A Multifunctional Shuttling Protein Nucleolin Is a Macrophage Receptor for Apoptotic Cells*

Early apoptotic Jurkat T cells undergo capping of CD43, and its polylactosaminyl saccharide chains serve as ligands for phagocytosis by macrophages. This suggests the presence of a polylactosaminoglycan-binding receptor on macrophages. Here we show that this receptor is nucleolin, a multifunctional shuttling protein present in nucleus, cytoplasm, and on the surface of some types of cells. Nucleolin was detected at the surface of macrophages, and anti-nucleolin antibody inhibited the binding of the early apoptotic cells to macrophages. Nucleolin-transfected HEK293 cells expressed nucleolin on the cell surface and bound the early apoptotic cells but not phosphatidylserine-exposing late apoptotic cells. This binding was inhibited by anti-nucleolin antibody, by polylactosamine-containing oligosaccharides, and by anti-CD43 antibody. Deletion of the antibody binding region of nucleolin resulted in loss of the apoptotic cell-binding ability. Moreover, truncated recombinant nucleolin in solution containing this region blocked the apoptotic cell binding to macrophages, and the blocking effect was cancelled by the oligosaccharides. These results indicate that nucleolin is a macrophage receptor for apoptotic cells.

Macrophages and other phagocytes recognize and ingest apoptotic cells in tissue, preventing their lysis and subsequent release of harmful or immunogenic intracellular components. Therefore, clearance of apoptotic cells by phagocytes is crucial in the maintenance of tissue turnover and homeostasis. Moreover, it has been suggested that the apoptotic cell-ingested phagocytes play a role in suppression or resolution of inflammation (1)(2)(3)(4)(5).
Cells undergoing apoptosis display a variety of "eat me" signals, namely cell-surface changes to be recognized by phagocytes. These include externalization of phosphatidylserine (PS), 3 as yet little identified alterations of carbohydrates, and unidentified alterations of other membrane components such as intercellular adhesion molecule-3 and "thrombospondin binding sites" (1)(2)(3)(4)(5). Among these, the most common and the best-characterized change is externalization of PS, although the mechanism of externalization has not been fully understood (6,7).
In contrast to the poor understanding of the nature of eat me signals, various proteins of phagocyte membrane or in extracellular fluid have been reported as receptors or bridging factors for apoptotic cells. For externalized PS on apoptotic cells, CD36 (8), CD68 (9), CLA-1 (10), LOX-1 (11), and PS receptor (12) have been reported as phagocyte receptors; a serum protein ␤ 2 -glycoprotein I (13), complement component C3bi (14), and milk fat globule-epidermal growth factor-factor 8 (15) have been reported to bridge apoptotic cells and phagocytes through exposed PS. For intercellular adhesion molecule-3 on apoptotic cells, CD14 was suggested to be a macrophage receptor (16). Thrombospondin is also known as a bridging protein between unidentified sites on apoptotic cells and CD36 or vitronectin receptor (␣ v ␤ 3 integrin) on phagocytes (17). In addition, it has been reported that some collectins, a family of proteins of the innate immunity with C-type lectin domains, such as C1q, mannose-binding lectin, and surfactant proteins A and D facilitate engulfment of apoptotic cells by bridging apoptotic cells and phagocytes through calreticulin-CD91 complex on phagocytes, although whether or not carbohydrates on apoptotic cells are involved is uncertain (18,19).
Involvement of carbohydrates in the recognition of apoptotic cells by phagocytes was suggested by inhibition studies with various carbohydrates (20 -24). However, little was known about carbohydrate ligands and receptor molecules involved, and how carbohydrate ligands become susceptible to the phagocytic recognition.
In a series of studies, we have demonstrated that moderately oxidized cells such as oxidized erythrocytes (25,26), neutrophils (27), and human T lymphocytic Jurkat cells (28) are recognized by macrophages and that the ligands on the oxidized cell surface recognized by macrophages are carbohydrate chains containing sialyl residues and polylactosaminoglycans, possibly sialylpoly-N-acetyllactosaminyl chains, of membrane glycoproteins.
Furthermore, we recently found that apoptotic Jurkat cells transiently undergo capping of a major membrane sialoglycoprotein CD43 at an early stage of apoptosis and are recognized and phagocytosed by macrophages through the cluster of sialyl residues and polylactosaminyl chains of the capped CD43 (29). This transient capping of CD43 and the carbohydrate-mediated recognition occurred earlier than an increase in PS exposure and the PS-mediated recognition (30). This carbohydratemediated new mechanism appears to play a role in the removal of dying cells at a very early stage of apoptosis.
