Senescent human neutrophil binding to thrombospondin (TSP): evidence for a TSP-independent pathway of phagocytosis by macrophages


Dr Joseph F.Murphy Department of Clinical Pharmacology, Royal College of Surgeons in Ireland, 123 St Stephen's Green, Dublin 2, Ireland.


This study investigated the interaction of apoptotic polymorphonuclear neutrophils (PMN) with thrombospondin (TSP), an important event mediating the clearance of apoptotic neutrophils by macrophages. We developed an in vitro assay to examine this interaction. Based on this assay, we found that apoptotic but not fresh PMN bound specifically to surface-immobilized TSP (33 ± 0.03 × 103 cells/well) compared to fibrinogen, fibronectin or laminin (8.0 ± 0.3 × 103 cells/well). Moreover, the binding was specific for surface bound but not soluble TSP and appeared to be divalent cation dependent, was not significantly inhibited by heparin and was sensitive to cycloheximide (CHX) treatment of senescent PMN (>90% inhibition at 10 μM CHX). In contrast to the binding studies, phagocytosis of senescent PMN by macrophages was not affected by EDTA or cycloheximide. Phosphatidyl- L-serine liposomes, phospho- L-serine, glucosamine, galactosamine, and the acetylated sugars had no effect on phagocytosis. We conclude that: (i) there was specific binding of senescent human PMN to immobilized TSP, which is divalent cation dependent and requires new protein synthesis in the PMN during senescence; (ii) in addition to the recently defined TSP-dependent pathway, there is a TSP-independent pathway mediating phagocytosis of senescent PMN by macrophages. The identity of this pathway remains to be defined.

Programmed cell death or apoptosis is a process whereby developmental or environmental stimuli activate a specific series of cellular events that culminate in the death and efficient disposal of a cell ( Wyllie et al, 1972 ; Ellis et al, 1991 ; Raff, 1992) and the mechanism by which senescent neutrophils die ( Savill et al, 1989a , b). Phagocyte recognition of cells undergoing apoptosis is a swift and efficient way of removing unwanted cells from tissues without release of potentially toxic cell contents which might otherwise damage neighbouring cells and elicit an inflammatory/immune response. At least three different recognition mechanisms have been identified in vitro, although each is only partly characterized ( Savill et al, 1993 ) and different pathways are involved for different cell types. (1) Changes in cell surface carbohydrate structures in the apoptotic cells, suggest a sugar-lectin type interaction; however, the importance of this pathway is uncertain. Phagocytosis of apoptotic murine thymocytes could be blocked by N-acetyl glucosamine or N-acetyl galactosamine ( Duvall et al, 1985 ), but significant differences in the binding of specific labelled lectins to apoptotic and fresh thymocytes were not found ( Morris et al, 1984 ). (2) Phosphatidylserine (PS), the negatively charged phospholipid normally restricted to the inner monolayer of cells ( Fadok et al, 1992b , 1993) but externalized during apoptosis ( Martin et al, 1995 ), is recognized by a putative PS receptor on macrophages. Apoptosis of irradiated thymocytes could be blocked by phospho- L-serine and phosphatidylserine liposomes ( Fadok et al, 1992b ). (3) Thrombospondin (TSP), the multifunctional RGD-bearing adhesive glycoprotein of wide tissue distribution, has been shown to act as a molecular bridge between the vitronectin receptor (αvβ3) and CD36 on macrophages and an as yet unidentified site on senescent but not fresh PMN, resulting in phagocytosis ( Savill et al, 1992 ).

TSP is found in many different cell types, including platelets, monocytes and macrophages ( Silverstein et al, 1986 ; Bornstein & Sage, 1994). It is a major α-granule protein of human platelets and is expressed on the cell surface upon platelet activation, where it plays a role in platelet aggregation ( Leung, 1984; Leung et al, 1992 ). Multiple cell surface moieties display TSP-binding properties, including CD36 (GPIV), the vitronectin receptor (αVβ3), sulphatides, heparin sulphate, and the GPIIb–IIIa complex (αIIbβ3 integrin). It has recently been shown that transfection and expression of CD36 on COS-7 cells conferred the capacity for phagocytosis of apoptotic neutrophils and lymphocytes ( Ren et al, 1995 ). This process could be blocked by monoclonal antibodies (Mabs) directed against TSP, αVβ3, CD36 and the β3 subunit of integrin, whereas it was not inhibited by excess phospho- L-serine. More recently we have identified a specific domain on CD36 that mediates the interaction with TSP and the phagocytosis of apoptotic neutrophils ( Puente Navazo et al, 1996 ). Therefore, substantial amounts of data support the notion that TSP plays an important role in mediating the uptake of senescent neutrophils by macrophages.

