The CD47 protein, a member of the membrane protein IG superfamily, is ubiquitously expressed on varieties of types of cellular membranes, especially on senile erythrocytes and cancer cells (Hayat et al., 2020). Its molecular structure includes a single IgV-like extracellular domain at the N terminus, a highly hydrophobic stretch with five membrane-spanning sections and an alternative splicing cytoplasmic domain at its C terminus (Brown and Frazier, 2001; Mushegian, 2002). By interacting with integrin and TSP-1, CD47 is involved in a variety of physiological processes, such as migration, adhesion, proliferation, differentiation (Lindberg et al., 1996; Liu et al., 2001; Lymn et al., 2002). As an inhibitory receptor, CD47 can bind with SIRPα to not only inhibit phagocytosis by phagocytes but also inhibit the activation and maturation of dendritic cells (DCs) (Oldenborg et al., 2000; Latour et al., 2001; Lutz and Bogdanova, 2013). In addition, CD47-SIRPα also regulates neuron development and bone remodeling (Barclay, 2009; Maile et al., 2011).

SIRPα, a member of the Ig superfamily (IgSF), consists of three Ig-like extracellular domains at its N terminus four tyrosine phosphorylation sites and two immunoreceptor tyrosine inhibitory motifs (ITIMs) in its cytoplasmic domain (Adams et al., 1998; Hayat et al., 2020). In contrast to the ubiquitous expression of CD47, SIRPα is limitedly expressed on macrophages, monocytes, granulocytes and neurons (Adams et al., 1998).

The N-terminal Ig-like extracellular domain of SIRPα binds with the N-terminal IgV-like extracellular domain of CD47, resulting in the phosphorylation of ITIM of SIRPα and recruitment and activation of protein tyrosine phosphatases, especially Src homology 2 including SHP-1 and SHP-2, dephosphorylation of the downstream molecule ITAM, the accumulation of myosin IIA damage in the phagocytic synapse, releasing the “don’t eat me” signal, leading to an inhibition of phagocytosis (Hayat et al., 2020; Logtenberg et al., 2020; Figure 1).

Homeostasis of cells and tissues depends on a balance of regulation of pro-phagocytic signals [calreticulin (CRT)-low density lipoprotein-receptor related protein-1 (LRP-1), Fcγ, complement receptor] and anti-phagocytic signals (CD47-SIRPα) (Oldenborg et al., 2001; Chao et al., 2010a). Chao et al. demonstrated that CRT plays a leading role in pro-phagocytic signals and is essential for anti-CD47 antibody therapy in multiple human cancers. In their study, the in vitro phagocytosis assays were performed by incubating primary human normal cells and cancer cells with human macrophages with a therapeutic dosage of anti-CD47 antibody, and showed that primary cancer cells were obviously phagocytized, whereas no phagocytosis of normal cells was observed, suggesting that blocking the CD47-SIRPα is not the only rationale for pro-phagocytosis (Chao et al., 2010a). Note that the anti-CD47 antibody with intact Fcγ should be utilized with caution given the pro-phagocytosis role of Fcγ, which could increase systemic toxicity by enhancing antigen sink effects (Ingram et al., 2017).

Increasing studies have concluded that the antitumor effect mediated by blocking the CD47-SIRPα interaction mainly owes to the activation of innate immune responses [including phagocytosis by macrophages and the antibody dependent cellular cytotoxicity (ADCC) by neutrophil granulocytes] (Jaiswal et al., 2009). However, it is important to note that the results of these works were based on xenograft models, which may favor innate immune responses to kill tumor cells with some unique features (Majeti et al., 2009; Chao et al., 2010b; Willingham et al., 2012). Liu et al. used syngeneic immune-competent mouse models to exclude these effects (Liu et al., 2015a). In this study, the mouse anti-CD47 antibody showed an evident antitumor effect, especially by intratumoral delivery, and the therapeutic effect was diminished when CD8⁺ T cells were depleted. In addition, CD47-SIRPα blockade activates the maturation of DCs and boosts DC-mediated antigen cross-presentation and cytotoxic T lymphocyte induction. Hence, the CD47-SIRPα signaling axis is an inhibitory checkpoint that bridges innate and adaptive immunity for tumor evasion.

