|Home | About | Journals | Submit | Contact Us | Français|
The distinction between physiological (apoptotic) and pathological (necrotic) cell deaths reflects mechanistic differences in cellular disintegration and is of functional significance with respect to the outcomes that are triggered by the cell corpses. Mechanistically, apoptotic cells die via an active and ordered pathway; necrotic deaths, conversely, are chaotic and passive. Macrophages and other phagocytic cells recognize and engulf these dead cells. This clearance is believed to reveal an innate immunity, associated with inflammation in cases of pathological but not physiological cell deaths. Using objective and quantitative measures to assess these processes, we find that macrophages bind and engulf native apoptotic and necrotic cells to similar extents and with similar kinetics. However, recognition of these two classes of dying cells occurs via distinct and noncompeting mechanisms. Phosphatidylserine, which is externalized on both apoptotic and necrotic cells, is not a specific ligand for the recognition of either one. The distinct modes of recognition for these different corpses are linked to opposing responses from engulfing macrophages. Necrotic cells, when recognized, enhance proinflammatory responses of activated macrophages, although they are not sufficient to trigger macrophage activation. In marked contrast, apoptotic cells profoundly inhibit phlogistic macrophage responses; this represents a cell-associated, dominant-acting anti-inflammatory signaling activity acquired posttranslationally during the process of physiological cell death.
Cell death is vital to the morphological shaping of tissues in development and to the careful sculpting of functionally appropriate cellular repertoires (Surh and Sprent, 1994 ; Cecconi et al., 1998 ; Yeh et al., 1998 ; Yoshida et al., 1998 ). Selective cell deaths continue to play a role in the homeostasis of mature tissues. For example, the deletion of immune cells in the attenuation of an immune response (Webb et al., 1990 ; Kawabe and Ochi, 1991 ) and the elimination of cells that have become functionally inappropriate, including virally infected and transformed cells (Kägi et al., 1995 ), depend on the selective induction of cell death. The cell death process generally assures both that cells triggered to die will cease to function and that they will be cleared in an orderly manner. Cells that die in these physiological contexts typically are removed rapidly by phagocytic cells, including macrophages (Duvall et al., 1985 ; Savill et al., 1989 ). Of primary significance, these cell deaths ensue without inflammatory consequence (Kerr et al., 1972 ).
Apoptosis is characterized by an orderly sequence of internal events, of which chromatin condensation is one, that precede the loss of cellular integrity (Russell, 1983 ; Wyllie et al., 1984 ; Harvey et al., 2000 ). Early studies also recognized that physiological cell deaths occur in a cell autonomous manner and that bystander cells are unaffected (Ucker et al., 1989 ; Dhein et al., 1995 ). Consistent with these observations, engulfing macrophages remove dying cells in the absence of infiltrating immune effectors. The efficient noninflammatory clearance of inappropriate T cells during development in the thymus illustrates this phenomenon dramatically (Surh and Sprent, 1994 ). In contrast, necrotic cell death, marked by rapid, disorganized swelling and rupture and associated with pathological tissue injury (Henson and Johnson, 1987 ), elicits inflammatory responses as well as clearance by phagocytic cells (Henson and Johnson, 1987 ; Stern et al., 1996 ).
These observations have led to the hypothesis that properties unique to the dying cell must determine the mode and outcome of phagocytic clearance. A variety of molecules have been implicated as putative recognition elements, including phospholipid ligands on the surface of the apoptotic corpse. Alterations in lipid composition occur early in the apoptotic process (Fadok et al., 1992b ; Schlegel et al., 1993 ). Phosphatidylserine (PS), an anionic phospholipid normally cloistered in the inner leaflet of the plasma membrane, is externalized during the cell death process (Fadok et al., 1992b ). Phospholipid flipping during cell death is linked to the activation of effector caspases (death-specific proteases) and occurs upstream of nuclear changes (Bratton et al., 1997 ; Harvey et al., 2000 ). The belief that this externalized phospholipid serves as a ligand for macrophage recognition follows from studies demonstrating that similar changes target aged erythrocytes for clearance (Schroit et al., 1985 ; McEvoy et al., 1986 ). The interaction of dying nucleated cells with macrophages is inhibited partially by phospho-l-serine and PS vesicles, which appear to bind to sites on the macrophage (Fadok et al., 1992b , 1998b ; Verhoven et al., 1995 ; Terpstra et al., 1998 ).
