SIRPα-Antibody Fusion Proteins Selectively Bind and Eliminate Dual Antigen-Expressing Tumor Cells Free

CD47 is a transmembrane protein, which is widely expressed on many cell types and serves as a ligand for signal regulatory protein alpha (SIRPα), a receptor on phagocytic cells including macrophages and dendritic cells (1). This interaction transmits a “don't eat me” signal by initiating signaling cascades that ultimately inhibit phagocytosis (2–5). Elevated CD47 expression relative to normal cell counterparts has been detected on acute myeloid leukemia stem cells (AML LSC), multiple subtypes of B-cell non-Hodgkin lymphoma (NHL), and many human solid tumor cells (6–9). CD47 contributes to cancer pathogenesis by enabling tumor cells to evade phagocytosis, and disruption of the CD47–SIRPα interaction is a therapeutic strategy to induce phagocytic elimination of tumor cells (10). However, wide distribution of CD47 on normal cells may represent potential sites of toxicity or an “antigen sink” that could prevent CD47-targeting agents from reaching tumor cells at therapeutically relevant doses.

The N-terminus of SIRPα contains an immunoglobulin superfamily V-like fold that interacts with the N-terminus of CD47 (11, 12). The CD47–SIRPα interaction can be antagonized with antibodies that bind to the interaction interface on either protein or with recombinant SIRPα (6, 13, 14). However, the native interaction between CD47 and SIRPα is relatively weak, and this may limit the utility of wild-type SIRPα as a therapeutic agent (13, 15). To address this issue, High-affinity variants of SIRPα were generated and shown to antagonize the interaction and induce phagocytosis in settings where a pro-phagocytic Fc domain was present (13).

Antibody Fc domains can serve as pro-phagocytic signals by binding to Fc receptors (FcR) on phagocytic effector cells and triggering antibody-dependent cellular phagocytosis (ADCP; ref. 16). CD47–SIRPα antagonists enable phagocytosis by blocking an inhibitory signal and have been shown to synergize with pro-phagocytic Fc-mediated signals. Synergy was initially shown with blocking anti-CD47 antibodies and rituximab, an anti-CD20 antibody known to activate FcRs (7). Synergistic cell elimination was also observed between trastuzumab and anti-CD47 or anti-SIRPα antibodies (17). High-affinity SIRPα monomers also enabled synergistic induction of phagocytosis when combined with tumor-specific monoclonal antibodies including trastuzumab, rituximab, and cetuximab (13). Taken together, these studies demonstrate the potential for CD47–SIRPα antagonists to synergize with tumor-specific monoclonal antibodies to induce potent elimination of tumor cells by phagocytosis.

We sought to develop a novel therapeutic antibody format that would disrupt the CD47–SIRPα interaction, deliver an Fc domain pro-phagocytic stimulus selectively to tumor cells, and lack toxicity against cells expressing CD47 alone. To achieve selective binding to tumor cells, our format uses a monoclonal antibody specific for an established tumor antigen as a scaffold. The N-terminal Ig domain of SIRPα was grafted onto the antibody scaffold to function as a CD47 antagonist. We hypothesized that the resulting “SIRPabodies” would deliver blockade of the CD47–SIRPα interaction specifically to cells expressing the tumor antigen as wild-type SIRPα interacts weakly with CD47, and that strong binding to tumor cells would depend on avidity contributions provided by binding to the tumor antigen. Moreover, by adding a CD47–SIRPα blocking component to an Fc-domain containing antibody, we predicted enhanced tumor elimination analogous to the synergy observed through combination of CD47–SIRPα blocking agents and FcR engagement. Here, we tested these hypotheses by generating SIRPabodies in a variety of formats using rituximab, an extensively studied anti-CD20 antibody, as the tumor-specific antibody for proof-of-concept studies. SIRPabodies were interrogated for binding specificity, therapeutic efficacy against human NHL in vivo, and toxicity in nonhuman primates.