According to the different signaling pathway blocking sites, the strategies for inhibiting the CD47-SIRPα interaction can be divided into three types: molecules that inhibit the CD47 protein on the tumor cells, molecules that inhibit SIRPα protein on the myeloid cells, and inhibitors of the glutaminyl-peptide cyclotransferase-like (QPCTL) enzyme, which is necessary for the maturation of CD47 protein (Logtenberg et al., 2020).

The strategy targeting CD47 on tumor cells has been the most commonly studied. For instance, Hu5F9-G4, a humanized anti-CD47 antibody with a human immunoglobulin G4, has been proved a potent antitumor effect in preclinical experiments and clinical trials. In a malignant pediatric brain tumor-bearing mouse model, administration of Hu5F9 evidently inhibited tumor growth and showed significant survival benefit (Gholamin et al., 2017). However, CD47 is not only over-expressed on cancer cells, but also expressed on normal cells, such as ethrocytes. Therefore, administration of CD47 blocking agents would lead to anemia and “antigen sink” effect. In the development of CD47 targeting agents, multiple approaches have been employed to solve these problems, such as change in the mode of administration (Liu et al., 2015b; Advani et al., 2018; Sikic et al., 2019); dual targeting bispecific antibodies of CD47 (Dheilly et al., 2017; Shi et al., 2020; Wang et al., 2021); CD47 antibodies/SIRPα fusion proteins (Petrova et al., 2017; Meng et al., 2019; Puro et al., 2020; Andrejeva et al., 2021). From the efficacy of view, dual targeting bispecific antibodies are more promising. Wang et al. designed a CD47-PD-L1 bi-specific antibody, named IB322 (Wang et al., 2021). As a dual inhibitor of innate and adaptive immune checkpoint, IB322 efficiently triggered the tumor cell phagocytosis by macrophages and killing effect by T cells and induced complete tumor regression in vivo. Moreover, IB322 showed negligible RBCs depletion and was well tolerated in non-human primates.

Compared with the wide expression of CD47, SIRPα is restrictedly expressed on myeloid cells and neurons (Adams et al., 1998). Hence, biologicals that target SIRPα do not suffer from amenia and “antigen sink” issues (Yanagita et al., 2017; Voets et al., 2019). For example, Ho et al. developed an engineered, high-affinity, CD47 variant (termed Vecro-CD47), which could remarkably increase the affinity to wild-type (WT) SIRPα and disrupt the CD47-SIRPα interaction, thereby promoting macrophage phagocytosis of tumor cells (Ho et al., 2015). Voets et al. developed a humanized mAb ADU-1805, which inhibits the CD47-SIRPα signaling pathway by closely binding with SIRPα, showing similar antitumor efficacy as the anti-CD47 antibody with good safety in vitro and in vivo (Voets et al., 2019). However, researches and clinical trials focused on SIRPα blocking target are fewer than that on CD47 blocking target.

Inhibitors of QPCTL enzyme is another promising strategy that does not result in anemia easily and “antigen sink” issues (Ingram et al., 2017). Logtenberg et al. reported that both genetically and pharmacologically blocking QPCTL activity enhanced antibody-dependent cellular phagocytosis (ADCP) and ADCC of tumor cells (Logtenberg et al., 2019). Moreover, the intervention of QPCTL activity can alter the immunosuppressive tumor microenvironment (monocyte skewed, myCAF, TGF-β) to a proinflammatory (macrophage skewed, iCAF, IFN) milieu, and enhances the therapeutic effect of anti-PD-L1 therapy (Bresser et al., 2022).

To date, over 20 CD47/SIRPα blocking agents have been employed in clinical trials (summarized in Table 1), involving in both hematological malignancies and solid tumors. However, the current clinical trial data of QPCTL inhibitors in antitumor treatment are lacking.

