Granulocytes undergoing apoptosis are recognized and removed by phagocytes before their lysis. The release of their formidable arsenal of proteases and other toxic intracellular contents into tissues can create significant damage, prolonging the inflammatory response. Binding and/or uptake of apoptotic cells by macrophages inhibits release of proinflammatory cytokines by mechanisms that involve anti-inflammatory mediators, including TGF-β. To model the direct effects of necrotic cells on macrophage cytokine production, we added lysed or apoptotic neutrophils and lymphocytes to mouse and human macrophages in the absence of serum to avoid complement activation. The results confirmed the ability of lysed neutrophils, but not lymphocytes, to significantly stimulate production of macrophage-inflammatory protein 2 or IL-8, TNF-α, and IL-10. Concomitantly, induction of TGF-β1 by lysed neutrophils was significantly lower than that observed for apoptotic cells. The addition of selected serine protease inhibitors and anti-human elastase Ab markedly reduced the proinflammatory effects, the lysed neutrophils then behaving as an anti-inflammatory stimulus similar to intact apoptotic cells. Separation of lysed neutrophils into membrane and soluble fractions showed that the neutrophil membranes behaved like apoptotic cells. Thus, the cytokine response seen when macrophages were exposed to lysed neutrophils was largely due to liberated proteases. Therefore, we suggest that anti-inflammatory signals can be given by PtdSer-containing cell membranes, whether from early apoptotic, late apoptotic, or lysed cells, but can be overcome by proteases liberated during lysis. Therefore, the outcome of an inflammatory reaction and the potential immunogenicity of Ags within the damaged cell will be determined by which signals predominate.
Apoptotic cell recognition and removal by phagocytes is critical for the restoration and/or maintenance of normal tissue structure and function. Macrophages engulf apoptotic cells before they lyse, thus preventing release into the tissue of potentially toxic and immunogenic intracellular substances. In addition, the binding and/or uptake of apoptotic cells not only fails to induce macrophage secretion of inflammatory mediators (1, 2), but actually inhibits their proinflammatory cytokine production following stimulation (3, 4). Human monocyte-derived macrophages (HMDM)3 and J774 mouse macrophages were found to produce TGF-β1 following exposure to apoptotic cells (3, 4), and this could be mimicked by PtdSer-containing liposomes and a mAb against a newly described PtdSer receptor (5). By incubating macrophages so treated with anti-TGF-β Abs, the inflammatory response to LPS was restored, suggesting that TGF-β1 induced by apoptotic cells was a major cause of the anti-inflammatory effect observed. Anti-IL-10 Abs were not effective, and, in fact, IL-10 production was down-regulated during exposure to apoptotic cells in the absence of serum (3), although Voll et al. reported that IL-10 appeared to play a role in the regulation of cytokine production by human monocytes following interaction with apoptotic peripheral lymphocytes (6). In fact, Freire-de-Lima and coworkers have shown that uptake of apoptotic cells by macrophages fuels the growth of trypanosomes in Chagas’ disease, in a TGF-β- and PGE2-dependent manner, thereby promoting disease progression (7). PGE2 has been reported to be released following the binding and uptake of apoptotic cells by HMDM (3). By contrast, Uchimura, Kurosaka, and coworkers have suggested that uptake of apoptotic cells is proinflammatory (8, 9). As phagocytic targets, they used apoptotic CTLL2 cells cultured in the absence of IL-2 for 12 or 28 h. Both populations contained high levels of necrotic cells, as indicated by trypan blue and propidium iodide positivity, which might explain the proinflammatory effects they observed.
