Cutting Edge: Eosinophils Undergo Caspase-1–Mediated Pyroptosis in Response to Necrotic Liver Cells Free

Apoptosis is a homeostatic process for eliminating cells in the absence of inflammation. In contrast, insults such as tissue injury (1, 2) or infections (3, 4) can trigger necrosis, a type of cell death characterized by the disorderly release of intracellular contents (5). Innate immune cells are equipped to respond to damage-associated molecular patterns (DAMPs) released during necrosis. Immune cell activation by DAMPs results in inflammatory cascade characterized by cytokine release and further recruitment of inflammatory cells (6). This response can be beneficial in facilitating clearance of infectious organisms or tumors, but can also lead to damage of healthy tissue if not well regulated. This paradigm can be generalized across a wide spectrum of diseases, making understanding the inflammatory response to necrotic cells an imperative area of investigation.

We investigated this paradigm by injecting necrotic liver cells i.p. and monitoring immune cell recruitment and cytokine release in the peritoneal lavage. It was not our intention to establish a model of any particular liver disease per se, but rather to use the model to explore potential mechanisms applicable to a number of diseases. Previous studies have demonstrated necrotic EL4 cells recruit neutrophils in an NF-κB–dependent manner (3). We found that necrotic liver cells not only induced neutrophil and macrophage recruitment, but also induced an eosinophil response.

Eosinophils usually associate with mucosal tracts during homeostatic conditions. During inflammation, they can be recruited from the bone marrow into the blood and ultimately to the site of tissue injury. These cells are frequently associated with Th2 responses in a number of pathological conditions, being major constituents of the inflammatory infiltrate in the lungs of asthmatics and in patients with chronic obstructive pulmonary disease (7–10). There is also a growing literature demonstrating eosinophilia in a number of liver diseases (1, 2, 11–15).

In this report we demonstrate that necrotic liver cells recruit eosinophils, which directly respond to the necrotic debris by undergoing a unique form of cell death termed pyroptosis (16). Pyroptosis is a programmed form of cell death that is mediated by caspase-1 and occurs with the simultaneous release of the cytokines IL-1β and IL-18 (17). The term pyroptosis means fiery cell death, and has been previously described to occur in macrophages (18) and dendritic cells (DCs) (19) in response to microbial infection. Our results demonstrate that eosinophils can also undergo pyroptosis, and that the response is not limited to infectious insults, but can occur during sterile inflammatory conditions.

C57BL/6, Dbl GATA, BALB/c, and Caspase-1/11^−/− mice were obtained from the Jackson Laboratory. The MedImmune animal care and use committee approved all mouse studies. All Schistosoma mansoni experiments were performed at the National Institutes of Health using mice bred and housed under specific pathogen-free conditions in an Association for Assessment and Accreditation of Laboratory Animal Care approved facility under Animal Study Proposal LPD-16E. The National Institute of Allergy and Infectious Diseases animal care and use committee approved all experimental procedures.

Ultra-pure LPS (Escherichia coli O55:B5) was from Alexis, ATP was from Teknova, and PMA was from Sigma. Caspase-1 inhibitor II YVAD-FMK was from Calbiochem and Alfa Aesar.

Mice (8–10 wk old) were injected i.p. with 1 μg LPS, the equivalent of 50 million necrotic liver cells, or PBS in a total volume of 500 μl, and lavaged at the indicated time points. Then 1 ml of lavage fluid was collected for cytokine analysis, with an additional 9 ml for cell analysis.

Livers from donor mice were collected and gall bladders removed prior to mechanical disruption through 70 μm filters. Analysis of the cell preparations at this stage indicated that ∼50% of the cells were already dead. Cells were washed in PBS and 50 million cells were resuspended in 500 μl of deionized water. Cells were homogenized for 1 min then resuspended in 1× PBS at a final concentration of 50 million cell equivalents per 100 μl.