Complementary studies have focused attention on scavenger receptors as the putative receptors of engulfing macrophages. Class B scavenger receptors, especially CD36 (Savill et al., 1992 ), constitute an attractive group of candidate recognition molecules that exhibit specificities consonant with the known apoptotic cell surface changes (Fadok et al., 1992b , 1998b ; Chang, M.-K. et al., 1999 ). CD36 has been proposed to function in concert with distinct macrophage integrins (Savill et al., 1990 ; Albert et al., 1998 ) as an anchor for a lectin-like thrombospondin bridge to the target (Savill et al., 1992 ). Other studies have attributed recognition specificity and engulfing activity to class A scavenger receptors and CD68 (Platt et al., 1996 ; Ramprasad et al., 1996 ; Terpstra et al., 1997 ). Of greatest interest, a novel cell surface molecule has been identified recently that is implicated, both by antibody blocking and gene transfer experiments, in PS-dependent clearance of apoptotic cells (Fadok et al., 2000 ). In addition to these, a number of other molecules have been identified that may be involved in selective cases of target cell interaction with the phagocytic cell (Hart et al., 1997 ; Schwartz et al., 1999 ). CD14, a lipopolysaccharide (LPS)-binding molecule normally associated with inflammation, is of particular intrigue (Devitt et al., 1998 ). Although an essential role for scavenger receptors in the clearance of dead cells is established through genetic studies in Drosophila melanogaster (Franc et al., 1999 ), similar definitive assignments for mammalian scavenger receptors and other molecules have not been made, in large part because specific inhibitors afford only incomplete interference. As well, loss-of-function mutations in the class A scavenger receptor (Suzuki et al., 1997 ; Terpstra et al., 1997 ; Platt et al., 1998 ) or CD36 (Hughes et al., 1997 ; Febbraio et al., 1999 ) confer only marginal reductions in phagocytic recognition in vivo. Of course, it may be that recognition activity cannot be ascribed to any one class of molecules exclusively. Genetic studies of developmental cell death in the worm Caenorhabditis elegans similarly reveal the absence of singularly essential molecules for engulfment, implying that multiple and redundant mechanisms operate for the clearance of dead cells (Ellis et al., 1991 ).
Technical limitations and differences in experimental approaches also have obscured a resolution of the issues of specificity in the phagocytic process. Few studies have discriminated binding from engulfment, and the low frequency of phagocytic activity in many cases has necessitated subjective evaluation of dead cell interactions with phagocytic cells. Variability also pertains to the different populations of targets and engulfing cells that have been used. Most importantly, the fates of apoptotic and necrotic cells have not been compared rigorously (Hirt et al., 2000 ); typically apoptotic cells have been contrasted with nonnative targets opsonized with immunoglobulin or complement molecules. Certainly, opsonized cells are phagocytic targets and inflammatory elicitors distinct from true apoptotic cells (Ravetch, 1994 ), but they are reflective of pathogenic invaders rather than endogenous cells that have suffered a pathological fate.
We have developed a quantitative and objective approach to explore the recognition by macrophages of native cells that have undergone physiological or pathological deaths. Here, we show that recognition of apoptotic and necrotic targets occurs by distinct and noncompeting mechanisms and that PS is not a specific ligand for apoptotic cell recognition. These distinct recognition processes are linked to opposing phlogistic responses. Necrotic cells enhance, but are not sufficient to initiate, macrophage activation. Most significantly, the process of physiological cell death imparts on apoptotic cells an anti-inflammatory activity that functions in a dominant manner to abrogate proinflammatory responses of engulfing macrophages.
We have developed an objective microwell assay to assess the extent of target cell recognition (as well as the consequent production of proinflammatory cytokines) by phagocytic cells. In essence, recognition of fluorescently labeled target cells can be quantified as the extent of (nonadherent) target cell fluorescence that becomes associated with (adherent) phagocytic cells after their interaction. We also have chosen to work with clonal lines of macrophages and target cells in which we induced physiological death or triggered pathological killing, in order to minimize issues of heterogeneity.
Target cells were labeled covalently, especially with the amine-reactive probe (5,6)-carboxyfluorescein diacetate succinimidyl ester (CFDA) or the lipophilic carbocyanine dye chloromethylated lipophilic carbocyanine dye DiIC18(3) (CM-DiI). CFDA labeled cells uniformly, whereas CM-DiI labeling was less homogenous and more “grainy.” These labels were retained in living cells and throughout the death process. Moreover, staining was maintained quantitatively after dead cell ingestion (our unpublished results). The uptake of propidium iodide (PI) by apoptotic cells was sufficiently stable to permit PI-stained corpses to be visualized after their ingestion; in contrast, PI leaked from necrotic cells during the phagocytosis assay, precluding its use in this analysis (our unpublished results). Of technical note, a cotransfected green fluorescent protein marker (Harvey et al., 2000 ) also was suitable to track the fate of transfectants upon interaction with macrophages. The intensity of tracker dye signals permitted the interactions of target cells with macrophages to be monitored on a microwell scale, facilitating their objective and quantitative assessment with a fluorescence plate reader. At the same time, culture supernatants could be collected and assayed for the release of cytokines. In the experiments reported here, we have monitored the secretion of TNF-α and IL-6 as a measure of proinflammatory outcome.
