Role of CD47 as a Marker of Self on Red Blood Cells

See allHide authors and affiliations

Science  16 Jun 2000:
Vol. 288, Issue 5473, pp. 2051-2054
DOI: 10.1126/science.288.5473.2051


The immune system recognizes invaders as foreign because they express determinants that are absent on host cells or because they lack “markers of self” that are normally present. Here we show that CD47 (integrin-associated protein) functions as a marker of self on murine red blood cells. Red blood cells that lacked CD47 were rapidly cleared from the bloodstream by splenic red pulp macrophages. CD47 on normal red blood cells prevented this elimination by binding to the inhibitory receptor signal regulatory protein alpha (SIRPα). Thus, macrophages may use a number of nonspecific activating receptors and rely on the presence or absence of CD47 to distinguish self from foreign. CD47-SIRPα may represent a potential pathway for the control of hemolytic anemia.

Natural killer (NK) cells eliminate target cells recognized by a range of activating receptors that bind ligands on many normal cells. However, expression of self major histocompatibility complex (MHC) class I molecules can protect a cell by binding to NK cell inhibitory receptors, which recruit and activatesrc-homology phosphatases (SHP-1 and SHP-2) that inhibit cell activation (1–4). NK cells thus spare cells that express “markers of normal self” in the form of MHC class I molecules, and eliminate them when these markers are absent or inadequately expressed. In contrast to what might be expected, MHC class I–deficient mice are not autodestructive. Rather, NK cell recognition adapts to the level of inhibitory ligand expressed in the NK cell environment (5). Expression of molecules related to NK cell inhibitory receptors on other leukocytes suggests that similar mechanisms are operative, for example, in macrophage activation (4, 6, 7). Although many of these molecules recognize MHC class I, the “marker of self” could in principle be any ubiquitously expressed surface molecule.

SIRPα is a unique receptor in that it appears to bind CD47 (integrin-associated protein, IAP), a ubiquitously expressed cell surface glycoprotein, rather than MHC (8–14). SIRPα is expressed on many cells, most abundantly on polymorphonuclear leukocytes, monocytes, and monocyte-derived cells, where it is an inhibitory receptor (8–14). In contrast to MHC, CD47 is present also on red blood cells (RBCs). We hypothesized that CD47 might function as a RBC marker of self, and that CD47-deficient cells would therefore be rapidly destroyed by SIRPα−expressing leukocytes.

CD47 / mice are viable and have normal RBC parameters (15, 16). To investigate whether CD47 / RBCs can survive in wild-type mice, we transfused CD47 / and control recipients with CD47 / and wild-type (CD47+) RBCs (17). Transfused CD47 / RBCs were rapidly eliminated from the circulation of wild-type recipients, but were not affected in CD47 / recipients (Fig. 1A). Wild-type RBCs were unaffected in both recipient types (Fig. 1B). Elimination of the CD47 / RBCs was not due to recognition of the null allele itself, but rather due to recognition of the CD47-deficient state, because heterozygotes were indistinguishable from the wild-type both as donors and as recipients. Antibody and classical T and B cells were not required for the elimination of RBCs, because clearance was rapid also in immune- and antibody-deficient Rag1 / mice (Fig. 1C). Furthermore, gene-targeted mice deficient in complement component 3 (C3) were similarly able to eliminate CD47 / RBCs, suggesting that complement activation was not required (Fig. 1D).

Figure 1

Wild-type mice rapidly eliminate CD47-deficient RBCs independently of antibody and complement. Fluorescently labeled RBCs were injected intravenously. At the times indicated, 5 μl of venous blood was sampled from a tail vein of the recipients and analyzed by flow cytometry for the fraction of fluorescent RBCs (17). Data were normalized to the level at 30 min after injection (usually 4 to 6% of total RBCs). (A) CD47 / RBCs were injected into wild-type (•) or CD47 / (○) recipients. (B) Wild-type RBCs were injected into wild-type (•) or CD47 / (○) recipients. (C) Wild-type RBCs (•) or CD47 / RBCs (○) were injected into CD47 wild-type rag1 / recipients. (D) CD47 / RBCs were injected into CD47 wild-type C3 / recipients. Data are the means ± SD for three to four recipients per group.

Clearance of RBCs occurs primarily in the spleen. To determine if this was the case for the rapid elimination of CD47 / RBCs, we compared elimination of CD47 / RBCs in splenectomized or sham-operated wild-type recipients to that in unoperated wild-type and CD47 / controls (Fig. 2A). Elimination of CD47 / RBCs was almost completely abolished by splenectomy, whereas sham operation had no effect. Thus, the spleen was required for the elimination of CD47 / RBCs.

