|Home | About | Journals | Submit | Contact Us | Français|
CD47, a ‘self’ recognition marker expressed on tissue cells, interacts with immunoreceptor SIRPα expressed on the surface of macrophages to initiate inhibitory signaling that prevents macrophage phagocytosis of healthy host cells. Previous studies have suggested that cells may lose the surface CD47 during aging or apoptosis to enable phagocytic clearance. In the present study, we demonstrate that the level of cell surface CD47 is not decreased but the distribution pattern of CD47 is altered during apoptosis. On non-apoptotic cells, CD47 molecules are clustered in lipid rafts forming ‘punctates’ on the surface, whereas on apoptotic cells, CD47 molecules are diffused on the cell surface following the disassembly of lipid rafts. We show that clustering of CD47 in lipid rafts provides a high binding avidity for cell surface CD47 to ligate macrophage SIRPα, which also presents as clusters, and elicit SIRPα-mediated inhibitory signaling that prevents phagocytosis. In contrast, dispersed CD47 on the apoptotic cell surface is associated a significant reduction of the binding avidity to SIRPα and failure to trigger SIRPα signal transduction. Disruption of lipid rafts with methyl-β-cyclodextrin (MβCD) disrupted CD47 cluster formation on the cell surfaces, leading to decrease of the binding avidity to SIRPα and a concomitant increase of cells being engulfed by macrophages. Taken together, our study reveals that CD47 normally is clustered in lipid rafts on non-apoptotic cells but is diffused in the plasma membrane when apoptosis occurs, and this transformation of CD47 greatly reduces the strength of CD47-SIRPα engagement, resulting in the phagocytosis of apoptotic cells.
CD47 is an immunoglobulin (Ig) superfamily transmembrane protein that is universally expressed on mammalian cells and tissues. Through its trans interactions with SIRPα on macrophages, CD47 triggers tyrosine phosphorylations in the SIRPα cytoplasmic immunoreceptor tyrosine-based inhibitory motifs (ITIMs) and recruitment of protein tyrosine phosphatases SHP-1/SHP-2, which further mediate negative signaling events that inhibit macrophage phagocytosis. For this, CD47 acts as a “self” marker and prevents macrophage engulfment of host cells (1, 2). This self-recognition system mediated by CD47-SIRPα interaction plays a critical role in restraining macrophages. Disruption of CD47-SIRPα interaction would lead to normal tissue damage (3–6) on one hand, while preservation of this self-recognition could result in failure of clearing apoptotic cells, pathogen-infected cells, or tumor cells (7) on other hand.
Recent studies of cell apoptosis and how apoptotic cells are cleared by macrophages suggest that there are three kinds of potential signals controlling macrophages to target apoptosis cells. The first signal is a ‘find me’ signal. The apoptotic cells release soluble factors such as lysophosphatidylcholine (LPC) (8) that act as chemoattractants for recruiting macrophages or other phagocytes. Following macrophages approaching, previous studies have shown molecules that are especially increased on apoptotic cells, such as phosphatidylserine (PS) (9) and calreticulin (10, 11), initiate the next ‘eat me’ signaling, the second class of signal (7,8). Meanwhile, CD47, through ligation of macrophage SIRPα, provides an additional control - the “don’t eat me” signal, which should restrain the process initiated by the first two classes of signaling. As apoptotic cells do indeed get engulfed by host macrophages, some explanations regarding the impotence of this usually effective final veto is required. Evidence suggests that apoptotic cells, as well as senescent cells, may lose their surface CD47 or change the cell surface localization pattern of CD47 (12–14), resulting in a dysfunction of “don’t eat me” signaling. However, the mechanism that governs the changes of both cell surface expression level and the pattern of CD47, and how the CD47 pattern change affects the CD47-SIRPα interaction during apoptosis is incompletely understood.
In the present study, we monitored the kinetics of the cell surface level and the pattern of CD47, and also the CD47-SIRPα interaction following UV-induced cell apoptosis or apoptosis induced by other means. Our results showed that cell apoptosis does not decrease the CD47 level on the cell surface but alters the cell surface pattern of CD47 from ‘punctate’ clusters into diffused distribution, which dramatically decreases the avidity of CD47-mediated cell binding to SIRPα and incapacitates SIRPα-mediated inhibitory signaling in macrophages. Our data further suggest that dispersion of surface CD47 is related to apoptosis-induced disruption of lipid rafts in the plasma membrane.