The V_H and V_L from the 2B8 clone from which rituximab was generated were synthesized using custom gene synthesis (MCLAB) and subcloned into the pCEP4 mammalian expression vector (Invitrogen) containing the human C_H1 or C_k genes, respectively, to create the heavy and light chains for expression of anti-CD20 (18). To create SIRPabodies, the N-terminal V-set Ig domain (residues 1-118) of the predominant allele of SIRPα (allele 2) was introduced onto the heavy chain of anti-CD20 via overlapping PCR with introduction of a linker (12, 19). The following linkers were used: CD20–2GL–SIRPα HC: (GGGGS)₂, CD20–4GL–SIRPα HC: (GGGGS)₄, SIRPα-γ-CD20 HC: ASTKGPSVFPLAP. Plasmids containing each chain were cotransfected into Freestyle 293 cells (Invitrogen) using 293 Fectin (Invitrogen) according to manufacturer's instructions. After 4 to 7 days, cell culture supernatant was prepared by spinning cell cultures at 300 × g. Supernatant containing antibody was subject to antibody purification by affinity chromatography on Protein A Sepharose (GE Healthcare Lifesciences). Purified antibody was dialyzed against PBS and analyzed by 10% SDS-PAGE (Invitrogen) under reducing and nonreducing conditions followed by Coomassie Brilliant Blue staining. Purified antibody was quantified by A280.

SIRPα-Fc (hIgG4) was described previously (13). Anti-CD47 (clone B6H12.2, mouse IgG1) was described previously (6). The V_L and V_H from B6H12.2 were subcloned into pCEP4 expression vector containing human C_k or C_H1 genes, respectively, to create anti-CD47 (hIgG1). SIRPα-Fc and anti-CD47 were expressed and purified as described above. Human IgG1 isotype control antibody used for in vitro experiments was purchased from Sigma. Additional antibodies used for in vivo experiments are as follows: Rituximab (anti-CD20, human IgG1) was purchased from the Stanford University Medical Center, and mouse IgG from Innovate Research.

YB2/0, Raji, Daudi, ST486, and Ramos cell lines were purchased from ATCC. Raji cells expressing modified luciferase and eGFP were described previously (7). YB2/0 cells were engineered to stably express human CD20 cDNA (Genecopoeia) via an engineered transposable element (20). YB2/0 cells were engineered to stably express human CD47 by lentiviral transduction with virus generated from human CD47 cDNA (Open Biosystems, Thermo Scientific) in the pCDH backbone (System Biosciences).

Cells were stained with primary antibodies at 10 μg/mL unless otherwise stated prior to staining with PE antihuman Fc secondary antibody (eBiosciences). Flow cytometry was performed using the BD FACSCanto II. For competition-based binding experiments, cells were stained with the indicated prior antibodies prior to staining with 10 μg/mL DyLight 488 anti-CD20 (2B8-hIgG1) or Alexa Fluor 647 anti-CD47 (clone B6H12.2; BD Biosciences). DyLight 488 anti-CD20 was generated by labeling purified antibody with the DyLight 488 Antibody Labeling Kit (Thermo Scientific) according to manufacturer's instructions.

The kinetics of antibody binding to recombinant CD47 antigen were determined by surface plasmon resonance-based measurements with a Biacore 2000 instrument performed by Biosensor Tools. The test antibodies were diluted in 10 mmol/L sodium acetate pH 5.0 and amine coupled onto a GLM sensor chip. The CD47 antigen was a His-tagged monomer to allow for affinity measurements in the absence of avidity contributions. The antigen was injected over reaction matrices in a threefold dilution series. Each sample was injected across the antibody surface for 460 seconds. The association and dissociation rate constant, k_a (M⁻¹s⁻¹) and k_d (s⁻¹), respectively, were monitored, and a K_D value was determined.

Biotinylated human CD47-Fc fusion protein was incubated with DyLight 650-conjugated neutravidin (Pierce) to form CD47 tetramers. CD47 tetramers were coincubated with YB2/0 cells engineered to express human CD20 and the experimental antibodies. Antibody binding to cells was detected with a secondary PE antihuman Fc antibody (eBiosciences) and flow cytometry analysis. Gating on live cells and double positive (PE+DyLight 650+) events indicated simultaneous binding of primary antibody to CD47 and CD20.

Human CLL samples were obtained from patients at the Stanford Medical Center with informed consent, according to an IRB-approved protocol (Stanford IRB # 6453). Human peripheral blood mononuclear cells (PBMC) and red blood cells (RBC) were from anonymous donors and were purchased from the Stanford Blood Center.