The above findings suggested the presence on macrophages of a lectin-like receptor that recognizes polylactosaminyl saccharide chains, especially those clustered on oxidized and early apoptotic cells. To identify this receptor, the membrane proteins of THP-1 macrophages differentiated from human monocytic THP-1 cells (THP-1 monocytes) were fractionated by successive affinity chromatographies using human lactoferrin and band 3 glycoprotein of erythrocyte membrane (31), both bearing sialyl residues and polylactosaminyl chains (32,33). After further fractionations, a receptor candidate protein with a relative molecular mass of 50,000 (M r 50 kDa) was obtained, and its tentative N-terminal amino acid sequence was determined (34).
In the present study, we further analyzed the N-terminal amino acid sequence of the 50-kDa protein, and finally identified it within human nucleolin, a multifunctional shuttling protein present in nucleus, cytoplasm, and on the surface of some types of cells (35)(36)(37). Here, we show that 1) nucleolin is present on the surface of monocytes and macrophages, 2) nucleolin on the surface of macrophages, and recombinant nucleolin expressed on the surface of HEK293 cells (HEK cells), work as early apoptotic cell-binding proteins with a specificity for polylactosaminoglycans, and 3) its amino acid residues 295-304 are required for its apoptotic cell-binding activity.

EXPERIMENTAL PROCEDURES
Cells and Materials-Jurkat, THP-1, and HEK293 cells (HEK cells) were obtained from Riken Cell Bank, Tsukuba, Japan, Japanese Cancer Research Resources Bank, Osaka, Japan, and Health Science Research Resources Bank, Osaka, Japan, respectively. THP-1 macrophages were obtained by treatment of THP-1 cells with phorbol myristate acetate as described previously (29). Mouse macrophages were isolated from the peritoneal cavities of adult male BALB/c mice (resident macrophages) or the mice injected with 2 ml of 3% thioglycolate medium 4 days before (thioglycolate-elicited peritoneal macrophages (TG macrophages)), by lavage with Hanks' balanced salt solution. An antibody against a synthetic octapeptide corresponding to the residues 295-302 of nucleolin (i.e. residues 1-8 of the 50-kDa protein) (Fig. 1, shaded sequence A) (anti-NUC295) was raised in rabbits and affinity-purified similarly to the work previously described (34). Anti-Fas mouse monoclonal antibody (clone CH-11), anti-CD43 mouse monoclonal antibody (clone DF-T1), and control rabbit IgG were purchased from Medical & Biological Laboratories, DAKO, and Genzyme-Tech, respectively. Anti-rNUC284 was raised in rabbits and affinity-purified using a HiTrap N-hydroxysuccinimide-activated Sepharose column (Amersham Biosciences) conjugated with a truncated recombinant nucleolin containing residues 284 -710 (rNUC284). Erythrocyte oligosaccharides were prepared as described before (29). PKH67 Green and PKH26 Red Fluorescent Cell Linker Kits were obtained from Sigma-Aldrich. Recombinant human annexin V was purchased from Medical & Biological Laboratories.
Apoptotic Cell Binding Assay-Apoptosis of Jurkat cells was induced by incubation of the cells (4 ϫ 10 6 cells/ml) in RPMI 1640 medium containing 5% fetal bovine serum with 10 M etoposide or 2 ng/ml anti-Fas antibody for 2 h at 37°C, the conditions under which the cells become susceptible to the carbohydrate-mediated macrophage recognition at their early stage of apoptosis (29). The cells thus prepared (i.e. early apoptotic cells) were washed twice with DPBS(Ϫ) and resuspended in RPMI 1640 medium buffered with 20 mM HEPES, pH 7.2 (RPMI 1640-HEPES) at 4 ϫ 10 6 cells/ml. Binding of apoptotic Jurkat cells to THP-1 macrophages or TG macrophages was assayed by coincubating the cells for 2 h as described before (29).
