CD47 and the 19 kDa Interacting Protein-3 (BNIP3) in T Cell Apoptosis*

CD47 is a surface receptor that induces either coactivation or apoptosis in lymphocytes, depending on the ligand(s) bound. Interestingly, the apoptotic pathway is independent of caspase activation and cytochrome c release and is accompanied by early mitochondrial dysfunction with suppression of mitochondrial membrane potential (Δψm). Using CD47 as bait in a yeast two-hybrid system, we identified the Bcl-2 homology 3 (BH3)-only protein 19 kDa interacting protein-3 (BNIP3), a pro-apoptotic member of the Bcl-2 family, as a novel partner. Interaction between CD47 and the BH3-only protein was confirmed by immunoprecipitation analysis, and CD47-induced apoptosis was inhibited by attenuating BNIP3 expression with antisense oligonucleotides. Finally, we showed that the C-terminal domain of thrombospondin-1 (TSP-1), but not signal-regulatory protein (SIRPα1), is the ligand for CD47 involved in inducing cell death. Immunofluorescence analysis of CD47 and BNIP3 revealed a partial colocalization of both molecules under basal conditions. After T cell stimulation via CD47, BNIP3 translocates to the mitochondria to induce apoptosis. These results show that the BH3-dependent apoptotic pathways, previously shown to be activated by intracellular pro-apoptotic events, can also be turned on by surface receptors. This new pathway results in a fast induction of cell death resembling necrosis, which is likely to play an important role in lymphocyte regulation at inflammatory sites and/or in the vicinity of thrombosis.

Multicellular organisms eliminate excess, damaged or infected cells by stereotypic programs of cell death (PCD). 1 In its classic form, apoptosis is characterized by well defined ultra-structural changes including cell shrinkage, exposure of phosphatidylserine at the outer leaflet of the cytoplasmic membrane, changes in mitochondrial permeability, membrane blebbing, caspases activation, and DNA degradation. Lymphocyte PCD plays an important role in controlling immune responses and occurs both in central and peripheral lymphoid organs. Disturbed PCD may contribute to multiple immune disorders such as cancer and autoimmune and degenerative diseases. Upon signaling, pathways that influence T cell proliferation and survival, CD95/CD95L and tumor necrosis factor receptor pathways have been extensively studied over the past few years (1). However, a number of others T cell surface receptors such as major histocompatibility complex class I and II (2), CD2 (3), CD4 (4), CD45 (5,6), or CD99 (7) can also trigger PCD. In contrast to the former pathways, they have all been described to act independently of any of the known caspases (8), and the molecular mechanisms and the physiological and/or the pathological relevance of these death pathways remains to be established. Among these molecules, recent reports implicate CD47 in the triggering of atypical cell death (9,10).
CD47 (also known as IAP for integrin-associated protein), expressed on all mammalian cells (11), displays an extracellular Ig-like domain with five transmembrane (TM) segments and a short C-terminal cytoplasmic tail. CD47 is associated with ␤ 3 integrins on several cell types. However, on other cell types such as lymphocytes, no association with integrins has been documented. More generally, CD47 has been shown to activate integrins, either through direct interaction or at a distance (11)(12)(13). The two natural ligands currently known for CD47 are thrombospondin-1 (TSP-1), a protein found in extracellular matrix and released in large amounts by platelets upon activation, and the signal-regulatory protein (SIRP␣1), expressed on the surface of macrophages and endothelial and dendritic cells. Moreover, we and others have shown that CD47 can trigger T cell activation and proliferation (14 -16) or induce T cell spreading (17). Therefore an important task has been to determine under which condition(s) CD47 induces T cell death and/or survival. Indeed, in addition to the cytoplasmic proteins Gi (18,19) and proteins linking IAP with cytoskeleton (PLIC) (20), we show in the present study that CD47 associates with the pro-apoptotic molecule 19 kDa interacting protein-3 (BNIP3).