Cynomolgus monkey whole blood used for ex vivo staining (Fig. 4B) was obtained from BioreclamationIVT. Toxicity studies in Cynomolgus macaques were performed by Charles River Laboratories (Reno, NV) in accordance with Association for Assessment and Accreditation of Laboratory Animal Care international guidelines. Male cynomolgus macaques weighing approximately 3 kg were administered a single dose of CD20–2GL–SIRPα HC by intravenous infusion. Flow cytometry analysis was performed by Flow Contract Site Laboratory (Kirkland, WA). Whole blood was stained with CD45 APC clone MB4606 (Miltenyi) and RBCs were discriminated as the CD45⁻ population. Binding of CD20–2GL–SIRPα HC to RBCs was determined ex vivo by staining with PE antihuman Fc secondary (eBiosciences). Median fluorescence intensity values were derived from cytogram plots and converted to molecules of equivalent fluorochrome (MOEF) by collecting beads of known fluorescence, generating a standard curve and extrapolating the value using a spreadsheet supplied by the bead manufacturer (Spherotech). To assess depletion of B cells upon treatment with CD20–2GL–SIRPα HC, whole blood was treated with ACK lysis buffer (Life Technologies) and stained with CD45 APC (described above), CD3 FITC clone SP34 (BD Biosciences), and CD21 V450 clone B-ly4 (BD Biosciences). B cells were identified as CD45⁺ CD3⁻ CD21⁺ cells.

A total of 1 × 10⁶ luciferase-labeled Raji cells were injected subcutaneously into the hind flank or intravenously into the tail vein of 6- to 10-week-old NOD. Cg-PrkdcscidII2rgtm1Wjl/SzJ (NSG) mice. Mice were monitored for engraftment via bioluminescent imaging and daily intraperitoneal injections of 200 μg mouse IgG control, rituximab, SIRPα–hIgG4, CD20–2GL–SIRPα HC, or 200 μg SIRPα–hIgG4 + 200 μg rituximab were administered. Tumor progression was monitored with weekly bioluminescent imaging analysis and mice were followed for overall survival. All experiments were performed according to the Stanford University institutional animal guidelines.

Luciferase imaging analysis was performed as described previously (7).

Serum samples were collected from treated mice and antibody levels were determined by ELISA. ELISA plates were coated with goat antihuman antibody, Fcγ fragment specific (Jackson ImmunoResearch) and blocked with Superblock (Pierce) prior to incubation with serial dilutions of serum samples. Detection was carried out with antihuman kappa horseradish peroxidase antibody (Bethyl) as described above. Concentrations of antibody in serum samples were determined by extrapolation from standard curves using the four-parameter logistic fit.

To create antibody-like molecules that block the CD47–SIRPα interaction and also require binding to a second tumor antigen for high avidity binding, we sought to exploit the low affinity of this interaction (11, 12, 15). CD20 was chosen as the target tumor antigen for initial proof-of-concept studies as rituximab, an anti-CD20 antibody, has been shown to synergize with CD47–SIRPα antagonists (7, 13). The N-terminal Ig domain of SIRPα was grafted onto either the N- or C-terminus of the heavy chain of rituximab (Fig. 1A). A 13 amino acid sequence derived from the N-terminal end of the CH1 domain was used for linkage to the N-terminus to create SIRPα-γ-CD20 HC. A flexible polyglycine-serine linker (GGGGS) of either two or four repeats was used to link SIRPα onto the C-terminus to make CD20–2GL–SIRPα HC or CD20–4GL–SIRPα HC, respectively. Each SIRPabody was produced as a single IgG-like species in mammalian cells and the increased size of the heavy chain relative to parental anti-CD20 is reflective of the addition of the SIRPα Ig domain (Fig. 1B).

