Introduction

Over the past decade, the use of blocking agents against inhibitory immune checkpoints has been one of the most significant advances in anticancer treatment [1]. The exciting results obtained with CTLA-4 or PD-1 blockade led to the evaluation of several innate immune checkpoints that could be targeted in anticancer treatment, including pathways regulating macrophage function. Macrophage receptors discriminate altered-self-molecules on aged or dying cells from normal self-markers on healthy cells, resulting in either activation or inhibition of the phagocytosis response [2]. Macrophages express SIRPα that interacts with CD47, a ubiquitously expressed protein that mediates a "don't eat me" signal [3]. This inhibitory CD47-SIRPα ligand–receptor pair has been identified as an important, though not universal, innate immune checkpoint regulator in the homeostatic clearance by macrophages [4]. Current biological understanding of the CD47-SIRPα axis indicates that two signals are necessary to engage macrophage-dependent phagocytosis, integrating both activating (FcγR, CRT, LRP-1) and inhibitory (SIRPα-CD47) signals [3]. Upon binding to target-bound antibodies, the Fc receptor on macrophages becomes phosphorylated and initiates a signaling cascade that promotes phagocytosis. Phosphorylation of the Fc receptor ITAMs is balanced by inhibitory signals from the CD47-SIRPα axis, mediated by SIRPα immunoreceptor tyrosine-inhibition motif (ITIMs) recruitment of the src-homology-2 domain containing tyrosine phosphatases, SHP-1 and SHP-2. These phosphatases can in turn inhibit accumulation of myosin II at the phagocytic synapse, thereby preventing phagocytosis [5].

Under physiological condition, CD47 is broadly expressed by many cell types in almost all normal tissues. Cancer cells have evolved to hijack the CD47-SIRPα interaction by upregulating the expression of CD47 on their cell surface, thus counterbalancing pro-phagocytic signals and increasing the chances of evading innate immune surveillance mediated by effector cells in the tumor microenvironment [4]. Many tumors, both solid and hematological malignancies, were reported to express high level of CD47 [6,7,8,9,10,11]. Overexpression of CD47 is often shown to be an adverse prognostic factor in cancer, indicating CD47-SIRPα axis might be a widespread mechanism of immune evasion for cancer cells. And blockade of the CD47-SIRPα interaction represents a promising immunotherapeutic strategy to activate the phagocytic clearance of tumor cells in various cancer types [6, 10, 11, 13]. Several SIRPα–CD47 blocking agents, including humanized and fully human anti-CD47 antibodies and anti-SIRPα antibodies [11, 14,15,16,17,18,19], soluble SIRPα dimers fused to the Fc portion of human IgG [20, 21], high-affinity monomeric SIRPα devoid of Fc portion and camelid-derived monomeric fragments of anti-CD47 antibodies (nanobodies) [21] and CD47/ tumor-antigen bispecific antibodies, have shown to induce and enhance phagocytosis. Additionally, anti-CD47 antibody enabled dendritic cells to cross-present tumor antigens via MHC class I molecules and activate CD8+ T cells against tumor cells [23]. Several of the anti-CD47 and anti- SIRPα antibodies are now being tested in clinical trials [24]. Many of these molecules are bifunctional, with F(ab)2 domain binding to CD47 to block CD47-SIRPα interaction and Fc domain to engage Fc receptor on macrophages. However, widespread expression of CD47 on normal cells including red blood cells (RBCs) raises concerns of unwanted toxicity of anti-CD47 antibodies due to Fc-mediated effector function [25]. In fact, dose-limiting toxicities related to anemia and thrombocytopenia have been observed in non-human primate models [26] and patients receiving anti-CD47 therapies [27]. Thus, development of a molecule which has minimal impact on normal healthy cells while maintaining high affinity to CD47 on tumor cells is a critical need to improve clinical efficacy and safety [13, 17]. Recent reports suggest some humanized anti-CD47 antibodies possessing such properties [18, 19].