Freshly cloned cells were grown at 37°C in a humidified, 5% (vol/vol) CO2 atmosphere in RPMI 1640 medium (Mediatech, Herndon, VA) supplemented with l-glutamine (2 mM), 2-mercaptoethanol (50 μM), and heat-inactivated fetal bovine serum (10% vol/vol; Hyclone Laboratories, Logan, UT). J774A.1 and RAW 264.7 are monocyte-derived macrophage cell lines derived from H-2d mice. Murine S49 thymoma (H-2d) and DO11.10 T-cell hybridoma (H-2d × H-2k) cells were used as targets. Physiological cell death (apoptosis) was induced by treatment of the T cells with the synthetic glucocorticoid dexamethasone (1 μM, 8–22 h) or the macromolecular synthesis inhibitor actinomycin D (200 ng/ml, 8–18 h; Ucker et al., 1989 ). Cells were killed pathologically (necrotic death) by incubation at 55°C for 10–15 min (until Trypan Blue uptake indicated compromise of membrane integrity).
Target cells were labeled with CFDA (Molecular Probes, Eugene, OR; Exλ = 490 nm; Emλ = 525 nm). Cells (1 × 106 cells/ml in PBS) were incubated with CFDA (5 μM) for 10 min at 37°C and then washed twice in complete medium. Cells were labeled before the induction of physiological cell death or heat killing. Target cells also were labeled with CM-DiI (Molecular Probes; Exλ = 530 nm; Emλ = 645 nm) under similar conditions (2 μM CM-DiI, 30 min at 37°C). For visualization of chromatin, cells were incubated with Hoechst 33342 (Sigma Chemical, St. Louis, MO; 1 μg/ml, 10 min, 37°C), and stained chromatin was visualized (Exλ = 355 nm, Emλ = 465 nm) with a Nikon Diaphot 200 microscope with epi-fluorescence (Nikon, Garden City, NY). The accessibility of phosphatidylserine was revealed by the binding of FITC-conjugated annexin V (PharMingen, San Diego, CA; Exλ = 488 nm, Emλ = 525 nm). Cells were harvested and washed twice with cold PBS. Cells were resuspended in 100 μl of binding buffer (10 mM HEPES, pH 7.4, 140 mM NaCl, 2.5 mM CaCl2) and incubated with 5 μl of FITC-conjugated annexin V for 15 min in the dark at 25°C. After incubation, 400 μl of binding buffer was added per sample, and cells were analyzed cytofluorometrically (FACSCaliber instrument and CellQuest software; Becton Dickinson, San Jose, CA). PI, which was used to assess plasma membrane integrity, was added to cells at 1 μg/ml immediately before cytofluorometric analysis (Exλ = 488 nm, Emλ = 610 nm). Cytofluorometric data were processed with WinMDI software (Joe Trotter, Scripps Research Institute, LaJolla, CA).
Two hours before the initiation of phagocytosis assays, macrophages were plated in 96-well flat-bottom tissue culture plates (Costar, Corning, NY) at a density of 2 × 104 cells/0.32-cm2 well to allow semiconfluent monolayer formation. Graded numbers of target cells were added to the macrophage monolayers, and cells were allowed to interact at 4°C (binding) or at 37°C (engulfment as well as binding). In typical interaction assays, the duration of incubation was 60 min. Wells then were washed twice with ice-cold PBS, and plate-bound fluorescence was analyzed on a Cytofluor 2350 Fluorescence Plate Reader (Millipore, Marlborough, MA). For quantitation, a standard curve was prepared with graded number of labeled target cells. The fluorescent labeling of target cells, which varied little (<10%) between experiments, yielded specific fluorescence intensities of ~1.25 × 103 fluorescence units per 2 × 104 cells (above a background of <50 fluorescence units). That is, in a well with 2 × 104 macrophages, target cell binding, even at level of 0.1 target/macrophage, could be quantified reliably. All data points are the means (± SEM) of triplicate determinations, and each of the experiments presented is representative of multiple (typically ≥10) repetitions. Binding and phagocytosis also were visualized microscopically. For detailed microscopic examination, targets were allowed to interact with macrophages that had been plated at lower density (1 × 105 cells/1.8 cm2 chamber) in microplates with coverslip bottoms (Fisher Scientific, Hanover Park, IL). Digital images were acquired with a SenSys CCD Camera (Photometrics, Tucson, AZ). Images were analyzed using Image-Pro Plus software (Media Cybernetics, Silver Spring, MD).
