Anti-leukemic activity and tolerability of anti-human CD47 monoclonal antibodies

Abstract

CD47, a broadly expressed cell surface protein, inhibits cell phagocytosis via interaction with phagocyte-expressed SIRPα. A variety of hematological malignancies demonstrate elevated CD47 expression, suggesting that CD47 may mediate immune escape. We discovered three unique CD47-SIRPα blocking anti-CD47 monoclonal antibodies (mAbs) with low nano-molar affinity to human and cynomolgus monkey CD47, and no hemagglutination and platelet aggregation activity. To characterize the anti-cancer activity elicited by blocking CD47, the mAbs were cloned into effector function silent and competent Fc backbones. Effector function competent mAbs demonstrated potent activity in vitro and in vivo, while effector function silent mAbs demonstrated minimal activity, indicating that blocking CD47 only leads to a therapeutic effect in the presence of Fc effector function. A non-human primate study revealed that the effector function competent mAb IgG1 C47B222-(CHO) decreased red blood cells (RBC), hematocrit and hemoglobin by >40% at 1 mg/kg, whereas the effector function silent mAb IgG2σ C47B222-(CHO) had minimal impact on RBC indices at 1 and 10 mg/kg. Taken together, our findings suggest that targeting CD47 is an attractive therapeutic anti-cancer approach. However, the anti-cancer activity observed with anti-CD47 mAbs is Fc effector dependent as are the side effects observed on RBC indices.

Figure 1

C47B161, C47B222 and B6H12.2 Fabs have steric clashes with SIRPα. (a) Structural overlay of Fab/CD47 complexes onto the SIRPα/CD47 complex showing regions of clash between Fab and SIRPα. The overlay was achieved by superposition of equivalent CD47 Cα atoms in both complexes. The membrane-proximal C15G mutation that was introduced to prevent CD47 aggregation is shown as a yellow sphere. (b) Regions of overlap between each epitope and the SIRPα binding site (red areas). The structure of CD47 from the C47B222 complex was used in (b). (c) Epitope regions of C47B161, C47B222 and B6H12.2. The epitope residues are shaded and SIRPα binding residues are marked with a dot above the human CD47 ECD sequence. The secondary structure of CD47 is shown above the sequence with arrows and cylinders representing β-strands and helices, respectively.

Figure 2

IgG1 anti-CD47 mAbs mediate ADCP and ADCC activity. ADCP of HL60 (a), Kasumi-3 (b), or MV4-11 (c) cells by human PBMC-derived macrophages in response to IgG1/IgG2σ anti-CD47 mAbs. (d) ADCC of human PBMC against HL60 target cells in response to increasing concentrations of IgG1 anti-CD47 mAbs. (e) CDC against WIL2-S target cells in the presence of human complement in response to increasing concentrations of IgG1 anti-CD47 mAbs.

Figure 3

In vivo activity of IgG1 and IgG2σ anti-CD47 mAbs in AML xenograft models. (a) Ten million HL60 cells were implanted into NSG mice (n=5/group) on day 0 and mice received twice weekly treatments for 3 weeks (10 mg/kg) starting day 6 after implant (final day 23; grey-shaded area). Control mice were killed on day 27. (b) Ten million Kasumi-3 cells were implanted into NSG mice (n=5/group) on day 0 and mice received twice weekly treatments (10 mg/kg) for 6 weeks, starting day 6 after implant (final day 44; grey-shaded area). (c) Five million MV4-11 cells were implanted into NSG mice (n=5/group) on day 0 and mice received twice weekly treatments (10 mg/kg) for 3 weeks, starting day 6 after implant (final day 23; grey-shaded area). Control mice were killed on day 34. (a–c) The graphs show peripheral blood leukemic burden (human CD45+ cells) over time, which was monitored weekly, starting on day 14 (HL60 and Kasumi-3) or day 21 (MV4-11).

Figure 4

In vivo activity of IgG1 and IgG2σ C47B222-(CHO) in primary AML and xenograft mouse models. (a) NSG mice were implanted with the indicated AML patient samples. Bone marrow aspirates were analyzed 5 weeks later for engraftment, mice were randomized and treatment was initiated (n=10/group). IgG1 and IgG2σ C47B222-(CHO) were administered twice weekly (10 mg/kg) for 3 weeks. BM and spleen were collected on day 21 and analyzed via FACS to assess leukemic cell burden (hCD45+CD3− cells). (b) Ten million HL60 cells were implanted into NSG mice (n=10/group) on day 0 and mice received twice weekly treatments (20 mg/kg), starting day 6 after implant until study end. (c) Five million MV4-11 cells were implanted into NSG mice (n=10/group) on day 0 and mice received twice weekly treatments (20 mg/kg), starting day 6 after implant until study end. For both (b) and (c), survival was monitored, and bone marrow and spleen were collected on day 28 (HL60) and day 35 (MV4-11) and analyzed via FACS to assess leukemia burden (hCD45+ cells). Increased life span (ILS) was calculated as follows: ((treatment median survival−control median survival)/control median survival) × 100; Statistical significance was assessed by unpaired t-test vs PBS control (*P<0.05, **P<0.01 and ***P<0.0001; NS, non-significant).

Figure 5

Non-human primate study with IgG1 and IgG2σ C47B222-(CHO). (a) Non-naive female cynomolgus monkeys were treated with anti-CD47 mAbs as detailed in schematic (n=4/group). Graphs summarize changes in red blood cell (b), hematocrit (c), hemoglobin (d), and reticulocyte levels (e) in response to treatment with IgG1 and IgG2σ C47B222-(CHO). Arrows above graphs indicate dosing with mAbs.