.05; **P. < 0.01. (G) Survival in mice injected with lethal E. coli (1010 CFU, i.p.) or CLP ± 4 hours CO initiated 6 hours later. P. <0.02, air + CLP versus CO + CLP; P. <0.02, air + E. coli versus CO + E. coli. (H) Survival of WT and Nalp3–/– mice treated as in G with or without CO (started 1 or 6 hours after bacteria). CO was unable to protect Nalp3–/– mice (P. < 0.01). (I.) Bacteria counts in peritoneum of E. coli–Infected mice. Results represent mean ± SD of 4 to 6 / group repeated twice. *P. <0.03, CO versus air; #P. < 0.05, Casp1–/– versus Casp1+/+ in both air and CO-treated mice. (J) Survival ± Sn-PP-IX to block HO-1 then ± 4 hours CO beginning 1 hour after bacteria. P. <0.05, 10 / group. (K) Survival of indicated mice ± CO as in G. P. < 0.03. n = 10-30 / group.

Mice infected i.p. with E. faecalis and exposed to air succumbed to lethal septic shock (75% mortality) within 24 to 48 hours (Figure 8G). Inhaled CO, when initiated 1 hour after infection, resulted in 100% survival, while exposure to CO initiated as long as 6 hours after bacterial injection resulted in 60% survival (Figure 8G). The experiments were stopped at 4 days, which was considered long-term survival. CO also conferred protection in a cecal ligation and puncture (CLP) model of sepsis. CO administered 1 hour after CLP increased survival to 60% compared with 5% in air-treated controls (Figure 8G).

CO requires NALP3 and caspase-1 to impart protective effects against lethal infection and is able to rescue mice deficient in HO-1. Our in vitro evidence supporting a role for the NALP3 inflammasome on the CO effects led us to next inoculate Nalp3–/– other Casp1–/– mice with E. faecalis and measure clearance. Mice deficient in either protein were unresponsive to CO in the lethal sepsis model or bacterial killing assay, and Casp1–/– mice showed a significant deficit in the ability to kill bacteria with or without CO present (Figure 8, H and I). These data suggest that CO requires activation of caspase-1 and the resulting IL-1β release to promote bacterial clearance, confirming our in vitro observations (Figure 4).

Animals lacking HO-1 activity are exquisitely sensitive to bacterial infection (32, 37). Importantly, CO was able to rescue a significant percentage of inoculated LyzM-Cre Hmox1fl / fl mice (HO-1 absent in myeloid cells) and WT mice in which HO activity was inhibited with Sn-PP-IX (Figure 8, J and K).