Here we describe preclinical characterization of CC-90002, a humanized anti-CD47 IgG4PE antibody designed to ensure CD47 blockade against tumor cells while minimizing impact on normal healthy cells. We show that CC-90002 antibody blocks CD47-SIRPα interaction with high affinity and activates macrophage-mediated killing of tumor cells. The antibody recognizes a unique conformational epitope on CD47 without promoting hemagglutination of RBCs. In addition, CC-90002 was shown to have no Fc-effector function to eliminate the risk of effector function-related toxicities (CDC, ADCC) in CD47-expressing normal tissues. CC-90002 demonstrated potent macrophage-mediated antitumor activity in xenograft models [26]. Importantly, in non-human primates, CC-90002 displays acceptable pharmacokinetic properties and a favorable toxicity profile. These data provided basis for clinical studies CC-90002-ST-001 (NCT02367196) and CC-90002-AML-001 (NCT02641002) with CC-90002.

Materials and methods

Antibody discovery and engineering

The parental 2A1 monoclonal antibody (mAb) was generated by immunizing mice with recombinant CD47-IgV (immunoglobulin-like variable-type) protein by a modified rapid immunization strategy [29]. Following immunization B-cells were isolated from lymph nodes and fused to the mouse myeloma cell line NS1. Hybridoma supernatants were screened for CD47 binding and the ability to block the CD47-SIRPα interaction. 2A1 was selected based on its high affinity to CD47 and non-hemagglutination feature. The 2A1 mAb was humanized using a structure-based modeling approach. CC-90002 was made by incorporating site-specific residue substitution and employing the IgG4PE (immunoglobulin G4 with S228P and L235E mutations; CD47.G4PE) Fc isotype to the humanized anti-CD47 mAb.

Affinity determination by surface plasmon resonance

SPR (Biacore 2000 and T200) was used to evaluate CC-90002 kinetics of binding to multiple recombinant CD47-extracellular domain (ECD) proteins. An optimized protocol of CD47 antibody immobilized on a CM5 chip was used to investigate the kinetics. The ability of CC-90002 to block the interaction of SIRPα with CD47 was also determined by immobilizing recombinant SIRPα protein (NOVOPROTEIN) on CM5 chip using amine coupling.

Cell lines and cell cultures

CCRF-CEM, HL-60, RPMI-8226, NCI-H929 and Raji cells were purchased from American Type Culture Collection (ATCC, Gaithersburg, MD) and cultured according to ATCC recommendations. NCI-H929 cells resistant to lenalidomide (NCI-H929 R1 and NCI-H929R3) were established in house by continuous exposure of the parental cell line to increasing concentrations of lenalidomide [30]. Cell samples from AML patients were obtained from Conversant Bio (Huntsville) and are described in Supplemental Table 1.

CC-90002 binding on cells and blockade of SIRPα to CD47 positive cells

Evaluation of CC-90002 binding was also conducted by flow cytometry on Raji human Burkitt’s lymphoma cells using Alexa 647, Dylight 647 or FITC-conjugated anti-human IgG secondary antibody (Jackson Immuno Research). The same assay was used to evaluate CC-90002 binding to cynomolgus monkey (cyno), mouse and rat CD47. In addition, CC-90002 was analyzed by flow cytometry for its ability to block recombinant SIRPα binding to CD47 on CCRF-CEM cells. Binding of recombinant SIRPα-IgV fused to a mouse Fc was monitored in the presence of increasing amounts of CC-90002 using a Dylight 647 conjugated anti-mouse secondary antibody (Jackson Immuno Research).

Hemagglutination assay

To evaluate the hemagglutinating capacity of CD47 antibodies, human RBCs were diluted to 10% in PBS and incubated at 37 °C for 2 to 6 h with a titration of CD47 antibodies in a round bottom 96-well plate. Evidence of hemagglutination was demonstrated by the presence of non-settled RBCs, appearing as a haze compared to a punctate red dot of non-hemagglutinated RBCs.

Phagocytosis assay

Macrophages were generated by incubating purified human peripheral blood mononuclear cells (PBMCs) in AIM-V medium containing M-CSF (50 ng/mL) and 10% FBS for 10 days and harvesting the adherent fraction. In some phagocytosis assay, M1 or M2 macrophages were obtained after incubation of macrophages with human IFNγ (50 ng/mL, 48 h) and lipopolysaccharide (100 ng/mL, the last 24 h) for M1 phenotype and with IL-4 (50 ng/mL, 48 h) and IL-13 (50 ng/mL, 48 h) for M2 phenotype in M-CSF-containing medium. Unpolarized (M0) and polarized M1 and M2 macrophages were stained with anti-CD38 and anti-CD206 antibodies, markers of M1 and M2 macrophages, respectively.