BNIP3 belongs to the Bcl-2 homology 3 (BH3)-only family, a Bcl-2-related family possessing an atypical Bcl-2 homology 3 (BH3) domain, which regulates PCD from mitochondrial sites by selective Bcl-2/Bcl-X L interactions (21,22). BNIP3 family members contain a C-terminal transmembrane domain that is required for their mitochondrial localization, homodimerization, as well as regulation of their pro-apoptotic activities (23). BNIP3-mediated apoptosis has been reported to be independ-* This work was supported by grants from the Institut National de la Santé et de la Recherche Médicale, the Association pour la Recherche sur le Cancer, the Etablissement Français des Greffes, and the Fondation pour la Recherche Médicale. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In the present work, we describe an original mechanism by which the BH3-only protein BNIP3 could be an important mediator of CD47-induced cell death. This pathway would be likely to act at sites where large amounts of soluble TSP-1 are available and could play an important role in T cell death in the vicinity of thrombotic events.
Assay for Apoptosis-Phosphatidylserine exposure and decrease in mitochondrial membrane potential (⌬⌿m) were measured by flow cytometry using a FACScan (BD Biosciences). For phosphatidylserine exposure, cells were double-stained with annexin V-FITC and propidium iodide as described by the manufacturer (Roche Applied Science). The decrease in ⌬⌿m was assessed by incubating Jurkat cells with 40 nM 3,3Ј-dihexyloxacarbocyanine iodide (DiOC 6 ) (Molecular Probes, Eugene, OR) for 30 min at 37°C.
Confocal Microscopy-Jurkat cells, treated or not with the pro-apoptotic 4N1K peptide for 2 h, were plated on polylysine glass slides (Menzel-Glaser, Freiburg, Germany). For mitochondrial staining, Texas Red MitoTracker (Molecular Probes) was added to the medium to a final concentration of 300 nM for 45 min. Cells were then fixed with 3.7% formaldehyde, permeabilized with 0.2% Triton X-100, and blocked for 20 min with phosphate-buffered saline containing bovine serum albumin (1%) and saponine (0.05%). Cells were stained for CD47 using SIRP␣1-Fc and FITC-conjugated mouse anti-human antibody (Dako) at 4°C before fixation, and for BNIP3 using the mouse monoclonal anti-BNIP3 antibody (21) followed by a Cyan V-conjugated rabbit antimouse antibody. Samples were mounted in Mowiol (Calbiochem) and observed with a confocal microscope (Leica TCS-SP, Heidelberg, Germany).
Cytochrome c Release-Cytosolic fractions were prepared by resuspending Jurkat cells (5 ϫ 10 6 ) in 100 l mitochondrial isolation buffer (MIB; 250 mM sucrose, 20 mM HEPES, 10 mM KCl, 2 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, and protease inhibitors). Cells were broken and homogenized by 10 passages through an ice-cold 26-gauge needle. Unlysed cells and nuclei were removed by centrifugation (5 min, 750 ϫ g). The supernatant was spun at 10,000 ϫ g for 30 min at 4°C, and the resulting supernatant was saved as the cytosolic extract. Cytosolic extracts were separated on 15% SDS-PAGE and transferred to polyvinylidene difluoride membrane (Amersham Biosciences). Blots were probed with mouse anti-cytochrome c antibody (65985A) (BD Biosciences) and anti-actin (Chemicon, Temecula, CA), incubated with horseradish peroxydase-conjugated secondary antibody (Dako), and developed using enhanced chemiluminescence (Amersham Biosciences). In a separate set of experiments, Jurkat cells were transfected with GFP-tagged cytochrome c, stimulated via CD47 or Fas 48 h after transfection, then labeled with Annexin V-PE as described by the manufacturer (Roche Applied Science). Cells were observed with a confocal microscope.