Human colonic epithelial cell HT-29, human mammary gland epithelial cells T47D, MCF7, MDA435 and HS578T, and primary cultured human foreskin fibroblasts (HFF-1) (all from American Type Culture Collection (ATCC)) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Human microvascular endothelial cells (HMEC-1) initially primarily cultured by Dr. E.W. Ades (Centers for Disease Control and Prevention, Atlanta) (15) were maintained in MCDB 131 medium with 10 mM/L L-glutamine, 10 ng/ml mouse epidermal growth factor (mEGF, BD Biosciences), 1 µg/ml hydrocortisone (Sigma) and 10% FBS and were used within 15 passages (16). Human acute monocytic leukemia cells THP-1 (from ATCC) were cultured in RPMI 1640 supplemented with 50µM β-mercaptoethanol and 10% FBS (16). Human peripheral polymorphonuclear leukocytes (PMN) were isolated from whole blood of healthy volunteers (16–18). Murine splenocytes were isolated from spleens of healthy C57BL6 mice. Mouse mAbs against the human CD47 extracellular domain, B6H12.2 and PF3.1, and the human SIRPα extracellular domain, SE5A5 and SE7C2, and a rabbit polyclonal antibody against the SIRPα extracellular domain (anti-SIRPα.ex) were used previously (19, 20). The IgG Fab fragment of PF3.1 was prepared by cleavage with papain protease using a Fab Preparation Kit (Thermo Scientific, Rockford, IL). Mouse mAb clone 4G10 (EMD Millipore) was used to detect protein tyrosine phosphorylation and anti-SHP-1 antibody was purchased from Santa Cruz Biotechnology. Rat mAb against mouse CD47, miap301, was purchased from BD Biosciences. The anti-actin antibody and FITC-conjugated cholera toxin B subunit (CT-B) were obtained from Sigma. A recombinant fusion protein containing the entire extracellular domain of human SIRPα and rabbit Fc, SIRPα.ex-Fc, was generated and used previously (21, 22). Another recombinant fusion protein containing the entire extracellular domain of human CD47 and alkaline phosphatase (AP), CD47-AP, which functionally ligates SIRPα, was generated in the lab (17, 22, 23). Octylglucoside (OG) was from Millipore. Other essential reagents including Methyl-β-cyclodextrin (MβCD), lipopolysaccharide (LPS), lectin from Psophocarpus tetragonolobus, PMA, protease inhibitors and cocktails were purchased from Sigma.
HT-29, T47D, HS578T, MCF7, HFF-1 and HMEC-1 were induced apoptosis by UV irradiation in 35 mm Petri dishes in the absence of cell culture medium. Specifically, after removal of culture medium, the dishes were placed under a UV lamp/cross-linker (model XL-1000, Spectronics Corporation, New York) and irradiated for various time periods, all within 200 seconds, to achieve induction of cell apoptosis starting at 2–4 h post-UV, but not necrosis or instant death (the total energy strength for different cell types are listed in Table 1). After UV irradiation, the culture medium was added back to the dishes and the cells were cultured for various time periods until analysis. Cell apoptosis was assessed by labeling with FITC-conjugated Annexin V (Sigma) for cell surface PS and YO-PRO-1 iodide (Life Technologies) for cell nuclei. Cell necrosis or dead cells were detected by staining with propidium iodide (PI). Cell apoptosis was also assessed by Western blot to detect cleavage of Poly ADP-Ribose Polymerase (PARP) using a rabbit polyclonal anti-PARP antibody (Roche).
Cell adhesion to immobilized recombinant proteins was performed according to previously described methods (22, 24) with modifications. Briefly, the purified SIRPα.ex-Fc (10 µg/ml), lectin (50 µg/ml), control BSA (5%), or anti-CD47 mAb PF3.1 (10 µg/ml) in HBSS were immobilized in 96-well microtiter plates by incubation for 2 h at 25°C or overnight at 4°C. After blocking with HBSS containing 2% BSA and 5% normal goat serum (1 h, 25°C), the wells were incubated with suspensions of cells (1–5 × 105) for 30 min (25°C) in the presence or absence of inhibitory antibodies or peptides. After incubation, the wells were gently washed (3x) using a 27-gauge needle connected to vacuum followed by analysis for cell adhesion microscopically. In some experiments, cells were labeled with a fluorescent dye 2”,7”-bis (carboxyethyl)-5,6- carboxyfluorescein (BCECF) prior to incubation with immobilized proteins. Cell adhesion was quantified using a fluorescence plate reader (Perkin Elmer) by reading the fluorescence intensity of the total load versus the after wash.
The binding avidity of cells to immobilized SIRPα was analyzed using the SPR technology. Briefly, a new CM5 chip (Biacore, GE Bioscience) with four binding surfaces/channels was placed into the Biacore T100 system (Biacore, GE Healthcare) followed by binding surface activation using 400 mM 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC) and 100 mM N-hydroxysuccinimide (NHS) according to a standard protocol in Biacore Sensor Surface Handbook (GE Healthcare). After activation, one channel that was used as the reference and three were for binding experiments. To immobilize SIRPα.ex-Fc on the chip, purified SIRPα.ex-Fc (100 µg/ml) in NaAc/HAc buffer (pH 5.0) was injected into activated channels. The final response unit (RU) change with immobilized protein was about 4500. After protein coating, 1 M ethanolamine was then injected into the channels to neutralize residual activated carboxyl groups. The reference channel was directly blocked with ethanolamine. For cell adhesion experiments, cells suspended in HBSS (1 × 107 cells/ml) were injected into the channels at a flow rate of 3µl/min, which was lasted at least 500 sec. Continual flow of HBSS at the same rare was given after the cell injection. To determinate cell binding, the channels were washed/regenerated by running 5 M LiCl in HEPES, pH 7.3, for 15 min at a flow rate of 3µl/min.