Although promising, the CD47-SIRPα blocking agents still face some challenges that restrict their translation to clinical settings. For instance, the ubiquitous expression of CD47 indicates that large dose or frequent administration of anti-CD47 antibodies is required (eg: antigen sink effect) (Chen et al., 2022), suggesting the efficacy of anti-CD47 antibodies treatment is relatively low. With regard to targeting bispecific antibodies of CD47 and CD47 antibodies/SIRPα fusion protein technology, while promising, it requires complex design and isolation (Chen et al., 2013; Labrijn et al., 2019). Therefore, the cost of these therapies is usually unaffordable for patients, which limits their clinical applications (Chen et al., 2013; Labrijn et al., 2019). As for SIRPα targeting strategy, it constantly fail to induce ADCP and ADCC against cancer cells when administrated alone due to immune cells target (Chen et al., 2013; Ring et al., 2017; Labrijn et al., 2019; Chen et al., 2022). The demand for safer and more efficient drug delivery is therefore increasing.

Nanomedicine, defined as the application of nanotechnology, can meet this need. Nanotechnology enables therapeutic drugs to target sites with high spatial and temporal resolution, prolonged half-life and great convenience for combination therapy. Therefore, CD47-SIRPα targeting based nanomedicine holds great potential in antitumor field, which will be reviewed in more detail in the following sections.

Nanoparticles (NPs) are important components of nanomedicine. The unique characteristics of NPs, such as large surface-volume ratio, small size, capacity to encapsulate various drugs, and tunable surface chemistry, provides themselves a large variety of advantages, including multivalent surface modification, efficient navigation in vivo, increased intracellular trafficking and sustained release of drug payloads (Xu et al., 2015). Currently, diverse types of NPs exist, including liposomes (Huynh et al., 2009; Wang et al., 2016; Olusanya et al., 2018; Yang et al., 2021), micelles (Torchilin, 2007; Tawfik et al., 2020), poly (lactic-co-glycolic acid) (PLGA) (Sadat Tabatabaei Mirakabad et al., 2014; Rezvantalab et al., 2018), graphene (Diez-Pascual, 2020), graphene oxide (Kinnear et al., 2017; Diez-Pascual, 2020), protein nanoparticles (Lohcharoenkal et al., 2014; Jain et al., 2018), extracellular vesicles (EVs) (S et al., 2013; Si et al., 2022; Logozzi et al., 2021), exosomes (De La Peña et al., 2009; Nie et al., 2020; Xia et al., 2020; Logozzi et al., 2021), magnetic NPs (MNPs) (Colombo et al., 2012; Wu et al., 2019; Farzin et al., 2020), mesoporous silica NPs (MSNPs) (Fu et al., 2013; Wang et al., 2015; Pelaz et al., 2017; Rastegari et al., 2021), and metal-organ frameworks (MOFs) (Zheng et al., 2016; Wu and Yang, 2017; Xing et al., 2020), Ferritin (Lee et al., 2017; Cho et al., 2018; Sun et al., 2021). Detailed information about the charaterizations, advantages and disadvantages of each type of NP is summarized in Table 2. As summarized in Table 2, while each nanocarrier possesses unique merits, they still face certain problems that restrict their optimal performance in the drug delivery system.

Targeted delivery capacity, including passive targeting and active targeting, is a key feature of nanomedicine and has been vastly studied.

Passive targeting is achieved by the enhanced permeability and retention (EPR) effect, prolonging the half-life of NPs in the body circulation (Maeda et al., 1999; Fang et al., 2003). Thus, prolonging the circulation time of NPs is a good strategy to increase their accumulation at the tumor site. The stealth modification of NPs is of great importance. Camouflaging the surface with polyethylene glycol (PEG) is the most common way to create a protective layer for encapsulated cargo by reducing the absorption of plasma proteins and extending the half-life of NPs (Gref et al., 2000). In addition to PEGylation, new biomaterials and drug delivery strategies have been developed to prolong the circulation time of NPs, including zwitterionic polymer-coating (Zhu et al., 2014), minimal self-peptides (e.g., CD47-derived self-peptides) (Rodriguez et al., 2013), and biomimetic membrane-coating (Hu et al., 2011). Nevertheless, passive targeting is far to reach the requirement of therapeutic efficacy. Due to the phagocytosis of mononuclear phagocytes, the majority (more than 90%) of NPs are inevitably entrapped by reticuloendithelial organs, such as liver and spleen (Albanese et al., 2012).

