Targeting CD47/SIRPα as a therapeutic strategy, where we are and where we are headed

Immunotherapy using PD-1 and CTLA4 inhibitors to stimulate T cell immunity has achieved significant clinical success. However, only a portion of patients benefit from T cell-based immunotherapy. Macrophages, the most abundant type of innate immune cells in the body, play an important role in eliminating tumor cells and infectious microbes. The phagocytic check point protein CD47 inhibits the phagocytic activity of macrophages through binding to SIRPα expressed on macrophages. Blockade of the interaction between CD47 and SIRPα could restore phagocytic activity and eliminate tumor cells in vitro and in vivo. In this manuscript, we review the mechanism of action and development status of agents (antibodies targeting CD47 and SIRPα, SIRPα-Fc fusion proteins, and bi-specific antibodies) that block CD47/SIRPα interaction in preclinical studies and in the clinical setting. In addition, small molecules, mRNA, and CAR-T/M that target the CD47/SIRPα axis are also reviewed in this article.


Introduction
Tumor cells evade immune destruction by transmitting inhibitory signals to lymphocytes and myeloid cells [1]. Blockade of these inhibitory molecules, which include CTLA4, PD-1, and PD-ligand 1 (PD-L1), could restore T cell function and promote elimination of tumor cells. Immune checkpoint inhibitors (ICIs) have improved outcomes for patients with multiple types of cancers. However, many patients do not respond to this type of immunotherapy [2,3]. In some cases, these therapeutic agents have been associated with disease progression [4,5], the cause of which is currently being investigated. Therefore, drugs that act on a novel class of targets to immobilize a broader immune cell population are needed to improve upon current therapeutic options.
Macrophages [6] are typically the first dedicated innate immune cells to detect the presence of infectious pathogens or tumor cells. Macrophages are derived from monocyte precursors that circulate in the blood and migrate into tissues, after which they differentiate into tissue macrophages such as Kupffer cells in the liver, alveolar macrophages in the lung, and microglia in the brain. Circulating monocytes and resident tissue macrophages can directly kill tumor cells via phagocytosis (innate immune response) and can activate the adaptive immune response. However, these immune responses can be inhibited by ligand binding to inhibitory receptors expressed on the macrophage cell surface [7]. Signal regulatory protein alpha (SIRPα) is a transmembrane protein expressed on all myeloid cells, including monocytes, macrophages, and neutrophils. SIRPα contains immunoreceptor tyrosine-based inhibition motifs (ITIMs) that can be phosphorylated, resulting in recruitment of inhibitory molecules such as Src homology 2 (SH2) domain-containing protein tyrosine phosphatase (SHP)-1 and SHP-2 [8]. Binding of CD47 to SIRPα triggers coupling of SIRPα to these phosphatases, resulting in inhibition of phagocytic activity [9,10]. CD47 is ubiquitously expressed on many types of cells to prevent phagocytosis by phagocytes. However, tumor cells overexpress CD47 to evade the immune system through inhibition of myeloid cell-mediated elimination [11]. Inhibition of CD47-SIRPα interaction restored the phagocytic activity of phagocytes in vitro and in vivo [12][13][14][15]. Targeting the CD47/SIRPα axis has become a promising strategy to promote tumor elimination through innate immunity. This review focuses on development, safety, and efficacy of agents that target the CD47/SIRPα axis in preclinical and clinical studies.