In this study we sought to determine the receptor on senescent PMN mediating phagocytosis by macrophage CD36 and αVβ3 via TSP. Consistent with previously published data, we found that senescent, but not fresh, PMN were phagocytosed by macrophages ( Savill et al, 1989a , b) and optimal phagocytosis occurred when cells were aged for at least 12 h but no more than 24 h. Our binding studies to purified TSP show that PMN synthesized a receptor during senescence which was specific for TSP and this receptor synthesis was completely inhibited in the presence of cycloheximide. This binding was also inhibited in the presence of EDTA. Despite this, phagocytosis was not affected, indicating that an alternative pathway, which was not mediated by PS/receptor or sugar/lectin interactions, two of the known phagocytotic pathways for certain cell types, exists but has not previously been detected or described.


All chemicals were from Sigma Chemical Co. (St Louis, Mo.) unless otherwise stated. Tissue culture medium RPMI and supplements [fetal calf serum (FCS), penicillin/streptomycin, glutamine, sodium bicarbonate] were from Gibco Life Technologies Ltd, Grand Island, N.Y. Hank's balanced salt solution (HBSS) was from Irvine Scientific, South San Francisco, CA 94080. Tissue culture flasks and plates were either from Falcon (Becton Dickinson, San Jose, Calif.) or Costar (Cambridge, Mass.).

Cell isolation and culture

Routine isolation of PMN was performed under sterile conditions using Ficoll-Hypaque gradient medium or endotoxin free histopaque (1077) in a gradient-density-centrifugation technique ( Boyum, 1968; English & Andersenn, 1974) followed by erythrocyte lysis. Whole blood was drawn from donors on acid/citrate/dextrose (ACD), carefully layered on gradient and centrifuged at 2500 rpm for 25 min at room temperature (RT). Lysing buffer (0.154 ammonium chloride, 10 m M potassium bicarbonate, 1 m M EDTA) was added to the pellet containing erythrocytes and PMN for 8 min, centrifuged at 1500 rpm for 5 min and the PMN pellet washed three times with PBS. Cells were incubated at 37°C in either 25 ml or 75 ml flasks (Falcon, Lincoln Park, N.J.) for varying periods of time, for most experiments incubation was overnight (16–20 h) in RPMI media supplemented with 10% decomplemented FCS. All cells were washed two or three times in PBS before final suspension in either RPMI, HBSS or PBS Ca2+ (0.9 m M) Mg2+ (0.45 m M) depending on the experiment to be carried out.

Human peripheral blood derived macrophages were prepared from fresh buffy coats (Stanford Blood Center) using routine methods ( Musson, 1983). Briefly, mixed mononuclear cells from density gradients were washed three times with PBS, suspended in RPMI 10% FCS (10 × 106/ml) and 0.5 ml added to each well of a 24-well plate or 0.25 ml to a 48-well plate (Costar, Cambridge, Mass.). Cells were incubated for at least one night at 37°C, 5% CO2, following which non-adherent lymphocytes were removed by gentle washing and aspiration with RPMI. Adherent monocytes were cultured for a minimum of 7 d in RPMI 10% FCS supplemented with GMCSF, the medium being changed at 3 d intervals.

Flow cytometric analysis

PMN were assessed for apoptosis by looking for the presence of PS on their surface by flow cytometry. Both fresh and aged PMN were washed three times with PBS and then resuspended in PBS Ca2+ (0.9 m M) Mg2+ (0.45 m M) at a cell density of 1–2 × 106. Annexin V–FITC conjugate (R + D systems, Minn.), which has been shown to bind PS ( Koopman et al, 1994 ), was added at a concentration of 5 μg/ml, incubated at 4°C for 30 min and PS expression was immediately assessed with an Epics Profile II FACScan (Coulter Electronics, Luton, U.K.).

DNA isolation and electrophoresis

DNA was isolated from both fresh and aged PMN using a TurbogenTM genomic purification kit (Invitrogen Corporation, San Diego, Calif.) according to the manufacturer's instructions. Isolated DNA was then subjected to electrophoresis as previously described ( Arends et al, 1990 ).

