Eosinophil extracellular DNA trap cell death mediates lytic release of free secretion-competent eosinophil granules in humans

Human eosinophils, leukocytes notably associated with allergic, anthelmintic parasite, and other immune responses,^1-3  contain an abundant singular population of crystalloid-bearing granules. As intracellular organelles, these granules are central to the functional responses of eosinophils in that eosinophil granules house preformed protein stores of (1) 4 major cationic proteins, including eosinophil cationic protein (ECP), major basic protein (MBP), and eosinophil peroxidase (EPO); (2) hydrolytic enzymes; and (3) over 4 dozen cytokines, chemokines, and growth factors.^4,5  Intact eosinophils may secrete their granule proteins by classic exocytosis (principally on the surfaces of large, multicellular helminths) or more commonly by piecemeal degranulation.⁴ 

Unlike granules from other granule-containing leukocytes, eosinophil granules have long been recognized to be present extracellularly in tissues and sputum associated with diverse eosinophil-associated diseases. Extracellular eosinophil granules have been detected by immunostaining for their cationic proteins and/or by ultrastructural studies that demonstrated intact granules still bound by their granule-delimiting membranes. Free extracellular eosinophil granules have been documented in diverse diseases, including atopic dermatitis and nasal allergy, and they have been correlated with the severity of urticaria.^4,6  Since the late 1800s, free eosinophil granules have been noted in the sputum of asthmatics, and clusters of free eosinophil granules (termed “Cfegs”) have been documented in human asthma and experimental guinea pig models of asthma.⁷ 

The release of intact, membrane-bound granules occurs via an enigmatic mode of eosinophil “degranulation” that arises from the cytolysis of eosinophils. Prior in vivo studies revealed that (1) lytic eosinophils, noted ultrastructurally by chromatolysis and loss of plasma membrane integrity, were frequently observed in human airway specimens rather than apoptotic cells,^8,9  (2) the numbers of free granules increased severalfold within an hour after allergen-induced airway provocation,⁷  and (3) released granules exhibited little evidence of loss of their granule contents.¹⁰  Collectively, the many in vivo findings of free extracellular eosinophil granules suggested that a process exists for human eosinophils to undergo nonapoptotic but cytolytic cell death that liberated intact extracellular granules.

Defining the means by which eosinophils may release extracellularly intact granules is more cogent with our recent findings, in which isolated eosinophil granules remain secretion competent. Cell-free eosinophil granules express ligand-binding cytokine, chemokine, and eicosanoid receptors on the surface of their delimiting membranes.^11,12  Cell-free human and mouse eosinophil granules with receptor-mediated activation of intragranule signaling pathways can directly secrete selected granule-derived proteins, including ECP, EPO, ribonucleases, and cytokines, for example, the interleukins IL-4 and IL-6.^12-14  Thus, the local tissue release of cell-free, secretion-competent eosinophil granules, secondary to eosinophil lysis, might constitute a means by which postmortem eosinophils could continue to provide immunoregulatory, pro-inflammatory, and other immunopathogenic stimuli. Despite these observations, the mechanisms of the eosinophil cytolytic release of their intact granules are not well known.

Recently, an active form of cell death, namely, extracellular DNA trap cell death (called ETosis¹⁵ ), has been recognized in neutrophils¹⁶  and mast cells.¹⁷  In these cells, ETosis develops with time, often over 1 or more hours, and is morphologically distinct from other classic cell death processes, including apoptosis and necrosis.¹⁸  In contrast to apoptosis, nuclear condensation and DNA fragmentation do not occur. Instead, nuclear chromatin decondenses in the cytoplasm.¹⁹  Finally, rupture of the plasma membrane releases nuclear DNA to form extracellular DNA nets, which for neutrophils bind free antimicrobial molecules such as granule proteins and histones.²⁰  Reduced NADP (NADPH)-oxidase-mediated production of reactive oxygen species (ROS) has an essential role in the activation of ETosis.¹⁵  To date, while it has been shown that IL-5– or interferon-γ−primed human eosinophils, in response to certain agonists, can very rapidly and noncytolytically release mitochondrial, and not nuclear, DNA to form extracellular DNA nets,²¹  eosinophil ETosis has not been recognized.