Macrophages were gently scraped off the plates using Accutase and were resuspended in fresh complete AIM-V medium and incubated overnight in 12-well plates at a density of 750,000cells/well. Medium was aspirated and serum-free AIM-V medium was added with different concentrations of CC-90002 as well as 1.5 × 106 CSFE-labeled tumor cells. Plates were briefly centrifuged at 500 rpm and incubated for 3 h at 37 °C. Supernatants were collected, and non-phagocytosed CFSE+ cells were counted by flow cytometry. Phagocytosis index was measured by analyzing macrophages stained with the anti-human CD14 antibody with flow cytometry. Cyno monocyte-derived macrophage were made following the same protocol for human monocyte-derived macrophage. Briefly, CCRF-CEM cells were treated with 10 μg/mL of CC-90002 or IgG4 isotype control and subsequently mixed with cyno macrophages for a final concentration of 5 μg/mL. The mixture was cultured for 3 h and phagocytosis was measured by FACS analysis.

Immunofluorescence staining and Immunohistochemistry (IHC)

Anti-human IgG conjugated with Alexa Fluor® was used to determine binding of CC-90002 to tumor cells. Frozen sections were fixed in 4% paraformaldehyde, washed in PBS, blocked and permeabilized with normal goat serum and Triton X-100. Sections were incubated with anti-human IgG conjugated with Alexa Fluor® 488 followed by counterstaining with DAPI. IHC was performed on the Bond-Max automated slide stainer (Leica Microsystems, Buffalo Grove, IL) and the associated Bond Polymer Refine Detection Kit. For IHC, formalin fixed paraffin embedded tumor sections were deparaffinized and antigen retrieval was performed with Epitope Retrieval 2 (pH 9.0). Anti-F4-80 antibody was used to detect the macrophages in the tumor. Horseradish peroxidase (HRP)-labeled polymer and diaminobenzidine tetrahydrochloride was used to visualize specific antibody localization. Negative controls included omission of the primary antibody.

ELISA

The potency of SIRPα blocking by CD47 antibodies was measured by an enzyme-linked immunosorbent assay (ELISA)-based format. Both cyno and human CD47 were produced as fusion proteins with human Fc and were coated on NUNC Medisorp plates in PBS at 2 μg/mL in 100 μL overnight at 4 °C. The sample plate was prepared by making a serial dilution of CC-90002 in a fixed concentration of 0.05 μg/mL SIRPα-Fc in PBST (PBS with 0.05% Tween-20). The cyno SIRPα was produced as a llama Fc protein and the human SIRPα was produced as a mouse Fc fusion protein. An anti-llama Fc HRP secondary (Bethyl Laboratories Inc.) or anti-mouse Fc HRP secondary (Jackson Immuno research Laboratories Inc.) antibody was used to detect the residual SIRPα bound. SureBlue TMB substrate was added to develop the assay after wash of secondary antibody. MCP-1, MIP-1α and IL-1β (R & D Systems) ELISA assays were conducted following manufacturer’s instructions and the plates were scanned using a microplate reader (VersaMax, Molecular Devices). Cytokine concentration in tumor was calculated as amount of cytokines per mg of total protein.