It is commonly believed that necrotic cells are proinflammatory. Certainly most pathological lesions characterized by necrosis are also characterized by inflammation. Stern and colleagues showed that following association with necrotic (postapoptotic) eosinophils, unstimulated HMDM produced equivalent amounts of thromboxane B2 and GM-CSF to those that had ingested opsonized eosinophils or opsonized zymosan; these results contrasted sharply with the lack of induction by apoptotic eosinophils (2). Therefore, it was important to determine whether necrotic cells had similar effects on stimulated macrophage production of cytokines, and whether the type of apoptotic cell affected the macrophage cytokine response. Therefore, we compared the direct effects of binding and/or uptake of apoptotic vs lysed neutrophils and lymphocytes on cytokine production by unstimulated or zymosan-stimulated macrophages. The experiments were performed in the absence of serum to rule out the potential contribution of complement fixation following exposure to intracellular organelles (10, 11).
Materials and Methods
HMDM were cultured as described previously (3, 12). In brief, mononuclear cells were isolated from the blood of normal donors, and plated at 4 million/well in 24-well tissue culture plates. Lymphocytes were removed following incubation on tissue culture plates for 1 h in DMEM (Irvine Scientific, San Diego, CA). This method yielded 1.46 ± 0.5 macrophages per well. After washing, the macrophages were cultured for 7 days in X-Vivo (BioWhittaker, Walkersville, MD) containing 10% human serum pooled from five donors; the medium was changed once after the first 3 days of culture. Bone marrow cells were obtained from C3H/HeJ mice and cultured as described (13) in DMEM containing 10% heat-killed FCS and 10% L cell-conditioned medium as a source of M-CSF. Bone marrow-derived macrophages (BMDM) were used 5–7 days after isolation and culture.
Induction of apoptosis
Human neutrophils and the human T cell line Jurkat were used as apoptotic targets. Neutrophils were induced to undergo apoptosis by exposure to UV irradiation for 5 min followed by culture for 3 h, as described previously (3). Jurkat T cells were irradiated for 10 min and cultured for 3 h. Apoptosis was assessed by morphological examination of cytocentrifuged cells and by flow cytometry, using FITC-conjugated annexin V. For neutrophils, the average percent apoptosis was 74.3 ± 5.4 (SEM); for Jurkats, 75.3 ± 4.2 (SEM); trypan blue positivity was <4% for either population; propidium iodide positivity was <6%. Results of annexin V positivity were not significantly different from apoptosis as assessed by morphology (data not shown). For late apoptotic cells, human neutrophils were irradiated as above, but cultured for 24 h; this yielded populations of cells that were 95% annexin V positive, 45% trypan blue positive, and 65% propidium iodide positive.
Human neutrophils and Jurkat T cells were rapidly frozen as cell pellets on dry ice, then stored at −70°C. They were thawed at room temperature without washing, and suspended in X-Vivo without serum before use. The average percent lysis, assessed by trypan blue positivity, for neutrophils was 92.6 ± 2.4% (SEM) and for Jurkat T cells 95.6 ± 1.4% (SEM). Staining with FITC-conjugated annexin V and propidium iodide revealed that 100% of each population were positive for both. One cycle of freeze/thaw was used to give intact but trypan blue-positive bodies. For preparation of membrane and cytosol fractions, the cells were subjected to five freeze-thaw cycles, homogenized in X-Vivo medium, and centrifuged at 100,000 × g. The membrane pellets or the membrane-free supernatant were added to macrophages as cell equivalents to compare them to the intact but leaky lysed cells. Membrane and cytosolic fractions were also made from apoptotic and viable granulocytes by homogenizing them in X-Vivo, and by ultracentrifugation.