Prior studies reported that the immobilized immunoglobulins (IgG, IgA) and the calcium ionophore A23187 induced eosinophil cell death morphologically, similar to lytic degranulation.^22,23  In the present study, we evaluated whether eosinophil cytolysis is mediated by ETosis. We have now shown that activated human eosinophils undergo NADPH-oxidase-dependent ETosis and that during ETosis-mediated cytolysis, eosinophil granules are liberated both when contained within plasma membrane–bound clusters and as cell-free granules. Cell-free eosinophil granules retained their cationic proteins, and some were functionally competent to secrete in response to the chemokine ligand CCL11 (also known as eotaxtin-1).

Eosinophils were purified from 320 mL of blood from mildly atopic and healthy donors by negative selection, as described in Neves et al.¹⁴  In accordance with the Declaration of Helsinki, informed consent was obtained under Institutional Review Board–approved protocols. Briefly, after blood was collected into a 6% dextran saline solution (Pharmacosmos, Holbaek, Denmark) and red blood cells (RBCs) were sedimented, the upper leukocyte-containing region was collected and centrifuged over Ficoll-Paque (GE Healthcare, Pittsburgh, PA). Eosinophils were isolated from granulocyte pellets by incubation with a depletion antibody (Ab) cocktail (StemSep, Stemcell Technologies, Vancouver, BC, Canada) followed by passage over magnetized columns (Miltenyi Biotec, Auburn, CA). The purity of isolated eosinophils was >97% of nucleated cells with ≤3% contaminating RBCs, and eosinophil viability was >99%, as determined by microscopic analyses. Separately, RBCs (>99% purity) were purified from the dextran sedimented fraction, as described in Aoshiba et al.²⁴  Tissue biopsy samples were obtained during clinical evaluations from the left frontal sinus and the skin from a patient with allergic sinusitis and a patient with hypereosinophilic syndrome (negative for FIP1-like 1/platelet-derived growth factor-α mutation), respectively.

Purified eosinophils were stimulated with A23187 (Sigma-Aldrich, St. Louis, MO), PMA (phorbol 12-myristate 13-acetate, Sigma-Aldrich), IL-5 (R&D Systems, Minneapolis, MN), GM-CSF (R&D Systems), and/or platelet activating factor (PAF; Enzo Life Sciences, Farmingdale, NY) in poly-L-lysine–coated chambered cover glasses (Nalge Nunc, Rochester, NY) or 96-well, flat-bottomed tissue culture plates in 0.1% bovine serum albumin (BSA, wt/vol; Sigma-Aldrich) or as stated in fetal bovine serum (FBS; Gibco, Grand Island, NY) containing phenol red-free HEPES-buffered RPMI 1640 medium. Human IgG or IgA (Sigma-Aldrich), diluted with phosphate-buffered saline (Gibco) in indicated concentrations, was coated on culture plates for 3 hours at 37°C. For detection of extracellular chromatin structure, eosinophils (2 × 10⁵ cells per well) were seeded on chamber slides and stimulated with the indicated conditions followed by fixation with 3% paraformaldehyde (PFO, Electron MIcroscopy Sciences, Hatfield, PA) without permeabilization. Eosinophils were incubated with mouse Abs against histones (recognizing histones H1, H2A, H2B, and H4, 1:500, 45 minutes at room temperature (RT); United States Biological, Swampscott, MA) followed by goat antimouse IgG (1:100, 30 minutes at RT; Invitrogen, Carlsbad, CA) and propidium iodide (PI; Invitrogen). For quantification of cell death, CYTOX green (Invitrogen) was added to the medium, and CYTOX intensity was measured using the GloMax-Multi detection system (Promega, Madison, WI). The increase of CYTOX intensity was concurrent with cell death (confirmed using lactate dehydrogenase release). In some experiments, diphenyleneiodonium chloride (DPI, 20 μM; Sigma-Aldrich) was added to the culture medium.