In the case of recognition by HEK cells and nucleolin-transfected HEK cells, cells in minimum essential medium Eagle containing 10% fetal bovine serum were plated in 24-well plates at 1 ϫ 10 5 cells/well, in which round coverslips (15-mm diameter) precoated with 100 g/ml human plasma fibronectin were placed. HEK cells were transiently transfected with a cDNA encoding for nucleolin (pcDNA4/HisMax-E/NUC) or for nucleolin deletion mutant (pcDNA4/HisMax-E/NUC(⌬295-304)) using Lipofectoamine2000 reagent (Invitrogen) in Opti-Mem I medium (Invitrogen) a day before each experiment. Protein expression was monitored by SDS-PAGE and immunoblotting. Apoptotic Jurkat cells were applied to the monolayer of HEK cells after additional 1-h incubation to adjust the time of apoptosis to that of macrophage assays, and co-incubated at 37°C for 1 h with gentle shaking. After removal of unbound cells, attached cells were fixed, stained, and counted as described (29). All the data are expressed as the number of bound Jurkat cells per 100 macrophages or HEK cells (cell binding) as counting more than 200 cells.
In the case of binding assay using nonadherent THP-1 monocytes, early apoptotic Jurkat cells and THP-1 monocytes were labeled with 10 M membrane-labeling fluorescent dyes PKH67 (green) and PKH26 (red), respectively, at room temperature for 5 min. The labeling was quenched by addition of RPMI 1640 medium containing 5% fetal bovine serum. The cells were then washed twice with Hanks' balanced salt solution. The labeled apoptotic cells were incubated with THP-1 monocytes at a cell number ratio of 10:1 in RPMI 1640-HEPES (phenol redfree) at 37°C for 2 h with gentle shaking. After the incubation, the binding of Jurkat cells and THP-1 macrophages was analyzed by flow cytometry. The binding index is defined as the percentage of THP-1 cells (red) having bound apoptotic cells (green). Mean values of triplicate experiments are shown with the indicated S.D. For comparison, binding of THP-1 macrophages and apoptotic Jurkat cells was similarly performed using THP-1 macrophages detached from culture substrate by incubation in Puck's EDTA solution (5 mM HEPES, 0.1 M NaCl, 5 mM KCl, 4 mM NaHCO 3 , 1 mM EDTA, 5.6 mM glucose) for 10 min.
Flow Cytometry-For measurement of cell-surface nucleolin, cells adherent to culture substrates (THP-1 macrophages, TG macrophages, and HEK cells) were detached by incubation in Puck's EDTA solution for 10 min. Before treatment with antibody, monocytes and macrophages in suspension were pretreated with 10 g/ml human IgG Fc fragment to block Fc receptors. Cells in suspension were treated with 10 g/ml of the primary antibodies (rabbit anti-NUC295 for monocytes and macrophages, and rabbit anti-rNUC284 for HEK cells), and stained by 10 g/ml of the secondary antibody (Alexa Fluor 488-conjugated goat anti-rabbit IgG, Molecular Probe), and immediately analyzed by a flow cytometer (FACSCalibur, BD Biosciences), gating for propidium iodide-unstained cells (alive cells).
Immunofluorescence Microscopy-Monocytes or macrophages were pretreated with 10 g/ml human IgG Fc fragment in RPMI 1640-HEPES-3% BSA at 0°C for 10 min. After washing with DPBS(Ϫ) at 0°C, cells were incubated with 10 g/ml anti-NUC295 or with 10 g/ml control IgG in RPMI 1640-HEPES-3% BSA at 0°C for 30 min. Bound antibody was detected by treatment of the cells with 10 g/ml Alexa Fluor 488-conjugated goat anti-rabbit IgG in RPMI 1640-HEPES-3% BSA at 0°C for 30 min. In the case of transfected HEK cells, cells were treated with 10 g/ml anti-NUC295 or rabbit control IgG in RPMI 1640-HEPES-3% BSA at 37°C for 30 min, and washed with DPBS(Ϫ). Bound antibodies were detected by treatment of the cells with 10 g/ml Alexa Fluor-488 goat anti-rabbit IgG (HϩL) conjugate (Molecular Probe) in RPMI 1640-HEPES-3% BSA at 37°C for 30 min. After washing with DPBS(Ϫ) at 0°C, the cells were resuspended in phenol red-free RPMI 1640 at 0°C and immediately subjected to confocal fluorescence microscopy ( Radiance, Bio-Rad).
Statistical Analysis-The data are presented as the mean Ϯ S.D. of at least triplicate experiments. The data were compared using a Student's t test, and statistical significance was determined. *, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.001.