Macrophage cytokine production
Macrophages were cultured in 24-well plates and each condition was run in duplicate. Five million apoptotic or necrotic (lysed) cells were added per well of macrophages for 1.5 h, then washed out. Fresh X-Vivo medium without serum was added, and supernatants were collected 24 h later. Replicate wells were used to assess uptake of the phagocytic targets. For neutrophils, the monolayers were stained for myeloperoxidase as previously described and shown; only those macrophages that had engulfed neutrophils were scored as positive for staining (3). Uptake of apoptotic neutrophils was associated with intracellular staining of discrete intact bodies, whereas uptake of necrotic (lysed) cells was associated with paler staining intracellular bodies and with diffuse staining of the macrophages. Uptake of lymphocytes was assessed by staining the macrophage monolayers with a modified Wright’s Giemsa stain. For some experiments, the lysed neutrophils were incubated with the macrophages in the presence or absence of 100 μM PMSF (Sigma, St. Louis, MO), 500 μM 4-(2-aminoethyl) benzenesulfonylfluoride (AEBSF) (Calbiochem, La Jolla, CA), 5 μg/ml aprotinin, 0.25 μg/ml leupeptin, 1 μg/ml E-64, 40 μg/ml bestatin, 0.7 μg/ml pepstatin, and 15 μg/ml calpain I inhibitor (Boehringer Mannheim, Indianapolis, IN), anti-human neutrophil elastase, anti-human cathepsin G, and isotype controls for 24 h before collection of supernatants for evaluation of cytokine concentrations. Preliminary dose responses were performed to insure a lack of toxicity to the macrophage monolayers. Viability at the concentrations used was verified by trypan blue concentration (≤4% positive cells) and by preservation of the stimulatory response to zymosan (assessed by measurement of macrophage-inflammatory protein 2 (MIP2), TNF-α, IL-10, and TGF-β). Anti-human elastase and anti-human cathepsin G Abs were purified sheep IgG purchased from Cortex Biochem (San Leandro, CA), and were used at 50 μg/ml. Zymosan (Sigma) was used as a stimulus for cytokine production; HMDM were stimulated with 25 μg/ml and BMDM with 75 μg/ml. In some experiments, 0.01 U/ml purified human neutrophil elastase and 5 U/ml purified human cathepsin G (both obtained from Calbiochem) were added to determine their direct effects on macrophage cytokine production. These concentrations were determined by lack of effects on macrophage viability and peak of cytokine production. Supernatants from duplicate wells were pooled, centrifuged to remove cellular debris, and stored at −70°C until analyzed. As a control, apoptotic or lysed cells were cultured for the same time periods in X-Vivo medium without macrophages to determine whether they produced cytokines. Neither apoptotic neutrophils nor apoptotic Jurkat T cells produced significant levels of any of the cytokines tested.
Analysis of cytokines
Cytokines assessed included TGF-β1, TNF-α, IL-10, and either MIP2 for mouse or IL-8 for human cells. Matched Ab pairs for the ELISAs were purchased from R&D Systems (Minneapolis, MN). For TGF-β1, supernatants were activated with HCl before analysis. Cytokines levels were detected following incubation with the biotinylated secondary Abs by incubation with avidin-conjugated HRP, then tetramethylbenzidine and H2O2 as substrate. The plates were read in a Bio-Tek EL309 ELISA reader (Biotek Instruments, Winooska, VT), and results were analyzed using the log/log curve fit option from Delta Soft 3 (BioMetallics, Princeton, NJ).
Measurement of elastase
Freshly isolated neutrophils were suspended at 10 million per ml in X-Vivo, then immediately centrifuged, separated into pellet and supernatant, and frozen at −70°C. Early and late apoptotic neutrophils were prepared as described above except that they were cultured at 10 million per ml in X-Vivo medium, then centrifuged. The samples were frozen as pellets, and the supernatant was derived from culture. Lysed neutrophils were prepared by suspending neutrophils at 10 million per ml in X-Vivo and freezing at −70°C. On the day of the analysis, the pellets and supernatants were thawed. The necrotic cell preparations were thawed, then centrifuged to separate the material into pellet and supernatant. All cell pellets were solubilized in lysis buffer containing 50 mM Tris (pH 7.4), 1% Triton X-100, 0.25% deoxycholate, 150 mM NaCl, and 1 mM EGTA. For elastase analysis, 20 μl of solubilized pellet or supernatant was added in triplicate to 96-well ELISA plates, followed by 55 μl elastase reaction buffer (0.1 M HEPES, 0.5 M NaCl, 10% DMSO pH 7.5). Then, 150 μl of 0.2 M Elastase substrate I (methoxysuccinyl-ala-ala-pro-val-p-nitroanilide; Calbiochem) in elastase reaction buffer was added, and the samples were incubated at 37°C for 1 h. Elastase levels were determined by measuring absorbance at 410 nm, using serial dilutions of porcine elastase (Calbiochem) as standards. Specificity was determined by incubating the samples and standards in the presence or absence of Elastase Inhibitor III (methoxysuccinyl-ala-ala-pro-val-chloromethylketone (Calbiochem).