Eosinophils, added to round coverslips coated with poly-L-lysine in culture plates, were stimulated with A23187 for 1 hour. Cells were fixed (1% PFO, 1.25% glutaraldehyde in 1 M sodium cacodylate buffer, pH 7.4) for 30 minutes at RT. Coverslip-adherent cells were postfixed with 1% osmium tetroxide in water for 1 hour and dehydrated in an ascending ethanol series from 50% (vol/vol) to absolute ethanol (10 minutes per step). Cells were then critical point dried in carbon dioxide. Coverslips were mounted on aluminum holders, sputtered with 5 nm gold, and analyzed in a scanning electron microscope (Quanta 200 FEG; FEI, Eindhoven, The Netherlands).

Eosinophils, seeded in 8-well Lab-Tek II chamber slides (Nalge Nunc) or immunoglobulin-coated Millicell EZ slides (Millipore), were stimulated for indicated time points and then fixed with 3% PFO for 10 minutes. For MBP staining, cells were permeabilized with 0.1% saponin for 10 minutes, incubated with primary mouse antihuman MBP Ab (5 μg/mL, clone AHE-2, 60 minutes at 4°C; BD Pharmingen), and followed by Alexa Fluor 488 conjugated Ab (goat antimouse IgG, 1:300, 30 minutes at RT; Invitrogen). Control Abs and PI were used for each experiment.

Blood eosinophils were added to chamber slides, stimulated with 2 μM of A23187 for 1 hour, and immediately fixed in a mixture of freshly prepared aldehydes (1% PFO, 1.25% glutaraldehyde) in 1 M sodium cacodylate buffer, pH 7.4, for 30 minutes at RT. To obtain optimal morphology, cells directly on the slide surface were postfixed in 1% osmium tetroxide and processed as described in Melo et al.^25,26  Biopsy samples were fixed for 4 hours at RT using the same fixative and processed for electron microscope as before.^25,26  Sections were mounted on uncoated 200-mesh copper grids (Ted Pella, Redding, CA) before staining with lead citrate, and then viewed with a transmission electron microscope (CM10; Philips, Eindhoven, The Netherlands) at 60 kV.

Eosinophils in chambered cover glasses were stimulated with immobilized IgG (1 mg/mL) for 40 minutes or 2 μM A23187 for 20 minutes in a carbon dioxide incubator. Chambered cover glasses were sealed with 37°C mineral oil before microscopic imaging, and morphological changes and granule release were evaluated using an inverted microscope (Eclipse TE300; Nikon, Tokyo, Japan) equipped with a digital camera in conjunction with iVision 4.0 image analysis software (BioVision, Exton, PA). The stage was preheated at 37°C, and phase-contrast images were recorded for 30 minutes (every 3 seconds) with a 100× objective. Images were sequenced into movies (every 0.03 seconds per image).

Through the use of a modified fluorescent bead-count–based assay for microparticles,²⁷  subcellular structures in eosinophil culture medium were counted and studied using FACScan (BD Biosciences, San Jose, CA). In some experiments, subcellular structures from eosinophils loaded with acridine orange (AO, 3 μM, 15 minutes; Sigma-Aldrich) and structures stained with annexin V (Invitrogen) were examined. Data were analyzed by Flowjo (version 9.03; Tree Star, Ashland, OR). Details are available in supplemental “Materials” (supplemental Figure 1).

After eosinophils were stimulated with 2 μM A23187 for 1 hour, culture medium was gently collected. Contaminating cells and DNA were removed by centrifugation at 200 × g for 10 minutes, followed by incubation for 10 minutes with DNase I (20 U/mL; New England Biolabs, Ipswich, MA). Thereafter, supernatants were centrifuged at 2500 × g for 10 minutes, and pelleted subcellular structures were stained as previously described in Neves et al.¹¹  The gating strategy for flow cytometry was the same as for supplemental Figure 1. CellMask, LysoTracker, and CellTracker (Invitrogen) were used for microscopic analyses as described in supplemental “Materials” (supplemental Figure 7).