RESULTS
Identification of 50-kDa Protein as a Fragment of Nucleolin-The N-terminal amino acid sequence of the 50-kDa protein isolated as a candidate for the macrophage receptor for oxidized cells, and for early apoptotic cells, was not identified within known proteins (34), although some fragments of the sequence were found in nucleolin. Then, we re-analyzed its N-terminal amino acid sequence and noticed some errors to be corrected. The corrected sequence was present in human nucleolin (Fig. 1, underlined residues 295-304). This suggests that the 50-kDa protein is a proteolytic fragment of nucleolin. Indeed, nucleolin is highly sensitive to proteolysis and tends to generate various sizes of fragments, including a 50-kDa fragment (38 -40).
Presence of Nucleolin on the Surface of Monocytes and Macrophages-We then investigated whether cell-surface nucleolin can function as a receptor for early apoptotic cells. First, expression of nucleolin on monocyte/macrophage surface was examined using anti-NUC295, an antibody against a synthetic octapeptide corresponding to the residues 295-302 of nucleolin ( Fig. 1, shaded sequence A). Nucleolin was detected on the surface of monocytes and macrophages in various ways. In flow cytometric analysis, anti-NUC295 bound to these cells, producing small but distinct shifts of the fluorescence peaks as compared with control IgG (Fig. 2A, left panels). Cell-surface nucleolin of THP-1 monocytes and THP-1 macrophages was also detected by immunofluorescence microscopy under impermeable conditions using anti-NUC295 (data not shown).
To detect the cell-surface nucleolin biochemically, the cell surfaces of THP-1 monocytes and THP-1 macrophages were biotinylated with a cell-impermeable biotinylation reagent. The biotinylated proteins were then isolated by adsorption to streptavidin-agarose and subjected to immunoblotting by anti-NUC295. The 110-kDa band of nucleolin was detected in the fraction of biotinylated cell-surface proteins (Fig. 2B).
Mouse resident peritoneal macrophages and thioglycolate elicited peritoneal macrophages (TG macrophages) also bound the antibody as detected by flow cytometry (Fig. 2A, right panels) and fluorescence microscopy (data not shown), indicating the presence of nucleolin on primary macrophages. These results confirmed that nucleolin is expressed on the surface of monocytes and macrophages.
Involvement of Cell-surface Nucleolin in the Binding of Early Apoptotic Jurkat Cells-Induction of apoptosis in Jurkat cells by 10 M etoposide or 2 ng/ml anti-Fas monoclonal antibody (clone CH-11) results in transient capping of CD43, a mucin-like major sialoglycopro-tein on various types of hematopoietic cells (51), at 2 h (29), which is earlier than PS exposure (30). At this early stage, the cells become susceptible to binding and engulfment by macrophages, and this is mediated by sialyl residues and polylactosaminoglycans, mainly those of the capped CD43, on Jurkat cells (29). Using this assay system for the carbohydrate-mediated macrophage recognition of early apoptotic cells, we next investigated whether the cell-surface-expressed nucleolin is involved in the recognition of early apoptotic cells. When THP-1 macrophages were pretreated with anti-NUC295, macrophage binding to etoposide-induced and anti-Fas-induced apoptotic cells was markedly inhibited, whereas control normal IgG had no effect (Fig. 3A). TG macrophages also recognized the etoposide-induced early apoptotic Jurkat cells, and this recognition was inhibited by anti-NUC295 (Fig.  3B). The result suggests that macrophage-surface nucleolin is involved in the macrophage recognition of the early apoptotic cells and that epitopes within the residues 295-302 or/and their neighbor structures are important for the receptor function.
To directly demonstrate the ability of cell-surface nucleolin to recognize apoptotic cells, recombinant nucleolin was expressed on the surface of non-macrophage cells HEK. Nucleolin-expression vector constructed in pcDNA4 was transiently transfected into HEK cells. Sixteen hours after transfection, anti-rNUC284, an antibody against a truncated form of recombinant nucleolin described in the later section, bound to HEK cells, producing a small but distinct shift of the fluorescence peak  , and mouse thioglycolate-elicited peritoneal macrophages (TG macrophages) were stained with anti-NUC295 (bold lines) as described under "Experimental Procedures." Thin lines represent control IgG staining. B, biotinylation and detection of cell-surface nucleolin. Cell surfaces of THP-1 monocytes and THP-1 macrophages were covalently labeled with sulfosuccinimidyl-6-(biotin-amide) hexanoate as described under "Experimental Procedures." Biotinylated proteins were collected using streptavidin-agarose, and separated by SDS-PAGE. Nucleolin was detected by immunoblotting using anti-NUC295. (Fig. 4A, middle), whereas untransfected HEK cells were not reactive to the antibody (Fig. 4A, top). Cell-surface expression of the transfected nucleolin was also detected by immunofluorescence microscopy using anti-NUC295, and the recombinant nucleolin molecules displayed a spotted distribution on the surface of HEK cells (data not shown) as was reported for nucleolin-expressing another type of cells (45,50).