SIRPabodies were evaluated for their ability to bind to antigen-expressing cells using rat cells engineered to express human CD20 or separately, human CD47. All SIRPabodies bound to CD20 in direct staining assays and blocked binding of labeled anti-CD20 in competition experiments with a similar potency as parental anti-CD20 (Fig. 2A and Supplementary Fig. S1A). The B6H12.2 anti-CD47 antibody has been extensively studied and served as a reference for robust CD47 binding (21). SIRPα-Fc and all SIRPabodies demonstrated weaker binding to CD47 than anti-CD47, which is indicative of the relatively low affinity of the CD47–SIRPα interaction (Fig. 2B). SIRPα-γ-CD20 HC, with SIRPα attached at the N-terminus, displayed comparable binding to CD47 as SIRPα-Fc control. In contrast, CD20–2GL–SIRPα HC and CD20–4GL–SIRPα HC each had reduced binding to CD47 relative to SIRPα-Fc, presumably due to steric hindrance of the CD47 binding site. All SIRPabodies were outcompeted by labeled anti-CD47 in competition experiments, consistent with their weaker interaction with CD47 (Supplementary Fig. S1B). Because weak binding to CD47 is a central component of our rationale, we quantitatively determined the affinity of each SIRPabody for monomeric CD47 using surface plasmon resonance analysis. All SIRPabodies had lower affinity for CD47 than B6H12.2, with SIRPα-γ-CD20 HC having an approximately 20-fold reduction and CD20–2GL–SIRPα HC and CD20–4GL–SIRPα HC having more than 50-fold reduction (Fig. 2C). Collectively, these data show that the SIRPabodies retain strong binding to CD20 and that the strength of CD47 binding can be modulated based on the location of SIRPα attachment to the antibody molecule.

To address whether a single SIRPabody molecule could simultaneously bind to CD20 and CD47, cells expressing CD20 but not CD47 were incubated with SIRPabodies along with complexes containing a recombinant biotinylated CD47 antigen bound to neutravidin-fluorescent conjugates. Antibody binding to cells was then detected by flow cytometry with a secondary antibody conjugated to a different fluorophore, and double positive live cell events indicated simultaneous binding to each antigen (Fig. 3A). All SIRPabodies demonstrated simultaneous binding to each antigen (Fig. 3B). To address whether SIRPabodies bind to both CD20 and CD47 on dual antigen-expressing cells, Raji cells were incubated with SIRPabodies prior to staining with labeled anti-CD47 or labeled anti-CD20 (Supplementary Fig. S2). All SIRPabodies bound CD47 on Raji cells as indicated by blocking of labeled anti-CD47 antibody (Supplementary Fig. S2A). Importantly, the blocking of anti-CD47 observed with the SIRPabodies is due to enhanced avidity contributed by simultaneous binding to CD20, as SIRPα-Fc did not block as strongly due to its relatively weak affinity for CD47 (Supplementary Fig. S2A). The reciprocal experiment demonstrated binding of SIRPabodies to CD20 on Raji cells (Supplementary Fig. S2B). Collectively, these data demonstrate that CD20 and CD47 each contribute to SIRPabody binding to dual antigen-expressing cells.

A major rationale for generating SIRPabodies targeting CD47 along with a tumor-associated antigen is to avoid binding to cells that express CD47 alone while retaining the therapeutic benefit of blocking the CD47–SIRPα interaction on tumor cells. RBCs highly express CD47 and are a likely “antigen sink” as they are abundant and accessible to antibody in the bloodstream (22). To determine whether SIRPabodies selectively bind dual antigen-expressing cells in the presence of excess CD47-only expressing RBCs, CFSE-labeled tumor cells expressing CD47 and variable levels of CD20 were mixed with a 20-fold excess of unlabeled RBCs prior to incubation with antibody (Fig. 4A). Primary cells from three patients with B-cell chronic lymphocytic leukemia (B-CLL) were used as dual antigen-expressing cells in this assay as high expression of CD47 has been observed in this tumor type (Fig. 4B; ref. 7). All SIRPabodies, and the control anti-CD47 and anti-CD20, bound to tumor cells to a degree dependent on CD20 expression levels (Fig. 4C and Supplementary Fig. S3). In contrast, only anti-CD47 bound to RBCs, as indicated by a complete shift relative to isotype control. Importantly, all SIRPabodies did not bind to RBCs, indicating that weak binding to CD47 successfully eliminated binding to RBCs. Analysis of the distribution of cell types within the population of cells bound by each antibody revealed that, similar to anti-CD20, all SIRPabodies bound selectively to CLL cells in the presence of an antigen sink (Fig. 4D). Importantly, SIRPabody binding to tumor cells blocked subsequent staining with labeled anti-CD47, indicating that SIRPabody binding to tumor cells engages CD47 and is not solely driven by high-affinity binding to CD20 (Fig. 4E). Notably, the degree of blocking of labeled anti-CD47 corresponded to CD20 expression levels on tumor cells. Collectively, these data suggest that CD20 expression drives SIRPabody binding to tumor cells and that there is a threshold of CD20 expression that must be present on tumor cells to achieve CD47 blockade. This finding confirms that the SIRPabody approach directs CD47 blockade to dual antigen-expressing cells and successfully avoids targeting of cells expressing CD47 alone.