ECP levels in supernatants and extraction buffer¹⁴  (Biosource, Camarillo, CA)-disrupted subcellular structures were quantified by ECP enzyme-linked immunosorbent assay kits (Medical and Biological Labs, Nagoya, Japan). Samples were frozen at −20°C until measurement.

Eosinophils (5 to 7 × 10⁶ cells) were stimulated with 2 μM A23187 for 1 hour in 0.1% BSA containing RPMI medium, followed by the addition of 20 U/mL DNase I for 30 minutes. Culture medium was collected, residual intact eosinophils were removed by centrifugation at 200 × g for 10 minutes, and granule-rich subcellular structures were pelleted by 2500 × g centrifugation for 10 minutes. Pelleted structures were washed and resuspended in 0.1% BSA RPMI medium. The absence of live, intact eosinophil contamination was confirmed by trypan blue staining. Released granule-associated structures were incubated with the indicated concentrations of CCL11 (R&D Systems) or 2 μM A23187 for 1 hour at 37°C.

After AO loading, eosinophils were stimulated for 1 hour with 2 μM A23187, and ETosis-released granule-rich subcellular structures were isolated by centrifugation. Cell-free AO-stained granules were stimulated as described previously in Neves et al with some modifications.¹⁴  After recovered structures were washed and resuspended in 0.1% BSA RPMI medium, they were spread on poly-L-lysine–coated slides and coverslipped. Stimuli (800 ng/mL CCL11 or 4 μM A23187) or control medium were introduced in 2-μl drops at the edge of the coverslips, enabling stimuli to diffuse under the coverslip. The responses of AO-stained granules were monitored by fluorescence microscopy using the fluorescein isothiocyanate band fluorescence filter. To prevent AO quenching during time-lapse microscopy, 3 neutral-density filters were used during data acquisition. Fluorescence images were recorded for 100 seconds (100 frames total) and acquired using a BX62 microscope (Olympus).

Results were expressed as mean ± SD. The differences of groups were evaluated using 2-tailed Student t test, with the level of statistical significance taken as P < .05.

Disruptions of nuclear and plasma membranes (PMs) in cytolytic eosinophils have been observed in varied allergic and parasitic diseases.^28,29  To ascertain the ultrastructural characteristics of eosinophil cytolysis in vivo, tissue-infiltrated eosinophils were evaluated using TEM (Figure 1). Eosinophils within biopsies from the sinus and the skin (from patients with allergic sinusitis and hypereosinophilic syndrome, respectively) showed similar characteristics of cytolytic degranulation with disrupted PMs and nuclear envelope membranes. In contrast to the nuclear chromatin condensation observed in apoptotic eosinophils,³⁰  both distinct nuclear heterochromatin decondensation and release of nuclear contents into the extracellular space (Figure 1, arrowheads) were observed. Moreover, eosinophil granules released in vivo were intact and exhibited little evidence of loss of their granule contents (Figure 1).

In neutrophils, ETosis releases nuclear contents as filamentous chromatin structures, and neither apoptosis nor necrosis induces DNA nets.¹⁹  To test the hypothesis that immobilized IgG or IgA²³  and A23187,²²  previously shown to elicit eosinophil cytolysis, elicit ETosis, eosinophils were stained with antibodies against histones and fixed-cell permeable PI. Using nonpermeabilized cells, we were able to distinguish intact cells and extracellular histones (supplemental Figure 2). Immobilized IgG and IgA, like A23187, elicited the release of histone-positive DNA from disrupted cells (Figure 2A). In addition, PAF, in combination with IL-5 or GM-CSF, induced EETosis (Figure 2A, supplemental Figure 2A). PAF itself was a less potent and somewhat donor-dependent EETosis inducer, and IL-5 or GM-CSF alone did not induce EETosis (data not shown). PMA, an ETosis inducer for neutrophils,¹⁶  also induced EETosis.