For Figs. 1–3 and 6–7, data were evaluated by ANOVA and the Tukey Kramer method. For Figs. 4 and 5, ANOVA and Dunnett’s method were used, designating macrophages treated with lysed neutrophils as the control for Fig. 4, and macrophages without target cells added as the control for Fig. 5. Significance was set at p < 0.05.
We first studied the effects of the binding and/or uptake of apoptotic vs lysed human neutrophils on HMDM. The percent macrophages positive for uptake of apoptotic neutrophils was 49 ± 8.5% and for uptake of lysed neutrophils was 57.5 ± 10.5%. There appeared to be no measurable cytokine production by the apoptotic or lysed cells when incubated without macrophages (data not shown); thus the cytokine levels shown here appear to reflect macrophage production. As shown in Fig. 1, exposure to apoptotic human neutrophils in the absence of serum had the expected effects: increased production of TGF-β1, with no stimulation of IL-10, TNF-α, or a chemokine (IL-8), and suppression of IL-8, TNF-α, and IL-10 when macrophages were costimulated with zymosan. In contrast, exposure to lysed neutrophils induced significantly lower levels of TGF-β1 than apoptotic cells, and stimulated production of IL-10, TNF-α, and IL-8 (p < 0.05). Although lysed neutrophils stimulated proinflammatory cytokine production, the levels were lower than those induced by zymosan. When the macrophages were stimulated with zymosan to produce inflammatory cytokines, both apoptotic and lysed neutrophils were able to significantly suppress TNF-α and IL-10 production, although apoptotic cells were significantly more effective (p < 0.05). Similar effects were seen when mouse BMDM were used (Fig. 2). Percent uptake of apoptotic cells by BMDM was 45 ± 7.8% and of lysed cells 52.9 ± 11.2%. If the serine protease inhibitor PMSF was added, the proinflammatory response of the unstimulated macrophage to lysed neutrophils was significantly inhibited (Fig. 2,A), as was the production of IL-10 (p < 0.05). When PMSF was added to the lysed neutrophils coincubated with zymosan-stimulated macrophages, the lysed neutrophils behaved like apoptotic cells (Fig. 2,B). Treatment with PMSF did not alter the effects of apoptotic cells or zymosan on macrophage cytokine production. Lysed Jurkat T cells had the same effects on macrophage cytokine production as apoptotic cells (Fig. 2). Both populations of the lymphocytes inhibited stimulated production of MIP2, TNF-α, and IL-10, while stimulating TGF-β production. These data suggested that serine proteases liberated from the granulocyte were critical to the induction of a proinflammatory response by lysed neutrophils.