In vivo tumor studies

All animal studies were conducted according to guidelines established by the internal Institutional Animal Care and Use Committee (IACUC). Female NOD-SCID mice were inoculated with RPMI 8226, NCI-H929, NCI-H929/ R1 tumor cells subcutaneously with a single-cell suspension. Animals with tumors of approximately 200 mm3 were randomly assigned to treatment groups. CC-90002 was formulated in 50 mM histidine, 150 mM NaCl, 2% Trehalose, 0.1% Tween20, adjusted to pH 6.0. The tumor volumes were calculated using the formula: width2 × length/2 and expressed in mm3. The percentage body weight change during the course of study was calculated using initial body weight measurements. For HL-60 disseminated model, female NOD-SCID mice were inoculated with HL-60 cells by tail vein injection (n = 9–10/group) and treated 2 weeks post-engraftment with vehicle, human immunoglobulin G subtype 4 (hIgG4) isotype control, cytarabine (AraC; 50 mg/kg daily for 10 days), fludarabine (60 mg/kg twice a day [BID] for 9 days) or CC-90002. Tumor burden from the blood of HL-60-bearing animals was assessed by FACS at Weeks 5 and 7 post-tumor cell inoculation. At the termination of the experiments, plasma and tumor samples were collected for cytokine/chemokine analysis using ELISA. Tumor samples were also used to determine the binding of CC-90002 to tumor cells and macrophage infiltration by IHC. For macrophage depletion studies, RPMI8226 tumor-bearing mice were randomized when average tumor volumes reached ~ 400mm3. Mice were treated with a combination of hIgG4, CC-90002, liposome or liposome-clodronate (FormuMax Scientific Inc., Sunnyvale, CA). Liposome or liposome-clodronate, at 50 mg/kg, were dosed IV Q3Dx8 and hIgG4 or CC-90002, at 1 mg/kg, was dosed IP QWx4. Depletion of F4/80 + CD115 + CD11b + monocyte population from bone marrow and peripheral blood were monitored by flow cytometry.

Human CD33 cell measurement from peripheral blood by flow cytometry

To track tumor burden in HL-60 model, the animals were bled at weeks 5 and 7 post tumor cell inoculation. After lysing the RBCs, samples were centrifuged and PMBCs were collected. The PBMCs were then washed and blocked with Anti-Mouse CD16/CD32 (FcIII/II R block, eBioscience) and anti-human Fc Receptor binding inhibitor (eBioscience), followed by staining with a human-specific anti-CD33 antibody conjugated with FITC (BD Biosciences). Cells were washed and recorded using BD Calibur Cytometer (BD Biosciences). The live cells were gated using a forward scatter/side scatter strategy. The percentage of human CD33+ cells within the live cell population was determined and data were analyzed using FlowJo software (FlowJo X 0.7). The PBMCs obtained from naïve NOD-SCID mice were used as negative controls.

Non-human primate toxicity and toxicokinetics studies

Single- and repeated-dose toxicity studies were conducted at Charles River Laboratories (Reno, Nevada, USA) and Huntington Life Sciences (Somerset, New Jersey, USA), respectively, each according to a written study protocol and facility standard operating procedures in compliance with IACUC criteria, national legal regulations on animal welfare, and accepted animal welfare standards. The repeated-dose study was conducted in compliance with the Food and Drug Administration (FDA) Good Laboratory Practice Regulations (GLP) as set forth in Title 21 of the US Code of Federal Regulations, Part 58. For the single-dose study, female cynomolgus monkeys (3 animals per dose) were administered a single intravenous (IV) infusion of CC-90002 (10, 30 or 100 mg/kg) or vehicle control (50 mM L-Histidine pH 6.0 containing 150 mM NaCl, 2% trehalose, 0.1% polysorbate 20) following a 20-day observation period. In the repeated-dose study, CC-90002 was administered by IV once weekly for a total of five doses to cynomolgus monkeys (5 animals/sex/group) at dose levels of 0 (vehicle: 10 mM citrate, 0.05% Polysorbate 20 and 9% sucrose, pH 6), 20, 60 or 150 mg/kg/dose. In-life sampling, monitoring or/and evaluations included clinical observations, body weight, food consumption, cardiovascular safety pharmacology (repeated-dose study only), clinical pathology and toxicokinetics. At the end of the treatment period for multi-dose groups (4 weeks after the first dose), 3 animals/sex/group were euthanized, necropsied and subjected to a complete histopathology evaluation. In the repeated-dose study, the remaining 2 animals/sex/group were held for a 5-week treatment-free recovery period and were then likewise euthanized, necropsied and subjected to histopathologic evaluation.

Statistical analysis

The mouse tumor data were analyzed by GraphPad Prism (GraphPad Software Inc.) and the results were presented as mean ± SE. Comparisons were performed using a One-Way ANOVA.