Binding of etoposide-induced and anti-Fas-induced early apoptotic cells to nucleolin-transfected HEK cells was increased as compared with untransfected cells (Fig. 4B). When nucleolin-transfected cells were pretreated with anti-NUC295, binding of the transfected cells to etoposide-induced early apoptotic cells was markedly inhibited, but control IgG had no effect, suggesting that the cell-surface-expressed nucleolin recognizes the early apoptotic cells (Fig. 4C).
To test whether the sequence, including residues 295-302 in nucleolin (Fig. 1, shaded sequence A), is required for the receptor activity for the early apoptotic cells, HEK cells were transfected with a deletion mutant of nucleolin, NUC(⌬295-304), which is devoid of 10 amino acids from Lys-295 to Thr-304, and tested for their ability of apoptotic cell recognition. As shown in Figs. 4A (bottom), the transfectant expressed NUC(⌬295-304) on cell surface as effectively as undeleted nucleolin. However, the deletion mutant-expressing cells were unable to recognize the early apoptotic cells (Fig. 4D), suggesting that the residues 295-304 are required for the receptor activity of nucleolin.
Ligands on Early Apoptotic Jurkat Cells for the Cell-surface-expressed Nucleolin-To confirm that the binding of the early apoptotic Jurkat cells to the nucleolin-transfected HEK cells is mediated by the carbohydrate chains of the apoptotic cells, inhibition studies were carried out. The binding was inhibited by oligosaccharides of human erythrocyte membrane mainly composed of poly-N-acetyllactosaminyl chains (Fig. 5A). The inhibitory activity of the oligosaccharides was abrogated when they were treated with endo-␤-galactosidase, an enzyme that specifically cleaves poly-N-acetyllactosaminyl structure (i.e. Gal␤1-4GlcNAc␤1-3 repeats) at the ␤-galactosidic bond, or with neuraminidase that removes sialyl residues from the nonreducing termini of carbohydrate chains (Fig. 5A). The results suggest that nucleolinexpressing HEK cells recognize carbohydrate chains containing sialyl residues and polylactosaminoglycans on the early apoptotic cells.
The carbohydrate chains recognized by nucleolin-expressing HEK cells are most likely to be those of CD43, because the binding of the HEK cells to the apoptotic cells was inhibited by anti-CD43 antibody (Fig.   FIGURE 3. Effect of anti-NUC295 on the binding of early apoptotic Jurkat cells to macrophages. A, effect on the binding to THP-1 macrophages. THP-1 macrophages were preincubated with 2 g/ml anti-NUC295 or control IgG at 4°C for 30 min, washed, and subjected to the binding assay using etoposide-induced and anti-Fas-induced early apoptotic cells as described under "Experimental Procedures." B, effect on the binding to TG macrophages. Experiments were done similarly to those described in the legend to A. Data are expressed as "cell binding," which is defined under "Experimental Procedures." Each column represents the mean Ϯ S.D. of triplicate experiments. 5B). It is thus likely that the sialic acid-and polylactosaminoglycancontaining saccharide chains of CD43 on early apoptotic cells are recognized by nucleolin on the transfected HEK cells.
To see whether the cell-surface-expressed nucleolin is also capable of recognizing PS on apoptotic cells, late apoptotic Jurkat cells exposing substantial amount of PS on cell surface were prepared by extending the time of etoposide treatment to 12 h as described (30) (data not shown). As shown in Fig. 5C (left panel), binding of the late apoptotic Jurkat cells to nucleolin-transfected HEK cells was as low as that of control cells, whereas that of early apoptotic cells was significantly increased. In contrast, macrophage binding of the same late apoptotic Jurkat cells was increased, and was prevented by pretreatment with annexin V that specifically binds to PS (right panel), confirming that the late apoptotic Jurkat cell preparation used here is susceptible to PS-mediated macrophage recognition. Therefore, nucleolin is unlikely to recognize PS on apoptotic cells.