Lysed neutrophils stimulated low levels of chemokines and TNF-α, but also exhibited some inhibitory effects on zymosan induction of these cytokines. The effects of the lysed neutrophils were significantly different from those of the apoptotic cells (p < 0.05). This information, with the observation that PMSF treatment inhibited the proinflammatory effects of the lysed cells, suggested the hypothesis that the proinflammatory effects of serine proteases could be opposed by the anti-inflammatory effects of the PtdSer-containing lysed cell membranes. To determine whether this was true, lysed neutrophils were homogenized and centrifuged to obtain a crude membrane fraction and a cytosolic fraction; each of these were added to mouse macrophages in the presence or absence of zymosan to determine their effects on cytokine production. We also used membranes and cytosol from lysed Jurkat cells. As shown in Fig. 3, the cytosolic fraction from neutrophils contained the activity that stimulated MIP2, TNF-α, and IL-10. The membranes were anti-inflammatory as they induced the secretion of TGF-β, and inhibited zymosan-induced MIP2, TNF-α, and IL-10. We also used membrane and cytosolic fractions from viable and apoptotic neutrophils; as for neutrophils lysed by freeze-thaw cycles, the cytosolic fraction stimulated MIP2, TNF-α, and IL-10, whereas the membranes stimulated TGF-β (data not shown).
In the next set of experiments, we used a variety of protease inhibitors, as well as Abs against human neutrophil elastase and human cathepsin G, to determine their effects on lysed neutrophil-induced cytokine production. In Fig. 4, we show that the serine protease inhibitors PMSF and AEBSF were strong inhibitors of the proinflammatory response. The serine protease inhibitors aprotinin and leupeptin were weak inhibitors. Inhibitors of cysteine proteases (E-64), metalloproteases (bestatin), aspartic proteases (pepstatin), and calpain had no effect. Anti-human elastase Ab (50 μg/ml) was a strong inhibitor, but anti-human cathepsin G (50 μg/ml) was only a weak inhibitor; increasing the concentration of the latter did not improve the inhibition, and the isotype control (sheep IgG) had no effect. Neither of these Abs at the concentration used stimulated cytokine production from control HMDM (data not shown). These results support the interpretation that neutrophil elastase plays a major role in the stimulation of macrophage cytokine production.
Next, the production of cytokines from mouse BMDM exposed to early apoptotic neutrophils (cultured for 3 h after UV irradiation, trypan blue positivity ≤4%), late apoptotic neutrophils (cultured for 24 h after UV irradiation, trypan blue positivity 45%), and lysed neutrophils was assessed. As shown in Fig. 5, late apoptotic cells are similar to early apoptotic cells in that they induce TGF-β secretion and do not significantly induce MIP2, TNF-α, and IL-10.
It was important to determine whether neutrophil elastase could induce the same pattern of cytokine production as lysed neutrophils. Therefore, HMDM and mouse BMDM were exposed to purified human neutrophil elastase and human cathepsin G. As shown in Fig. 6, neutrophil elastase at 0.01 U/ml was a potent stimulator of mouse macrophage TNF-α, MIP2, and IL-10 production; cathepsin G virtually no activity, requiring 5 U/ml to see a small amount of IL-10 production only. Although not shown, the effects on HMDMs were identical.
Last, elastase concentrations in the pellet and medium from early and late apoptotic neutrophils were compared with those from freshly isolated and lysed neutrophils. Fig. 7 shows that early and late apoptotic neutrophils released virtually no elastase into their medium. Lysed neutrophils, as expected, released a large amount of elastase, and the proportion remaining in the cell pellet was significantly lower than that in the pellets from the apoptotic cells (p < 0.05). Lysed Jurkat T cells, as expected, were negative for elastase (data not shown).