Results

Antibody engineering and characterization of CC-90002

CC-90002 was generated from a murine parental antibody 2A1, which had a high affinity toward the endogenous cell surface CD47 and significantly blocked the interaction between CD47 and SIRPα. The humanized antibody maintained the same binding affinity as the murine mAB 2A1 but was further engineered to lower the Fc-mediated effector function (Fig. 1). The heavy chain variant Q was engineered to remove a potential asparagine deamidation site within CDR2 (N55Q) and the light chain variant N was engineered to remove a potential aspartate isomerization site within CDR2 (D56S). The variable domains of the QN.IgG1 were grafted to the IgG4PE (IgG4 with S228P and L235E mutations; CD47.G4PE; CC-90002) isotype. The CD47.G4PE was designed to stabilize the IgG4 hinge and have significantly lower Fc-mediated effector function. The heavy chain constant region, γ4, contains two non-standard amino acid substitutions, S228P and L235E. Serine 228 was changed to proline to reduce the level of “half-antibody” that is commonly observed in the production of the IgG4 antibody subclass [31]. Leucine 235, one of the critical amino acids involved in heavy chain interactions with Fcγ receptors, was changed to glutamic acid. The L235E substitution reduced the interaction of γ4 chain to Fcγ receptors (Fig. 1A) and minimized Fc-mediated antibody-dependent cellular cytotoxicity (ADCC) (Fig. 1B). In addition, inherent lack of complement binding by γ4 heavy chain minimizes the risk of complement-dependent cytotoxicity (CDC) function.

Fig. 1
figure 1

Engineering of CC-90002 reduced its binding to activating FcγRs and ADCC function. The binding of CC-90002 to activating and inhibitory human Fcγ receptors [FcγRI, FcγRIIA (H131), FcγRIIA (R131), FcγRIIB and FcγRIIIA (V158)], including specific genetic polymorphisms was examined (A). A CC-90002 antibody variant on an IgG1 framework (referred to as IgG1, CD47.G1 or CD47-IgG1) shown to have effector function was used as a positive control. In addition to the FcγR binding, ADCC activity was determined using a co-culture assay with human PBMCs and the CD47-expressing cell line CCRF-CEM at a 50:1 ratio (B). Cells were incubated for 3 h in the presence of human IgG4 isotype control, CC-90002 or CD47 IgG1. Antibody-dependent cellular cytotoxicity activity was measured using CytoTox Glo™ and % cytotoxicity was determined

Full size image

Studies were conducted using SPR to evaluate the binding of CC 90002 to CD47 (Fig. 2A). CC-90002 bound to the human CD47-ECD with a dissociation constant (KD) value of 0.54 ± 0.37 nM. The ability of CC-90002 to bind endogenous CD47 was evaluated in Raji cells by fluorescence-activated cell sorting (Fig. 2B). The half maximal effective concentration (EC50) was approximately 0.035 nM (Fig. 2A). The ability of CC-90002 to block the interaction of SIRPα with CD47 was determined using SPR with immobilized human recombinant SIRPα and a fixed concentration of CD47-ECD (Fig. 2C). The 50% inhibitory concentration (IC50) value was determined to be 0.23 nM. CC-90002 demonstrated potent blocking of the CD47– SIRPα interaction with an IC50 value of 0.39 nM in the cell-based SIRPα blocking assay using recombinant SIRPα-IgV-mouse Fc and CCRF-CEM human ALL cells as the source of CD47 (Fig. 2D).

Fig. 2
figure 2

source of CD47. Data were normalized to a no antibody control. E. CC-90002 does not promote hemagglutination of human RBCs. Human RBCs were treated with a titration of CD47 antibodies. Evidence of hemagglutination was demonstrated by the presence of non-settled RBCs, appearing as a haze compared to a punctate red dot of non-hemagglutinated RBCs. The anti-CD47 antibody 9E4 was a positive control. 2A1m and 2A1-xi (chimeric): murine anti-CD47 antibodies. QN-IgG1 and QN-IgG4P: variants of CC-90002