35S-methionine incorporation

PMN were examined for the synthesis of new proteins during senescence by the determination of 35S-methionine incorporation in the presence or absence of cycloheximide according to established methods ( Harlow & Lane, 1988). Briefly, freshly isolated PMN were suspended in methionine free RPMI (prewarmed to 37°C). Cells were divided into 25 ml flasks, 5–10 × 106/ml and cycloheximide added to test flasks (not to control flask) at final concentrations of 10 μM and 100 μM. Flasks were then incubated at 37°C, 5% CO2 overnight, removed and centrifuged at 1500 rpm for 5 min, washed twice in methionine free RPMI and the cell pellets lysed with 100 μl 100 × Triton containing protease inhibitors (1 m M EDTA, 1 m M PMSF, 10−5 M pNGB and 50 μg/ml SBTI). Radioactivity, quantitated as counts per minute (cpm), was measured using a liquid scintillation counter (LKB Wallac, 1219 Rackbeta).

TSP purification and binding assays

Thrombospondin was isolated from fresh platelets as previously described ( Lawler et al, 1978 ). The concentration of the purified TSP was determined using a Bio-Rad assay (Richmond, Calif.) and either used immediately or stored at 4°C for a maximum of 2 weeks. Binding assays were carried out on 48-well plates (Costar, Cambridge, Mass.). Wells were coated with either TSP, fibrinogen, fibronectin or laminin at concentrations between 20 and 40 μg/ml in carbonate buffer (0.2 M, pH 9.6) for 2 h at RT, washed 3 × PBS and post-coated with 1% gelatin (endotoxin negative) for 1–2 h and again washed 3 × PBS. Aged PMN were then split and used either in the binding or phagocytosis assay (see below). PMN used in the binding assay were added to each well of the protein-coated 48-well plate (0.25–0.5 × 106/well) in PBS in presence or absence of Ca2+ (0.9 m M) Mg2+ (0.45 m M). After incubation for 40 min, non-adherent cells were aspirated, the wells were washed three times with PBS and adherent cells lysed and assayed for myeloperoxidase (MPO) activity ( Newman et al, 1982 ). Lysed cells were added with HBSS (0.3 ml), phosphate buffer (pH 6.4, 0.2 ml), dimethoxybenzidine (O-dianisidine HCl (50 μl) and H2O2 (50 μl). The colour product was measured spectrophotometrically (Titertek, Multiscan) at 450 nm. Standard curves were plotted for each assay using PMN ranging between 0.03 × 105 and 2 × 105, from which the number of adherent cells in the assay was quantitated.

Phagocytosis assay

Recognition and ingestion of aged PMN was assayed using previously described methods ( Newman et al, 1982 ) with minor modifications. Aged PMN were washed three times in PBS and suspended in RPMI or HBSS at 5 × 106 cells/ml. Macrophages were washed twice with RPMI and then 0.5 ml PMN suspension added to each well of a 24-well or 0.25 ml to a 48-well plate. Inhibitors were added in the same volume of RPMI or HBSS and left in wells throughout assay. PMN were seen to rapidly settle into a carpet in close association with the macrophage monolayer. The interaction was allowed to proceed for 2–3 h, during which time wells were monitored by light microscopy at low power and PMN adherence to, and ingestion by, macrophages could be clearly observed. Following interaction, all wells were washed thoroughly four times with PBS. Wells were checked by light microscopy to ensure no residual non-adherent or non-ingested PMN remained. Macrophages with ingested PMN were then lysed for 20 min with 2% Triton X-100 (Bio-Rad, Richmond, Calif.) and assayed for MPO activity as described above. Aged PMN were 100% MPO-positive whereas 7-d-old macrophages were 100% MPO-negative. Interassay variation was dependent either on the number of adherent macrophages which varied between samples or the number of PMN added.