CD47/SIRPα: the molecules and biology
SIRPα [16,17], also named SHPS-1 or CD172a, is a transmembrane glycoprotein mainly expressed on neurons and myeloid cells that is particular enriched on macrophages. Human SIRPα is coded by the SHPS-1 gene located at human chromosome 20p13. The open reading frame region is composed of eight exons, including a signal peptide, extracellular domain, a transmembrane segment, and three parts of one cytoplasmic domain. The extracellular domain consists of three Ig-like regions, an NH2-terminal immunoglobulin (Ig) variable (V) region (domain 1, D1), and two Ig constant (C) regions (domain 2 and 3). The cytoplasmic region contains two immunoreceptor tyrosine based inhibitory motifs (ITIMs) and a proline-rich region (YYYY), which bind to Src homology (SH2) domain-containing molecules.
CD47 is a 52 kD transmembrane glycoprotein belonging to the immunoglobulin superfamily. Human CD47 is encoded by the CD47 gene located at the q13.12 region of chromosome 3. Human CD47 contains an NH 2 -terminal Ig variable-like extracellular domain (ECD), a 5-transmembrane spanning helical bundle domain, and a short intracellular COOH-terminal domain (CTD) [18]. CD47 is an essential component of the innate immune system, and binding of its extracellular domain with its ligands αVβ3, SIRPα, and thrombospondin-1 (Tsp-1) activates different signaling pathways that control cell proliferation and differentiation, angiogenesis, and immune regulation. The CTD is alternatively spliced and can exist as four isoforms, ranging from 4 to 36 residues. A schematic diagram of the compositions of CD47 and SIRPα proteins are shown in Fig. 1.

Therapeutic strategy
The critical role of the CD47/SIRPα axis in the innate immune response suggests that these two proteins may be attractive therapeutic targets. Antagonists targeting the innate immune checkpoint CD47/SIRPα pathway are currently in clinical development. These antagonists include 1) monoclonal antibodies targeting CD47 or SIRPα, 2) SIRPα-Fc fusion proteins, 3) bispecific antibodies (BsAb), 4) small molecules to down-regulate CD47 on tumor cells, 5) RNAi and, 6) CD47-chimeric antigen receptor-T cell/Macrophages.

Monoclonal antibodies and fc fusion proteins
Three types of agents targeted to the CD47/SIRPα axis were developed: antibodies, SIRPα-Fc fusion proteins targeted to CD47, and antibodies targeted to SIRPα. The mechanisms of CD47-SIRPα blocking agents are summarized in Fig. 3. Agents targeted to CD47 should block the CD47-SIRPα interaction to remove the anti-phagocytic signal and restore the phagocytic activity of macrophages [27]. In addition, engagement of FcRs to limit activity is considered to be necessary for agents targeted to CD47 [28]. In addition, anti-SIRPα antibodies using inert Fc to prevent toxicity resulting from SIRPα expressed on myeloid immune cell perhaps have therapeutic potential.
The second therapeutic mechanism is bridging innate and acquired immunity [29]. Tumor cells are recognized, taken up by antigen presenting cells (APCs, such as dendritic cells and macrophages), and presented to naive T cells, resulting in T cell activation. Antibodies targeted to CD47 could induce direct killing of tumor cells by inhibiting protein kinase A via Giα, resulting in clustering of CD47 in the membrane, and caspase-independent programmed cell death [30].