To ascertain differences among apoptosis, necrosis, and EETosis, we identified conditions that led to ∼20% dead cells (supplemental Figure 3A) mediated by necrosis, apoptosis, or EETosis by inducing cell death by brief heating, anti-Fas Abs,³¹  or A23187, respectively. Morphological changes were evaluated using phase-contrast microscopy and concomitant staining for annexin V and PI (supplemental Figure 3B). Anti-Fas–activated cells showed typical apoptotic morphology (cytoplasmic and nuclear condensation, budding) with bright annexin V–positive membranes and abundant released apoptotic bodies. Heated cells showed bleb formation, characteristic of necrotic cells.³²  EETosis exhibited completely different appearances, releasing filamentous DNA and cell-free granules, with only weak staining for annexin V relative to Fas-activated cells. In heated or EETosis cells, surface annexin V single-positive cells were rarely observed, indicating that the phosphatidylserine (PS) redistribution typical during apoptosis was lacking. Of note, extracellular filamentous DNA was never observed in heated or Fas-activated cells. As assessed by the release of lactate dehydrogenase and histone staining, A23187-stimulated eosinophils exhibited cytolytic EETosis in an A23187 concentration-dependent manner (supplemental Figure 3C).

To study the process of nuclear shape changes during EETosis, we loaded live eosinophils with Hoechst 33342 dye, which stained nuclear DNA; cells were fixed at different time points. As shown in supplemental Figure 4, supplemental Video 1, and supplemental Video 2, initially the bilobular form of the eosinophil nucleus started to lose its shape, and thereafter the nuclear envelope disintegrated and DNA filled the cytoplasm. Finally, the PM ruptured, allowing the extrusion of the DNA net (confirmed by DNase treatment; data not shown). This process of nuclear shape change and dissolution was entirely consistent with neutrophil ETosis.¹⁹  A23187-induced EETosis developed more rapidly when compared with immobilized IgG and PMA (supplemental Figure 4). High-resolution SEM showed that DNA nets were associated with lytic cells (Figure 2B, left, arrows). DNA nets formed weblike structures that consisted mostly of fibers with diameters of 25 to 35 nm in conjunction with larger fibers (Figure 2B, right). These results indicate that activated eosinophil cell death is EETosis, which is distinct from necrosis and apoptosis.

Neutrophil ETosis is dependent on ROS and inhibited by the NADPH oxidase inhibitor DPI.¹⁸  To monitor the kinetics of cell death and the effect of DPI, eosinophils were stimulated with immobilized IgG and A23187, and cell death was measured every 15 minutes (Figure 3). CYTOX intensity progressively increased 45 minutes after immobilized IgG stimulation in a concentration-dependent manner, and DPI completely abolished the effect of IgG- and A23187-stimulated eosinophils to undergo DPI-inhibitable cell death. Similar inhibition was observed with DPI when stimulated with other EETosis-inducing stimuli we tested (data not shown). In line with the time course of nuclear shape change (supplemental Figure 4), A23187-induced EETosis developed ∼30 minutes earlier when compared with IgG-induced EETosis (Figure 3B). As reported for ETosis in mast cells and neutrophils,^17,19  both fetal bovine serum and BSA had concentration-dependent protective effects on A23187-induced EETosis (supplemental Figure 5A), likely based on their capacities to scavenge ROS. To further evaluate whether the generation of extracellular ROS was critical for EETosis, eosinophils were stimulated in the presence of the cell-impermeable ROS scavenging enzyme, catalase. Catalase inhibited A23187-induced eosinophil cytolysis and death (supplemental Figure 5B). RBCs, also known to scavenge ROS,²⁴  likewise inhibited A23187-induced eosinophil cytolysis (supplemental Figure 5C).