Phosphatidyl- L-serine liposomes

Liposomes (small unilamellar vesicles) were prepared by sonication of pure phospholipids. Powdered liposomes were dissolved in CHCl3 (1 ml) and dried under N2 gas until a homogenous solid formed and then placed under vacuum to remove residual CHCl3. An appropriate amount of PBS was then added to yield a molarity of 25 m M. The mixtures were subsequently sonicated at 60–80% power level until the solution cleared to transparency (5–10 min) then centrifuged at 10 000 rpm for 2 min to pellet the titanium dust particles and multi-lamellar vesicles. The supernatant was used as the liposome vesicle solution. Liposomes contained 70 molar percent phosphatidylcholine an 30 molar percent phosphatidylserine. Further dilutions were made in serum-free RPMI immediately before each experiment. Phospha- L-serine, phospha- D-serine and L-serine were from Sigma and diluted in RPMI before each experiment.

Presentation of data

Each experiment was carried out in replicate as described and the mean ± SD calculated. Data for most experiments is presented as actual data (cell numbers) from one representative experiment with each experiment carried out at least three times. In experiments where comparisons were made between binding to TSP and phagocytosis, control data was normalized to 100% and the reduction in binding or phagocytosis caused by inhibitor calculated as a percentage of this.


Specific binding of senescent PMN to immobilized thrombospondin

To determine the binding site on senescent PMN for TSP, an in vitro binding assay was developed. PMN were incubated at 37°C for 16–24 h following isolation and examined for PS expression by annexin V binding. Significant binding of annexin V to aged PMN but not freshly isolated PMN was demonstrated, indicating that the aged PMN had undergone apoptosis (Fig 1). This was further confirmed by demonstration of DNA fragmentation in the aged but not fresh PMN as previously described ( Savill et al, 1989b ) and results not shown.

Figure 1.

6–20 h) and added to annexin V/FITC conjugate. The same procedure was carried out simultaneously with freshly isolated PMNs and both samples were analysed by FACs immediately afterwards. The figure shows binding of annexin V to aged but not fresh PMNs. Aged PMNs without addition of annexin V were included as a negative control.

When senescent PMN were layered onto protein-coated plates, there was significant binding to immobilized TSP (33 ± 0.03 × 103 cells/well) compared to fibrinogen, fibronectin or laminin (8.0 ± 0.3 × 103 cells/well) (Fig 2). Fresh PMN did not specifically bind to TSP (11.1 ± 1.5 × 103, 9.12 ± 1.7 × 103, 9.3 ± 0.7 × 103 and 11.85 ± 1.3 × 103 for TSP, laminin, fibronectin and fibrinogen, respectively) which was consistent with previously reported results ( Suchard et al, 1991 ). Soluble TSP did not inhibit this binding in competition assays (20–30 μg/ml) nor was it seen to bind to aged PMN as determined by FACS (results not shown). Binding was inhibited by EDTA (∼70% inhibition at 0.5 m M EDTA), suggesting that it was divalent cation dependent (Fig 3, dotted bar). In contrast, the addition of heparin was not significantly inhibitory, suggesting that it was not mediated by the heparin-binding domain on TSP (Fig 4, dotted bar).

Figure 2.

+, incubated at RT for 40 min and non-adherent cells removed by thorough washing. The number of adherent cells was quantitated as described in Methods. Data represent mean ± SD of quadruplicate determinations from one experiment, similar results were obtained in at least two other separate experiments.

Figure 3.

 × 105 bound TSP control and 0.19 ± 0.02 × 105 cells were phagocytosed in control wells. Data represent mean ± SD of quadruplicate determinations from one experiment; similar results were obtained in at least two other separate experiments.

Figure 4.

Fig 4. Effect of heparin on senescent PMN binding to TSP and phagocytosis. The assay was carried out as described in the legend to Fig 2 in the presence of heparin at the indicated concentrations. Data represent mean ± SD of quadruplicate determinations from one experiment; similar results were obtained in at least two other separate experiments. *P < 0.005; **P < 0.001.

Effect of EDTA and heparin on phagocytosis

Having observed that senescent PMN bound specifically to TSP and that this binding appeared to be divalent cation dependent, we then carried out phagocytosis assays on the senescent cells in parallel. Freshly isolated PMN were not phagocytosed by macrophages. Optimal phagocytosis of senescent PMN occurred when cells were aged between 16 and 24 h and incubated with macrophages for 2–3 h. In contrast to the binding of senescent PMN to TSP, EDTA did not inhibit phagocytosis (Fig 3, shaded bar), whereas the addition of heparin inhibited phagocytosis by >50% (Fig 4, shaded bar). The inhibitory effect of heparin on phagocytosis was consistent with previously reported results ( Savill et al, 1989a ).