Anti-CD47 monoclonal antibodies
More than ten anti-CD47 antibodies are in different phases of clinical development ( Table 1). All of these antibodies are based on human IgG4-Fc, except AO-176, which is based on human IgG2-Fc [31]. Clinical study results of these antibodies were as follows (ordered by clinical trial developmental state).
Magrolimab (Hu5F9-G4) Magrolimab, previously known as Hu5F9-G4, was the first anti-CD47 antibody to enter clinical trials, and is currently in Phase III development. The clinical trials for magrolimab are listed in the Table 1. In a preclinical study [32], combined treatment with magrolimab and rituximab showed significant clearance of Raji cells in vitro and elimination of AML tumor cells in vivo. Magrolimab blocks the CD47/ Fig. 2 Biology of CD47/SIRPα interaction. CD47 binds to SIRPα to transmit inhibitory signals to macrophages and to inhibit or lessen phagocytic activity through uncoupling of receptor binding and signal transduction. Therapeutic agents block the interaction between CD47 and SIRPα to remove the inhibitory signal and restore the phagocytic activity SIRPα interaction and the anti-phagocytic signal on macrophages, and binding of rituximab to FcRs initiates pro-phagocytic signaling. Activation of pro-phagocytic signaling is a beneficial effect of anti-CD47 antibody treatment. Magrolimab caused significant hemagglutination and phagocytosis of RBCs in vitro, which indicated potential toxicity. Non-human primate pharmacokinetic and toxicology studies showed dose-dependent anemia. To overcome treatment-related anemia and thrombocytopenia, a low priming dose was given to stimulate production of new RBCs and to facilitate tolerance of subsequent higher maintenance doses. The same strategy (1 mg/kg priming dose on day 1) effectively controlled anemia during subsequent infusion of magrolimab in a clinical trial. The saturation concentration (receptor occupancy) on circulating white and red cells was 30 mg/ kg [33]. In a phase Ib [34] (NCT02953509) study evaluating relapsed or refractory non-Hodgkin's lymphoma, patients with diffuse large B-cell lymphoma (DLBCL) or follicular lymphoma were treated with magrolimab in combination with rituximab. The results showed that 50% of the patients had an objective (i.e., complete or partial) response, with 36% having a complete response. Specifically, the rates of objective response and complete response were 40 and 33%, respectively, among patients with DLBCL, and 71 and 43%, respectively, among those with follicular lymphoma. Calreticulin (CRT) is a member of the endoplasmic reticulum lectin chaperone family of proteins that plays important biological roles in Ca 2+ homeostasis (in the endoplasmic reticulum, ER), integrin-dependent cell adhesion (in the cytoplasm), and immune response activation (on the plasma membrane) [35]. CRT translocates from the ER to the cell surface during immunogenic cell death in response to various stress factors such as chemotherapy, irradiation, photodynamic therapy, and cytokines. CRT on the surface of stressed or dying cells acts as an "eat me" signal by binding to LRP1 on macrophages [36].
Azacitidine is a chemotherapeutic drug that could induce ICD through translocation of CRT to the cell surface. Interaction of CRT with LRP1 transmits "eat me" signals to macrophages and promotes phagocytosis of dying cells [37]. Combined treatment with magrolimab and azacitidine in a preclinical study resulted in significantly increased phagocytic activity in vitro, and elimination of HL60 in vivo [38]. Preliminary data from a trial (NCT04778397) of magrolimab combined with azacitidine for treatment of AML resulted in a 65% OR and a 40% CR in 34 patients. Transient on-target anemia was observed, and 56% of patients with AML became red blood cell transfusion independent in response to this    Lemzoparlimab (TJC4) Lemzoparlimab (TJC4), from I-MAB, is a human IgG4 antibody targeted to CD47 that was screened using a phage display system. The crystal structure of the TJC4/CD47 complex (straighter headto-head orientation) showed that TJC4 binds to a different epitope than does magrolimab (tilted head-to-head orientation) [42]. In addition, TJC4 showed minimal binding to RBC and no hemagglutination was observed at 100 μg/ml in vitro. No significant erythrocyte toxicity was observed in cynomolgus monkeys dosed with 10-100 mg/kg QW. A phase I trial (NCT03934814) [43] evaluating treatment of R/R advanced solid tumors and lymphoma with TJC4 alone, or in combination with pembrolizumab or rituximab, is ongoing. No DLT or SAE (severe adverse events) were observed, and treatmentassociated anemia occurred in 30% of patients (6 of 20, 2 receiving 3 mg/kg, 2 receiving 10 mg/kg, 1 receiving 20 mg/kg, and 1 receiving 30 mg/kg). Anti-drug antibody (ADA) events occurred in 25% of patients, but there were no concerns regarding safety or pharmacokinetics (PK). Maximal saturation of peripheral T cells was achieved at a dose of 20 mg/kg administered weekly.
AO176 AO176, developed by Arch Oncology, is a humanized IgG2 subclass anti-CD47 antibody [31]. AO176 was shown to bind to integrin-β1 expressed on tumor cells, but not on RBC. Interestingly, AO176 blocked the CD47/SIRPα interaction to stimulate phagocytosis of tumor cells, and also directly killed tumor cells (non-ADCC). A tolerability and hematologic study on Cynomolgus monkeys showed no anemia. In phase I/ II clinical trials [44], Grade 3 TRAE (treatment-related adverse events) were observed in > 10% of patients. In addition, DLT was observed at 20 mg/kg. Studies evaluating AO176 monotherapy and combination therapy with paclitaxel are ongoing.
CC-90002 CC-90002 is a humanized monoclonal IgG4 CD47 antibody. An early clinical trial (NCT02641002) that evaluated treatment of Acute Myeloid Leukemia (AML) and high-risk myelodysplastic syndrome (MDS) was terminated due to poor activity and safety profiles [45,46]. Another phase I trial (NCT02367196) evaluated treatment of advanced solid (alone) and hematologic malignancies (in combination with rituximab). In a combination trial with patients with R/R NHL [47], the ORR (overall response rate) was 13% with 25% achieving stable disease. However, 50% had anemia (of any grade), 33% had thrombocytopenia, and DLTs were observed in 3 subjects (1 subject infusion-related reaction at 15 mg/ kg Q2W and 2 subjects had grade 3 thrombocytopenia requiring platelet transfusion occurring at 30 mg/kg).