Bone marrow–derived eosinophils (BMDEs) were prepared according to the method of Dyer et al. (20). On day 16, BMDEs were plated at 1.0 × 10⁶ cells per ml in 24-well tissue culture plates and treated with LPS at 200 ng/ml, PMA at 200 ng/ml, necrotic-liver at 5–30 × 10⁶ cells per ml. Supernatants collected for IL-1β (eBioscience), IL-18 (eBioscience), and Eosinophil Peroxidase (CUSABIO or LifeSpan BioSciences) were measured by ELISA. Cells were stained with Siglec F PE (BD Pharmingen), CCR3 APC (BD Pharmingen), CD107a Pacific Blue (eBioscience), aqua live/dead Amcyan (Life Technologies), FAM-FLICA Caspase-1 (ImmunoChemistry Technologies), and FLICA 660 Caspase 3/7 far-red fluorescence (ImmunoChemistry Technologies). Cells stained for eosinophil markers were monitored on an LSRII.

Caspase-1 was immunoblotted with rat anti-mouse caspase-1 (Clone 5B10; eBioscience), which recognizes the 42 kDa proform and the p20 active form.

C57BL/6 infected livers were kindly provided by Thomas Wynn. S. mansoni eggs for i.v. injections were extracted from the livers of infected mice (Biomedical Research Institute, Rockville, MD). Female 6–8 wk-old mice on a C57BL/6 background were infected with 25–30 cercariae of S. mansoni eggs through percutaneous injections of the tail (21). Infected fibrotic livers were harvested 8 and 12 wk later and eosinophils were purified within hours of harvest at MedImmune. Single-cell suspensions were prepared by placing livers in HBSS with CaCl₂, MgSO₄, collagenase I (40 μg/ml), and DNase I (2 μg/ml) into C tubes from Miltenyi Biotec. Samples were dissociated on the GentleMACS Octo Dissociator. The sample was passed through a 70-μm strainer, resuspended in 1.051 g/ml Optiprep, and layered on 1.078 g/ml Optiprep. After centrifugation, the cells were washed in FACS buffer, blocked using 2.4G2, and stained as above.

Previous studies have shown immune infiltration into the peritoneum following injection of necrotic cells (22). These studies have primarily used necrotic material from cell lines as a stimulus and focused on neutrophil recruitment. We reasoned that the response might vary in different organ systems or in response to different cell types. Therefore, we began by characterizing the cellular immune response in vivo by injecting necrotic liver cells into the peritonea of recipient mice and analyzing the infiltrating cell types in the peritoneal lavage. As another means for inducing inflammation, we injected LPS. As previously reported from other necrotic stimuli, neutrophils were recruited early, peaking around the 4 h time point, and tapering off afterward. There was macrophage infiltration at the 48 h time point, as well as a steady increase of DCs, a small number of CD8⁺ T cells, and an influx of CD4⁺ T cells. B cell numbers remained unaffected (Supplemental Fig. 1A). There was a greater number of infiltrating macrophages and neutrophils in response to necrotic liver than to LPS (Supplemental Fig. 1A).

Based on the previously described presence of eosinophils in liver diseases, we assessed and found large numbers of eosinophils 24 h after peritoneal injection of necrotic liver (Fig. 1A), showing increased cell counts over time as well as them being a sizable percentage of the overall infiltrate. Eosinophil infiltration was not induced in response to LPS, indicating that the recruitment was not a general response to an inflammatory reaction, but displays specificity to the type of stimulus.