CD47 purification from various cells was described previously (17). Briefly, non-apoptotic cells and cells that were 6 h post-UV (> 5 × 108 each) were lysed in a cold buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM MgCl2, 1% NP-40, 1% TritonX-100, 0.5% sodium deoxycholate, protease inhibitor cocktail (Sigma) and 1 mM PMSF). After pre-clearing by a control IgG-conjugated Sepharose, the lysates were incubated with Sepharose conjugated with PF3.1 (2 mg of IgG/ml). After incubation, the beads were washed with the same lysis buffer but with the detergent changed to 1% octylglucoside. After washing, CD47 was eluted in a buffer containing 50 mM TEA, pH 10.5, 1% octylglucoside and 100 mM NaCl, followed by neutralization. To assay CD47 binding to SIRPα.ex-Fc, purified CD47 (~100 µg/ml in buffer containing 1% OG) was diluted 10-fold in HBSS in ELISA plate wells and allowed protein binding for 2 h (25°C). The wells were then blocked, and incubated with SIRPα.ex-Fc (5 µg/ml) in HBSS for 30 min in the presence or the absence of anti-CD47 or anti-SIRPα mAb (5µg/ml). After washing, SIRPα.ex binding was detected using a peroxidase-conjugated goat anti-rabbit Fc antibody and the substrate 2,2’-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (ABTS).
Trypsinized cells were blocked with 5% normal goat serum followed by cell surface labeling of CD47 with antibody B6H12.2 or PF3.1 (30 min). Cells labeled with mouse normal IgG were used and controls. To label cells growing in the culture dishes, cells were briefly treated with 1 mM EDTA to dissemble the intercellular junctions (21) before blocking and CD47 labeling. After primary antibody labeling, cells were washed and labeled with AlexaFluor–conjugated anti-mouse secondary antibody (Invitrogen) followed by FACS, fluorescence microscopy or confocal microscopy in the presence of anti-fade reagent (Molecular Probes). CD47 granularities and particle fluorescence intensities were analyzed by NIH software Image J and Nikon camera software NIS-Elements B.R. 4.20.00. In some experiments, CD47-expressing cells were also incubated with SIRPα.ex-Fc (5 µg/ml) in HBSS (30 min, 4°C). After washing, SIRPα.ex-Fc binding was detected with an AlexaFluor–conjugated anti-rabbit Fc antibody. Plasma membrane lipid rafts were labeled with FITC–conjugated cholera toxin B (30 min, 4°C).
Cells were lysed by an ice-cold lysis buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP-40, 1% TritonX-100, 0.5% sodium deoxycholate, a cocktail of protease inhibitors (Sigma) and 1 mM PMSF. Clear cell lysates were obtained by centrifugation at 14,000 rpm (10 min). For WB of CD47, cell lysates were supplemented with NaIO4 (10 mM, to enhance CD47 detection) prior to mixing with non-reducing SDS-PAGE sample buffer and heating (5 min, 60°C). After transferring proteins onto nitrocellulose, the membrane was blocked with 5% nonfat milk followed by detection using anti-CD47 mAb B6H12.2 (2 µg/ml, 1 h, 25°C). For WB of actin and PARP, SDS-PAGE was performed under reducing conditions. For WB of SIRPα, anti-SIRPα.ex was used. For IP of SIRPα, the THP-1 macrophage lysates were incubated with 1–2 µg anti-SIRPα.ex and protein A-conjugated Sepharose for 4 h (4°C). After washing, the beads were heated in 1x SDS-PAGE sample buffer to release proteins, and samples of pre-bound lysates, post-bound lysates and IP were separated by SDS-PAGE followed by WB analyses for SIRPα. To detect SIRPα phosphorylation, THP-1 macrophages were treated with freshly prepared pervanadate (2 mM, 3 min, 37°C) prior to cell lysis and IP. The tyrosine phosphorylated SIRPα was WB detected using mAb 4G10, and SIRPα associated SHP-1 was detected using an anti-SHP-1 antibody.
Cultured THP-1 cells were induced to differentiate into phagocytic macrophages using PMA (100nM) for 3 days (37°C). After induction, THP-1 changed morphology and adhered to the culture dishes. To assay THP-1 macrophage phagocytosis, healthy cultured cells (non-phagocytosis control) and cells that were induced apoptosis by UV or other methods were loaded with carboxyfluorescein diacetate succinimidyl ester (CFSE) or labeled with YO-PRO-1 iodide (for apoptotic cells) prior to incubation with THP-1 phagocytes at 37°C. After 30 min, the dishes were gently washed to remove floating cells, and THP-1 phagocytes were stained with a PE-conjugated anti-CD11b antibody followed by fluorescence microscopy. To determine phagocytosis, macrophages (CD11b+, red) in 5–6 view fields with a total cell number > 200 were analyzed for ingestion of target cells that were stained with CSFE (green) or YO-PRO-1 (apoptotic marker, green). Successful phagocytosis were those macrophages in which green cells were ingested or partially ingested hence showing green staining, while macrophages without green staining within the cells were not considered phagocytic. Alternatively, a fluorescence plate reader (Perkin Elmer, Waltham, MA) was used to assess phagocytosis and in those experiments only target cells were labeled green. After co-incubation with THP-1 macrophages (without labeling), the wells were washed to remove cells that were not phagocytized or ‘stuck’ to phagocytes, followed by reading fluorescence intensities that reflected cells captured by macrophages. This method is normally used in combination with microscopic determination. To detect cytokines produced during phagocytosis, THP-1 macrophages (2 × 105) were incubated with healthy cells or cells that were treated by UV or other methods (1 × 105 each cell sample). At various time points, the cell-free media from the THP-1-target cell co-incubation were collected and used for ELISA detection of IL-10, IL-6 and TNFα using capture and detection antibodies (27, 28).