Effect of Truncated Recombinant Nucleolin on the Binding of Early Apoptotic Jurkat Cells to Macrophages-To further characterize nucleolin as a receptor molecule for early apoptotic cells, rNUC284, a truncated recombinant nucleolin containing residues 284 -710 corresponding to the C-terminal two-thirds of the molecule (Fig. 1, box B), was produced in E. coli and purified (Fig. 6A). rNUC284 includes the sequence of the 50-kDa protein originally isolated as a receptor candidate having the carbohydrate-binding activity (34). As was the case for the 50-kDa protein, rNUC284 was obtained as a soluble protein. For control, rCAT, a recombinant chloramphenicol acetyl transferase, was also produced (Fig. 6A). When etoposide-induced and anti-Fas-induced early apoptotic cells were pretreated with rNUC284, the binding of macrophages to the apoptotic cells was effectively blocked, whereas control rCAT had no effect (Fig. 6B). When etoposide-induced early apoptotic cells were pretreated with rNUC284 in the presence of the erythrocyte oligosaccharides, the blocking effect of rNUC284 was abrogated (Fig. 6C), indicating that rNUC284 binds to carbohydrate chains of the apoptotic cell surface. Interestingly, rNUC305, another truncated recombinant nucleolin containing residues 305-710 (Fig. 1, box C, and  Fig. 6A), did not block the binding (Fig. 6D). This again indicates requirement of some residues within Met-284 to Thr-304 for the receptor activity of nucleolin.
Comparison of Monocytes and Macrophages for Their Ability to Recognize Early Apoptotic Jurkat Cells-Because the quantity of cell-surface nucleolin on THP-1 monocytes appeared to be higher than that of THP-1 macrophages ( Fig. 2A), the ability of THP-1 monocytes to recognize early apoptotic Jurkat cells was examined. To measure the binding of THP-1 monocytes and Jurkat cells in suspension by flow cytometry, THP-1 monocytes and early apoptotic Jurkat cells were labeled with fluorescent dyes PKH26 (red) and PKH67 (green), respectively. For comparison, THP-1 macrophage monolayers were detached from culture substrates, suspended, and similarly stained in red. After incuba-  NOVEMBER 25, 2005 • VOLUME 280 • NUMBER 47 tion of THP-1 monocytes or THP-1 macrophages in suspension with early apoptotic Jurkat cells (cells incubated with 10 M etoposide for 2 h) at 37°C for 2 h with gentle shaking in the absence of serum, the cell suspension was subjected to flow cytometric analysis, and the binding index of monocytes or macrophages (the mean percentage of THP-1 monocytes or macrophages (red) having attached Jurkat cells (green) Ϯ S.D.) was determined.

Nucleolin Is a Receptor for Apoptotic Cells
The indices of THP-1 monocyte binding to control Jurkat cells (Jurkat cells similarly incubated without etoposide) and to early apoptotic cells were 7.1 Ϯ 0.2% and 8.4 Ϯ 0.2%, respectively, whereas those of THP-1 macrophage binding to control and to early apoptotic cells were 36.8 Ϯ 0.9% and 42.6 Ϯ 0.3%, respectively. From these data, THP-1 monocytes appear to be much less active in recognizing early apoptotic cells than THP-1 macrophages. Thus, the ability of the phagocytes to recognize early apoptotic cells may not be necessarily dependent on the quantity of the cell-surface-expressed nucleolin.

DISCUSSION
We have thus identified the macrophage cell-surface receptor for early apoptotic cells as nucleolin. Using macrophages, nucleolin-transfected HEK cells, and truncated recombinant nucleolin in solution, we demonstrated that nucleolin is expressed on macrophage surface and binds early apoptotic cells through their carbohydrate chains containing sialyl residues and polylactosaminyl structures, most likely those of Binding assays were performed in the absence or presence of erythrocyte oligosaccharides (100 g/ml) or those digested with endo-␤-galactosidase (EndoGAL, 100 milliunits/ml) or neuraminidase (NA, 100 milliunits/ml) as described (29). B, involvement of CD43 on early apoptotic Jurkat cells in their binding to nucleolin-transfected HEK cells. Apoptotic Jurkat cells were treated with anti-CD43 as described (29). C, PS-exposing late apoptotic Jurkat cells are not recognized by nucleolin (left panel). Late apoptotic Jurkat cells were prepared by extending the time of incubation of Jurkat cells with 10 g/ml etoposide to 12 h as described (30), and subjected to binding assay using nucleolin-transfected HEK cells as Fig. 4B. The binding of late apoptotic Jurkat cells to macrophages is mediated by PS (right panel). The same late apoptotic Jurkat cell preparation used in the left panel was treated with or without 10 g/ml recombinant human annexin V in 10 mM Hepes, 150 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , pH 7.2 at 0°C for 30 min, washed twice in DPBS(Ϫ), and subjected to the macrophages binding assay as described (30).  NOVEMBER 25, 2005 • VOLUME 280 • NUMBER 47 transiently capped CD43. Nucleolin is known as a multifunctional protein interacting with various RNA/DNA/protein targets within nucleus, cytoplasm, and on cell surface, although it was originally identified as a major nucleolar RNA-binding protein (35,37).