CC-90002 binds to CD47 and blocks CD47/ SIRPα interaction without promoting hemagglutination. A. CC-90002 binding to extracellular domain of CD47. CC-90002 was immobilized on a CM5 chip and used for binding assessment. Plot shows binding (y-axis) vs time 10 (x-axis). The colored lines represent the different concentrations of CD47-ECD and the curve fits are shown in black. In this experiment, CC-90002 is bound to the human CD47-ECD with Kd = 1.02 nM, koff = 4.5 × 10–3 s-1, kon = 4.43 × 106 M−1 s−1, with SPR maximum response (Rmax) = 67.46 and X2 = 2.42. B. Binding of the three humanized variants of anti-CD47 to endogenous CD47 on Raji cells. CC-90002 is shown in red. C. CC-90002 blocks SIRPα binding to CD47 ECD. The determination of the fifty percent inhibitory concentration (IC50) of CC-90002 for blocking the interaction of CD47 with SIRPα was based on the data presented. In this experiment, the IC50 was 0.23 nM. D. CC-90002 blocks recombinant SIRPα binding to CD47 on cell surface. CC-90002 blocking of CD47 and SIRPα interaction using CCRF-CEM cells as the

Full size image

CD47 antibodies have been reported to cause hemagglutination of human erythrocytes [32], and several of more recently developed humanized antibodies were shown to have reduced RBC binding properties [18, 19]. To evaluate the hemagglutinating capacity of CC-90002, human RBCs were treated with a titration of CD47 antibodies. Figure 2E illustrates the hemagglutination assay with different antibodies. The anti-CD47 antibody 9E4 was a positive control and elicited hemagglutination while the murine anti-CD47 2A1, 2A1-xi (chimeric) and the three humanized variants, including CC-90002, did not promote hemagglutination.

CC-90002 species cross-reactivity to CD47 was evaluated using human, cyno and canine PBMCs or mouse and rat splenocytes. CC-90002 bound to cyno CD47 with an EC50 value of 0.94 nM, which was similar to the EC50 value for human PBMCs of 0.54 nM (Supplementary Fig. S1A). As shown in Supplementary Fig. S1A-D, CC-90002 did not bind to mouse, rat or canine CD47. CC-90002 blocked cyno CD47 binding to cyno SIRPα with an IC50 value of 7.0 ± 2.6 nM, which was similar to the IC50 value (19 ± 10 nM) for CC-90002 blockade of human CD47 binding to human SIRPα, indicating functional relevance of its binding to cyno CD47 (Supplementary Fig. S1E). In addition, CC-90002 induced an increase in phagocytosis of CCRF-CEM cells over the isotype control using cyno monocyte-derived macrophages. The % phagocytosis index (42.6 ± 14.5) was comparable to the human monocyte-derived macrophage phagocytosis index (Supplementary Fig. S1F).

CC-90002 enhances antibody-dependent phagocytosis of tumor cell lines and patient samples

Studies were conducted to demonstrate the effect of CC-90002 on the phagocytosis of hematological cancer cell lines and primary AML patient samples in vitro. CC-90002 enabled antibody-mediated phagocytosis of a panel of hematological cancer cell lines, including ALL, MM, and AML cells. All three variants of the CD47 antibodies (CD47.IgG1, CD47.IgG4P and CD47.IgG4PE) demonstrated similar phagocytic activity and all showed superior phagocytic activity compared to the benchmark mouse antibody mB6H12 (Supplementary Fig. S1G). Antibody concentration–response studies indicated that the effect of CC-90002 was concentration-dependent with an EC50 value for CCRF-CEM of 185.4 ± 61.3 ng/mL (n = 5 different PBMC donors; phagocytosis index ~ 60% at 1 μg/mL). A representative data set of CC-90002 effect on phagocytosis of CCRF-CEM cells is shown in Fig. 3A.