Effect of cycloheximide on aged PMN binding to immobilized TSP and phagocytosis by macrophages

We next sought to determine whether the TSP-specific binding required the synthesis of new proteins during senescence. We observed that 35S-methionine incorporation into senescing PMN was inhibited 7-fold when cells were aged in the presence of 100 μM cycloheximide. When PMN that had been aged in the presence of cycloheximide were added to TSP-coated wells, binding was completely inhibited (Fig 5). In contrast, these same cells, when added to macrophage monolayer in parallel assays, were phagocytosed. Senescent PMN binding to control proteins fibrinogen, fibronectin, laminin and gelatin was not affected by cycloheximide. The data suggested that senescent PMN binding to immobilized TSP was mediated by a cell surface protein(s) synthesized during senescence which specifically interacted with TSP. On the other hand, the absence of these receptors in cycloheximide-treated senescent PMN did not prevent phagocytosis by macrophages, suggesting the presence of a TSP-independent pathway of recognition.

Figure 5.

% relative to TSP control and was not affected by cycloheximide. Data represent mean ± SD of quadruplicate determinations from one experiment; similar results were obtained in at least two other separate experiments.

Effect of phosphatidyl- L-serine, aminosugars, and related compounds on phagocytosis

As PS was expressed on senescent PMN (Fig 1) and has previously been shown to mediate interactions between macrophages and apoptotic lymphocytes ( Fadok et al, 1992b , 1993), we investigated whether it was involved in the phagocytosis of senescent PMN. PS liposomes when added to the phagocytosis assay were found to have no inhibitory effect (Fig 6). Aqueous soluble phospho- L-serine, phospho- D-serine and L-serine were likewise non-inhibitory (results not shown). The other reported mechanism by which macrophages may ingest senescent or apoptotic cells is by a lectin–sugar interaction ( Duvall et al, 1985 ; Morris et al, 1984 ). A series of experiments was performed where aminosugars and their acetylated forms were added to the phagocytosis assay and they did not inhibit PMN uptake (Fig 7). We concluded that neither the phosphatidylserine/receptor nor the lectin/sugar pathways are involved in the phagocytosis of senescent PMN by macrophages.

Figure 6.

Fig 6. Effect of phosphatidyl- L-serine and related compounds on phagocytosis. Phosphatidy- L-serine (PS) liposomes were prepared as described in Methods and added to phagocytosis assay. PS dissolved in DMSO and the aqueous solutions of phospha- L-serine, phospha- D-serine and L-serine, which were assayed simultaneously at a concentration of 1 m M, were not inhibitory. Data represent mean ± SD of triplicate determinations from one experiment; similar results were obtained in three other separate experiments.

Figure 7.

Fig 7. Effect of sugars and their acetylated forms on phagocytosis. Amino sugars and their acetylated forms were diluted in RPMI and added to the phagocytosis assay. Data represent mean ± SD of triplicate determinations from one experiment; similar results were obtained in three other separate experiments.


Clearance of apoptotic PMN by circulating macrophages has been the focus of several studies, and it has been shown that TSP mediates binding between αvβ3 and CD36 on macrophages and an unidentified receptor on the PMN surface ( Savill et al, 1992 ). Using an in vitro binding assay, we have demonstrated that senescent, but not freshly isolated, PMN bound specifically to surface immobilized TSP. This binding appeared to be divalent cation dependent as it was significantly reduced in the presence of EDTA but had no effect on phagocytosis. Savill et al (1989a ) reported a novel charge sensitive recognition mechanism; however, in their study cationic sugars were used to modulate macrophage uptake of apoptotic PMN as opposed to EDTA. Binding was insensitive to heparin, suggesting that it was not mediated by the heparin-binding domain on TSP (Fig 4). Perhaps most interestingly, binding was sensitive to cycloheximide treatment of senescent PMN, suggesting that a TSP-binding protein was synthesized during senescence. A previous study has shown that PMN adherence to TSP occurs through a CD11/CD18-independent mechanism and that binding to vitronectin, laminin and fibronectin occurs via the β2 integrins ( Suchard et al, 1991 ). They did not, however, compare the binding of fresh and senescent PMN to these proteins.