HT-29 cells were stained with FITC-conjugated cholera toxin B subunit to identify lipid rafts (29). To disrupt the lipid rafts on cell surfaces, cells were incubated with 15 mM MβCD for 20 min at 37°C as described previously (30).
All images of immunofluorescence labeling, FACS and Western blots represent the results of at least three independent experiments. Data are presented as the mean ± SD or SEM for three or more independent experiments. Differences were considered statistically signiﬁcant at P<0.05, as analyzed using Student’s t-tests for paired samples and one-way ANOVA for k >2 samples.
It is generally believed that the interaction between the phagocyte surface SIRPα and the self-recognition marker protein, CD47, controls the phagocytosis of self-cells by phagocytes such as macrophages, and that during cell aging or apoptosis, cells may lose surface CD47 and/or the interaction with SIRPα in order to be phagocytized (5, 6). To test whether cells lose their surface CD47 when apoptosis occurs, we examined the CD47 expression level on multiple types of cells prior to and after induction of apoptosis. For the experiments, human colonic epithelial cell HT-29, breast cancer cells T47D and MCF7, and other types of cells were exposed to UV (energy levels of UV used for different cells were listed in Table 1) or other ways to induce apoptosis. As shown in Figure 1A, the UV-treated HT-29, MCF7 and T47D cells started to display apoptosis at about 4 h post-UV, indicated by the cell surface labeling of PS by a FITC-conjugated Annexin V, which started to show positive staining at 4 h and grew stronger thereafter. Cell apoptosis was also confirmed by detecting the cleavage of poly ADP-ribose polymerase (PARP), a DNA repairing polymerase that is cleaved under apoptosis (31) (Figure 1B). As cell necrosis (secondary necrosis) generally occurs following apoptosis if the apoptotic cells were not immediately phagocytically removed (32–34), co-staining with YO-PRO-1 iodide (green) and PI (red) was applied to differentiate cells that were apoptotic (YO-PRO-1 positive but PI negative) or necrotic (PI/YO-PRO-1 double positive). As shown in Table 1, cells began apoptosis 2 h after UV irradiation and, at 8 h, all cells were positive of YO-PRO-1 staining. Cell necrosis started to appear 6 h post-UV, and gradually increased with the time extension. At 18 h post-UV, all cells became necrotic and positively stained by PI. Therefore, in our analyses of apoptotic cells, cells within 8 h post-UV were generally used and YO-PRO-1/PI staining distinguishing apoptosis from necrosis was included in experiments.
As shown in Figure 1 (A-C), despite that cells had undergone apoptosis, cell surface labeling of CD47 with mAbs against the CD47 extracellular domain (B6H12.2 and PF3.1) showed that CD47 was continuously expressed on the cell surface without significant reduction until later time points when cells were necrotic (18 h post-UV, Figure 1C). Detailed flow cytometric analyses of cells after UV irradiation (Figure 1D) showed that the cells can be separated into two populations, the FSChigh large cells (gated as R1), and the FSClow small cells (gated as R2). The non-apoptotic cells (0–2 h) were all FSChigh large cells in R1 region, whereas the PI-positive necrotic cells (18 h post-UV) were FSClow small cell particles. As shown, at 4–8 h post-UV, although cells were apoptotic (positive YO-PRO-1 staining), they generally maintained the cell sizes (FSChigh, R1) and the surface level CD47. In contrast, when cells became necrotic, they significantly shifted towards the low values of FSC, becoming FSClow cells with smaller size (R2 region). At 18 h post-UV, nearly all cells were FSClow and also had significant reduced CD47 on the surface (~40% reduction).
Furthermore, we observed that membrane blebs emerged around the cell bodies when cells started apoptosis and then necrosis. This phenomenon, which has been reported previously, likely reflected vigorous dynamics and/or destabilization of the plasma membrane during apoptosis, resulting in partial plasma membrane dissociation from the underlying cytoskeleton cortex (35–38). For cell apoptosis especially at early stages (4–6 h post-UV), the membrane blebs were small relative to the cell body, transient and deficit of CD47 (Figure 1E, marked by white arrowheads). However, when apoptosis progressed to necrosis, large membrane blebs formed (relative to the remnant cell body) and on which CD47 staining was positive (Figure 1E, marked by white arrowheads). Shedding of these large membrane blebs might serve as a reason for loss of both the CD47 molecules and the cell size for necrotic cells.