Cytokine production by macrophages was assessed after incubation with target cells for 4 h at 37°C. Where indicated, LPS (Escherichia coli O111:B4; Sigma Chemical) was added to macrophages 2 h before the addition of targets. Culture supernatants were withdrawn from wells and assessed quantitatively for secreted TNF-α and IL-6. Cytokines were assayed by ELISA (Endogen, Woburn, MA), using matched-pair, cytokine-specific capture and biotinylated reporter antibodies. The reporter reaction was developed with HRP-conjugated streptavidin and quantified spectrophotometrically at 450 nm (corrected for turbidity at 550 nm; Microplate Autoreader model EL311; Bio-Tek Instruments, Winooski, VT).
Small unilamellar vesicles were prepared by sonication from egg phosphatidylcholine (PC; Sigma Chemical) alone or mixed (at a molar ratio of 7:3) with brain PS (Avanti Polar Lipids, Alabaster, AL) as described (Pradhan et al., 1997 ). Arg-Gly-Asp-Ser (RGDS) and Arg-Gly-Glu-Ser (RGES) tetrapeptides were purchased from Sigma Chemical.
We explored the ability of macrophages to recognize dying cells by challenging them in a dose- and time-dependent manner with distinct populations of targets. Physiological cell death was induced in T-cell targets by treatment with dexamethasone or actinomycin D (Ucker et al., 1989 ). These cell death responses are associated with the typical hallmarks of apoptosis, including cell shrinkage and chromatin condensation (exemplified in Figure Figure1,1, B, C, J, and K, for the T-cell hybridoma DO11.10), as well as caspase activation and genome digestion. Cells were killed with heat to generate pathological cell death targets. Incubation of DO11.10 target cells at 55°C for 10–15 min resulted in an immediate loss of viability, as assessed by the uptake of PI through a compromised plasma membrane (Figure (Figure1H).1H). Heat-killed cells exhibited no apoptotic hallmarks; rather, they were swollen (Figure (Figure1D),1D), and their chromatin was uncondensed (Figure (Figure1L)1L) and not digested (our unpublished results). We examined target cells that had been induced to die physiologically and had suffered an equivalent loss of plasma membrane integrity as well as cells at an earlier stage of the physiological cell death process, when fewer cells had yet lost substantial membrane integrity (cf. Figure Figure1F1F and and1G).1G). For brevity, we hereafter use the terms apoptosis and necrosis to reflect the consequences of physiological and pathological cell deaths, respectively.
Most studies of the interactions between phagocytic cells and their targets have not discriminated between recognition (binding) and engulfment. However, the early study of Duvall et al. (1985) suggested that binding might occur in the absence of engulfment when interactions are allowed to occur at 4°C. Apoptotic, necrotic, and viable CFDA-labeled DO11.10 cells were incubated at 4°C with a monolayer of J774A.1 macrophages without agitation. Graded numbers of targets were added to a constant number of freshly plated macrophages, at input target to macrophage ratios ranging as high as 50:1. At various times of incubation, unbound target cells were removed by washing (with ice-cold buffer), and the extent of target cell binding to the macrophage monolayer was quantified fluorometrically.
Figure Figure2A2A presents the results of one representative experiment, in which target cells were allowed to bind to macrophages for 60 min. Live cells were not bound appreciably to macrophages, whereas apoptotic and necrotic targets were bound to comparable and significant extents. The magnitude of target binding per macrophage displays a roughly linear dependence on target cell dose; in these assays, about 10% of all targets were bound. Most notably, macrophage binding capacity is high: one macrophage can bind more than one target. Cells at early and late stages of the physiological cell death process were bound to similar extents, suggesting that determinants for macrophage recognition appear well before obvious manifestations of cell death.
Typical kinetics by which macrophages bound apoptotic targets cells are shown in the inset of Figure Figure2A2A (at a target to macrophage ratio of 40:1). Binding was saturable kinetically, and the same kinetic pattern was followed (with proportionately lower maximum extents of binding) in experiments in which reduced numbers of target cells were added. Scatchard analysis of these data suggest that, on average, each macrophage has five to eight sites for binding apoptotic target cells (but see next). The extents and kinetics of necrotic and apoptotic cell binding to J774A.1 macrophages were equivalent (see Figure Figure2A2A and our unpublished results).
To visualize these interactions microscopically, target cells were stained with Hoechst 33342 to mark cells with apoptotic chromatin condensation (see Figure Figure1K),1K), and macrophages were labeled with CFDA. The peripheral association of apoptotic targets (blue, with fragmented nuclei) with macrophages (green) confirms that binding without internalization occurs at 4°C (Figure (Figure3A).3A). Under these conditions, macrophages were similarly decorated by bound necrotic (blue, unfragmented) targets (Figure (Figure3B).3B). Surprisingly, approximately 30% of the macrophages were responsible for all target cell binding. Apparently, within the clonal macrophage population, some macrophages are more competent than are others at any time. Fluorometrically derived averages incorporate a mix of macrophages that have bound no targets and others that are “jackpots.”