Subcellular Fractionation-Jurkat cells (1 ϫ 10 8 ) were stimulated with the mAb against CD47 Ad22 (1 g/ml) for 1 h at 37°C or incubated in medium alone, washed once, and resuspended in chilled isotonic buffer (250 mM sucrose, 20 mM HEPES, 20 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, and protease inhibitors). Cells were broken using a Dounce homogeneizer (80 strokes with a tight-fitting pestle). The cell homogenate was first centrifuged at 1000 ϫ g for 10 min. The postnuclear supernatant was then centrifuged at 10,000 ϫ g for 30 min. The pellet containing the mitochondrial fraction was resuspended in homogeneization buffer and further purified on a discontinuous metrizamide gradient (27). The microsomal fraction contained in the 10,000 ϫ g supernatant was collected by centrifugation (200,000 ϫ g, 2 h). Cell extracts were analyzed by Western blotting using rabbit polyclonal anti-BNIP3 antibody, anti-translocase, a marker of the outer mitochondrial membrane (40 kDa; TOM40) (Interchim), and anti-placental alkaline phosphatase (Dako).
Determination of Caspase 3 Activity-Each assay was performed in quadruplicate with 50 g of protein as described previously (28). Briefly, 4 ϫ 10 6 cells were lysed in 0.2% Triton X-100 isotonic buffer (50 mM HEPES, pH 7.6, 150 mM NaCl, 20 mM EDTA) with freshly added protease inhibitors. Cellular extracts were then incubated in a 96-well plate with 10 mM dithiothreitol and 200 M of Ac-DEVD-pNA as substrate for various times at 37°C. Caspase activity was measured by the release of pNA at 410 nm in either the presence or absence of 1 M Ac-DEVD-CHO, an irreversible inhibitor of caspases. The specific caspase activity represents the Ac-DEVD-CHO inhibitable activity and is expressed as nanomoles of substrate hydrolyzed per min and per mg of protein.
BNIP3 Antisense Oligodeoxynucleotides-Two M of BNIP3 antisense (sequence number 2.04252) or scrambled phosphorothioate oligodeoxynucleotides (Biognostik, Göttingen, Germany) were added to the cells at the start of culture and every time the culture medium was renewed during 72 h.
Yeast Two-hybrid System-The C-terminal 145 amino acids of type 2 CD47 corresponding to the multispan and cytoplasmic portion of the molecule were fused to the LexA DNA-binding domain in pLex-9 (pLex-CD47 multiply membrane spanning (MMS)) and used as a bait for screening a human cDNA library (29,30). Standard techniques were used for nucleic acid manipulations and preparation of DNA constructs. Primers used to generate the DNA fragments were 5Ј-GGAATTCGGT-ATTAAAACACTTAAATATAGATCCGGT-3Ј (sense oligonucleotide, EcoRI site underlined), and 5Ј-CGGGATCCTCAGTTATTCCTAGGAG-GTTG-3Ј (antisense oligonucleotides, BamHI site underlined, mutation introduced in bold). A second round of screening was done using a LexA-lamin plasmid as a negative control. CD47 MMS domain and BNIP3 interactions were assayed for growth on the basis of histidine prototrophy on DOBA Ura Ϫ Trp Ϫ Leu Ϫ His Ϫ plates. Subsequent testing for ␤-galactosidase activity was performed. Deletion mutants for BNIP3 in pcDNA3 vector were previously described (21): BNIP3⌬N (⌬1-49), BNIP3⌬BH3 (⌬104 -119), BNIP3⌬TM2 (⌬164 -194), and BNIP3⌬C (⌬184 -194). cDNAs were cloned into the EcoRI/BamHI sites of the yeast two-hybrid expression vector pActII.