To overcome this severe drawback, measures that make nanomedicine actively target the disease site should be taken to increase the accumulation of drugs at the target site and subsequently enhance the therapeutic efficacy. Active targeting is a strategy to achieve the goal of orientation in space and simultaneously eliminates the off-target effect in normal tissues by intentionally guiding NPs to the disease site. A common approach to active targeting is to decorate the appropriate ligands to the surface of NPs. These ligands interact with the surface receptor of target cells inducing receptor-mediated endocytosis (Chen et al., 2017). Target agents can be broadly categorized as proteins (mainly antibodies and their fragments), nucleic acids (aptamers), or other receptor ligands (Peer et al., 2007).

To release the drug at a specific site and time, various efforts have been made to develop stimuli-responsive NPs, which further enhance the therapeutic efficacy (Li et al., 2020). These stimuli-responsive NPs can be stimulated by either endogenous stimuli-responsive strategies, such as pH variation, redox, enzyme, hypoxia, or exogenous stimuli-responsive strategies, such as light, ultrasound, magnetic field, temperature (Tawfik et al., 2020). However, these single stimuli-responsive strategies still face some challenges. For example, several temperatue- and light-responsive agents can damage normal cells and even tissues and organs (Wu et al., 2018). Due to insufficient H₂O₂ levels in tumor tissues, the nonspecificity and low therapeutic efficiency of H₂O₂-responsive nanoplatforms are also key challenges for clinical translation (Chang et al., 2017). Nanoassemblies activated by both exogenous stimuli and endogenous stimuli have gained tremendous attention by virtue of the enhanced encapsulated payload and the higher accuracy of spatiotemporal release. For instance, chen et al. developed a photothermal-PH-hypoxia responsive multifunctional nanoplatform (TENAB NP) for cancer photochemotherapy, for synergistic chemo-phototherapy with minimized skin photosensitization (Chen et al., 2019a). In this multistimuli responsive drug delivery system, tirapazamine, the hypoxia-specific prodrug, and ENAB, the pH-responsive photosensitizer, were encapsulated into the phase change materials (LASA), a mixture of linoleic acid and stearyl alcohol.

The upregulated CD47 on tumor cells increases its interaction with SIRPα on macrophages, resulting in an evasion of immunological surveillance and the muturation of DCs (Willingham et al., 2012; Jaiswal et al., 2009; Liu et al., 2015a; Liu et al., 2017). This inhibitory checkpoint paves the way for therapeutic strategies involving blocking this interaction to enhance innate and adaptive immunity for tumor killing (Figure 2). For instance, Koh et al. constructed an exosome through surface engineering with SIRPα variants termed SIRPα-exosomes, which can bind to both human and mouse CD47 as antagonists. The therapeutic efficacy of these engineered exosomes was verified in HT27 tumor-bearing mice (including immunodeficient and immunocompetent mice). Systemic administration of SIRPα-exosomes induced significant regression of tumor growth in immunocompetent mice, while tumor growth was slightly reduced in immunodeficient mice, suggesting that T-cell immunity might be essential to maximize the antitumor effect of CD47 blockade therapy (Koh et al., 2017). Ramesh et al. developed a multivalent lipid-based phagocytosis nanoenhancer with the conjugation of anti-CD47 and anti-SIRPα antibodies (LPN). LPN treatment showed remarkable tumor growth suppression and increased survival in B16F10 tumor-bearing mice with no systemic toxicity (Ramesh et al., 2020).

However, cancers are very complex diseases involving multiple pathways, and their progression is associated with various continuous mutations in cell lines. In addition, in order to survive, tumor cells mutate as chemotherapy progresses, thereby resulting in intrinsic and acquired resistance to chemotherapeutics (Iyer et al., 2013). Hence, the inhibition of the CD47-SIRPα signaling pathway is not sufficient to fight against tumor, and often requires a combination of blockades of different pathways, genes or chemotherapeutics (Xu et al., 2015). The advantages of nanomedicine, such as the capacity of encapsulating different drugs, targeted delivery and controlled release, offer a great opportunity for combination therapy for tumors. Some examples to show how the combination of blockade of the CD47-SIRPα interaction and other therapeutics or modalities works in nanomedicine are listed in Table 3 and explained in the next section.