We next evaluated cytokines induced by necrotic liver with a particular interest in how they may contrast to those induced by LPS. As expected, LPS induced detectable levels of IL-1β in the peritoneal lavage. Necrotic liver also induced this cytokine, albeit at a slightly lower level (Fig. 1B). The necrotic liver lysate itself contained only trace amounts of IL-1β at levels insufficient to account for the amount detected in the lavage (data not shown). Necrotic liver also induced large amounts of IL-6, although LPS was much more potent (Fig. 1B). Attempts to evaluate the response to live liver cells revealed that the dissociation step itself induced a large degree of cell death. To circumvent this, and to ask whether this response was specific to necrotic liver cells or was more generalizable to any necrotic cell, we injected both live and necrotic spleen cells. Neither of these induced detectable levels of TNF-α or IL-1β, indicating the response has specificity for DAMPs released by the liver (Supplemental Fig. 1B). LPS induced very high levels of a number of cytokines, including IL-6 (Fig. 1B), IP10, KC, and MCP-1 (Supplemental Fig. 1C), all of which were induced to a much smaller degree by necrotic liver. This evaluation highlighted the extent to which LPS and necrotic liver induce qualitatively different responses.

Necrotic liver induced large amounts of both eotaxin and TARC. The levels of eotaxin induced by necrotic liver trended higher than those induced by LPS (Fig. 1C, not significant). However, necrotic liver induced significantly more TARC than did LPS (Fig. 1C). Taken together, the relatively higher levels of both eotaxin and TARC induced by necrotic liver as compared with LPS may in part explain the recruitment of eosinophils in response to necrotic liver but not LPS.

To determine the relative contribution of eosinophils to the overall response observed in vivo to necrotic cells, we used GATA1^−/− double GATA mice (23, 24), which are deficient in eosinophils. Although the previous experiments were performed in C57BL/6 mice, the double GATA mice are on the BALB/c background. We observed markedly different kinetics of eosinophils between the two strains with BALB/c mice showing eosinophil responses much earlier than in B6 mice (Supplemental Fig. 1D). Double GATA and BALB/c mice had identical kinetics of neutrophils and macrophages (Supplemental Fig. 1D), although macrophage kinetics differed markedly between these two strains and B6 (compare with Supplemental Fig. 1A). Double GATA mice had significantly less EPO and IL-1β in their peritonea than did wild type mice, indicating a direct response of the eosinophils (Fig. 1D).

The markedly different kinetics of eosinophils between the BALB/c and B6 mice make it difficult to directly extrapolate the results obtained in one strain to the other. However, the significantly lower levels of IL-1β in the double GATA mice as compared with the wild type BALB/c controls is highly suggestive that eosinophils respond to necrotic liver in vivo by releasing IL-1β (Fig. 1D).

To address whether eosinophils directly respond to necrotic liver cells we derived BMDEs (20). In vitro stimulation of BMDEs with necrotic liver led to a rapid release of IL-1β, IL-18, IL-1α, and IP10. Notably, LPS did not induce a similar response, nor was it required to prime for the response (Fig. 2A). BMDEs were, however, highly responsive to LPS, releasing high levels of IL-6 and KC. Intriguingly, whereas the response of BMDEs to necrotic liver was rapid, occurring within minutes, the response to LPS was much more delayed.

Historically, degranulation of murine eosinophils has been difficult to induce in vitro, often requiring powerful stimuli such as phorbol esters (25). We evaluated eosinophil degranulation in vitro by stimulating BMDEs with necrotic liver cells and measuring EPO in the supernatant (Fig. 2B), and found that necrotic liver cells induced EPO secretion to a higher level than PMA, and to an amount equal to the complete lysis of the cells. Although there were background levels of EPO in the supernatant of unstimulated cells, there was an almost undetectable level in LPS-treated cells, again highlighting the different responses induced by necrotic cells and LPS. Staining for surface expression of CD107a (26), a molecule that translocates from inside the cell to the cell surface upon degranulation, further confirmed this finding (Fig. 2B).

Necrotic liver cells, but not LPS, induced a high degree of eosinophil cell death by 2 h (Fig. 2C). This cell death did not appear to occur through apoptosis as the cells were almost negative for caspase-3 and 8 (Fig. 2D). However, virtually all of the cells were positive when stained with the pan-caspase activation indicator ZVAD (Fig. 2D).