When DO11.10 cells were incubated with J774A.1 macrophages at 37°C, phagocytosis of the targets ensued. Virtually all targets appeared to be internalized (Figure (Figure3,3, C and D). The data in Figure Figure2B 2B demonstrate that macrophages interacted stably with greater numbers of target cells when they were able to engulf as well as bind targets. Selectivity also was evident: macrophages bound and engulfed apoptotic and necrotic targets, but not viable cells, to a similar extensive degree. As visualized microscopically, engulfment appeared to involve the fragmentation of all target cells (Figure (Figure3,3, C and D).
These macrophages exhibited similar binding and engulfing activity when presented with different target cells and with apoptotic targets that had been induced to die by treatment with different stimuli. The data in Figure Figure22 (C and D), for example, detail the fate of S49 thymoma targets. In this case, apoptotic targets resulted from treatment with dexamethasone. Even fully allogeneic and xenogeneic apoptotic and necrotic targets, but not viable cells, were recognized specifically and efficiently by J774A.1 macrophages (our unpublished results). For consistency, the data presented below are derived from experiments with a single (semiallogeneic) target cell line, DO11.10.
That the extents of uptake of apoptotic and necrotic targets by macrophages were essentially indistinguishable suggested the possibility that a common mechanism exists for the recognition of all native (nonopsonized) corpses. The direct inference of that view is that necrotic corpses should compete with apoptotic ones for macrophage binding. To test this prediction, target cell binding was examined at early times, under conditions where competition for kinetically saturable sites was detectable (see Figure Figure2A,2A, inset). The ability of the unlabeled (or differentially labeled) targets to interfere with binding of the labeled ones was assessed. (To enhance the magnitude of the signal, the competition was performed at 37°C to allow macrophages to begin to engulf bound targets.) The data in Figure Figure4A4A document the ability of apoptotic targets to compete with themselves in a dose-dependent manner. In contrast, the binding of apoptotic targets was unaffected by necrotic cells, although necrotic competitors interfered effectively with labeled necrotic cell binding (Figure (Figure4B).4B). Of importance, microscopic examination revealed that apoptotic and necrotic targets (labeled distinctly with CFDA and CM-DiI) bound to the same macrophage (our unpublished results). The failure of apoptotic and necrotic cells to compete with each other implies that independent mechanisms for the recognition of apoptotic and necrotic targets are involved.
Binding studies with a different macrophage cell line confirmed this conclusion. RAW 264.7 macrophages bound and engulfed apoptotic targets at a similar rate and to a comparable extent as J774A.1 cells (Figure (Figure5).5). An estimate of this binding by Scatchard analysis would suggest that the average number and “avidity” of apoptotic sites on RAW 264.7 cells and J774A.1 cells differ by less than threefold. Remarkably, RAW 264.7 cells exhibited no ability to interact with necrotic targets prepared from DO11.10 (Figure (Figure5),5), S49, or other cell lines (our unpublished results). The dramatic dissociation of apoptotic and necrotic cell recognition by RAW 264.7 macrophages establishes that functionally distinct mechanisms pertain. Moreover, these data suggest that functional as well as morphological attributes of physiological cell death are retained throughout the death process and suggest that arbitrary distinctions drawn between early and late apoptotic cells (e.g., apoptosis versus secondary necrosis) may not be consequential.
The ability of PS vesicles to compete with apoptotic cells for binding to macrophages has been taken to suggest that exposure of PS on the outer leaflet of the plasma membrane is a sentinel event in the apoptotic cell death process (Fadok et al., 1992b ; Verhoven et al., 1995 ). Our characterization of apoptotic and necrotic targets, however, evidenced that the externalization of PS is not a feature unique to apoptotic cells (see Figure Figure1).1). PS exposure, detected by the binding of FITC-conjugated annexin V (Raynal and Pollard, 1994 ), preceded plasma membrane disintegration, marked by PI uptake, of necrotic as well as apoptotic cells. The annexin V+ PI− cells indicated in Figure Figure1H1H represent this intermediate stage of necrotic death. Equivalent apoptotic cells are present at early stages of physiological cell death (Figure (Figure1F)1F) and are less abundant at late stages (Figure (Figure11G).