Figure 2

Clodronate-sensitive cells in the spleen of wild-type mice are necessary for the clearance of CD47 / RBCs. (A) Lack of clearance of CD47 / RBCs in splenectomized wild-type recipients. Wild-type mice were splenectomized (▪) or sham-operated (□). Fourteen days later, the clearance of CD47 / RBCs in these mice was compared with that in unoperated wild-type and CD47 / recipients (17). For comparison, the clearance in wild-type (•) or CD47 / (○) recipients is shown. Data are the means ± SD for one to three recipients per group. (B) Elimination of clodronate-sensitive cells abolishes the clearance of CD47 / RBCs in wild-type recipients. Wild-type mice were injected intravenously with clodronate liposomes (▪) or, as a control, liposomes prepared without clodronate (□) (20) at 48 and 24 hours before the examination of CD47 / RBC clearance (17). Data are the means ± SD for five clodronate-treated mice and four mice treated with control liposomes.

Dichloromethylene diphosphonate (clodronate)–loaded liposomes injected intravenously kill splenic macrophages (18). Different repopulation kinetics make it possible to determine the relative contribution of various splenic macrophage subsets to a phenotype (19). Treatment of wild-type mice with clodronate liposomes (20) prevented the elimination of CD47 / RBCs, strongly suggesting that clearance was effected by macrophages (Fig. 2B). Recovery was already substantial 10 days after clodronate injection and virtually complete at day 28. This is consistent with the repopulation kinetics of red pulp macrophages, but not with those of marginal metallophilic macrophages (about 3 weeks) and marginal zone macrophages (about 2 months) (20). Thus, splenic macrophages were required for CD47 / RBC elimination, and among them, red pulp macrophages were sufficient. The identification of red pulp macrophages as the primary site of CD47 / RBC clearance was confirmed by immunohistochemistry (21) (Fig. 3) and by flow cytometry.

Figure 3

Splenic F4/80+ red pulp macrophages are responsible for clearance of CD47 / RBCs in wild-type mice. CD47 / (A and C) or wild-type (B andD) recipients were injected with fluorescent red (PKH26) CD47 / RBCs, and spleens were harvested 20 hours later. Frozen spleen sections were stained green for either MOMA-1 (a marker for marginal metallophilic macrophages) (A and B) or F4/80 (a marker for red pulp macrophages) (C and D) (21). RP, red pulp; WP, white pulp.

SIRPα, a receptor for CD47, is thought to generate macrophage-inhibitory signals (8, 22,23). It is highly expressed on myeloid cells, including red pulp macrophages (22–24). If SIRPα is the relevant receptor for RBC clearance, antibodies to SIRPα (anti-SIRPα) that block SIRPα-CD47 interaction should augment the phagocytosis of wild-type RBCs to the same extent as that seen with CD47 / targets. Indeed, isolated splenic macrophages (mainly F4/80+ red pulp macrophages) in vitro phagocytosed CD47 / RBCs at a high rate, whereas wild-type RBCs remained almost completely unaffected (24) (Fig. 4A). Antibody P84 is directed against SIRPα and blocks CD47-SIRPα interaction (9). This antibody had no effect on the phagocytosis of CD47 / RBCs, but increased phagocytosis of wild-type RBCs to the same extent as that seen with CD47 / targets (Fig. 4A). Thus, inhibition of phagocytosis by CD47 on the target RBCs required SIRPα and SIRPα-CD47 interaction. The observation that both types of RBCs were phagocytosed equivalently when SIRPα was blocked strongly suggests that CD47 / RBCs are cleared because of the absence of CD47 on their surface, rather than because of a possible secondary effect of CD47 deficiency on the RBCs.