Western blot analysis identified the 50 kDa procaspase-1 in untreated BDMEs, but only background levels of the active 20 kDa form of the enzyme. Necrotic liver rapidly activated caspase-1 in BMDEs as evidenced by the appearance of the 20 kDa band (Supplemental Fig. 1E). As a positive control, we treated BMDEs with ATP (25), a known activator of caspase-1, and observed the expected 20 kDa band. In this assay, it was conceivable the active caspase-1 was coming from the necrotic liver cells. However, there was no active caspase-1 in the necrotic liver, only controls, indicating the enzyme was being activated in the eosinophils. Furthermore, there was no caspase-1 activation in LPS-treated BMDEs (Supplemental Fig. 1E).

We further characterized the signaling pathways induced by necrotic liver, focusing on how they differed from those induced by LPS. As both ATP and necrotic liver induced caspase-1 activation to a similar degree in BMDEs, we included ATP in this analysis as an additional comparator. Necrotic liver induced a rapid and transient upregulation of pERK, whereas ATP induced a rapid and sustained pERK response (Fig. 3A). LPS did not upregulate pERK (Fig. 3A). LPS induced upregulation of pJNK and pNFκB (Fig. 3B, 3D). ATP alone upregulated pp38 (Fig. 3C). Although ATP induced sustained expression of pJNK, necrotic liver induced neither pJNK nor pNFκB (Fig. 3B).

The divergent cellular responses induced by LPS and necrotic liver can therefore be understood in terms of the distinct signaling pathways induced by these stimuli. The confluence of observations led us to hypothesize that eosinophils respond to necrotic liver by undergoing pyroptosis. Pyroptosis is a recently characterized form of nonapoptotic, non-necrotic, programmed and inflammatory cell death, which has been described in macrophages (18) and more recently DCs (27, 28) in response to microbial infection. It is defined as caspase-1 dependent, caspase-3/7 independent cell death, which leads to IL-1β and IL-18 release. Thus far we have demonstrated IL-1β and IL-18 release and cell death with concurrent caspase-1 activation in addition to degranulation.

To confirm pyroptosis, we needed to demonstrate that these events were caspase-1 dependent and explored this using a small molecule inhibitor approach. Caspase-1 inhibitors were able to block IL-1β and IL-18 secretion (Fig. 3E), degranulation (Fig. 3F), and cell death (Fig. 3G), showing that all of these processes were indeed caspase-1 dependent. Furthermore, we confirmed that inhibition blocked the activation of caspase-1 itself and dampened the pan-caspase activation within the cell (Fig. 3H). To confirm these results, we generated BMDEs from caspase-1/11 knockout mice, which produced neither IL-1β nor IL-18 and showed greatly reduced levels of eosinophil peroxidase release (Fig. 3I). Caspase-1/11 knockout mice were also greatly deficient in degranulation and underwent greatly reduced cell death in response to necrotic liver (Fig. 3J). These results confirm that pyroptosis occurs in eosinophils, and extends this cellular process to include degranulation in capable cells.

To determine if we could identify eosinophils undergoing pyroptosis in a disease setting, we analyzed eosinophils from S. mansoni–infected mouse livers. S. mansoni infection results in extensive liver fibrosis with fulminant eosinophilia (21, 29). Eosinophils isolated from the fibrotic livers expressed caspase-1, IL-1β, and CD107a, and were in large part positive for aqua live/dead (Fig. 4A), confirming that they were undergoing pyroptosis in vivo. Eosinophils cultured from fibrotic liver spontaneously released IL-1β and IL-18 in the absence of stimulation (Fig. 4B).

Although eosinophils are most commonly associated with Th2-type responses, we observed no evidence of a Th2 response in the peritoneal inflammation model, but rather a highly proinflammatory environment, with the recruitment of neutrophils, macrophages, and the release of IL-1β. We further demonstrated that eosinophils contribute to this response, and can in fact respond directly to necrotic debris. To our knowledge, this is the first report directly demonstrating eosinophil recruitment in response to cell death. Importantly, although LPS elicited neutrophil infiltration and cytokine secretion in this model, it did not induce eosinophil migration, demonstrating a specificity to the stimulus.