Neutrophils, diverse granule-derived proteins, many with antimicrobial activities, have been recognized to be bound as free proteins to extracellular DNA nets.^16,33  Likewise, for the elicited release of mitochondrial DNA from eosinophils, eosinophil granule proteins have been found bound to extracellular DNA nets.²¹  To assess whether specific eosinophil granule proteins were released free or were otherwise bound to eosinophil DNA nets, EETosis cells were immunostained for MBP (Figure 4A). Notably, many regions of extracellular DNA nets lacked linear immunostaining for free MBP, and immunostaining was localized to smaller punctate foci or cell-associated domains, suggestive that eosinophil granule proteins might still be within granules. Immunostaining for EPO showed similar results (data not shown). Indeed, TEM analyses of EETosis revealed the presence of clusters of intact eosinophil granules (Figure 4B, arrowheads), and some were associated with DNA nets. In residual intact eosinophils, granules exhibited various degrees of electron density, likely indicating secretion of some of their contents. Ultrastructurally, cell-free eosinophil granules released from cytolytic cells retained their typical morphologies with intact granule-delimiting membranes and intact crystalloid cores and granule matrices (Figure 4B, insert).

To study the EETosis processes that liberate intact eosinophil granules extracellularly, we monitored eosinophils with time-lapse, phase-contrast microscopy (Figure 5A, supplemental Video 1, and supplemental Video2). Eosinophils undergoing ETosis showed common morphological behaviors, starting with the arrest of their chemokinetic migration and with enhanced, rapid intracytoplasmic movements of their granules. As shown in supplemental Figure 4, the eosinophil’s bilobed nucleus progressively condensed into a single round shape. PM protrusions developed, and at times 1 to 3 granules were released together with enveloping PMs by budding off from the PM protrusions. In addition, larger collections of granules were released from eosinophils into PM-bound clusters. Thus, the budding release of small numbers of granules and the extrusion of larger clusters of granules resulted in the production of a variety of sizes of PM-bound, granule-packed structures. The release of granule structures typically precedes by two to five minutes nuclear membrane disintegration and nuclear chromatolysis. With the subsequent lysis of the eosinophils and the disruption of their PMs, the remaining intracellular granules were liberated.

In other cells, activated cells can induce vesicle-shedding without affecting cell viability.³⁴  To verify that the release of intact granules from eosinophils was associated with EETosis, we counted the absolute number of released subcellular structures (≤4 μm with high granularity) in the culture medium by using a flow cytometer (supplemental Figure 1). Notably, the release of extracellular free granule structures with A23187 was completely inhibited by DPI (Figure 5B). Enhanced release of granule structures was associated with increasing EETosis, as evidenced by concentration dependency (supplemental Figure 6A) and time course (supplemental Figure 6B). Moreover, extracellular structures were not liberated at 4°C (supplemental Figure 6C), indicating an active cellular process. To confirm the stability of released cell-free granules, we loaded eosinophil granules with AO, a lysosomal granule dye, followed by stimulation with A23187. After 1 hour, abundant cell-free, AO-retained granules were found in culture medium (supplemental Figure 7A). Consistent with Figure 5A, granules were found as single or grouped structures as well as PM-enveloped clusters containing different numbers of granules. Released subcellular structures were further subjected to flow cytometry; and in concert with fluorescent microscopy, most subcellular granule structures retained the granule dye (Figure 5C). Importantly, in contrast to apoptotic bodies derived from anti-Fas Ab–treated cells (supplemental Figure 3B), EETosis-induced subcellular structures lacked PS expression on their surfaces (Figure 5D).

To obtain granule-rich subcellular structures, any shed vesicles, membranes, and residual contaminating cells were removed by differential centrifugation. In concert with their identities as eosinophil granules, as shown in Figure 5E, isolated structures were positive for their contents of granule MBP. Moreover, there was bimodal staining for PM-derived major histocompatibility complex (MHC) class I protein, indicating that some, but not all, granule-containing structures lacked evidence of an associated PM. To target single granules (∼0.5 to1 μm in diameter), we also gated for small structures (≤1 μm), and similar results were obtained (Figure 5E, lower panels). Indeed, both PM-bound granules (supplemental Figure 7B,C) and PM-free granules (data not shown) were observed microscopically. To assess the intactness of PM-bound granules, live eosinophils were loaded with LysoTracker for labeling granules and CellTracker for cytoplasm and were stimulated with A23187. Intact granules were found both with and free of CellTracker-positive cytoplasm (supplemental Figure 7D). Similarly staining structures were observed following immobilized IgG- or PMA-induced EETosis (data not shown).