Nucleolin Is a Receptor for Apoptotic Cells
Expression of nucleolin on cell surface has been reported for some types of cells, including HeLa cells (52), lymphoblastoid T cells (52), breast carcinoma cells (46,53), lung (49), and laryngeal epithelial cells (50), and hepatocarcinoma cells (43). Nucleolin was also reported to be expressed on the surface of endothelial cells in angiogenic blood vessels (53). For monocytes and macrophages, there are some studies reporting the presence of nucleolin on the surfaces of monocytes (52), monocytic cell line U937 (40), and monocyte-derived macrophages (54). However, it was uncertain whether it is present on the cells we were using in the present study, because its expression has been suggested to be dependent on cell differentiation and cell types (52,53). In the present study, its presence on THP-1 monocytes, on THP-1 macrophages, and on mouse peritoneal macrophages were confirmed. Its quantity on these cell surfaces was very little as was the case with other cell types listed above. As seen in the flow cytometry profiles ( Fig. 2A, left panels), expression of nucleolin on THP-1 macrophages was lower than that on THP-1 monocytes. This is consistent with the previous observations that expression of nucleolin is lowered when cells differentiate or cell growth slows (55,56). However, in the present study, the ability of THP-1 monocytes to recognize early apoptotic Jurkat cells was suggested to be much lower than that of THP-1 macrophages. This may not be surprising, because it is known that phagocytic activities of monocytes are generally lower than those of macrophages (57,58). Therefore, the ability of phagocytes to recognize the early apoptotic cells may be determined not only by the presence of cell-surface-expressed nucleolin but by other membrane components that might interact or co-work with nucleolin, or by topology or the dynamics of the phagocyte membrane.
The carbohydrate-binding property of nucleolin was suggested in the present study. Nucleolin does not seem to have any known typical carbohydrate recognition domains but seems to have carbohydrate-binding activity as demonstrated by the inhibition studies using polylactosaminoglycan-containing oligosaccharides (Figs. 5A and 6C). Further analysis of the interaction between nucleolin molecule and various carbohydrate molecules is important for understanding of the nature and the specificity of the interaction. Considering that nucleolin interacts with anionic structures such as DNA and RNA, sialyl residues may be required as anionic residues. Conceivably, sialyl residues and the polylactosamine-containing residues may be separately interacting with different sites on nucleolin. In this context, the interaction of various polyanions with macrophage scavenger receptors (59) may have some similarity.
The polyanion-binding property of nucleolin also suggested the possibility of having affinity for PS or PS clusters. However, this was not true because nucleolin-expressing HEK cells were not able to bind to late apoptotic Jurkat cells, whereas THP-1 macrophages bound to the cells depending on their surface PS (Fig. 5C).
A preliminary analysis of nucleolin molecule with a view to identifying the residues involved in the recognition of apoptotic cells (i.e. carbohydrate chain clusters) suggested some residues to be required for the recognition. First, anti-NUC295, an antibody against the sequence Lys 295 to Glu 302 , was effective in blocking the binding of early apoptotic cells to macrophages (Fig. 3) and to nucleolin-transfected HEK cells (Fig. 4C). Second, NUC(⌬295-304), a deletion mutant of nucleolin devoid of 10 amino acids from Lys-295 to Thr-304, failed to recognize apoptotic cells when transfected into HEK cells (Fig. 4D). Third, a truncated recombinant nucleolin rNUC284, containing residues 284 -710 but lacking the N-terminal one-third, blocked the binding of apoptotic Jurkat cells to macrophages with a specificity to carbohydrate chains (Fig. 6, B and C), whereas rNUC305, another truncated recombinant nucleolin lacking the residues 284 -304 as compared with rNUC284, did not (Fig. 6D). From these results, some or all residues within the sequence 295-302 are likely to be required for the recognition. In addition, the N-terminal one-third (residues 1-283) appears not to be necessary for the recognition.