Similar results were obtained when studying apoptosis of other cell types, such as human fibroblasts HFF-1 and human microvascular endothelial cells HMEC-1 (both non-transformed cells), the breast cancer cells MDA435, MDA468, MDA231 and HS568A, murine cell lines B16 and MC38, murine freshly isolated splenocytes (> 90% being lymphocytes), and freshly isolated human neutrophils (PMN). As shown in Figure 2, A and B, and Table 1, UV irradiation induced apoptosis of HFF-1, HMEC-1 and various breast cancer cells, indicated by the positive YO-PRO-1 labeling and the cleavage of PARP, while without noticeably affecting CD47 expression on the cell surface as determined by anti-CD47 labeling, as well as immunoblot analysis. Apoptosis in murine splenocytes was induced by treatment with H2O2 (5 mM) and this apoptosis was not associated with cell surface CD47 reduction as assessed by a murine CD47 extracellular domain-specific antibody (mAb miap301) (Figure 2C). Similar results of no CD47 reduction were also observed in B16 and MC38 cells when these cells were induced apoptosis by UV irradiation (not shown). Different from other cells, freshly isolated human PMN without stimulation reportedly express only low level CD47 on the cell surface but store majority CD47 in intracellular specific granules (18); chemoattractant stimulation induces PMN degranulation and hence mobilizes CD47 to the cell surface (18). In this study, we treated PMN with the chemoattractant fMLF for 15 min to increase cell surface CD47 (Figure 2D), prior to inducing apoptosis by incubating the PMN on ice in the presence of 20 mM EDTA (EDTA also prevents PMN aggregation). As shown in Figure 2D, PMN that were incubated on ice overnight (> 14 h) were apoptotic and displayed positive of YO-PRO-1 staining; however, such change of PMN also did not result in reduction of CD47 on the cell surface. In conclusion, our results indicate that cell apoptosis, per se, does not lead to loss of CD47 on the cell surface. However, when apoptosis proceeds to secondary necrosis the cell sizes are reduced and as such CD47 proteins remained on the remnant cell surfaces are coordinately decreased.
Given that CD47 serves as a ‘self’-marker and its interactions with SIRPα on phagocytes inhibit macrophage phagocytosis, we asked whether apoptotic cells with CD47 expression could be phagocytized by macrophages. For these experiments, HT-29 cells were loaded with a fluorescent tracer CFSE prior to the induction of apoptosis by UV irradiation. At various time points post-UV, HT29 cells were incubated with THP-1 differentiated macrophages to allow phagocytosis. As shown in Figure 3, A and B, THP-1 macrophages (labeled with anti-CD11b antibody, red) displayed aggressive phagocytosis towards HT-29 cells 4 h post-UV despite the fact ample CD47 remained on the target cells. Apoptotic cells (post-UV 4–8 h) were rapidly grabbed and engulfed by macrophages and thus formed a wide area of orange labeling in merged images. THP-1 macrophages also aggressively ingested necrotic cells but completely avoided non-apoptotic cells. We also assayed cytokines produced by THP-1 macrophages during phagocytosis. As shown in Figure 3C, cell-free supernatants collected from the THP1 macrophage-HT29 incubation (37°C, 3 h) confirmed that phagocytosis of the CD47-expressing apoptotic HT29 cells (4–8 h post-UV) was associated with induction of the anti-inflammatory cytokine IL-10 while suppression of IL-6 and TNFα. Conversely, phagocytosis of necrotic cells collected at 18 h post-UV was associated with significant induction of proinflammatory IL-6 and TNFα. Similar results were obtained when testing THP-1 phagocytosis of apoptotic and necrotic HFF-1, HMEC-1, and various breast cancer cells (data not shown). Together, these results suggest that, while CD47 expression remains on apoptotic cells, it does not prevent macrophage clearance, and that phagocytosis of CD47-expressing apoptotic cells is a physiologically relevant, immunosuppressive event, whereas phagocytosis of necrotic cells is associated with proinflammatory conditions.
Although CD47 molecules were on the surface of apoptotic cells, they were incapable of ligating SIRPα and inducing the SIRPα-mediated inhibitory signaling in macrophages. Addition of a SIRPα-binding CD47 extracellular domain (CD47-AP) and a SIRPα-ligating mAb (SE7C2), both capable of triggering SIRPα-mediated inhibitory signaling, inhibited macrophage phagocytosis of apoptotic cells (Figure 3B). As shown in Figure 3D, direct incubation of apoptotic cells with THP-1 macrophages failed to trigger SIRPα ITIM tyrosine phosphorylation (SIRPαpY) and/or recruitment of SHP-1. The same manner of co-incubating healthy, non-apoptotic cells with macrophages resulted in SIRPα phosphorylation and SIRPα association with SHP-1, suggestive of the successful induction of SIRPα-mediated signaling in macrophages. Consistent with their inhibition in phagocytosis, addition of CD47-AP and SE7C2 into the THP-1-apoptotic cell co-incubation led to SIRPα phosphorylation and SHP-1 association.
Since CD47 molecules on the apoptotic cells do not prevent macrophage phagocytosis, we speculate that CD47 on apoptotic cells might not interact with SIRPα on macrophages. To test this possibility, we first evaluated whether CD47 molecules in apoptotic cells were intact and could bind to SIRPα. CD47 was affinity-purified from various, non-apoptotic and apoptotic cells, and then tested for binding to SIRPα using a recombinant SIRPα extracellular domain fusion protein (SIRPα.ex-Fc). As shown in Figure 4, CD47 proteins purified from normal cultured, non-apoptotic cells and UV-induced apoptotic cells displayed equal binding to SIRPα.ex-Fc. The binding interactions between the CD47 and SIRPα extracellular domains were confirmed by including the extracellular domain-specific, inhibitory mAbs for CD47 (B6H12.2) or SIRPα (SE5A5) that abrogated the binding.