Fig. 3
figure 3

CC-90002 enabled antibody-mediated phagocytosis of a panel of hematological cancer cell lines (ALL and MM) and AML patient samples. A. CCRF-CEM were fluorescently labeled with carboxyfluorescein diacetate succimidyl ester (CFSE) and incubated with macrophages in the presence of CC-90002 or isotype control (Hu.IgG4). Phagocytosis index was calculated as described in Material and Methods. The histogram is representative of five independent experiments performed with macrophages from 5 different donors. Data are presented as the mean with error bars representing standard deviation. B. CCRF-CEM were fluorescently labeled with CFSE and incubated with macrophages in the presence of CC-90002 or 1 μg/mL isotype control (Hu.IgG4). Phagocytosis index was calculated as described in Material and Methods. The histogram shown is representative of 4 independent experiments performed with macrophages from 4 different donors. Data are presented as mean values and error bars represent standard deviations. C. Histograms show the effect of CC-90002 and CD47.IgG1 on the phagocytosis of the lenalidomide-sensitive (H929-D1) and resistant (H929/R1, H929/R3) cell lines. Macrophages were derived from two different donors (donor #21 and #22). Untreated cells (No Ab) and cells treated with 1 μg/mL human IgG1 served as the negative and isotype controls, respectively. Data are presented as the mean with error bars representing the standard error of the mean (N = 2). D. Three (AML-1, AML-2 and AML-3) primary cells from AML patients were fluorescently labeled with CFSE and incubated with macrophages in the presence of CC-90002 isotype control (Hu.IgG4). Phagocytosis index was calculated as described in Material and Methods. **p < 0.01, ***p < 0.001

Full size image

Macrophages display different phenotypes depending on the microenvironment [33]. By analogy to the Th1/Th2 classification of T cells, two distinct polarized states of macrophages have been described: M1 and M2 macrophages [34]. Classically activated M1 macrophages are considered proinflammatory as they secrete proinflammatory cytokines and reactive oxygen species and display increased killing activity. In contrast, the alternatively activated M2 macrophages mainly participate in Th2 response. M2 macrophages have also been described as residing in tumors, participating in tumor progression [35]. A concentration dependence study of CC-90002-mediated phagocytosis was performed with M0, M1 and M2 macrophages. Figure 3B shows that M0, M1 and M2 were able to phagocytose tumor cells in response to CC-90002 treatment. Average phagocytosis index (%) was similar for M0 and M2 macrophages at 1 μg/mL CC-90002 (67.8 ± 14.1 for M0 and 51.6 ± 10.2 for M2 macrophages), and approximately 2 times superior to M1 macrophages (28.1 ± 11.7).

Treatment with CC-90002 also enabled the phagocytosis of MM cell lines, RPMI 8226 and NCI-H929 (Supplementary Figure S1H) and two cell lines derived from NCI-H929 cells made resistant to lenalidomide (H929/R1 and H929/R3) (Fig. 3C). The phagocytosis index at 1 μg/mL CC-90002 or CD47.IgG1 ranged from approximately 20% to 40% across MM cell lines. Importantly, CC-90002 treatment enabled phagocytosis of both lenalidomide-sensitive and lenalidomide-resistant H929 cells with the same level of efficiency (Fig. 3C). In addition to exhibiting effectiveness in ALL and MM, CC-90002 treatment significantly enabled the phagocytosis of primary AML cells from three patients (Fig. 3D; Supplementary Table S1).

CC-90002 demonstrated macrophage-dependent antitumor activity in vivo

The in vivo antitumor activity of CC-90002 was evaluated in human MM and AML cell line-derived xenograft models. RPMI8226 and Parental and lenalidomide-resistant NCI-H929 MM cells were transplanted subcutaneously into NOD-SCID mice and allowed to establish the tumor to approximately 200 mm3 prior to initiation of therapy. Significant dose-dependent antitumor activity was observed with CC-90002 treatment in RPMI 8226 (Fig. 4A) xenograft model. The lowest dose of CC-90002 tested (0.1 mg/kg, dosed once a week [QW]) inhibited 57% tumor growth when compared with vehicle control. At 1 mg/kg dosed QW, the majority of animals were tumor free by the end of the study in all three models. More importantly, CC-90002 exhibited similar antitumor activity in the parental NCI-H929 and lenalidomide-resistant NCI H929 (H929/R1) xenograft models (Fig. 4B, C). At the doses tested, there was no significant body weight loss or other clinical observations noted in animals treated with CC-90002.