RESULTS
CD47 Induces a Rapid Mitochondrial Dysfunction and Apoptosis-As previously described (9, 10), CD47 antibody Ad22 in a soluble form induces a fast cell death with an early phosphatidylserine exposure (Fig. 1A) and mitochondrial abnormalities. Indeed, using DiOC 6 to measure ⌬⌿m following cell stimulation via CD47 (Fig. 1B), we observed ⌬⌿m drop within 1 h, indicating that the mitochondrial dysfunction occurred as early as cell death. CD47 stimulation was nearly as efficient as CD95 at suppressing ⌬⌿m, as measured after optimal times. Despite early mitochondrial dysfunction following CD47 stimulation, we observed only a late and low release of cytochrome c which was undetectable before 3 h, in marked con-trast with the effects of CD95 stimulation (Fig. 1C). Therefore, cytochrome c release appears to be a secondary event. To further ensure that cytochrome c release occurs after cell death, FIG. 1. CD47 induces a fast mitochondrial dysfunction without cytochrome c release and caspase activation. A, time course of Jurkat cell death induced by the soluble CD47 antibody Ad22 (1 g/ml). Cell death was monitored by Annexin V-FITC binding and uptake of propidium iodide by flow cytometry. B, Jurkat cells, either stimulated with the soluble CD47 antibody Ad22 (1 g/ml, 1 h), with the CD95 (Fas) antibody CH11 (0.5 g/ml, 3 h), or incubated in medium alone, were stained with DiOC 6 and analyzed by flow cytometry. Data represent one of three independent experiments with similar results. C, Jurkat T cells were treated with the CD47 antibody for 0, 1, 3, and 6 h or the CD95 antibody for 0, 1, and 3 h at 37°C. The cytosolic fraction was isolated as described in "Experimental Procedures" and was subjected to Western blot analysis with an anti-cytochrome c antibody. The same membrane was blotted with a mouse anti-actin antibody as control for equal loading and transfer. D, Jurkat cells, transfected with GFP-tagged cytochrome c, were stimulated via CD47 or Fas or incubated in medium for 3 h prior to their labeling with Annexin V-PE to detect apoptosis and analysis by confocal microscopy. This image is representative of multiple fields. Scale bar represents 10 m. E, caspase 3 activity was measured after various times in lysates of Jurkat T cells incubated at 37°C with the CD47 antibody or the CD95 antibody. Ac-DEVD-pNA hydrolysis was determined in quadruplicate at different times in the presence or absence of Ac-DEVD-CHO to measure specific caspase 3 activity. Results are expressed as nanomoles of substrate hydrolyzed per min and per mg of protein and represent the mean of three different determinations. In the same experiment cell death was measured by cytofluorometric analysis of Annexin V-FITC binding and uptake of propidium iodide.
we transfected Jurkat cells with GFP-tagged cytochrome c and observed its mitochondrial release after CD47 stimulation while simultaneously measuring Annexin binding (Fig. 1D). Quite clearly, mitochondrial release was a later event. No role for caspases could be demonstrated by attempts to inhibit PCD with the broad-spectrum caspase inhibitor z-VAD-fmk (result not shown). Moreover, no activation of caspase 3 could be detected while PCD occurred in the same experiment, in marked contrast with controls measuring the effect of a CD95 antibody (Fig. 1E).
CD47 Interacts with the Pro-apoptotic Molecule BNIP3-To determine the apoptotic pathway triggered by CD47 stimulation, we searched for CD47 interacting molecule(s) in the yeast two-hybrid system. We used the CD47 MMS domain fused to Gal4 as a bait for screening a human lymphocyte cDNA library. We ascertained the requirement for the MMS domain of CD47 to induce apoptosis by using a chimeric form of CD47, composed of the extracellular domain of CD47 and the membrane and cytoplasmic domains of CD7 (CD47ec-CD7) (15). This chimera was unable to transmit an apoptotic signal when transfected in the CD47-deficient T-cell line which we previously generated (JIN cells) (26), (Fig. 2A). CD47ec-CD7 mutants and wild-type CD47 transfected cells showed comparable levels of Fas-induced apoptosis (result not shown). Note that the use of MMS domain as bait to perform the two-hybrid screen has already been successively performed (31)(32)(33)(34)(35)(36). In the present screen, five positive clones were selected of an estimated 5 ϫ 10 6 colonies, for their ability to reconstitute a functional Gal4 transcription complex. Among them, we found a clone encoding the pro-apoptotic BH3-only family protein BNIP3 (Fig. 2B). To determine the domain of the BNIP3 molecule involved in binding to CD47, we tested a series of BNIP3 deletion mutants (21) by using the two-hybrid system. Whereas full-length BNIP3 showed a strong interaction with CD47, BNIP3⌬TM2 was unable to interact with CD47 (Fig. 2B) indicating that the transmembrane domain of BNIP3 is required for interaction with CD47. Mutants lacking the NH 2 terminus, the BH3 domain, or the COOH terminus did not bind strongly to CD47 (Fig. 2B), suggesting that other regions in the full-length structure of BNIP3 are necessary for the tight interaction with CD47.
