CD47 is a membrane receptor glycoprotein composed of a heavily glycosylated N-terminal IgV domain and five transmembrane helices domain with a short cytoplasmic tail (1–3). It was originally identified as a tumor antigen on human ovarian cancer specimens (4). This transmembrane protein is well recognized for its role in “don’t eat me” anti-phagocytic signals by binding to SIRPα (signal regulatory protein alpha) (5). Additionally, TSP-1 (thrombospondin-1) and integrin α2β1 and αvβ3 have also been shown to be CD47 ligands (1, 6) ( Figure 1 ).

CD47 is involved in multiple fundamental cellular functions, including cell migration, apoptosis, and axon development (7–9). It is widely expressed on the surfaces of normal cells, especially hematopoietic cells and is best known for its interaction with SIRPα (10). SIRPα is a myeloid-specific immune checkpoint and has been classically described as counterbalanced by a variety of activating membrane receptors (11).The interaction between CD47 and SIRPα emerges as a key regulatory component of innate immune checkpoint by regulating signal transduction and cell clearance. In addition, CD47 is recognized as pivotal for erythrocyte homeostasis, during which both SIRPα and TSP ligands have been shown to be involved (12).

In this article, we will discuss basic functions of CD47 and its associated ligands, review the current knowledge about the involvement of CD47 in host cell phagocytosis, and survey current strategies for developing CD47-based immunotherapy for tumors. In addition, side effects and future research directions for CD47-based immunotherapy will be evaluated.

The most investigated binding partner of CD47 is SIRPα, which is highly expressed in neurons and a subset of myeloid hematopoietic cells such as dendritic cells (DCs) and macrophages (13). SIRPα is a transmembrane protein comprised of cytoplasmic region with four immunoreceptor tyrosine-based inhibition motifs (ITIMs) and extracellular region with three immunoglobulin (Ig)-like domains (14). The NH₂-terminal V-like domain of SIRPα is responsible for the binding the extracellular Ig-domain of CD47, whereas ITIMs provide the binding sites for the src homology-2 (SH2)-domain-containing protein tyrosine phosphatases SHP-1 and SHP-2 (15). Recruitment of SHP-1 and SHP-2 phosphatases prevents myosin-IIA accumulation at the phagocytic synapse (16).

By combining different ligands, different signaling pathways can be triggered: When CD47 interacts with SIRPα on the surface of phagocytes, it promotes phosphorylation of the intracellular ITIMs and activates the inhibitory phosphatases SHP-1 and SHP-2, which will inhibit the activation of immune cells by dephosphorylation of proteins containing immunoreceptor tyrosine-based activation motifs (10) ( Figure 1 ). Meanwhile, the dephosphorylation cascade initiated by CD47 inside phagocytes lead to deactivation of myosin-II and thereby preventing contractile engulfment (16). Molecular modeling revealed that a VMM motif in C-terminus of TSP-1 is an optimal locus to bind to CD47 (17). TSP-1-CD47 interaction regulates multiple biological functions such as cellular migration, angiogenesis, adhesion, cell aging (18). 4N1K, a peptide corresponding to the CD47 binding region of TSP on IL-12 and TNF-α production, was reported to mediate significant biological effects that were CD47‐independent (19–21). PKHB1, the first‐described serum‐stable soluble CD47‐agonist peptide, directly induced T‐leukemic cell death by engaging the CD47 receptor (20).

CD47 was originally also known as integrin-associated protein (IAP) in earlier studies because of its interaction with aIIbβ3 and a2β1 and αvβ3 integrins (22). In this manner, CD47 functions as a key regulator of migration of smooth muscle cell and platelet activation, etc. (6, 23).

There are other SIRP-family members sharing significantly conserved amino acid sequences within the extracellular domains but different signaling potentials, which are recognized as “paired receptors” (10, 24). In addition to aforementioned SIRPα, the SIRP family have the other members, SIRPβ1, SIRPβ2, SIRPγ, and SIRPδ (13, 25). Among these members, only SIRPγ exhibit a lower affinity with CD47 than that for SIRPα ( Figure 2 ).