We assessed the role of PS in the binding of apoptotic and necrotic targets. PS vesicles were able to inhibit apoptotic target binding (our unpublished results) and engulfment (Figure (Figure6A)6A) by J774A.1 and RAW 264.7 macrophages. Inhibition was significant although incomplete; vesicles composed of PC, a nonanionic phospholipid, failed to inhibit macrophage interactions with target cells (Figure (Figure6A).6A). PS vesicles, and not PC vesicles, also were effective at blocking necrotic cell interactions with macrophages, consonant with the presence of externalized PS on the necrotic cells. In contrast, PS vesicles were not effective at blocking interactions of macrophages with immunoglobulin-opsonized cells (our unpublished results).
Some studies have suggested that activated macrophages rely particularly on a PS-inhibitable mode of target cell recognition, whereas unactivated macrophages use an integrin-dependent process (Fadok et al., 1992a , 1998b ; Pradhan et al., 1997 ). Diagnostically, the integrin-dependent mechanism is inhibitable with a specific integrin-binding tetrapeptide, RGDS (Savill et al., 1990 ). This prompted us to ask whether the interacting macrophages identified visually as jackpots (Figure (Figure3)3) are a minor, spontaneously activated (PS-inhibitable and RGDS-uninhibitable) subpopulation within the J774A.1 culture. Contrary to this view, RGDS tetrapeptide also was able to inhibit macrophage binding of apoptotic target cells (Table (Table1).1). RGDS peptide was equally effective at inhibiting the binding of necrotic target cells (Table (Table1). 1).
We characterized the release of the proinflammatory cytokines TNF-α and IL-6 after target cell interactions with macrophages as an indicator of inflammatory outcome. Necrotic and apoptotic targets alone did not elicit J774A.1 and RAW 264.7 macrophages to secrete TNF-α and IL-6 (Figure (Figure77 [note the absence of cytokines when LPS is absent] and our unpublished results). LPS (Figure (Figure7)7) and opsonized targets (our unpublished results), on the other hand, were able to activate macrophages to secrete those proinflammatory cytokines.
We investigated whether necrotic target cells could augment a proinflammatory LPS signal. Indeed, when macrophages were primed with suboptimal concentrations of LPS, their incubation with necrotic cells resulted in a modest elevation of cytokine secretion relative to macrophages treated with LPS alone (Figures (Figures77 and and8).8). Secreted cytokines were the products of the macrophages, because necrotic cells did not secrete detectable TNF-α or IL-6, even when treated with LPS (our unpublished results). Of note, cytokine release does not appear to be a consequence of macrophage rupture, because engulfing macrophages remained viable and adherent throughout these assay (see Figure Figure3). 3).
Engulfment of necrotic cells in vivo is believed to result in an inflammatory response. Our results establish this phenomenon in a simplified cell culture system and also suggest that the interaction between a necrotic corpse and its engulfing macrophage is not sufficient for this response. These data demonstrate that neither do macrophages need to be activated to bind and engulf targets, nor do they become activated simply because they do engulf. Rather, necrotic cells trigger the enhanced secretion of proinflammatory cytokines from independently activated macrophages.
In contrast to the augmentation of cytokine secretion afforded by necrotic targets, macrophages that were incubated with apoptotic targets did not release TNF-α or IL-6, even when primed with optimal doses of LPS (Figure (Figure7).7). In fact, apoptotic cells (induced to die with actinomycin D [Figure 7] or dexamethasone [our unpublished results]) strongly inhibited the secretion of proinflammatory cytokines from macrophages. Macrophages pretreated with LPS retained their ability to bind and engulf target cells (Figure (Figure66B).
The absence of a cytokine response associated with apoptotic cell interaction might reflect the absence of a proinflammatory stimulatory signal on apoptotic targets. However, the observation that IL-6 and TNF-α secretion by LPS-activated macrophages is abrogated after engulfment of apoptotic targets strongly suggests that apoptotic cells actively antagonize (LPS-derived) proinflammatory signals. To test this model, a mixture of apoptotic and necrotic targets was presented to LPS-activated macrophages. The inhibitory effect of the apoptotic targets was completely dominant to the stimulatory effect of the necrotic cells (Figure (Figure8).8). The absence of proinflammatory cytokines is not due to simple absorption by the apoptotic cells, because the levels of already-secreted cytokines after LPS stimulation were not reduced by the addition of apoptotic targets (our unpublished results). Neither is this inhibition mediated through soluble factors released from apoptotic cells. In all experiments, apoptotic cells were washed twice before they were presented to macrophages; furthermore, apoptotic cell supernatants were not inhibitory (our unpublished results). We conclude that during the process of physiological cell death, apoptotic cells acquire a cell-associated, dominant-acting, anti-inflammatory signaling activity that overrides proinflammatory macrophage responses.