To confirm the association between CD47 and BNIP3 in vivo, we performed co-immunoprecipitation assays (Fig. 2C). BNIP3 was specifically co-immunoprecipitated with CD47 from Jurkat T cells. Nevertheless, as expected, no co-immunoprecipitation was observed from JIN cells or cells expressing the CD47ec-CD7 mutant. Thus, BNIP3 interacts with the MMS domain of CD47 both in yeast and mammalian cells.
Next, we investigated whether, by blocking BNIP3 synthesis with antisense oligodeoxynucleotides, the CD47-generated apoptotic signal could be interrupted. Jurkat cells were incubated with phosphorothioate-derivatized antisense oligodeoxynucleotides, and apoptosis in response to soluble CD47 antibody Ad22 or anti-Fas antibody CH11 was measured (Fig.  3A). Antisense oligodeoxynucleotides corresponding to BNIP3 blocked CD47-induced cell death, showing a 2-fold reduction in cell death, whereas a scrambled sequence of the same oligodeoxynucleotide had no effect. Moreover, oligodeoxynucleotides had no effect on either survival of control cells or on apoptosis induced by the anti-Fas antibody. Immunoblot analysis confirmed a loss of BNIP3 expression from cells cultured with the BNIP3 antisense oligodeoxynucleotides (Fig. 3A). Thus, reduction of BNIP3 expression specifically leads to a reduction of cell death induced via CD47. Finally, because it has been previously shown that BNIP3 binds Bcl-2 at the mitochondrial membrane, an event likely to play a role in BNIP3-dependent PCD (24), we induced overexpression of Bcl-2 in Jurkat cells. Over-expressing cells were less susceptible to PCD induced either via CD47 than their normal counterparts (Fig. 3B). Thus, reduction of BNIP3 expression or overexpression of Bcl-2 leads to a reduction of cell death induced via CD47, demonstrating a FIG. 2. CD47 interacts with the BH3-only protein BNIP3. A, CD47-deficient JIN cells were either transfected with the wild-type CD47 or with the chimeric molecule CD47ec-CD7. Clones expressing similar levels of CD47ec-CD7 and wild-type CD47 were selected by cytofluorometry. Cells were incubated for an optimal period of time with the soluble CD47 Ad22 antibody (1 g/ml, 1 h) or with medium alone. Cell death was monitored as above. B, a yeast two-hybrid screen of a human lymphocyte library using LexA-CD47MMS as a bait was performed as described in "Experimental Procedures." Deletion mutants for BNIP3, described elsewhere (21), were cloned into the yeast twohybrid expression vector pActII and were assayed for their interaction with the MMS domain of CD47, on the basis of histidine prototrophy and ␤-galactosidase activity. C, lysates from Jurkat cells, JIN cells, or CD47ec-CD7-expressing mutants were immunoprecipited (IP) with the CD47 antibody Ad22. Immunoblots were performed with the indicated antibodies. IP control with Jurkat cell lysates was done using the irrelevant mouse IgG1 antibody MOPC21. The single asterisk denotes Ig heavy chain. The double asterisk denotes protein G. Data are from a representative experiment performed three times. functional interaction between CD47 and BNIP3.