SGN-CD47M
SGN-CD47M, developed by Seagen (Seattle Genetics), is a CD47 targeting probody drug conjugate (PDCs). Probody therapeutics [48] are antibody prodrugs designed to remain inactive until proteolytically cleaved and activated in the tumor microenvironment. Probody drug conjugates [49] can be activated by multiple proteases in the tumor microenvironment, but remain inactive in the circulation and in normal tissues.
The efficacy of PDCs depends on multiple factors including binding affinity and specificity for the antigen, efficiency of cleavage in the tumor microenvironment, lack of cleavage in normal tissues, and internalization efficiency. Studies focused on PDCs are in the early clinical stage, and safety and efficacy have yet to be determined. Clinical trial (NCT03957096) of SGN-CD47M for treatment of advanced solid tumors was terminated based on portfolio prioritization.
Other drugs in early clinical development for safety and dosing evaluation are listed in Table 1.
SIRPα-fc Among the 10 allelic variants of SIRPα, SIRPα V1 and V2 are the most prevalent variants [50]. The affinity of wild type SIRPα binding to CD47 is in the micromolar range, which is 1000-fold weaker than that of anti-CD47 antibodies. Soluble SIRPα binding to CD47 on tumor cells could block inhibitory signals and enhance phagocytic activity. Six SIRPα-fusion proteins are currently in phase I or phase II clinical trials (Table 2, ordered by clinical trial developmental state).

Evorpacept (ALX148)
Evorpacept is comprised of SIRPα variant 1 domain 1 (V1D1) and inactive human IgG1-Fc [51]. Evorpacept (CV1) was selected from mutant libraries and includes 9 mutations (V6I, A271, I31F, E47V, K53R, E54S, H56P, L66T, V92I), which resulted in a 50,000-fold increase in affinity compared with that of wild type SIRPα. Preclinical data showed that evorpacept augmented macrophage antitumor activity in vitro and in vivo in combination with tumor-opsonizing antibodies (trastuzumab, obinutuzumab and cetuximab). However, no phagocytosis or antitumor activity was observed following treatment with evorpacept alone. A trial [52] in which patients with NHL received evorpacept alone at 10 mg/kg or at 15 mg/kg in combination with rituximab resulted in ORRs of 40.9 and 63.6%, respectively. The CD47 receptor occupancy on RBC and CD4 T cells was approximately 90% at 10-15 mg/ kg. A phase I study [53] of evorpacept in combination with pembrolizumab, trastuzumab, or zanidatamab, and/or chemotherapeutic agents, evaluating treatment of advanced solid malignancy is ongoing. Preliminary results showed anti-cancer activity of evorpacept in combination with pembrolizumab (AP) and/or chemotherapy (5FU + platinum) in patients with second line or greater HNSCC (head and neck squamous cell carcinoma) with prior platinum therapy. The ORR in patients with checkpoint inhibitor-naïve HNSCC (n = 10) treated with AP was 40%, but 0% in patients with HNSCC who had previously received checkpoint inhibitors (n = 10). A phase II study of evorpacept in combination with pembrolizumab for treatment of HNSCC was recently initiated. Treatment with evorpacept in combination