Although the above assays using purified CD47 suggested that the binding capacity of CD47 to SIRPα was not affected by apoptosis, the cell adhesion assays suggested otherwise. As shown in Figure 5, A-C, stationary incubation of immobilized SIRPα (SIRPα.ex-Fc-coated wells) with healthy cultured, non-apoptotic cells resulted in effective, CD47-SIRPα interaction -mediated cell adhesion. In support that the cell adhesion was specifically mediated by CD47-SIRPα interaction, the extracellular domain inhibitory mAbs against CD47 (B6H12.2) and SIRPα (SE5A5) were applied and these mAbs completely blocked healthy cells from adhesion to immobilized SIRPα.ex-Fc. However, the same manner of SIRPα.ex-Fc-coated wells completely failed to mediate adhesion of apoptotic cells (Figure 5, A-C). Interestingly, apoptosis did not diminish cell adhesion mediated by lectin or anti-CD47 antibody (Figure 5D). We further analyzed the differences of apoptotic cells and non-apoptotic cells in adhesion to immobilized SIRPα by SPR. For these experiments, healthy cultured cells and apoptotic cells of the same type were set to flow through CM5 chip surfaces coated with SIRPα.ex-Fc. As shown in Figure 5E, non-apoptotic cells exhibited rapid attachment/retention to the chip and demonstrated effective, CD47-mediated cell adhesion to SIRPα. In contrast, apoptotic cells displayed no evident retention/attachment to the SIRPα.ex-coated chip under flow. These results conclude that apoptotic cells lost the capability to mediate strong (high avidity) cell surface binding to immobilized SIRPα via CD47, resulting in being washed away by the flow shear force.
Multiple additional types of cells were tested in cell adhesion assays and all yielded the similar results indicating that apoptosis led to loss of high avidity, CD47-mediated binding to SIRPα (Figure 5F). However, despite of failure to mediate cell adhesion, incubation of SIRPα.ex-Fc in solution with apoptotic cells resulted in SIRPα-CD47 binding (Figure 5G) in a manner similar to that of anti-CD47 antibody labeling. This result again indicates that failure of apoptotic cells to mediate the firm adhesion to the immobilized SIRPα is not due to depletion of CD47 on apoptotic cells or the change of CD47 extracellular domain structure, but the protein binding avidity.
Our previous study found that SIRPα forms clusters on the macrophage surface and such formation of SIRPα plays a key role in mediating high avidity cell interactions with CD47 (16). In this study, we asked whether CD47 also clusters on the cell surface and if such distribution is essential for SIRPα ligation and triggering SIRPα-based inhibitory signaling. We compared the cell surface distribution patterns of CD47 on non-apoptotic cells and apoptotic cells. To avoid possible antibody cross-linking, we used Fab fragments of both anti-CD47 and fluorophore-conjugated secondary antibodies to label cell surface CD47. As shown in Figure 6A, immunofluorescence microscopy revealed that CD47 on non-apoptotic HT29 cells was distributed in an uneven, punctate pattern, suggestive of cluster formation (Figure 6A, left panel, arrowheads). In contrast, CD47 on apoptotic cell surfaces were diffused resulting in relatively weak visual effects when microscopically analyzing the staining images (Figure 6A, right panel), despite that the actual level of CD47 was not significantly changed. Analyses of multiple other types of cells found the same CD47 distribution change before and after apoptosis (Figure 6B). Using NIH Image J and Nikon camera software (NIS-Elements B.R. 4.20.00) to analyze cell surface CD47 distribution and CD47 clusters/granules revealed remarkable differences between normal non-apoptotic cells and apoptotic cells. As shown in Figure 6C, Image J analysis showed that both the number and the sizes of CD47 granules/clusters on the cell surfaces were significantly decreased after apoptosis. In particular, CD47 granules/clusters larger than 50 pixel2 were distributed on non-apoptotic cells but deficit on apoptotic cells. In contrast, apoptotic cells were chiefly distributed with small CD47 ‘dusts’ of sizes < 20 pixel2. Figure 6D shows that on non-apoptotic cells, cluster-formed CD47 emitted high intensity, scattered peaks of fluorescence, whereas on apoptotic cells diffused CD47 emitted broad but weak fluorescence. As depicted in Figure 6E, this marked change of CD47 distribution on the cell surface before and after apoptosis correlates with the alteration of CD47 binding avidity to SIRPα in adhesion assays, suggesting that while individual CD47 is capable of binding, the overall low binding avidity of CD47 on apoptotic cells is insufficient to hold the cell in place. For our previous studies demonstrated that SIRPα clusters on macrophages, thus it is essential that CD47 also forms clusters in order to mediate high avidity cell surface interactions and strong inhibitory signal transduction effectively suppressing macrophage phagocytosis. Apoptosis-associated dispersal of CD47 significantly reduces the avidity of SIRPα ligation and hence facilitates macrophage phagocytosis.