In addition, several cytoplasmic binding partners of CD47 have been identified, including protein linking IAP and cytoskeleton-1 (PLIC-1), ubiquilin-2 (PLIC-2), vascular endothelial growth factor receptor (VEGFR2), and exportin-1 (18, 29–31). These binding receptors are being exploited in the design of new immune-therapeutic approaches and more robust data are yet to follow on clinical significance of them.

In earlier times, CD47-SIRPα interaction was best known for its role as a negative regulator during erythrocytes clearance (32, 33). The original evidence for this came from compelling research in which CD47 knockdown red blood cells, which were quickly cleared by spleen macrophage (33). In contrast, normal erythrocytes can survive for more than 40 days in mice by the reason of CD47 expression (33). Subsequently, it was found that aging erythrocytes have some conformational changes (possibly induced by oxidative modification) in CD47, which change from a “don’t-eat me” configuration into an “eat-me” signal (12). By this token, CD47 not only serves as a pro-phagocytic signal but also behaves as a molecular switch regulating erythrocyte homeostasis ( Figure 3 ). This process is thought to be highly associated with Gaucher disease, which is a genetic disorders with glucocerebroside accumulations in cells (40). Comparative analyses of membranous CD47 expression showed that erythrocytes from untreated Gaucher patients have week CD47 expression but can be overturned upon enzyme-replacement remedy (41).

Kusakari and colleagues’ study revealed that CD47-SIRPα interaction lead to the internalization of partial membrane by contiguous cells (42). Then it is proved that CD47 did not inhibit, but rather facilitated both phosphatidylserine-independent and phosphatidylserine-dependent uptake of apoptotic cells in the murine system (43). These results confirm prior evidence that CD47 is responsible for removal of some cells rather than just one negative regulator of phagocytosis.

In the current era in oncology, checkpoints immunotherapy of hematopoietic and solid malignancies are becoming a promising mode of treatment for cancer patients (74, 75). But the fact that not all cancer patients benefit from adaptive checkpoints immunotherapy catalyzed enormous interests in the targeting of novel immune checkpoint receptors. Considering innate immune system is the first line of defense directly target cancer cells, harnessing innate immunity represents another potential therapeutic avenue for cancer treatments. Innate checkpoints have recently received some attention regarding their possible roles in tumor-mediated immune escape. A list of innate checkpoints of interest are under investigation, including CD47-SIPα axis, TAM family (Tyro3, Axl, and MerTK), Siglec-9, CD24-Siglec-10 axis (76–79).

Given the ubiquitous expression of CD47 on normal cells, potential side effects such as anemia will need to be considered though several clinical trials are underway and have produced impressive clinical results in recent years ( Figure 6 ).

In several preclinical solid tumor models, tumor-specific delivery of CD47 blockade demonstrated superior accumulation and fewer side effects at tumor sites compared with systemic administration (118) (119). Novel drug delivery system has emerged as a critical modulator in the development of anti-CD47 therapeutics, such as nanomedicine and gel matrix (117, 120–122).

Bispecific antibodies are helpful for the purpose to limit bystander effects due to widespread CD47 expression. Heretofore, some bispecific antibodies have been developed with one arm binding and blocking CD47 to prohibit its interaction with SIRPα, and the other arm binding tumor-associated antigens such as CD19, CD20, and PD-L1 (123). In comparison to other CD47 antibodies, bispecific antibody was engineered to have a lower affinity towards CD47, limiting binding activity with normal cells expressing CD47. Here, proof was provided by an anti-CD47/CD19 bispecific antibody that selectively block the CD47-SIRPα on malignant cells expressing CD19 and interact weakly with normal cells expressing CD47 (124). Such a bispecific design could be applied to limit the negatively impacts caused by “antigen sink”.

In addition to drug delivery system and bispecific antibody, there are many other strategies to prevent anemia caused by anti-CD47 therapeutics: gradually increase therapeutic dose (125); differentiate erythrocytes CD47 and tumor cell-expressing CD47 (114, 126–130); sacrifice antibodies mediated ADCP (antibody-dependent cellular phagocytosis) (131–133); pro-body technology (134) (120). Apart from anemia, issues with thrombocytopenia, hyperbilirubinemia, and neutropenia have also limited the use of anti-CD47. For further reading on this field, we recommend to the curious reader one excellent review by Yuchen W et al. (135). The review also high-light recent advances in tumor therapy targeted on the CD47-SIRPα axis and provides ideas for further clinical transformation.