The purpose of the experiments presented here was to determine whether lysed cells, used as a model for necrotic cells, could directly stimulate macrophages to release proinflammatory mediators in the absence of serum. We found that only lysed neutrophils could do so, and that the activity could be attributed to released neutrophil elastase. The binding and/or uptake of apoptotic neutrophils and lymphocytes by either mouse or human primary macrophages suppressed zymosan-induced production of proinflammatory cytokines as expected, exemplified by decreased levels of TNF-α, IL-10, and either MIP-2 or IL-8, as we have shown previously (3, 4). This down-regulation results in large part from secretion of TGF-β1, at least part of which is bioactive (3, 4), and the effects of apoptotic cells can be mimicked by treatment of macrophages with PtdSer-containing liposomes or a mAb against a PtdSer receptor (5). Thus, uptake of apoptotic cells before their lysis not only prevents the release of potentially toxic or immunogenic intracellular contents but also induces an anti-inflammatory phenotype in the macrophage. We also determined that late apoptotic neutrophils, even though becoming permeable to propidium iodide and trypan blue, behave more like early apoptotic cells (Fig. 5) because they induced production of TGF-β but did not significantly stimulate TNF-α, IL-10, or the chemokine MIP2. Furthermore, late apoptotic cells released very little elastase into their culture medium (Fig. 7). These results are in keeping with the earlier observations of Ren and Savill, who observed that apoptotic neutrophils beginning to become permeable to trypan blue were not proinflammatory (14). The response to lysed cells in the unstimulated macrophages was expected (2) in that they induced TNF-α and chemokine production; however, the levels were low relative to those seen with a strong proinflammatory stimulus such as zymosan or LPS (the latter not shown). Lysed cells also consistently induced IL-10 production, in contrast to apoptotic cells.
When macrophages were stimulated with zymosan, they produced robust levels of MIP-2 or IL-8, TNF-α, and IL-10. Interestingly, lysed neutrophils were found not to have additive effects in the absence of serum; rather, they partially (but significantly, p < 0.05) inhibited zymosan-stimulated production of TNF-α and IL-10, although having no effect on the chemokines MIP-2 or IL-8, both of which are CXC chemokines chemotactic for neutrophils. Lysed neutrophils also induced significantly less TGF-β1 than did early or late apoptotic cells. Thus, based on these results alone, the presence of lysed (fully necrotic) cells in vivo would likely cause continuing influx of neutrophils, particularly following macrophage activation. The net effect would be prolongation of an inflammatory response.
As expected, lysed neutrophils appeared to be more potent inducers of macrophage cytokine production in the absence of serum than were lysed Jurkat T cells, which behaved, for the most part, like apoptotic cells. Given that granulocytes produce high levels of several different types of proteases, it seemed reasonable to suggest that the differing effects of lysed neutrophils compared with lysed lymphocytes were related to protease release. The serine protease inhibitors PMSF and AEBSF significantly decreased the proinflammatory effects of lysed neutrophils on macrophage cytokine production, while leading to an up-regulation of TGF-β1, permitting them to have anti-inflammatory effects similar to those induced by apoptotic cells. The strong inhibitory effect of the anti-elastase Ab supports the notion that neutrophil elastase is a major contributor to the induction of proinflammatory cytokines and IL-10. This interpretation is supported by the observations that if lysed neutrophils were homogenized and separated into membrane and soluble fractions, the effects of the membranes were indistinguishable from apoptotic cells, which express phosphatidylserine on their outer leaflets, and that the proinflammatory signal resided in the soluble fraction. Furthermore, purified neutrophil elastase directly stimulated IL-10, TNF-α, and chemokine production, and only lysed neutrophils released elastase into the medium.