Antibodies targeting SIRPα on myeloid cells
Anti-SIRPα antibodies induce weak or no phagocytic activity alone, but induce significantly increased phagocytic activity when combined with opsonizing antibodies (rituximab, cetuximab) [58][59][60]. Several issues with anti-SIRPα antibodies are important to consider. First, since SIRPα is expressed on myeloid cells, anti-SIRPα antibodies using inactive human IgG-Fc to avoid Fc effector-mediated toxicity on these immune cells may be advantageous [60]. Second, SIRPγ expressed on T and NK cell shares 74.37% amino acid similarity with the extracellular domain with SIRPα [61]. SIRPγ on T cells binds to CD47 on APCs to mediate cell-cell adhesion and enhances antigen presentation, resulting in T cell proliferation and cytokine secretion [62]. Development of a SIRPα targeting antibody with specificity toward SIRPα to avoid interference with the interaction between CD47 and SIRPγ may preserve T cell activity. Third, antibody internalization could lead to rapid clearance of antibody in vivo. Higher doses or multiple dosing is required to ensure that levels remain therapeutically relevant [63,64]. Internalization of SIRPα decreases the inhibitory signal and may enhance the ability of antibodies to restore phagocytic activity [59,65]. Finally, Although SIRPα has ten known variants, V1, V2, and V8 are the most prominent (over 90%) haplotypes in the human population. Antibodies targeting all three of these variants may be more potent than those that target a single variant [66,67]. Antibodies targeting SIRPα currently being evaluated are listed in the Table 4. Trials evaluating treatment with SIRPα antibodies in combination with immunotherapies are in the early clinical stage of development, and include OSE-172 [68] from OSE Immunotherapeutics (co-developed with Boehringer Ingelheim), CC-95251 from Celgene, and FSI-189 from Gilead [59].

Bi-specific molecules
Bi-specific molecules bind two targets or two distinct epitopes of one target. The antigen binding sites of bispecific molecules could consist of two antibodies or proteins (ligand or receptor), or could consist of one antibody and one protein. Bi-specific antibodies can bind two target antigens in-cis and in-trans. Rational design of bispecific molecules based on biological activity may result in distinct effects or improved efficacy when compared to combination treatments. Four bispecific antibodies, catumaxomab [69] (withdrawn in 2017), blinatumomab [70], emicizumab [71], and amivantamab-vmjw [72] have been approved by EMA or FDA. Bi-specific molecules can be constructed from CD47 targeting antibodies or SIRPα and other antigen-targeting molecules (Fig. 4). Target antigens could include: A) tumor associated cell surface antigens (PD-L1, CD20, CD19, MSLN (Mesothelin), Claudin18.2, and Her2), B) immune checkpoint proteins (PD-1, CD40, 41BB), and C) cytokines or receptors (CSF-2 receptor, VEGF). Reduced affinity for CD47 and increased affinity for the second target may reduce toxicity and enhance efficacy. For type A antibodies, IgG1-Fc was selected to enhance antibodymediated killing of tumor cells (ADCC, ADCP, and CDC). However, inactivated Fc is preferred when used in type B bi-specific molecules. Use of CD47-targeting biologics has shown significant clinical efficacy for treatment of R/R AML, NHL, and MDS. The combination of Hu5F9 and rituximab showed particular promise as a treatment approach for R/R NHL [34]. Several CD47related BsAb are in early-stage clinical trials. NI1701 [73], which targets CD47 and CD19, is an IgG-like BsAb constructed using modified knobs-into-hole technology [74], and contains an IgG1-Fc. Preclinical data showed that NI1701 [75] selectively binds to CD47 and CD19 co-expressing cells, but interacts poorly with normal healthy cells (CD47 + CD19 − ), resulting in avoidance of normal cells acting as sinks for binding of the antibodies, and reduced toxicity. In vitro and in vivo studies showed that NI1701 more potently killed tumor cells than did anti-CD47 and anti-CD19 antibodies alone, or in combination. A BsAb named NI1801 [73] targeted to CD47 and MSLN showed similar preclinical activity as NI1701. IMM0306, a bispecific antibody fusion protein targeted to CD20 (rituximab) and CD47 (SIRPα) with wild type IgG1-Fc [76], is in a phase I trial evaluating treatment of R/R CD20-positive B-cell non-Hodgkin's lymphoma. SL-172154 [77] is a fusion protein targeted to CD47 with SIRPα and CD40 with CD40L, and is in a phase I trial for treatment of solid tumors. A BsAb, HX009, comprised of the extracellular region of SIRPα V2D fused with an anti-PD1 antibody (HX008) was evaluated in a phase I trial. a antibody in clinical trial The study showed that HX009 blocked both CD47/SIRPα and PD-1/PDL-1 interactions, and interacted with CD47 on tumor cells and PD1 on T cells to help present tumor antigens to T cells, resulting in activation of the innate and acquired immune responses. Several BsAbs have been developed to inhibit CD47/SIRPα and PD-1/PD-L1 interactions. One such BsAb, IBI322 [78], was designed as a selective CD47 binding CD47/PD-L1 bispecific antibody. IBI322 showed no negative hematological effects in cynomolgus monkeys, but the binding affinity of IBI322 to cynomolgus CD47 was not disclosed, and IBI322 showed dose-dependent binding to RBC. Other BsAb antibodies listed in Table 2 are in the preclinical proof-ofconcept stage.