The notion that apoptotic cells can inhibit proinflammatory macrophage responses is consistent with previous work, including studies of LPS-activated, monocyte-derived macrophages (Voll et al., 1997 ; Fadok et al., 1998a ; McDonald et al., 1999 ). Henson and coworkers (Fadok et al., 1998a ; McDonald et al., 1999 ) have argued that apoptotic inhibition, observed after long periods (14–18 h) of incubation, is effected in a paracrine manner, through the induced release by macrophages of antagonistic factors, including platelet-activating factor, prostaglandin E2, and especially TGF-β. A distinct issue is raised by the observation that the inhibition imposed by apoptotic cells is rapid. The secretion of TNF-α and IL-6 by macrophages, which was detectable after less than 2 h of LPS stimulation (see Figure Figure7),7), was halted by the addition of apoptotic cells, even when incubation in the presence of LPS was continued (Figures (Figures77 and and88 and our unpublished results). These results suggest that a virtually immediate and direct anti-inflammatory effect of apoptotic cells must be exerted on the engulfing macrophage, proximal to proinflammatory signaling. It remains to be determined on what level this blockade is enforced.
Cell death is critical in normal organismal development and homeostasis, particularly for shaping and maintaining appropriate cellular networks. We have addressed previously the fundamental question of whether a common cell-autonomous effector mechanism pertains in distinct cases of cell death. That work has led to the identification of a thematically conserved, ordered pathway for cellular destruction. The ability of a dying cell to trigger phagocytosis without eliciting an inflammatory response likely is the overriding biological purpose of the physiological cell death process. The crucial questions in this context are how recognition without inflammatory response is assured and where within the apoptotic process signals for phagocytic clearance are expressed.
Our data demonstrate that macrophages discriminate innately between cells that have undergone a physiological death and those that have suffered a pathological death. The distinction drawn between apoptotic and necrotic corpses may have seemed arbitrary, especially in light of the similar binding behaviors and inhibitor (including phospholipid vesicle) profiles that we observed for the two classes of targets. However, competition experiments between different targets and binding studies with the RAW 264.7 macrophage cell line establish that macrophage recognition of the products of physiological and pathological cell deaths occur by distinct mechanisms. Clearly, it will be important to extend these finding to primary macrophage populations. At the same time, the RAW 264.7 macrophage cell line, which lacks an intact recognition mechanism for necrotic cells, is an intriguing subject for genetic reconstitution. Of greater interest would be a complementary macrophage lacking apoptotic cell recognition function.
Altogether, our data confirm that the distinction between apoptotic and necrotic cells is real and that independent, noncompeting modes of binding are involved in the recognition of native cells that have died by distinct modes of death. It is significant that clearance and phlogistic outcomes attributed to the processes of physiological cell death and pathological death in vivo are recapitulated reliably in a simple cell culture setting. These observations imply that the interaction of a macrophage with its target cell alone is sufficient to effect these responses.
It is surprising that even within clonal populations of macrophages, only a minority of cells is competent to bind and engulf targets at any time. Previous studies with primary macrophages (of peritoneal or monocytic origin) have established that phagocytic activity increases after in vitro culture (“maturation”; Newman et al., 1982 ), and that treatment with particulate stimuli also enhances subsequent activity (Fadok et al., 1993 , 1998b ). Binding that is inhibitable with PS vesicles, moreover, is reportedly restricted to activated macrophages (Pradhan et al., 1997 ; Fadok et al., 1998b ). Consistent with this view, expression of the candidate PS-dependent receptor as well as scavenger receptors and CD68, molecules that could be responsible for anionic phospholipid binding, is elevated after activation (Fukasawa et al., 1996 ; Ramprasad et al., 1996 ; Murao et al., 1997 ; Fadok et al., 2000 ). Integrin-dependent (RGDS-inhibitable) interactions, in comparison, have been attributed to unactivated cells (but see Savill et al., 1990 ; Pradhan et al., 1997 ; Fadok et al., 1998b ). Although J774A.1 cells have been characterized as resembling “unactivated” macrophages (Pradhan et al., 1997 ), the assays we use here evince substantially higher levels of binding and engulfment than reported previously (also see McDonald et al., 1999 ) and reveal significant inhibition of their target cell interactions by PS vesicles. These data suggest that PS-inhibitable interactions may not be restricted to activated macrophages and that modes of binding inhibitable by RGDS and by PS are not mutually exclusive.
More importantly, our data indicate that PS is unlikely to be involved specifically in the recognition of apoptotic target cells. Both necrotic and apoptotic cells display externalized PS, yet relative to apoptotic cells, necrotic cells are not recognized equivalently (and not at all in one case). It seems clear that PS exposure cannot be sufficient for recognition (see also Pradhan et al., 1997 ). That PS vesicles, presumably binding to the macrophage (Terpstra et al., 1998 ), inhibit partially the recognition of both classes of corpses suggests that PS-specific binding sites on the macrophage may facilitate interactions with target cells or that bound vesicles may interfere sterically with accessibility of targets for other recognition molecules.