TSP-1 Promotes CD47-induced Apoptosis and BNIP3 Translocation to Mitochondria-CD47 acts as a receptor both for the C-terminal domain of TSP-1 and for SIRP␣1 (37,38). TSP-1 is a protein expressed on endothelial cells, found in large amounts in extracellular matrix and released by activated platelets, whereas SIRP␣1 is expressed on the surface of macrophages and endothelial and dendritic cells. To determine which natural ligand induces cell death, we treated Jurkat cells or JIN cells with the CD47-binding agonist peptide from TSP-1 named 4N1K (37) or with a recombinant SIRP␣1-Fc fusion protein (25) (Fig. 4A). At high concentrations (400 M), soluble 4N1K induced a rapid death of Jurkat T cells, but had no nonspecific cytotoxic effect on CD47-deficient cells. Control peptide 4NGG failed to induce cell death in the two cell types. By contrast, the soluble SIRP␣1-Fc fusion protein, which efficiently bound CD47 (results not shown), had no apoptotic effect, regardless of the amount of protein or the incubation time we used. It is worth noting that only high concentrations of 4N1K (400 M) are able to induce cell death, whereas lower concentrations have been reported to sustain clonal expansion of inflammatory T cells (39) or to induce anergy of naive T cells (40,41). This apoptotic effect of 4N1K on T cells is not due to a nonspecific cytotoxicity, because CD47-deficient cells are resistant to cell death induced by 4N1K.
BNIP3 has been described to localize at and to act on the mitochondria when overexpressed (21,23,42). The finding of a potential association between the cell surface CD47 molecule and BNIP3 prompted us to examine the subcellular localization of native BNIP3. We detected BNIP3 at the inner leaflet of the plasma membrane in addition to mitochondrial localization (results not shown). Because it was shown that BNIP3 is active when integrated in the outer membrane of the mitochondria, we have investigated whether the plasma membrane-associ-ated BNIP3 molecule would relocalize to the mitochondria. To do so, we performed subcellular fractionations by ultracentrifugation of Jurkat cells stimulated with the CD47 antibody Ad22 or incubated in medium alone (Fig. 4B). In control cells incubated in medium alone, BNIP3 was detected both in the mitochondrial fraction, as assessed by the presence of TOM40, a protein localized on the outer mitochondrial membrane, and in the microsomal fraction including the plasma membrane, as assessed by the presence of placental alkaline phosphatase. Note that it is not possible to directly compare the microsomal and the mitochondrial fractions because the dilution factors of these fractions cannot be adjusted. When Jurkat cells were stimulated by the death-activating antibody, BNIP3 was no longer detected in the microsomal fraction, whereas its presence in the mitochondrial fraction appeared to be increased. To  Fig. 1A. B, Jurkat cells were stimulated 1 h with the soluble Ad22 antibody (1 g/ml) or incubated with medium alone. Cells lysates were subfractionated by ultracentrifugation and each fraction was blotted with appropriate antibodies. Mitochondrial and microsomal fractions were monitored by immunoblotting for the presence of translocase (TOM40) and placental alkaline phosphatase (PLAP), respectively. Note that dilution factors of the mitochondrial and microsomal fractions cannot be adjusted and are quite different, the mitochondrial fraction being the most concentrated. C, Jurkat cells were stimulated with the soluble 4N1K peptide (400 M) or incubated with medium alone. CD47 was labeled with SIRP␣1-Fc using FITC-conjugated second antibody, BNIP3 with anti-BNIP3, and Cyan V-conjugated second antibody, and mitochondria were stained with the Texas Red MitoTracker. The three images were merged (BNIP3 ϩ CD47 ϩ mitochondria). Scale bar represents 10 m. This image is representative of a larger field of view, and data are from a representative experiment performed three times. confirm that BNIP3 translocates from the plasma membrane to the mitochondria upon CD47 stimulation, we performed double labeling experiments in which T cells were stimulated or not with the pro-apoptotic 4N1K peptide and then double labeled for either CD47 or BNIP3, and mitochondria were revealed with the MitoTracker (Fig. 4C). When T cells were pretreated with the pro-apoptotic 4N1K peptide, BNIP3 was seen to localized to mitochondria, whereas substantial amounts of BNIP3 remained localized at the cell membrane in basal conditions (Fig. 4C). DISCUSSION Our data demonstrate that a BH3-only protein can be pivotal in transducing an external signal via a surface receptor. So far BH3-only proteins have been mainly regarded as sensors of intracellular or stress-induced damages, and BNIP3 was shown to mediate hypoxia-induced apoptosis (43). However, the latter process appears to be slow, likely due to a requirement for de novo synthesis of BNIP3, because the molecule is quickly degraded by the proteasome (44,45). It has been previously observed (21,42) that, when overexpressed, BNIP3 localizes to mitochondria. Although we made the same observation, we also collected several lines of evidence demonstrating that under basal conditions a significant amount of BNIP3 is localized at the plasma membrane where it might associate with CD47. In a yeast two-hybrid system using the membranespanning domain of CD47 as bait, we identified BNIP3. Moreover, immunoprecipitation of CD47 was accompanied by coprecipitation of BNIP3. Finally, double immunofluorescence visualization of CD47 and BNIP3 and subcellular fractionation revealed that a significant proportion of both molecules colocalize. The use of deleted forms of BNIP3 allowed us to identify the transmembrane domain of BNIP3 as the region necessary for interaction with the TM domains of CD47.