CD20–2GL–SIRPα HC was chosen as a lead candidate for scaled up production and in vivo studies as the addition of SIRPα to the C-terminus resulted in weaker CD47 binding than addition to the N-terminus. Furthermore, we reasoned that the shorter linker used in CD20–2GL–SIRPα HC would be less likely to be targeted for proteolytic cleavage in vivo compared to the longer linker of CD20–4GL–SIRPα HC. To test the in vivo therapeutic efficacy of CD20–2GL–SIRPα-HC, immunodeficient NSG mice were transplanted either subcutaneously or intravenously with luciferase-expressing Raji cells to model localized or disseminated lymphoma, respectively, as previously described (7, 23). Mice were treated with daily injections of control IgG, rituximab, SIRPα-Fc fusion protein, SIRPα-Fc plus rituximab, or CD20–2GL–SIRPα HC. In the subcutaneous model, SIRPα-Fc had no therapeutic effect relative to control IgG and did not augment rituximab activity when administered in combination (Fig. 5A). However, rituximab and CD20–2GL–SIRPα HC each reduced disease burden and extended survival relative to IgG control. CD20–2GL–SIRPα HC induced a statistically significant reduction in the median tumor volume compared to rituximab or rituximab plus SIRPα-Fc as measured by bioluminescence signal, which was further reflected in an increase in overall survival (Fig. 5B). In the disseminated lymphoma model, luciferase-labeled Raji cells were intravenously injected into NSG mice and followed for disease by bioluminescent imaging and for overall survival. Treatment with CD20–2GL–SIRPα HC significantly reduced tumor burden and extended survival relative to treatment with SIRPα-Fc, rituximab, or rituximab plus SIRPα-Fc (Fig. 5C and D). These data demonstrate that the CD20–2GL–SIRPα HC SIRPabody exhibits significantly improved in vivo efficacy compared to parental rituximab alone. Importantly, CD20–2GL–SIRPα HC and rituximab had similar pharmacokinetic profiles allowing direct comparison of therapeutic efficacy in vivo, and suggests that CD20–2GL–SIRPα HC is a viable therapeutic molecule for further development (Supplementary Fig. S4).

To rigorously test the potential for CD20–2GL–SIRPα HC to bind to CD47 on RBCs, a titration of antibody was tested to see if significant binding to RBCs occurred at high concentrations (Fig. 6A and B). No binding to RBCs from humans or cynomolgus macaques was detected up to 500 μg/mL, in contrast to anti-CD47 control which demonstrated strong binding as low as 0.5 μg/mL. To address the potential for toxicity and binding to an antigen sink in vivo, cynomolgus macaques, which express a CD47 ortholog that is identical to human CD47 at the SIRPα interaction interface, were administered a single dose of CD20–2GL–SIRPα HC at 3, 10, or 30 mg/kg and followed for 2 weeks. Transient binding of CD20–2GL–SIRPα HC to RBCs was observed with the two highest doses, but binding did not lead to RBC depletion or substantial reduction in hemoglobin levels (Fig. 6C and E). As a pharmacodynamics marker, the percentage of B cells within the leukocyte compartment was determined by flow cytometry. Depletion of B cells was observed with all doses, indicating successful antibody delivery and cell targeting in vivo (Fig. 6D). Together, these data show that CD20–2GL–SIRPα HC effectively overcomes a physiologic antigen sink and depletes target cells in vivo with no observed RBC toxicity.

We report here the generation of SIRPabodies, antibody-derivative candidate therapeutic molecules with wild-type SIRPα grafted onto an established tumor-targeting monoclonal antibody. The rationale for SIRPabodies is twofold: (i) to create a therapeutic molecule that delivers CD47–SIRPα blockade specifically to tumor cells and (ii) to recapitulate the synergistic effects of combining CD47–SIRPα blockade with a pro-phagocytic signal in the form of an Fc domain. Using the anti-CD20 antibody rituximab in proof-of-concept studies, we show that SIRPabodies achieve both of these goals, establishing them as promising novel cancer-targeting agents.