Engineered T cells and macrophages
T cells with chimeric antigen receptors (CARs) showed promising therapeutic efficacy against hematologic malignancies, and several CAR-T therapeutics have been approved [79][80][81][82][83]. However, low treatment response rates of solid tumors to CAR-T treatment were observed. Golubovskaya et al. [84] showed that CD47-CAR-T cells effectively killed ovarian, pancreatic, and other cancer cells, and induced production of high levels of IL-2, which correlated with expression of CD47 antigens. Treatment with CD47-CAR-T cells may be a novel strategy for treating different types of cancers. Huyen [85] designed a third generation of CD47-CAR-T cell that could effectively kill lung cancer cells (A549) and inhibit lung cancer cell metastasis. A dual CAR-T targeting CD47 and TAG-72 (tumor-associated glycoprotein 72) generated by Shu [86] showed promising results against ovarian cancer in preclinical experiments. An Anti-PD-L1 (A12) CAR-T with the ability to secrete anti-CD47 VHH (variable heavy domain of heavy chain antibodies or nanobodies) (A4), developed by Xie [87], represented a novel strategy for cancer treatment. An A12-A4 CAR-T showed better anti-cancer activity than an A12 CAR-T plus soluble A4 in C57BL/6 PD-L1-KO mice bearing B16F10 cells. Each of these CD47-CAR-T cells are in the preclinical stage, and efficacy and safety should be further investigated. Development of CAR-M1 (M1: classically activated macrophages) is an emerging therapeutic strategy [88], and several studies of engineered CAR-M cells [89][90][91] showed tumor cell elimination activity in vitro and in vivo. These studies showed that CAR-M induced phagocytosis and induced M2 to M1 polarization through secretion of pro-inflammatory factors and chemokines [90]. CD47 is ubiquitously expressed on the surfaces of multiple hematopoietic and solid tumor cells. However, adverse events due to cytokine secretion by macrophages during immune checkpoint activation [92,93] and CAR-T [94,95] treatment was common. Technological improvements for preparation and production of CAR-M are needed to offer scalable and reproducible manufacturing processes.