It is not known how nucleolin is attached to or embedded in the cell membrane. Nucleolin has a nuclear localization signal but does not have transmembrane domains. Thus, the mechanisms of its externalization from cytoplasm to cell surface and its interaction with membrane components are important questions to be elucidated. rNUC284 in solution inhibited the binding of early apoptotic Jurkat cells to macrophages. If rNUC284 retains not only the binding site for the carbohydrate chains on apoptotic cells but also that for macrophage membrane supposed to be present on a whole nucleolin molecule, it may have behaved as a bridging molecule leading to enhanced binding of apoptotic cells to macrophages, like the known bridging factors such as thrombospondin, milk fat globule-epidermal growth factor-factor 8, and collectins. Therefore, rNUC284 may not retain the binding site for the macrophage membrane, and the site may be located in the N-terminal onethird of nucleolin. However, another possibility that the supposed nucleolin binding sites on macrophages are saturated with the pre-existing nucleolin molecules cannot be ruled out.
Another important question is how external ligands attaching to cellsurface nucleolin, including apoptotic cells, are internalized into the cells. Because cell-surface nucleolin is reported to be associated with actin cytoskeleton (52), nucleolin-ligand complexes may be translocated into the cells by such machineries. A recent report (46) demonstrated that lactoferrin binds to cell-surface nucleolin and internalizes with it. In addition, a tumor-homing peptide F3 binds to cell-surface nucleolin and is internalized into the cells, and subsequently transported into the nucleus (53). Furthermore, the internalization of midkine, a 13-kDa cytokine, by cells has been reported to be cell-surface nucleolin-dependent, and attachment of midkine to cell-surface nucleolin is suggested to induce capping of nucleolin together with lipid raft components CD90 and CD59 (45).
Interestingly, cell-surface nucleolin has been shown to act as an adherence receptor for bacterium, enterohemorrhagic E. coli O157:H7 (50), and as an entry cofactor for viruses such as human immunodeficiency virus-1 (48), coxsackie B viruses (47), and human parainfluenza virus type 3 (49). These findings, coupled with our present finding that it is a receptor for early apoptotic cells, may suggest that nucleolin plays a role as a molecular chaperon in the cellular transport and in phagocytosis of extracellular particulate ligands into the cytoplasm and the nucleus.
We are also interested in the mode of nucleolin interaction with the carbohydrate ligands on apoptotic cells. To answer why pre-existing and unlikely cryptic carbohydrate ligands on unoxidized or non-apoptotic cells are not recognized by macrophages, and why those carbohydrate ligands become susceptible to the macrophage recognition when cells are oxidized or apoptotic, we hypothesized that, although the binding affinity of the receptor molecule on macrophages to the carbohydrate ligands is too low to make firm junctions for the cell-cell adhesion, membrane glycoproteins of target cells aggregate to form clusters upon cell oxidation or initiation of apoptosis, and the resultant clusters of their extracellular carbohydrate chains provide multivalent high "avidity" ligands for the receptors on macrophages (25-28, 60). Indeed, membrane glycoproteins on oxidized or apoptotic cells have been found to form clusters (29,61). On the other hand, distribution and movement of cell-surface nucleolin are not known. If nucleolin molecules form clusters on macrophage surface, that would facilitate their interaction with carbohydrate ligands on apoptotic cells. In the present study, we noticed that the transfected nucleolin usually showed spotted distribution on HEK cells. This may suggest that nucleolin works in clusters. Similar spotted distribution of membrane nucleolin was also observed on the surface of other types of cells (45,50). Therefore, nucleolin molecules on HEK cells and macrophages may tend to cluster, and thereby can form stable and firm multivalent junctions with apoptotic cells by the nucleolin-carbohydrate interaction. Other possibilities, including involvement of other membrane components of macrophages in the recognition process, also have to be investigated.
The present work showed two unexpected and novel functions of nucleolin: a cell-surface receptor for early apoptotic cells, and the carbohydrate-binding property. Our recent work using mouse splenic lymphocytes and peritoneal macrophages indicated that nucleolin is also a macrophage receptor for early apoptotic primary lymphocytes. 4 Further studies of nucleolin, including its molecular dissection, as a macrophage receptor for apoptotic cells and as a carbohydrate-binding protein is necessary for better understanding of the mechanisms of apoptotic cell clearance in tissue homeostasis.