Disruption of the CD47–SIRPα interaction has been explored as a therapeutic strategy through several approaches including monoclonal antibody targeting of CD47 or SIRPα and recombinant SIRPα proteins that antagonize the interaction (6, 13, 14, 17). However, although these strategies exploit the elevated CD47 expression observed on tumor cells, they do not address the need for tumor cell selectivity. CD47 is widely expressed at low levels on most normal cell types, including RBCs and leukocytes. Normal cells expressing CD47 create an “antigen sink” that may sequester CD47 targeting agents, preventing these agents from binding to tumor cells in vivo. Moreover, undesired binding of CD47 targeting agents to normal cells may lead to toxicity, particularly anemia due to RBC phagocytic elimination. To overcome these limitations and direct the benefit of CD47–SIRPα blockade specifically to tumor cells, we created SIRPabodies with wild-type SIRPα engineered onto an established tumor-specific monoclonal antibody. We hypothesized that the low affinity of wild-type SIRPα for CD47 would facilitate weak binding of SIRPabodies to cells expressing CD47 alone. In contrast, on dual antigen-expressing tumor cells, SIRPabodies would bind strongly to the tumor-associated antigen and utilize CD47 binding as an additional source of interactions leading to avid binding. Consistent with this hypothesis, SIRPabodies bound weakly to cells expressing CD47 without CD20 (Fig. 2B and Supplementary Figs. S1B, S4A, and S4B) and bound to both antigens on dual antigen-expressing cells (Fig. 3B and Supplementary Fig. S2). These properties enabled SIRPabodies to selectively bind to tumor cells in the context of competition with 20-fold excess of RBCs expressing CD47 alone (Fig. 4). RBCs represent a large antigen sink and potential source of toxicity with CD47 targeting agents as CD47 expression regulates RBC clearance in vivo (22, 24). Surprisingly, SIRPabodies and wild-type SIRPα did not exhibit any detectable binding to RBC, even at high concentrations (Figs. 4C, 4D, 6A, and 6B). This may be due to a lack of crosslinking of CD47 by SIRPα on the surface of RBCs, as preincubation of RBCs with antibody that crosslinks CD47 has been shown to increase subsequent SIRPα binding to RBCs (25). Importantly, CD20–2GL–SIRPα HC did not cause anemia, or other toxicity, in nonhuman primates. As previously reported, we have developed a novel anti-CD47 humanized monoclonal antibody, Hu5F9-G4 (26). With single-dose administration of Hu5F9-G4 at 0, 0.1, 0.3, 1, 3, 10, and 30 mg/kg in nonhuman primates, Hu5F9-G4 caused a transient and dose-dependent anemia. Administration of 1, 3, or 10 mg/kg of Hu5F9-G4 in NHPs resulted in a nadir of hemoglobin level below 10 g/dL between days 5 and 7. One of the NHPs administered 30 mg/kg of Hu5F9-G4 developed severe anemia with a hemoglobin level that dropped below 8 g/dL. In contrast, none of NHPs that received CD20–2GL–SIRPα HC up to 30 mg/kg developed a hemoglobin level below 10 g/dL, indicating that the SIRPabody strategy successfully avoids toxicity and functional binding to a potential antigen sink created by circulating RBCs (Fig. 6).