We included LPS in a number of our experiments to highlight the uniqueness and specificity of the response to and signaling pathways induced by necrotic cells. Although there are many ligands capable of stimulating the innate immune system through a wide assortment of various pattern recognition receptors, LPS is one of the mostly highly studied and well characterized. The contrast of the widely divergent signaling pathways induced between necrotic cells and LPS highlights the spectrum of responses eosinophils can undergo. The fact that eosinophils did not require LPS priming demonstrates they have the cellular machinery to directly respond to necrotic cells, and the signaling pathways induced by that machinery have specificity for the inciting stimulus.

We did not observe eosinophil recruitment in response to necrotic cells from other organs such as the spleen or pancreas. There have been numerous liver diseases with associated eosinophilia reported in the literature. As any of these disease states is likely to involve cell death, it supports the hypothesis that dying liver cells release signals that recruit eosinophils.

A large number of DAMPs involved in liver diseases have been identified (30), any of which are likely candidates. It is unlikely the results we obtained were in response to a single molecular entity. When cells die, they release a wide variety of molecules, and the overall response is an integration of the responses to all of these signals. For this reason, we have focused our attention on determining the nature of the overall response, and uncovering the underlying mechanisms. The fact that we did not observe pyroptosis in response to necrotic spleen suggests the responsible DAMPs are specific for the liver and indicates a strategy for determining their identity. One intriguing possibility is that ASC prionoid-like specks are released from the dying liver cells and induce the response (31).

We have established the unique finding that eosinophils react to necrotic cells by undergoing pyroptosis. Meaning fiery cell death, pyroptosis is a programmed, inflammatory form of cell death that has been reported to occur in macrophages and DCs (16), and more recently CD4⁺ cells (32). Our results extend these findings to include eosinophils as well. Mechanistically, pyroptosis is a caspase-1–mediated cell death concurrent with IL-1β and IL-18 release. We not only observed cell death and IL-1β and IL-18 release, but also caspase-1 dependency, establishing that caspase-1 has a central and nonredundant role in this process. In this regard, our use of LPS as a comparator is particularly relevant. LPS is a potent inducer of pyroptosis in macrophages (33), but not in eosinophils. Importantly, previous reports have only shown pyroptosis to occur in response to infection. To our knowledge, this is the first report to document pyroptosis in sterile inflammation.

Murine eosinophils are considered to be somewhat resistant to degranulation in vitro (25). That necrotic cells are able to induce eosinophil degranulation on a relatively fast timescale suggests that pyroptosis may be a major pathway leading to degranulation in eosinophils. Also, the observation of degranulation being caspase-1 dependent indicates the definition of pyroptosis should include degranulation in capable cells.

The peritoneal inflammation model we used should be interpreted as a mechanistic model and not a recapitulation of any particular disease. The fact that numerous liver diseases are associated with eosinophilia begged the question of whether our results were applicable to a bona fide liver disease. To address this, we used the S. mansoni–induced liver fibrosis model, which results in fulminant eosinophilia. Direct ex-vivo analysis of these infiltrating eosinophils identified a large proportion undergoing pyroptosis, demonstrating that eosinophil pyroptosis does indeed occur in a natural liver disease.

How eosinophil pyroptosis influences the overall pathogenesis of this disease would be an intriguing avenue for further study. The highly inflammatory nature of pyroptosis suggests that modulating eosinophil responses in these diseases may lead to clinical benefit by reducing tissue damage and possibly fibrosis.

We thank Miguel A. Sanjuan for comments on this manuscript, Bhargavi Rajan and Radhika Rayanki for assistance with flow cytometry, and Dan Rowe for assistance with Western blot analysis.

D.P.M., J.C., and T.S.D. are current employees of MedImmune. T.M. is a former employee of MedImmune. The other authors have no financial conflicts of interest.