Collectively, we conclude that (1) EETosis releases eosinophil granules, (2) the extracellular granules retain specific granule proteins even after being released in culture medium, (3) some released granules may singularly or in clusters be surrounded by PS-nonexposed intact PM, and (4) some cytolytically released individual granules were devoid of PM.

We previously demonstrated that human eosinophil granules, as isolated free of PMs principally by subcellular fractionation, express functional CCR3 chemokine receptors and secrete ECP in response to the ligand CCL11 by using their intragranular signaling and membranotubular network-based secretion systems.^12,14  We evaluated whether EETosis-released granules would possess similar functional secretory properties. Following the stimulation of eosinophils with 2 μM A23187 for 90 minutes, ∼4% of the total ECP was recovered within the released granule structures. Isolated granules were stimulated with A23187 or CCL11 for 60 minutes, and secreted ECP was measured. Interestingly, the released granules did not respond to A23187. In contrast, the granules secreted granule-derived ECP in response to CCL11 in a concentration-dependent manner (Figure 6A).

The activation of AO-loaded eosinophil granules releases, into nonacidic media, monomeric AO that exhibits a spectral shift to green, yielding transient green flashes from secretion-responding granules.¹⁴  To obtain direct evidence showing that EETosis-released granules were responsive to CCL11, after AO-loaded cells were treated with 2 μM A23187, isolated granules were stimulated with CCL11. Some EETosis-released free granules exhibited rapid activation in response to CCL11, as evidenced by fluorescent flashes attributable to the release of AO (Figure 6B, upper panels). In contrast, some granules, including both free and those likely clustered within PMs, failed to exhibit secretory AO flashes in response to CCL11 (Figure 6B, lower panels). Stimulation of isolated granules with medium or A23187 did not induce AO secretion (data not shown). These results indicate that some granules extruded during A23187-mediated EETosis retained their capacity for CCL11-elicited secretory responses.

Eosinophils as granulocytes contain distinct cytoplasmic granules, notable by their ultrastructurally unique crystalline cores and by their content of proteins, including specific cationic proteins as well as preformed cytokines. In diverse human diseases in which eosinophils may participate, there have been recognitions over many years that cell-free eosinophil granules, released from eosinophils, are found in tissues or secretions.⁴  These findings include ultrastructural studies recognizing intact free eosinophil granules;^8,10  immunofluorescent demonstrations of eosinophil cationic proteins in extracellular granule foci; and, especially in asthma, clusters of extracellular eosinophil granules.⁷  Disruptions of PMs and nuclear membranes along with chromatolysis have been observed in eosinophils associated with allergic and anthelmintic responses.^28,29  Our studies aimed to examine the mechanisms and consequences by which human eosinophils could cytolytically and extracellularly release their intact granules.

Ultrastructural examination of eosinophils in the sinus tissue of a patient with allergic sinusitis and the skin of a patient with a hypereosinophilic syndrome, respectively, (Figure 1) revealed the liberation of free extracellular granules as well as nuclear changes (disruption of nuclear membranes and decondensation of nuclear chromatin) that were distinct from apoptosis or necrosis. Apoptotic eosinophils exhibit condensed compaction of chromatin against the nuclear envelope and shrinkage and budding of the whole cell to form membrane-bound, PS-exposed apoptotic bodies. Neither anti-Fas–elicited eosinophil apoptosis nor heat-elicited eosinophil necrosis yielded extracellular release of nuclear DNA or cell-free eosinophil granules (supplemental Figure 3). To investigate mechanisms of eosinophil cytolysis, we used immobilized immunoglobulins (IgG, IgA), PAF with either IL-5 or GM-CSF, and nonphysiologic stimuli, calcium ionophore A23187 and PMA. Each of these stimuli elicited eosinophil lytic death (EETosis) that was distinct from apoptosis or necrosis. EETosis occurred progressively over 30 to 120 minutes and was dependent on the generation of ROS, being inhibitable by DPI. These stimuli are known to induce/prime ROS production by eosinophils.^35-37 