How neutrophil elastase directly promotes macrophage proinflammatory cytokine production is not yet known. It may bind directly to protease-activated receptors. For example, Ishihara and coworkers recently demonstrated specific binding of neutrophil elastase to macrophages; binding was accompanied by enhanced production of chemokines, which was inhibited by PMSF (15). In vivo, the proinflammatory actions of the proteases likely result not only from their actions on macrophages but on other cell types and on extracellular matrix. Alternatively, or in addition, the proteases may cleave a critical membrane signal or signals that mediate the down-regulation of proinflammatory cytokines associated with the uptake of apoptotic cells. In support of this notion is our observation that the PtdSer receptor can be cleaved off the cell surface by trypsin (5); in addition, we have preliminary data to suggest that neutrophil elastase may cleave it as well (W. Vandivier, V.A.F., and P.M.H., unpublished data). In fact, the predicted extracellular domain of the PtdSerR has several potential cleavage sites for neutrophil elastase, given that the preference for P1 is A>V>T>I (16). This receptor, when stimulated by apoptotic cells, phosphatidylserine-containing membranes, or a stimulatory mAb, induces the release of TGF-β and the down-regulation of proinflammatory cytokines (5). Our studies suggest that the cellular carcass of a cell progressing to the late stages of apoptosis can retain anti-inflammatory activity. With regard to the observations of Ren and Savill that neutrophils that have progressed from early to late apoptosis (i.e., becoming permeable to trypan blue) fail to stimulate macrophage cytokine production (14), we suggest that the anti-inflammatory signal of the exposed phosphatidylserine on the late apoptotic cell predominates, as the cells have not released proteases to cleave the PtdSer receptor. In any event, we suggest that activated, necrotic, or lysed cells can, by releasing proteases, overwhelm the anti-inflammatory effects of macrophage PtdSerR interactions with PtdSer-expressing apoptotic cells or membranous debris, and that prolonged inflammation would be the predicted end stage of cell death in the absence of clearance by phagocytes. In vivo, endogenous anti-proteases also contribute to the complexity of the final response.
One of the interesting observations from this study was that lysed cells induced release of IL-10, which could be attributed to the effect of serine proteases, particularly neutrophil elastase. Others have observed that IL-10 is up-regulated early and concomitantly with TNF-α and chemokines in a variety of inflammatory conditions (17, 18, 19, 20, 21). Although IL-10 is not believed to affect the constitutive rate of neutrophil apoptosis, it enhances apoptosis of neutrophils at inflammatory sites, promotes the survival of macrophages, and enhances macrophage removal of apoptotic neutrophils (22, 23, 24, 25, 26). Thus, the protease-mediated release of IL-10 in inflammatory sites may represent a protective mechanism designed to promote resolution of inflammation by enhancing neutrophil apoptosis. Certainly, in its absence, inflammation is more severe in a number of models (27, 28, 29).
Both TGF-β (resulting from stimulation with early or late apoptotic cells) and IL-10 (resulting from stimulation with lysed cells) have anti-inflammatory activities, including the down-regulation of inflammatory mediator production by macrophages (30, 31, 32, 33, 34, 35). However, in our experiments, down-regulation of chemokines and TNF-α was associated with TGF-β production; the pattern of IL-10 secretion in the absence of serum was similar to that for proinflammatory cytokines. Both TGF-β and IL-10 can be immunosuppressive by exerting effects on both lymphocytes and APCs (36, 37, 38, 39, 40). However, TGF-β has been shown to inhibit full maturation of dendritic cells, even in the presence of inflammatory stimuli such as TNF-α, effectively preventing the development of potentially harmful immune responses from a resolving inflammatory site (41). TGF-β also mediates bystander suppression associated with physiological self-tolerance in vivo, thereby preventing the activation of autoreactive lymphocytes (42). These activities may be critical when apoptotic cells are removed during physiological cell death or with apoptotic cell uptake associated with resolution of inflammation.
Exposure to necrotic vs apoptotic cells also has differential effects on maturation of dendritic cells for Ag presentation to lymphocytes. Sauter et al. recently showed that exposure of dendritic cells to necrotic tumor cells or to supernatants derived from these cells enabled them to mature into fully functional APCs (43). They also found differences in cell type in that only necrotic tumor cells, but not necrotic primary cells, could induce this effect. They could not define the active factor; however, released intracellular proteases may be worth evaluating in this system as well.
