Response: mechanisms of targeting CD47-SIRPα in hematologic malignancies

Zhao et al¹  raise several issues from our prior work^2,3  characterizing the mechanism of targeting CD47 in human hematopoietic malignancies. We reported that a blocking anti-CD47 antibody can eliminate multiple human tumor types and that a major mechanism is through phagocytosis via blocking CD47-SIRPα ligation. Zhao et al argue that tumor elimination likely occurs through antibody-dependent cellular cytotoxicity (ADCC) and that targeting CD47 may not be an effective strategy given widespread expression of CD47 in normal tissues. Here we respond to these issues.

First, Zhao et al argue that anti-CD47 antibody may not only disrupt CD47-SIRPα interactions but also opsonize cells for ADCC or antibody-dependent cellular phagocytosis (ADCP). They state that the distinction between these 2 mechanisms has not been clear because of the use of intact IgG anti-CD47 antibodies. While we primarily used intact anti-CD47 antibodies in our therapeutic experiments, we demonstrated in a proof-of-concept study that anti-CD47 antibody (clone B6H12) enables tumor elimination not through the Fc receptor (FcR; eg, ADCC, ADCP), but specifically through blocking CD47-SIRPα ligation.³  We first used anti-CD47 F(ab′)₂ fragments (as Zhao et al suggest), which enabled phagocytosis of lymphoma cells similarly to intact anti-CD47 antibody, demonstrating that the FcR was not required for phagocytosis.³  Second, intact anti-CD47 antibody was incubated with macrophages deficient in the FcγR (thus unable to engage in FcR-mediated ADCP), which still enabled phagocytosis, in contrast to rituximab, an FcR-dependent effector antibody. Third, while we found that anti-CD47 antibody did not enable human NK cell-mediated ADCC of 2 lymphoma cell lines,³  Zhao et al found that the same antibody induced ADCC of neutrophils and monocytes toward 2 cancer cell lines.⁴  This difference could be because of the use of different cancer target cells or effector cells to measure ADCC.

Recently, we demonstrated that disruption of CD47-SIRPα led to inhibition of lymphoma dissemination.²  Since we used intact anti-CD47 antibody in these experiments, we cannot rule out Fc-mediated effector functions. However, this was not a focus of the study given our previous work investigating this possibility.³  Nevertheless, we agree that disruption of CD47-SIRPα may not be the sole mechanism and may include FcR-dependent mechanisms: ADCC, ADCP, complement-dependent cytoxicity or even direct apoptosis.

In a second inquiry, Zhao et al question whether targeting CD47 could selectively eliminate tumor cells without toxicity (including autoimmune disease). While CD47 targeting may lead to significant toxicity given its ubiquitous expression, preclinical data from our prior work suggest that anti-CD47 antibody not only has minimal toxicity but also eliminates tumor cells while sparing normal cells.^5,6  First, we showed that the anti–human CD47 antibody (clone B6H12) enabled phagocytosis of hematologic cancers but not normal cell counterparts in vitro.^3,5  Second, anti–mouse CD47 antibody administered to normal immunocompetent mice led to only a mild neutropenia with no evidence of autoimmune disease.⁵  One would expect greater toxicity if anti-CD47 antibody induced cell killing through FcR-mediated effector functions. We then showed that selective targeting of tumor cells by anti-CD47 antibody was because of the presence of a pro-phagocytic signal, calreticulin, which is expressed on tumor but not normal cells.⁶  While preclinical data are favorable, the toxicity profile of a human antibody must be tested in a clinical setting, which is ongoing in our laboratory.⁷ 

Alternatively to targeting CD47, Zhao et al propose that targeting its macrophage ligand SIRPα may be more efficacious. We agree that targeting SIRPα is a viable therapeutic strategy and warrants further exploration. Compared with CD47, SIRPα expression is restricted,⁸  thus antibody targeting of SIRPα may lead to less toxicity. Both we and Zhao et al have shown that blocking anti-SIRPα antibodies eliminate tumor cells in vitro.^3-5  In addition, a recombinant SIRPα Fc-fusion protein may also be a viable therapy.⁹  However, 2 challenges to targeting SIRPα exist. First, therapeutic inhibition of SIRPα on all SIRPα-expressing phagocytes may lead to nonselective phagocytosis. Second, the human SIRPα gene contains many polymorphisms.¹⁰  Thus, anti-SIRPα antibodies may differentially interact with differing SIRPα alleles resulting in variable therapeutic efficacy, which Zhao et al has preliminarily shown.⁴  Agents targeting SIRPα may need to be individualized according to each patient's SIRPα allotype to achieve a response. This is in contrast to targeting CD47 whereby no human polymorphisms at the contact interface with SIRPα have yet been identified.

In conclusion, we previously demonstrated that anti-CD47 antibody can eliminate tumor cells through the FcR-independent mechanism of CD47-SIRPα blockade, while not excluding the possibility of FcR-dependent ADCC. Initial preclinical toxicity studies are limited. Further investigation will determine whether targeting CD47 and/or SIRPα represents a viable cancer therapeutic strategy.