Prior studies demonstrated that cytokine-primed human eosinophils, in response to certain agonists such as lipopolysaccharide, C5a, and CCL11, can rapidly (within seconds) and noncytolytically release not nuclear, but mitochondrial DNA to form extracellular DNA nets.²¹  Our findings identify another means by which eosinophils can form DNA nets, more akin temporally and mechanistically to ETosis in neutrophils¹⁶  and mast cells.¹⁷  As in neutrophils,¹⁸  eosinophil ETosis involves a more delayed, cytolytic release of nuclear DNA to form extracellular DNA nets. The temporal sequence of changes within eosinophils undergoing cytolysis is presented in Figure 7. Following stimuli-elicited initiation of eosinophil cytolysis, the typically bilobed nuclei of eosinophils lost their lobulation to coalesce into single nuclear structures. During this time, some eosinophil granules were released extracellularly as PM-enveloped structures. Thereafter, nuclei disintegrated intracellularly to form intracellular DNA nets. Subsequently, the eosinophils’ PMs rupture, releasing both nuclear DNA-derived (histone-positive) nets and free eosinophil granules. Eosinophil ETosis-mediated cytolysis leads to the extracellular liberation of intact eosinophil granules.

In neutrophils undergoing ETosis, granule-contained proteins can associate intracellularly with nuclear DNA even before PMs rupture, and granule proteins are bound to extracellular DNA nets.^19,38  Intact neutrophil granules are not recognized as bound to extracellular DNA nets. In contrast, in eosinophils undergoing ETosis, granules are released as intact structures. Our findings indicated that eosinophil granule proteins were not predominantly free and bound to DNA nets, but rather they remained principally within granules that themselves could bind to DNA nets. Moreover, we showed that some released cell-free eosinophil granules exhibited secretion competency in response the eosinophil-active chemokine CCL11 and secreted ECP and granule AO (Figure 6). Other granules, including clusters of likely PM-enveloped granules, did not exhibit individual CCL11-elicited granule secretory responses, at least in short-term in vitro assays. Whether the clusters of PM-enveloped granules might exhibit secretory competence in vivo requires further study.

In conclusion, our results reveal a process by which eosinophils undergo cytolysis to liberate intact cell-free granules, as is well recognized in tissues and secretions in association with diverse eosinophil-associated diseases. The process of ETosis in eosinophils is akin to that in neutrophils¹⁸  but notably differs in that intact eosinophil granules are released extracellularly and still contain the bulk of their protein contents. At least some EETosis-released granules remain secretion competent. Thus, EETosis results in the generation of histone-bearing nuclear DNA extracellular nets and cell-free granules, both of which may exert biological activities for eosinophils postmortem.

The authors are grateful to Kristen Young and Jason Xenakis for their outstanding technical assistance and Dr. Praveen Akuthota and Professor Hélio Chiarini-Garcia for helpful discussions.

This work was supported in part by Uehara Memorial Foundation (S.U.); Conselho Nacional de Desenvolvimento Científico e Tecnológicoand Fundação de Amparo à Pesquisa do Estado de Minas Gerais (R.C.N.M.); National institute of Health: National Heart, Lung and Blood Institute R01-HL096795 (I.G.); R01-HL095699 (L.A.S.); and grants from the National Institutes of Health: National Institute of Allergy and Infectious Diseases (R37-AI020241 and R01-AI051645, P.F.W.).

National Institutes of Health

Contribution: S.U. designed and performed experiments and wrote the manuscript. R.C.N.M. and A.M.D. performed electron microscopic experiments and wrote the manuscript. I.G. provided technical assistance and edited the manuscript. L.A.S. contributed scientific advice and discussions. P.F.W. supervised the research and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Peter F. Weller, MD, 330 Brookline Ave, CLS-943, Boston, MA 02215; e-mail: pweller@bidmc.harvard.edu.