In summary, we have learned that exposure to apoptotic cells, whether neutrophils or lymphocytes, inhibited macrophage proinflammatory cytokine by a mechanism involving TGF-β (data presented herein and in Refs. 4, 5). In contrast, lysed cells, particularly granulocytes, stimulated production of proinflammatory cytokines and chemokines, as well as IL-10. Necrotic cells partially down-regulated zymosan-induced production of TNF-α and IL-10, but had no effect on chemokines. In the presence of a serine protease inhibitor, necrotic cells behaved like apoptotic cells, inhibiting proinflammatory cytokines (including chemokines), and increasing release of TGF-β1. These data suggest that the release of serine proteases, particularly those from neutrophils, can be proinflammatory by directly stimulating macrophages to produce cytokines. Therefore, lack of removal of apoptotic cells before lysis will prolong inflammation, increasing the potential for tissue damage, and thus providing signals to promote an immune response.
It is important to note that the experiments presented here have been conducted in the absence of serum to avoid the confounding effects of complement activation. Giclas and colleagues showed many years ago that mitochondria can activate the complement cascade, thereby contributing to the proinflammatory effects of lysed cells (10, 11). Future work will focus on the effects of serum in our in vitro system and how it influences the interaction between apoptotic cells and macrophages; however, preliminary data suggest that exposing macrophages to either apoptotic neutrophils or Jurkat T cells in the presence of serum still causes down-regulation of TNF-α and chemokine production, as is seen in the absence of serum; TGF-β production is also up-regulated (V.A.F., unpublished data). These observations suggest that the anti-inflammatory effects of apoptotic cells are dominant to any proinflammatory effects of complement, as apoptotic cells have been shown to activate both the classical and alternative complement pathways (44). Interestingly however, in the presence of serum, we found that IL-10 production by macrophages was enhanced by apoptotic cells of either type, whereas in the absence of serum, its production was inhibited. Whether IL-10 contributes to the anti-inflammatory effects of apoptotic cells in our system remains to be determined. Furthermore, when serum was present, necrotic Jurkat T cells demonstrated proinflammatory activity in that unstimulated macrophages secreted TNF-α and chemokines. These preliminary results are in keeping with Giclas’ observations that intracellular organelles can fix complement (10, 11), particularly because heating serum to 56°C abolished these effects (V.A.F., unpublished data).
In conclusion, it seems reasonable to hypothesize that apoptotic neutrophils are anti-inflammatory for two reasons: first, they fail to release their serine proteases even when they begin to become leaky (as assessed by uptake of propidium iodide and trypan blue), and because the phosphatidylserine exposed on their surfaces is significantly anti-inflammatory. One can predict that the outcome of inflammation will be determined by the balance between proinflammatory signals generated by the release of granulocyte proteases and anti-inflammatory signals generated by exposure to phosphatidylserine-containing membranes associated with apoptotic cells or membranous cellular debris. The balance between proteases and anti-proteases will also contribute significantly to whether an inflammatory lesion resolves or not. Importantly, the anti-inflammatory potential of apoptotic neutrophils appears to be maintained even through the late stages, providing multiple opportunities for resolution of inflammation.
Note added in proof.
Cocco and Ucker very recently showed that the inhibitory effects of apoptotic cells on macrophage cytokine production were dominant to the proinflammatory effects of necrotic cells (45). Jaffray and coworkers recently showed that pancreatic elastase induces macrophages to produce TNF-α in a NF-κB-dependent manner, supporting the hypothesis of surface receptors for elastase on macrophages (46).
Particular thanks are due to Jay Westcott for his invaluable help with and provision of the cytokine ELISA, Azzedine Dakhama for his assistance with development of the neutrophil elastase assay, and Linda Remigio and David Riches for their gift of mouse BMDM.
This work was funded by the National Institutes of Health (GM 60449, GM 48211, and HL 60980) and by the Cancer Center at University of Colorado Cancer Center (Grant P30 CA 46934).
Abbreviations used in this paper: HMDM, human monocyte-derived macrophage(s); MIP2, macrophage-inflammatory protein 2; BMDM, bone marrow-derived macrophage(s); AEBSF, 4-(2-aminoethyl) benzenesulfonylfluoride; TMB, tetramethylbenzidine.