Apoptotic corpses derived from syngeneic and congenic cells induced by disparate suicidal stimuli are recognized and engulfed with equivalent specificity and efficiency in this system. Human targets also are recognized efficiently by these murine macrophages (our unpublished results; also see McDonald et al., 1999 ), implying that apoptotic cell determinants for macrophage recognition are widely conserved. Among the apoptotic stimuli we have used, inhibitors of macromolecular synthesis illuminate further that the appearance of these recognition determinants is not dependent on de novo synthesis (also see Flora et al., 1996 ).
Distinct macrophage mechanisms for the recognition of apoptotic and necrotic cells are associated with opposing phlogistic outcomes. A critical question is whether the anti-inflammatory signals acquired by an apoptotic cell are distinct from the recognition determinants it expresses. We have begun to explore this issue by asking where within the ordered physiological cell death process signals for macrophage recognition are expressed. Our initial experiments map the expression on the apoptotic cell of determinants for both recognition and inhibition of inflammatory response downstream of the action of Bcl-2 (our unpublished results). Target cells treated with a death-inducing concentration of actinomycin D and spared from death by transfected Bcl-2 were not recognized by J774A.1 cells and were unable to inhibit the LPS-mediated induction of TNF-α from those macrophages. These results extend earlier findings that Bcl-2 spared cells from phagocytic recognition (Flora et al., 1996 , but see Lagasse and Weissman, 1994 ).
In a larger context, the issue of an obligatory linkage between apoptotic death and noninflammatory outcome is posed profoundly with the death of macrophages themselves. Macrophages do not die when they engulf (Meagher et al., 1992 ; Bellingan et al., 1996 ), but they can be triggered to undergo physiological cell death, for example, by pathogens or immune effectors (Richter-Dahlfors et al., 1997 ; Oddo et al., 1998 ). This macrophage death is associated with the processing and release of the potent proinflammatory cytokine IL-1β (Hogquist et al., 1991 ). It will be of great importance to understand whether the apoptotic death of a macrophage is necessarily proinflammatory. Perhaps it is just this unusual apoptotic death that signals immunological “danger” (Matzinger, 1994 ).
It is striking that the anti-inflammatory effect of apoptotic targets is dominant to the inflammatory enhancement of necrotic cells. This dispels the commonly held notion that the phagocytosis of apoptotic cells occurs prelytically so as to circumvent the inflammatory response that ensues after the release of noxious intracellular contents from ruptured corpses (Ren et al., 1995 ; Stern et al., 1996 ; Fadok et al., 1998b ). Although the failure of apoptotic cells to promote the maturation of dendritic cells has not been associated with a dominant inhibitory effect (Sauter et al., 2000 ), the absence of an inflammatory macrophage response to apoptotic cells clearly reflects an active inhibitory signal, not simply the absence of a proinflammatory one. The acquisition of this anti-inflammatory signal represents a gain-of-function that occurs independently of de novo macromolecular synthesis. Like the induction of death-associated effector activities (SH Chang, KJ Harvey, M Cvetanovic, and D.S. Ucker, unpublished results), the appearance of determinants for recognition and inhibition of inflammation must occur primarily on a posttranslational level. If macrophages need receive this signal affirmatively in order not to produce proinflammatory cytokines, the loss of recognition function, for example, by mutation, likely would manifest a chronically uninhibited inflammatory response, akin to the loss of TGF-β (Shull et al., 1992 ; Kulkarni et al., 1993 ). This provides a stringent criterion in the evaluation of macrophage receptors for apoptotic cells.
Finally, the direct abrogation by apoptotic cells of proinflammatory cytokine release raises the issue of whether these inflammatory responses, like target cell interactions, are limited to a fraction of the macrophage population. Inhibition exerted independently of soluble factors could only be manifest if macrophages that secrete cytokines are exclusively the ones that interact with targets. It will be intriguing to determine whether that assessment of cytokine production on the level of the individual cell independently identifies macrophages visualized as jackpots by virtue of their target cell interactions.
We are grateful to Lin Tao (University of Illinois, College of Dentistry) and Richard Ye (University of Illinois, College of Medicine) for providing J774A.1 and RAW 264.7 cells, respectively, and to Hayat Onyuksel (University of Illinois, College of Pharmacy) for preparing liposomes. We thank our colleagues Oscar Colamonici, Jim Cook, Daniel Floryk, Amy Kenter, Dunja Lukovic, Navreet Nanda, Bellur Prabhakar, William Walden, and Richard Ye for their constructive comments. This work was supported by grants to D.S.U. from the National Institutes of Health and a generous fellowship to R.E.C. from the International Foundation for Ethical Research.