There is mounting evidence that therapeutic agents that block the CD47–SIRPα interaction may require an Fc component to achieve maximum therapeutic potency. Antibodies directed against CD47 have been shown to synergize with rituximab or trastuzumab, which are known to engage FcR (7, 17). As a proof-of-principle study, we developed bispecific antibodies (BsAbs) that cotarget CD47 and CD20 with reduced affinity for CD47 relative to the parental antibody, but retaining strong binding to CD20 (27). These characteristics facilitated selective binding of BsAbs to tumor cells, leading to phagocytosis. Treatment of human NHL-engrafted mice with BsAbs reduced lymphoma burden and extended survival while recapitulating the synergistic efficacy of anti-CD47 and anti-CD20 combination therapy. High-affinity SIRPα monomers developed as antagonists of the CD47–SIRPα interaction induced phagocytosis when presented as Fc fusion proteins, but failed to induce phagocytosis in monomeric form or when presented as a dimer lacking the Fc domain. This property allowed SIRPα monomers to function as adjuvants to increase the efficacy of Fc-containing tumor-specific antibodies including trastuzumab, rituximab, and cetuximab (13). Collectively, studies with high-affinity SIRPα monomers demonstrated a requirement for a pro-phagocytic FcR-activating signal in conjunction with CD47–SIRPα antagonism. We sought to introduce the benefits of CD47–SIRPα antagonism and FcR interaction into a single SIRPabody molecule. This strategy differs from monoclonal antibodies directed against CD47 or SIRPα and SIRPα-Fc fusion proteins as the SIRPabody presents the opportunity for selective binding to tumor cells through the introduction of additional tumor-specific binding regions. We hypothesized that SIRPabodies would be more potent than rituximab or SIRPα-Fc alone, as CD47–SIRPα blockade and FcR binding present two strategies for phagocytic induction, and incorporation into a single molecule should provide avidity contributions to overcome the low affinity of wild-type SIRPα for CD47. CD20–2GL–SIRPα HC demonstrated a significant increase in efficacy relative to rituximab in vivo, particularly in the more aggressive disseminated lymphoma model (Fig. 5). The observed therapeutic effect is most likely due to phagocytosis, as experiments were performed in NSG mice, which lack functional B, T, and NK cells and are therefore devoid of other effector mechanisms (7, 23).

All SIRPabodies were able to bind CD20 and achieve the desired selectivity for dual antigen-expressing tumor cells. Importantly, this selectivity permitted potent depletion of target cells with no observed toxicity in nonhuman primates, which possess a large antigen sink. Moreover, SIRPabodies, which incorporate multiple pro-phagocytic functions into a single molecule, are more efficacious than monospecific agents that utilize a single pro-phagocytic mechanism for tumor reduction. Thus, SIRPabodies demonstrate highly selective binding properties, potent therapeutic efficacy in vivo, and lack of toxicity in nonhuman primates, establishing this approach as a promising strategy to direct the therapeutic benefit of CD47–SIRPα blockade directly to tumor cells.

K. Weiskopf holds ownership interest in Alexo Therapeutics, Forty Seven Inc., and Stanford University for patent applications; and is a consultant/advisory board member for Alexo Therapeutics and Forty Seven Inc. R. Majeti holds ownership interest (including patents) in and is a consultant/advisory board member for Forty Seven Inc. No potential conflicts of interest were disclosed by the other authors.

R. Majeti is a co-inventor of U.S. Patent No. 8,562,997 entitled “Methods of Treating Acute Myeloid Leukemia by Blocking CD47” and U.S. Patent No. 8,758,750 entitled “Synergistic Anti-CD47 Therapy for Hematologic Cancers.” R. Majeti and E.C. Piccione are co-inventors of a U.S. patent application entitled “SIRP-Alpha Antibody Fusion Proteins.”

Conception and design: E.C. Piccione, L. Wang, K. Weiskopf, R. Majeti

Development of methodology: E.C. Piccione, J. Liu, L. Wang, K. Weiskopf, R. Majeti

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.C. Piccione, S. Juarez, S. Tseng, M. Stafford, C. Narayanan, R. Majeti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.C. Piccione, S. Juarez, J. Liu, L. Wang, R. Majeti

Writing, review, and/or revision of the manuscript: E.C. Piccione, J. Liu, K. Weiskopf, R. Majeti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Majeti

Study supervision: J. Liu, R. Majeti

The authors thank Adriel Cha, Stephen Willingham, Aaron Ring, and Ryan Corces-Zimmerman for reagents, Feifei Zhao for laboratory management, and Irv Weissman for discussion.

This research was supported by a Translational Research Program Award from the Leukemia and Lymphoma Society (R. Majeti), the New York Stem Cell Foundation, funding from the J. Benjamin Eckenhoff Memorial Foundation, and funding from the Virginia and D. K. Ludwig Fund for Cancer Research. E.C. Piccione is supported by an NIH T32 Ruth L. Kirschstein National Research Service Award (AI07290). R. Majeti holds a Career Award for Medical Scientists from the Burroughs Wellcome Fund and is a New York Stem Cell Foundation - Robertson Investigator.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.