Recent advances in tumor immunology suggest that the antitumor effect of blocking the CD47-SiRPα signaling pathway may be discounted by the immunosuppressive tumor microenvironment (TME) (Noy and Pollard, 2014; Chen et al., 2019b). In particular, colony-stimulating factors, secreted by tumor cells, are abundant in the TME, polarizing TAMs to the tumorigenic M2 phenotype (Noy and Pollard, 2014; Chen et al., 2019b). M2 TAMs can recruit regulatory T cells (Tregs) and secrete proinflammatory cytokines, all of which impair the activation of CD47 blockers against tumor T-cell immunity (Mantovani et al., 2017; Kulkarni et al., 2018; Feng et al., 2019). In this context, blocking the CD47-SIRPα signaling axis while polarizing tumorigenic M2- to anti-tumor M1-phenotype TAMs can improve the antitumor effect of CD47 immune checkpoint inhibitors (Ramesh et al., 2019; Rao et al., 2020a; Rao et al., 2020b; Chen et al., 2020; Nie et al., 2020; Chen et al., 2021; Huang et al., 2021). For example, Rao et al. have developed a genetically engineered cell-membrane-coated magnetic nanovehicle (gCM-MNs). The gCM shell genetically overexpressing SIRPα variants with prominent affinity and efficiently inhibits the CD47-SIRPα signaling axis, which can also protect the MN core from macrophage phagocytosis. The MN core promotes M2 TAM repolarization, synergistically triggering potent macrophage immune responses. Moreover, the MN core delivers the gCMs into tumor tissues under magnetic navigation, effectively promoting their tumor accumulation. In melanoma and triple-negative breast cancer models, gCM-MNs remarkably extended overall survival by inhibiting local tumor growth and distant metastasis (Rao et al., 2020a; Figure 3).

Many studies have investigated the efficacy of chemotherapy as an adjuvant to immunotherapy, suggesting that the most significant potential mechanism of such adjunct is immunogenic cell death (ICD) (Zitvogel et al., 2008; RA and BW, 2005; Hou et al., 2013). ICD has a number of clearly defined physiological characteristics, including cell surface CRT expression, release of DAMPs such as adenosine triphosphate and heat shock proteins, and release of high mobility group box 1 (Liu et al., 2016). Among these factors, the surface expression of CRT is considered to be the single most important element of ICD (Abdel-Bar et al., 2021). As previously mentioned, the effect of CRT exposure, serving as an “eat me” signal, is considered to be counterbalanced and potentially dampened by CD47 expression (Chao et al., 2010a). Moreover, the upregulated expression of CD47 on the surface of tumor cells makes it an active targeting site for tumor cells, facilitating nonspecific distributed ICD inducing drugs to target tumor tissues and reduce systemic toxicity (Tang et al., 2021). Therefore, codelivery, which simultaneously removes an inhibitory signal and introduces an activating signal, can produce an enhanced antitumor effect (Abdel-Bar et al., 2021; Lee et al., 2021). For example, Abdel-Bar et al. reported a stable nucleic acid-lipid particle (SNALP) formulation with the simultaneous delivery of an ICD inducing drug (Dox) with small interfering RNA (siRNA) knocking down CD47 (siCD47) for synergistic enhancement of ICD. In a CT-26-tumor-bearing mouse model, SNALPs synergistically inhibited tumor growth and prolonged the survival (Abdel-Bar et al., 2021).

EGFR is overexpressed in various solid tumors, such as breast, renal, colon, head and neck cancer (Normanno et al., 2001; Gazdar, 2009; Yu et al., 2013) Hence, EGFR targeting strategy is a promising way for antitumor treatment (Mendelsohn and Baselga, 2003; Yu et al., 2013; Sabbah et al., 2020). CD47 is also overexpressed on the surface of multiple tumor cells. Therefore, dual targeting to EGFR and CD47 strategy can efficiently target and inhibit tumor growth. For example, Si et al. constructed anti-EGFR/CD47 mAb marked EV which showed a high anti-TNBC efficacy with negligible toxicity in both 4T1 tumor-bearing mouse models and TNBC patient-derived xenograft models (Si et al., 2022).