It has been widely reported that clustering of cell membrane receptors requires the location of receptors in lipid rafts in the plasma membrane and such lipid raft-supported receptor clustering plays an essential for maintaining the structure and function of receptors (39–41). To test whether the clustering of CD47 on healthy cell surfaces is associated with lipid rafts and whether the dynamic distribution of lipid rafts is the mechanism underlying CD47 aggregation and/or disassembly, we performed several experiments. First, we double stained cell surface CD47 and lipid rafts using anti-CD47 Fab and FITC-conjugated cholera toxin B subunit, respectively, and studied their distribution patterns and potential co-localization. As shown Figure 7 (A-B), confocal microscopy taking images of HT29 cells at different focus layers (panels A & B) revealed that CD47 is largely co-localized with lipid rafts marked by cholera toxin B subunit in large punctate particles on healthy, non-apoptotic cell surface. This suggests that CD47 clustering is likely due to the assembly of CD47 within lipid rafts. Next, we disrupted lipid rafts with the cholesterol depletion agent MβCD and then examined the cell surface pattern of CD47. As can be seen, treating healthy HT-29 cells with MβCD (15 mM) led to a diffused staining pattern of FITC-conjugated cholera toxin B subunit, suggestive of lipid raft disruption in the plasma membrane (Figure 7, A-B). Clustering of CD47 on the cell surface also disappeared following MβCD treatment and CD47 staining displayed a diffuse pattern. This result therefore suggests that clustering of CD47 on the cell surface may be supported by lipid rafts. Analyses of the cell surface CD47 labeling using Image J and Nikon camera image software revealed that MβCD treatment caused CD47 distribution changes in a way similar to that of apoptotic cells, mainly diminishing large CD47 clusters that emitted high intensity of fluorescence (data no shown).
We next performed two additional experiments to confirm that non-apoptotic cells treated with MβCD would lose the high avidity CD47 binding to SIRPα and would promote macrophage phagocytosis. Cell adhesion assays were performed and cells with and without MβCD treatment were tested for adhesion to immobilized SIRPα.ex-Fc. As shown in Figure 7C, in contrast to HT-29, MCF7, T47D without MβCD treatment readily bound to SIRPα.ex-Fc, the same cells treated with MβCD completely lost stable cell adhesion to immobilized SIRPα.ex-Fc, indicating that raft disruption abrogated the high avidity cell surface CD47 binding. Phagocytosis assays (Figure 7D) by incubating MβCD-treated cells with THP-1 macrophages also showed enhanced phagocytosis of MβCD-treated cells compared to non-treated cells, and the presence of CD47-AP and anti- SIRPα mAb SE7C2, both ligating SIRPα, significantly inhibited THP-1 macrophage phagocytosis towards MβCD-treated cells.
As a marker protein for self-recognition, CD47 plays a critical role in protecting host cells/tissues from being attacked by host phagocytes. Accumulating evidence has shown that ligation of macrophage SIRPα by CD47 expressed on the encountered cells prohibits macrophage phagocytosis, whereas failure of SIRPα engagement by CD47, or deficiency of SIRPα ITIM-mediated signaling, promotes macrophage engulfment of the host cell (1, 3, 42–45). Through over-expressing CD47 on the surface, certain types of cancer cells can escape from macrophage-mediated immune clearance (46, 47). It has been reported that cells likely lose their surface CD47 when aging or undergoing apoptosis, and this change results in abrogation of the CD47-SIRPα-mediated inhibitory signaling and hence facilitates macrophage phagocytic clearance.
However, contrary to previous reports, we demonstrate in the present study that the level of CD47 on the cell surface is not reduced during cell apoptosis; instead, the localization pattern of CD47 on the cell surface is drastically altered. On non-apoptotic cells, CD47 is largely clustered into numerous punctates on the cell surface. After apoptosis, the punctate pattern of CD47 disappears and replaced with a diffuse CD47 distribution on the cell surface. The shift of cell surface CD47 from ‘clustering’ punctates into diffuse distribution is a critical factor for the loss of CD47 capacity to suppress macrophage engulfment through ligating SIRPα. Our cell adhesion data derived under both static and flow conditions showed that, although apoptotic cells maintain the expression of CD47 on the cell surface, these CD47 proteins have a low binding avidity and are incapable of mediating firm cell adhesion to immobilized SIRPα. On the contrary, the CD47 molecules expressed on healthy, non-apoptotic cells display a high binding avidity and strongly hold cells to immobilized SIRPα even under a flow shear force (SPR technique). Consistent with these results, apoptotic cells showed failure to elicit CD47-SIRPα interaction–mediated inhibitory signaling indicated by the SIRPα cytoplasmic ITIM phosphorylation and association with SHP-1 when co-incubated with macrophages, whereas macrophages co-incubating with non-apoptotic cells resulted in SIRPα signaling (Figure 3D). Moreover, we show that loss of CD47 binding avidity on apoptotic cells is not due to biochemical modifications or defect of CD47 molecules, for purified CD47 from either healthy or apoptotic cells displayed the same binding activities towards SIRPα. As SIRPα also present as clusters on macrophages (16), it only makes sense that both these counter-receptors cluster on the cell surfaces in order to mediate high-density ligations effectively driving a strong signal transduction leading to inhibition of macrophage phagocytosis. In addition, a previous study by Burger et al. (13) showed that CD47 on erythrocytes undergo a conformational change that alters CD47 binding avidity to TSP-1. Those interesting results together with our results in the present study reveal a more complex role for CD47 in cell apoptosis and/or senescence, as well as the determination of phagocytic clearance of apoptotic or senescent cells by phagocytes.