Although early data with agents targeting either CD47 or SIRPα have highlighted the therapeutic potential by sending a potent “don’t eat me” signal to prevent phagocytosis, no Phase III study convincingly demonstrated the efficacy of these novel therapeutic targets for cancer treatment so far. ENHANCE (NCT04313881), a randomized, double-blind, placebo-controlled multicenter Phase III study from Forty-Seven (Gilead Sciences) aimed to compare the effects of treatment with magrolimab plus azacitidine and placebo plus azacitidine in untreated patients with myelodysplastic syndrome, has recently resumed after FDA lifted the partial clinical hold placed on the item. This is the only Phase III clinical trial related to CD47 blocking therapy. Inhibitors and antibodies against CD47 in immunotherapy have several limitations to emerge, and some features of the therapy must be considered.

The first limitation is to search for safe potent antibodies that are not highly bound to erythrocytes at early step in terms of safety. Earlier detection of the potential toxicity during antibody development can help cut costs and improve success rate. Dual targeting bi-specific antibodies of both TAA (tumor-associated antigens) and CD47 can also be designed to resolve this issue, which can direct CD47 blockade selectively to cancer target cells, thereby improving the safety of CD47 blocking therapy (136) (137).

Secondly, immunotherapy with CD47 blockade has achieved impressive successes in treatment of hematological malignancies. But the clinical efficacy of CD47 blockades in solid tumors has been much less rewarding due to the tumor heterogeneity and intricacies of the tumor microenvironment. Compared with macromolecule agents, small molecule inhibitors are more penetrative to the targets and have broad prospects for development and application in solid tumors (138). For example, a humanized CD47/GPC3 (Glypican-3) BsAb enhances the Fc-mediated effector function and has significantly better efficacy than the combined treatment using CD47 and GPC3 monoclonal antibodies in the hepatocellular carcinoma model of xeno-transplantation (139). Apart from bispecific antibodies, nanobodies have also received widespread attention due to their deeper penetration in solid tumors. HuNb1-IgG4, a new type of CD47 nano-antibody that has a low affinity for human erythrocytes, enhances tumor phagocytosis mediated by macrophages in vitro, and shows strong anti-tumor activity in ovarian cancer and lymphoma (103). Thirdly, CD47 can also be used in other therapeutic areas, including atherosclerosis, neurological disorders, and autoimmune diseases. By administration of CD47-blocking antibodies, the atheroprone mice would develop significantly smaller atherosclerotic plaques compared to IgG controls (140). A potential role for CD47 in Alzheimer’s disease has been identified with studies showing CD47 facilitates Aβ oligomers internalization by microglia (141, 142). Even in the infancy of understanding its diverse function and potentiality, CD47 will be further revealed as an important mediator in these therapeutic areas.

Inspiring progress has been made with respect to the understanding of the role of CD47 in normal cell growth cycle as well as in the molecular pathogenesis of diverse diseases. CD47 serves as a receptor for TSP, integrins, and SIRP family members and has been shown to trigger a wide variety of cellular functions. For example, several lines of evidence suggests CD47 ligation appears to co-stimulate T cell proliferation and induce their arrest (54).

Utilizing anti-CD47 antibodies open up exciting avenues in cancer, but there are still unsolved questions in this field. First, the underlying mechanisms need to be elucidated. What is the importance of FcR engagement during CD47 blockade? How to avoid an uncontrolled immune response since the ubiquitous expression of CD47? The cell specific function of CD47 is not fully understood so far and its downstream signaling is still unclear. Secondly, there exists more than one binding receptor of CD47. It, therefore, would be challenging to decide what are the specific contribution of each partner in this intricate network? Thirdly, the question remains whether NK cells or other immune cells contribute to the effect of CD47 blockade rather than macrophage. Finally, some researchers asserted that anti-CD47 antibody did not induce apparent side effects in animal models, but whether if it is equally safe in human body still needs further elucidation. In conclusion, though there is still long way to go about mechanistic issue of CD47 blockade, immunological and clinical studies are yielding impressive results in this field. An exciting new era of innate checkpoint strategy targeting CD47 is likely to come especially when we fully understand underlying mechanisms.

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.