Programmed cell death-ligand 1 (PD-L1) blockade therapy has achieved exciting success in the clinic (Pardoll, 2012). PD-L1, which is highly expressed in many tumor cells, sends a “don’t find me” signal to the adaptive immune system, inhibiting T-cell activation by engaging the PD-1 receptor (Parsa et al., 2007). CD47 sends a “don’t eat me” signal to the innate immune system, inhibiting the phagocytosis of macrophages by engaging SIRPα (Hayat et al., 2020; Logtenberg et al., 2020; Zhang et al., 2020). Hence, dual-blockade of PD-L1 and CD47 can activate potent antitumor effects via both innate and adaptive immune responses (Chen et al., 2019c; Lian et al., 2019; Meng et al., 2021; Pham et al., 2021; Zhao et al., 2021). For example, Meng et al. designed genetically programmable fusion cellular vesicles (Fus-CVs), which were fused by SIRPα variants and PD-1 variants. This bispecific targeting design improves the targeting of tumor cells while reducing the adverse off-target effect on normal cells. In malignant melanoma and mammary carcinoma models, Fus-CVs synergistically suppressed postsurgery tumor recurrence and metastasis, thereby improving overall survival (Meng et al., 2021).

PTT is a promising cancer treatment modality. PTT-induced hyperthermia can be controlled through the local use of photosensitizers and minimally invasive near-infrared (NIR) radiation to reduce damage to untargeted tissues (Sica et al., 2006; Chu and Dupuy, 2014; Chen et al., 2015). Recently, increasing studies have demonstrated that hyperthermia can induce dying tumor cells to release massive amounts of cytokines, such as IL-1β and TNF-α, promoting the immune responses of macrophages, NK cells and T lymphocytes. Yet, it is difficult to completely eradicate large tumors with conventional PTT due to residual tumor mass at the treatment margins (Mantovani et al., 2006; Chu and Dupuy, 2014; Shim et al., 2017). Therefore, researchers combined CD47 blockers with PTT to synergistically enhance the antitumor effect (Wu et al., 2015; Guo et al., 2019; Xie et al., 2020; Zhan et al., 2021). For example, Guo et al. reported bismuth selenide nanoparticles conjugated with anti-CD47 antibody and coated with PEG (Ab-PEG-Bi2Se3). In the 4T1 tumor-bearing model, Ab-PEG-Bi2Se3 plus PTT synergistically eradicate the tumor (Guo et al., 2019).

Reducing the capture of NPs by reticuloendothelial organs (such as liver, spleen, and lung) and extending their circulation time in the blood to accumulate more NPs in tumor tissues have been challenges. Currently, there are many approaches to prolong the half-life of NPs in blood, such as PEG surface modification and bionic membrane coating techniques described above. However, these approaches still have some disadvantages and limitations. Therefore, more suitable alternatives are urgently needed. The CD47 protein, as a “self” marker, can evade phagocytosis by the CD47-SIRPα interaction (Logtenberg et al., 2020). With regard to the pivotal role of CD47 in the regulation of immune responses, the present paper outlines emerging methods for the production of bioinert biomaterials and NPs using CD47 (Gheibi Hayat et al., 2019; Figure 4) Examples of stealth functionalization by CD47 mimicry utilized for antitumor nanomedicine are listed in Table 4 (Kamerkar et al., 2017; Shim et al., 2017; Jiang et al., 2018; Song et al., 2019; Wang et al., 2019; Belhadj et al., 2020; Cheng et al., 2021; Tang et al., 2021; Xie et al., 2021; Zhang et al., 2021). For instance, Tang et al. designed a precise delivery nanomedicine to M2 macrophages by combining “eat me/don’t eat me” signals and verified its role in antitumor therapy in an A20 subcutaneous tumor mouse model. In this delivery system, CD47-derived self-peptide ligand and galactose ligand were introduced on liposomes to reduce the phagocytosis of M1 macrophages and enhance the uptake of M2 macrophages, respectively. Cleavable phospholipid-PEG covering on the surface of liposomes can be removed by the redox microenvironment upon transcytosis through the tumor endothelium and re-expose the self-peptide and galactose. Therefore, this nanocarrier can precisely target M2-type TAMs. In addition, DOX loaded into liposomes further enhances its antitumor effect (Tang et al., 2021; Figure 5)