Small molecules, peptides, and microRNA
RRx-001 [96] is an anticancer agent designed to induce M2 to M1 polarization and to promote recovery of phagocytic activity of macrophages toward tumor cells.
The anti-phagocytic inhibitory signal was removed or reduced through downregulation of both CD47 and SIRPα gene expression on tumor cells and macrophages, respectively. Elimination of tumor cells was shown in in vitro and in vivo. Phase III clinical trials (NCT03699956, NCT02489903) against small cell lung cancer [97,98] are ongoing.
D4-2 [99], a macrocyclic peptide targeted to mouse SIRPα was designed to inhibit the interaction between CD47 and SIRPα and promote macrophage-mediated phagocytosis of tumor cells when combined with rituximab. PKHB1 [100], a TSP-1-derived CD47 agonist peptide, induced cell death (CRT exposure and DAMP release) in chronic lymphocytic leukemia cells.
MicroRNAs (miRNAs), which are 20-22 nucleotides in length, play important roles in cancer pathogenesis and progression since they can repress the target gene at the translational level by directly binding to the 3'untranslated regions (3'UTRs) [101].
Overexpression miR-378a [102] in mice peritoneal macrophages downregulates SIRPα mRNA expression. Phagocytosis of Ishikawa cells by macrophages-miR-378a and macrophages was carried out in vitro. Phagocytic index in macrophages-miR-378a group is 3 times than that in macrophages group.
Zhao [103] reported that miR-200a inhibited the expression of CD47 by directly targeting the 3'UTR of the CD47 mRNA. MicroRNA 200a suppressed nasopharyngeal carcinoma (NPC) cell proliferation, migration, and invasion, and promoted phagocytosis of NPC cells by Fig. 4 Types of bis-specific molecules targeted CD47 and other molecules. Bispecific Ab-type 1: KIH format, targeted to CD47 (yellow) and other tumor associated antigens (CD19, CD20, and MSLN, blue); Bi-specific Ab-type 2: KIH format, targeted to CD47 (yellow) and immune checkpoint molecules (PD-1, CD40, and 41BB, green); Bi-specific Ab-type 3: SIRPα (cyan) fusion to N-terminus of H-chain of IgG (targeting tumor associated antigen, blue); Bi-specific Ab-type 4: ligand/receptor to modulate TME (GM-CSF, and VEGFR2, pink) fusion to the C-terminus of the H-chain of IgG (yellow); Bi-specific Ab-type 4: SIRPα fusion to the N-terminus of IgG-Fc and ligand/receptor to modulate TME (GM-CSF, VEGFR2, pink) fusion to the C-terminus macrophages through down-regulation of CD47 expression on NPC cells.
MicroRNA 708 [104] was directly targeted CD47 and resulted in downregulation of CD47 on T cell acute lymphoblastic leukemia cell line. MicroRNA 708 expression in the T-ALL cell line was sufficient to promote phagocytosis by macrophages in vitro, and inhibited tumor engraftment in vivo.

Conclusions and future perspectives
Following the clinical success of therapeutic antibodies targeting T cell checkpoint molecules, combination therapies using checkpoint inhibitors with other agents have been a major theme of clinical oncology studies. Specifically, inhibitors of the CD47/SIRPα pathway have emerged as promising therapeutic candidates. Overexpression on tumor cells makes CD47 an ideal target for cancer therapy. Antibodies targeting CD47 showed promising results against MDS and AML [105][106][107]. However, side effects such as anemia, hyperbilirubinemia, thrombocytopenia, and lymphopenia induced by CD47-targeting molecules are of specific concern and need to be addressed with development of new therapeutic agents.
In addition, the therapeutic effects of agents targeting the CD47/SIRPα axis on solid tumors are limited. Additional therapeutic strategies, including combination therapy and bi-specific antibodies, may be promising. Combination therapy with opsonizing antibodies, immune checkpoint inhibitors, chemotherapeutic agents to activate FcR on macrophages, and T cell sensitizers that induce immunogenic cell death to stimulate a more potent immunological effect all have potential. Agents targeted to the CD47-SIRPα axis should not only block the CD47/SIRPα interaction, but also activate signaling on macrophages (FcγR, CRT expression). Other agents including small molecules, mRNA, and CAR-T/M that block the CD47/SIRPα interaction are also in development. Many promising strategies targeting the CD47-SIRPα axis are in development and offer a great deal of hope to patients with cancer.