On the mechanism of CD47 targeting in cancer

The report by Willingham et al. (1) is the most recent in a series of publications from the same group, in which the interaction between the broadly expressed surface molecule CD47 and the myeloid inhibitory receptor SIRPα was implicated as a potential therapeutic target in a variety of malignancies. The authors used xenotransplantation models in which various human solid tumor cells were engrafted into immunocompromised mice, and showed that cancer growth and metastasis were inhibited by antibodies against human CD47 that block interactions with SIRPα, but not by nonblocking anti-CD47 antibodies. In similarly designed studies they had already demonstrated elimination of various hematopoietic malignancies with the blocking anti-CD47 antibody and also that anti-CD47 treatment synergizes with the therapeutic anti-CD20 antibody Rituximab in non-Hodgkin lymphoma. They suggested that targeting of CD47-SIRPα interactions, either in the absence or in the presence of a cancer therapeutic antibody such as Rituximab, facilitates the eradication of tumor cells by promoting their phagocytic clearance by macrophages (1).

In addition to xenogeneic tumor models, Willingham et al. also presented experiments in which antibodies against mouse CD47 are tested in syngeneic cancer models. Their experiments that were aimed to demonstrate safety and efficacy of CD47 targeting showed that one anti-CD47 monoclonal antibody, but not another, significantly inhibited the growth of established mouse breast cancer cells. Importantly, minimal toxicity was detected, which is of course good news with respect to potential targeting of CD47 in human patients with cancer. However, the antibody that caused substantial inhibition of tumor growth (clone MIAP410) has previously been documented as a reagent that does not block CD47–SIRPα interactions (2), whereas the one that did not provide significant effects on tumor growth (clone MIAP301) has been shown to be an effective inhibitor of CD47–SIRPα interactions (3). The question is how to explain this result, which rather argues against the proposed model that CD47–SIRPα interactions limit tumor cell clearance by macrophages and that blocking of this interaction would enhance tumor cell destruction. It seems clear, at least, that things are more complicated.

Of relevance, CD47 does interact not only with SIRPα, but also in cis with certain integrins and also with thrombospondin-1, and these interactions may also have been affected by the antibodies used. Another problem in interpreting the experiments is that the anti-mouse CD47 was used as intact IgG, which may also have caused direct Fc-mediated effects, like antibody-dependent phagocytosis or cytotoxicity. The same problem also complicates interpretation of the experiments with anti-human CD47 in xenogeneic models (4). The only way to exclude this problem is to use antibodies without Fc regions [i.e., F(ab′)₂ or Fab fragments] or otherwise dysfunctional with respect to effector function or to use mice lacking functional Fc receptors.

Finally, our own recent studies (5) applying the well-established syngeneic B16 melanoma cell model to SIRPα-signaling death mutant mice did not demonstrate any effect of the CD47–SIRPα signaling axis on tumor metastasis and outgrowth. However, when therapeutic antibodies against the melanoma cells were injected, the SIRPα-mutant mice cleared the tumor cells much more effectively than wild-type animals. Furthermore, in vitro ADCC experiments of human neutrophils toward Trastuzumab-opsonized breast cancer cells showed an enhancing effect of CD47–SIRPα antagonists in the presence, but not the absence, of cancer therapeutic antibodies. This result suggests that targeting CD47–SIRPα interactions may be beneficial in combination with antibody therapy in cancer. However, the evidence that targeting the CD47–SIRPα axis may also work in the absence of therapeutic antibody seems incomplete and contradictory.