Interestingly, we observed that while apoptosis does not lead to CD47 loss, the subsequent cell necrosis following apoptosis (‘secondary necrosis’) is associated with reduction of CD47. However, necrosis is also associated with significant reduction of the cell sizes. Therefore, decreases in CD47 on necrotic cells could be the result of reduction of the cell sizes and the loss of plasma membrane, even though CD47 molecules / per membrane area are not reduced. Our data also suggest that the loss of cell sizes during necrosis is likely due to the progressive membrane destabilization resulting in formation of large membrane blebs that eventually detach from the main cell bodies. Indeed, membrane blebs instantly appear when cells start apoptosis, and this result is consistent with observations by others (38, 48). As shown in our study, formation of blebs around the cell body appeared at 4 h post-UV, the same time point when other apoptotic signs such as surface PS and PARP-1 cleavage became apparent. To us, the appearance of these blebs becomes another indicator, which is easily visualized, for detecting apoptosis in addition to PS and PARP. As suggested by Charras et al. (37), membrane blebs are a result of membrane dynamics during which the plasma membrane transiently dissociates/ruptures from the underlying cytoskeleton cortex due to non-uniform intracellular hydrostatic pressure. Formation of membrane blebs is not apoptosis-specific. Certain active cellular processes, e.g. cell migration and cell proliferation, or when the cytoskeleton cortex integrity is compromised (e.g. filamin depletion), are also associated with cell blebbing (49–51). We observed that from apoptosis to the subsequent secondary necrosis, plasma membrane blebs are increased in size (from small to large blebs), suggestive of increased levels of plasma membrane destabilization and membrane dissociation from the underlying cortex. We consider that the shrinkage of cell sizes during necrosis is likely due to progressive damage/diminishment of the membrane cortex and detachment of large membrane blebs from the cell body. Along these dramatic cellular changes, CD47 at the stage of apoptosis shows remaining on the cell body without distribution to blebs, in a way very similar to the pattern of actin as reported by Charras et al (37), suggesting that CD47 may through its cytoplasmic domain connect to the underlying cytoskeleton cortex. Not only are they CD47-negative, the blebs at the stage of apoptosis are small and dynamic (continuously appearing and disappearing). In contrast, at the stage of necrosis membrane blebs are large and tend to detach from the cell body, and these blebs are CD47-positive. Therefore, the decrease in CD47 on necrotic cells could be the result of shrinkage of the cell size and loss of the plasma membrane, even though CD47 molecules per membrane area may not reduce. This explains that even under necrosis the CD47 level on the remnant cell body is never diminished completely. In our study, we specially utilized YO-PRO-1 and PI to distinguish apoptosis from necrosis. To confirm our results, we also compared the cytokines released by macrophages during their phagocytosis of apoptotic or necrotic cells. As shown by our data, macrophage phagocytosis of CD47-expressing apoptotic cells is accompanied with release of IL-10, while phagocytosis of those CD47-reduced necrotic cells is associated with IL-6 and TNFα production.
Our data further revealed that the functional clustering of CD47 on the cell surface is supported by lipid rafts in the plasma membranes. Immunofluorescence staining showed that cell surface CD47 is largely co-localized with lipid rafts, and disruption of lipid rafts also leads to diffusion of CD47. Clustering of CD47 within lipid rafts is in agreement with the previous finding by McDonald et al. (52), who showed that CD47 is associated with cholesterol-rich lipid rafts on human ovarian carcinoma cells, and also recruits αvβ3 integrin and its associated signaling molecules into these microdomains in the plasma membranes. An original study by Green et al. (53) showed that cholesterol directly binds to the multiple membrane-spanning domain of CD47 and is an essential component of the αvβ3/CD47/G protein-signaling complex.
In conclusion, this study demonstrates that cell apoptosis does not reduce CD47 expression on the cell surface but disperses CD47 from clustering in lipid rafts into the plasma membrane, and that this change of CD47 distribution on the cell surface results in loss of the high binding avidity of CD47 to SIRPα and renders CD47 incapable to trigger SIRPα-mediated inhibitory signaling in macrophages and hence facilitates phagocytic clearance. For healthy cells in tissues, the clustering of CD47 on the cell surface is a functional form supported by lipid rafts in plasma membranes and this form of CD47 enables effective ligation to SIRPα and triggers SIRPα-mediated inhibition in macrophages.
The authors thank Dr. Jill Leslie Littrell (Georgia State University, Atlanta, GA) for critical reading and constructive discussion of the manuscript.
This work was supported in part by Research Scholar Grant from the American Cancer Society (to Y.L.), National Institutes of Health Grant AI106839 (to Y.L.), a fellowship from the American Heart Association (to Z.B.), and a fellowship from the Chinese Scholarship Council (to Z.L.).