Figure 4

CD47 on RBCs inhibits phagocytosis through binding to macrophage SIRPα. (A) Adherent wild-type splenic macrophages were incubated without antibody (open bars), with 5 μg of anti-SIRPα mAb P84 (9) (closed bars), or with 5 μg of the isotype-matched anti-murine CD14 control antibody (rmC5-3) (hatched bars) for 15 min before the addition of wild-type or CD47 / RBCs (24). After lysis of noningested RBCs, the number of ingested RBCs per 100 macrophages was determined under each condition. Data shown are normalized to the level of phagocytosis of CD47 / RBCs in the absence of antibody (63 ± 15 RBCs per 100 macrophages). Data are the means ± SD for three separate experiments. (B)Macrophage SIRPα1 tyrosine phosphorylation upon contact with CD47 on RBCs. RBCs were allowed to sediment onto bone marrow–derived macrophages in ice-cold buffer. Cells were then rapidly warmed to 37°C in the presence of 2 mM pervanadate and lysed in sample buffer at the times indicated. Anti-SIRPα mab P84 immunoprecipitates were separated by SDS-PAGE and analyzed by Western blot for phosphotyrosine with mAb 4G10 and by using enhanced chemoluminescence (27). Phosphorylation is more pronounced and occurs earlier in contact with wild-type RBCs than with CD47 / RBCs, indicating that CD47 on the target induces a SIRPα signal.

SIRPα contains intracellular immune receptor tyrosine-based inhibitory motifs, which upon binding and aggregation are phosphorylated by src-family kinases, leading to the recruitment and activation of tyrosine phosphatases SHP-1 and/or SHP-2 (mainly SHP-1 in macrophages) (25). This in turn inhibits signaling in tyrosine kinase–dependent activation pathways (8, 26). In contrast to splenic macrophages, bone marrow–derived macrophages (BMMs) do not phagocytose CD47 / RBCs, allowing study of the inhibitory signal in isolation. To investigate whether CD47 on RBCs could induce a macrophage SIRPα signal, we assayed SIRPα tyrosine phosphorylation (8,27) in BMMs after contact with either CD47 / or wild-type RBCs. Wild-type RBCs induced markedly greater SIRPα tyrosine phosphorylation than CD47 / RBCs (Fig. 4B). The increase in SIRPα phosphorylation in the presence of CD47−/ RBCs was similar to that seen without RBCs and was probably due to interactions between macrophage CD47 and SIRPα or to the temperature shift in the experiment. Thus, RBC CD47 can induce a macrophage SIRPα signal.

Because CD47 binds SIRPα, regulation of CD47 expression might serve as a mechanism to control elimination or uptake of damaged or senescent RBCs. However, RBC CD47 expression is very narrowly distributed, with a negligible overlap between wild-type and heterozygous RBCs (15). Yet heterozygous RBCs showed normal survival in wild-type mice. Although this result argues against a gradual removal of CD47 as the mechanism regulating RBC life-span, it remains possible that control of CD47 expression might be used elsewhere as a means to regulate cell fate or cell-cell interactions.

It is unclear at what level of RBC CD47 expression the negative signal is sufficiently attenuated to allow phagocytosis. Rh polypeptides are expressed as a complex with CD47 on the cell surface. Rhnull individuals fail to express Rh antigen–bearing polypeptides on their RBCs (28), and consequently they express CD47 at ≤25% of normal levels. In contrast, CD47 levels are normal on Rhnull leukocytes and presumably also on splenic macrophages of Rhnull individuals. Rhnull individuals have hemolytic anemia, reticulocytosis, and stomatocytosis—all of which can be corrected by splenectomy (28). This parallels our transfusion model in which RBCs with reduced levels of CD47 are cleared in an environment of CD47+ cells. It is tempting to speculate that the elimination of Rhnull RBCs is secondary to reduced CD47-SIRPα signaling to splenic macrophages.

Although CD47 / RBCs are recognized and eliminated in wild-type recipients, they survive normally in CD47-deficient mice. This suggests that the macrophage is regulated by the expression of CD47 on its surface or in its environment. This adaptation does not appear to be at the level of inhibitory receptor expression, because the level of SIRPα is virtually identical on CD47 / and wild-type macrophages (29).

The data presented here demonstrate that CD47 is a crucial marker of self on RBCs. Preliminary data suggest that this is also true for other cells, such as lymphocytes (29). Red pulp macrophages specifically bind to circulating cells, and only CD47 expression prevents their elimination. Thus, splenic macrophages do not have to rely on activating receptors alone to differentiate between self and foreign, but can use the presence or absence of CD47 to make that distinction. Ovarian cancer cells express high levels of CD47 (13, 30), and CD47 analogs are encoded by smallpox and vaccinia viruses (11–13). In both of these instances, it is possible that the pathogen is taking advantage of SIRPα signaling to disable normal defenses. Similarly, down-regulation of SIRPα might confer selective advantages to malignancies of myeloid origin (31).

  • * To whom correspondence should be addressed. E-mail: lindberg{at}


Stay Connected to Science

Navigate This Article