The observation that soluble TSP did not bind senescent cells and did not inhibit their binding to immobilized TSP suggests that senescent cells recognize a conformation- dependent domain on TSP which becomes exposed only upon TSP immobilization on a surface. This may be analogous to the situation where the platelet αIIbβ3 (GPIIb–IIIa) integrin receptor recognizes and interacts with surface immobilized fibrinogen but not soluble fibrinogen ( Du et al, 1991 ). Previously, we have identified a specific anti-TSP monoclonal antibody that binds to immobilized TSP only in the presence of a CD36 peptide, indicating that TSP is capable of assuming different conformational states ( Li et al, 1993 ). Several groups have reported adhesion of cell lines to TSP-coated surfaces, and several receptor systems have been implicated ( Roberts et al, 1987 ; Varani et al, 1988 ; Asch et al, 1991 ).

In contrast to the in vitro findings of senescent PMN binding to purified TSP, phagocytosis of senescent PMN by peripheral blood monocyte derived macrophages exhibited different parameters. It was not divalent cation dependent (Fig 3) and, consistent with published results ( Savill et al, 1989a ), partially inhibited by heparin (Fig 4). Surprisingly, phagocytosis was also insensitive to cycloheximide treatment. The data suggest that the TSP-binding protein synthesized during cell senescence, although necessary for TSP recognition by the senescent PMN, was unnecessary for phagocytosis by macrophages and that there is an alternate pathway of macrophage recognition of senescent neutrophils. Our results do not contradict those obtained in a previous study ( Ren et al, 1995 ) which has shown that phagocytosis of apoptotic lymphocytes and PMN by COS-7 cells transfected with CD36 could be blocked by mAbs directed against TSP, αVβ3 and the β3 subunit of the integrin. The COS-7 cells apparently possessed only one recognition pathway and its inhibition by mAbs led to reduced phagocytosis. However, a number of different pathways may exist in a professional scavenger cell such as a macrophage and interference of one pathway does not necessarily affect engulfment. Studies using aminosugars and PS liposomes suggest that this alternate pathway is not mediated by a mannose/fucose type lectin or by the exposed PS on senescent cells. Of note, not all macrophage populations utilize the same recognition pathways, e.g. murine macrophages derived from the peritoneal cavity used a putative PS receptor to bind apoptotic cells whereas macrophages derived from bone marrow or peripheral blood used the TSP/VnR/CD36 pathway ( Fadok et al, 1992a , 1993). The recognition pathways may also vary and depend on the apoptotic target cell populations, for example the recognition signals on an apoptotic neutrophil may be quite different from that of an apoptotic lymphocyte ( Savill et al, 1993 ).

The recognition and engulfment of apoptotic cells by macrophages are probably mediated by multiple pathways that are partially overlapping and redundant in nature. This is not surprising considering the central importance of apoptosis. Detailed analysis of the apoptotic process in the nematode C. elegans has identified seven genes that are involved in the engulfment phase (ced-1, ced-2, ced-5, ced-6, ced-7, ced-8 and ced-10) ( Ellis et al, 1991 ), indicating the complexity of the process. Mutations in any of these genes can result in impairment of ingestion of apoptotic cells, suggesting that a number of different proteins may be involved. Thus, multiple pathways may act in a co-ordinated and partially overlapping fashion during the apoptotic process.

Finally, the recognition pathways for apoptotic and senescent PMN may not be identical. Studies with transgenic mice that over-expressed Bcl-2 in mature neutrophils showed that although Bcl-2 blocked PMN apoptosis, the aged PMN were still engulfed by macrophages ( Lagasse & Weissman, 1994). This data is particularly interesting given the tight correlation reported for PMN exhibiting apoptotic features with the ability of macrophages to bind and engulf these cells ( Savill et al, 1989a , b). This suggests that Bcl-2-expressing PMN exhibit cell surface changes during senescence that allow macrophage recognition and engulfment independent of apoptotic changes. The identity of these senescent signals remains to be determined.

In conclusion, we have shown in this study that although TSP may mediate recognition between macrophages and apoptotic PMN, there is an as yet undefined mechanism by which macrophages recognize and phagocytose senescent PMN. This mechanism is not mediated by PS/receptor or lectin/sugar interactions. Future studies will focus on elucidating this pathway.


We thank Diana Thompson for her excellent technical assistance. We are grateful to Dr Jie Yuan and Prashant Shah for liposome preparation and to Lisa Ma for assistance with FACS analysis. This work was supported in part by a National Institute of Health Grant 1R01-HL42943 (to L.L.K.L.).