A second series of experiments showed that BNIP3 is required for the pro-apoptotic effect of CD47. First, we observed a substantial reduction in CD47 pro-apoptotic effect when BNIP3 expression was markedly reduced, using an antisense oligodeoxynucleotide of BNIP3, or when the anti-apoptotic molecule Bcl-2, another BNIP3 interactor, was overexpressed. Because it has been shown that BNIP3 exerts its pro-apoptotic effect at the mitochondria membrane (22,23,44), this suggests a simple model based on the translocation of BNIP3 from the inner surface of the cell membrane to the mitochondria. Consistent with this model, we showed that, after a pro-apoptotic signal transmitted via CD47 upon binding of a mAb or a peptide mimicking TSP-1, BNIP3 translocates from the plasma membrane to the mitochondria as assessed both by subcellular fractionation and immunofluorescence studies. This model accounts for a fast apoptosis and is consistent with the "rheostatic" effect exerted by the Bcl-2 family proteins (46 -49). It is worth noting that several other BH3-only proteins were reported to also be subjected to fast post-translational modification and translocation based effect (50). It must be emphasized, however, that BNIP3, together with Nix, forms a peculiar and quite conserved subfamily (45,51) within the BH3-only proteins. Their BH3 domain was found to be non-functional in terms of association with the anti-apoptotic proteins Bcl-2 and Bcl-X L . Rather it was demonstrated that their transmembrane C-terminal segment as well as N-terminal residues are necessary for these associations. Although a series of observations have shown that "classical" BH3-only protein associates with BAX and BAK to induce cell death (46,52), the mitochondrial and post-mitochondrial mechanisms leading to cell death with the BNIP3 subfamily remain to be established. Moreover, whether other BH3-only proteins would mediate apoptotic signals transduced via distinct transmembrane receptors remains to be determined.
T cell PCD via CD47 can be triggered by TSP-1 and not by its other natural ligand SIRP␣1. The fact that only high amounts of TSP-1 can trigger apoptosis also fits with the above model. By contrast, lower amounts of TSP-1 and 4N1K can sustain clonal expansion of T cells (14 -16). It can therefore be assumed that this pro-apoptotic pathway could quickly limit the inflammatory response, which tends to develop after thrombosis. We have collected in vivo evidence, in an inflammation model, for such a role of CD47. 2 Consistent with this view, it must be kept in mind that the CD47 pro-apoptotic pathway acts only on activated, but not resting, normal T cells (9). Interestingly, an overexpression of TSP-1 has recently been observed after myocardial infarction, forming a belt around injured tissues (55). Moreover, cell types other than T cells could be subjected to an apoptotic regulation via the TSP-1/CD47 pathway, as both TSP-1 and CD47 have been reported to be strongly up-regulated on vascular endothelial cells upon disturbance of blood laminar flow and can induce their apoptosis (53,54).