Fig. 4
figure 4

Antitumor activity of CC-90002 in MM and AML xenograft models. A-C. RPMI8226 (A), parental NCI-H929 (B) and lenalidomide-resistant NCI-H929 referred to as NCI-H929-1051 (C) tumor cells were inoculated subcutaneously into the flanks of NOD-SCID mice. Mice with tumors of approximately 200 mm3 were randomized into treatment groups and administered with vehicle (10 mM citrate buffer, 9% sucrose, 0.05% PS20, pH 6.0), human immunoglobulin G subtype 4 (hIgG4) isotype control, CC-90002 or vincristine 1 mg/kg once in 4 days (Q4D). Vehicle, isotype control and CC-90002 were administered one a week (QW; arrows). Results are presented as the mean and error bars represent standard error of the mean. Statistical analysis was with one-way ANOVA with Dunnett’s post hoc test. D & E. Antitumor activity of CC-90002 in disseminated HL-60 promyelocytic leukemia model. HL-60 cells were inoculated intravenously via tail vein in NOD-SCID mice. Two weeks post-tumor cell inoculation, the animals were randomized into groups treated with vehicle (10 mM citrate buffer, 9% sucrose, 0.05% PS20, pH 6.0), human immunoglobulin G subtype 4 (hIgG4) isotype control, CC-90002, cytarabine 50 mg/kg daily (QD) for 10 days or fludarabine 60 mg/kg twice a day (BID) for 9 days. CC-90002 was administered once a week for 4 weeks (QWx4) or intra-subject dose escalation (0.1 or 0.3 mg/kg on day 1 (D1), 0.3 or 1 mg/kg on days 4 and 8 (D4 and D8) followed by once a week for 2 weeks (QWx2). D. Tumor burden assessment on weeks 5 and 7 post-tumor cell inoculation. CD33+ cells in the peripheral blood was analyzed by FACS. Data were calculated as the relative to hIgG4 control cohort. Results are presented as the mean and error bars represent standard error of the mean. Statistical analysis was with one-way ANOVA with Dunnett’s post hoc test. E. Survival analysis. Kaplan–Meier plot of overall survival of HL-60 inoculated mice treated as indicated. The p-values for all CC-90002 treated cohorts were < 0.0001 derived by log-rank test

Full size image

An additional study was conducted to determine the antitumor activity and tolerability of CC-90002 in a disseminated HL-60 human promyelocytic leukemia cell line-derived xenograft model. In this model, HL-60 human AML cells when injected intravenously engraft in host bone marrow and internal organs and animals develop progressive peripheral leukocytosis. Two weeks post-engraftment, the animals were treated with vehicle, human IgG4 isotype control, cytarabine (AraC; 50 mg/kg daily for 10 days), fludarabine (60 mg/kg twice a day [BID] for 9 days) or CC-90002 (Fig. 4D, E). The dosing groups for CC-90002 included 0.1, 0.3, and 1 mg/kg weekly (QWx4), and two additional cohorts with an intra-subject dose escalation (e.g., 0.1 mg/kg followed by 0.3 mg/kg on Day 4 and Day 8 and weekly thereafter for two weeks [QWx2]). Tumor burden from the blood of HL-60-bearing animals was assessed by FACS at Weeks 5 and 7 post-tumor cell inoculation (Fig. 4D).

Human CD33+ tumor cells were detectable in the peripheral blood at 5 weeks post-engraftment (% hCD33+ cells in the isotype and vehicle groups were 7.9% and 10.4%, respectively). Significant tumor burden reduction was observed in all CC-90002 dose groups at 5 weeks post-engraftment compared to the hIgG4 isotype control (72–86% reduction, p < 0.05–0.01) and at Week 7 post-engraftment (90–95% reduction, p < 0.0001). All CC-90002 dosing regimens were well tolerated. Mice from hIgG4 or vehicle-treated groups started to succumb to disease burden starting on Day 44 and Day 38 with a median survival of 49 and 46 days, respectively (Fig. 4E). The median survival for 0.1 mg/kg and 0.3 mg/kg (QWx4) CC-90002 treatment groups was 74.5 and 112 days, respectively, indicating a dose-dependent effect. At study termination on Day 143, the 1 mg/kg CC-90002 treatment group and the two intra-subject dose escalation treatment groups had not reached their median survival endpoint. All CC-90002 dose levels significantly prolonged survival (p < 0.0001; log-rank test) in the HL-60 disseminated tumor model compared with the hIgG4 isotype control group (Fig. 4E).

In addition to hematological tumors, antitumor activity of CC-90002 was evaluated in one cell line-derived (MDA-MB-231) and one patient-derived (AA2116) triple negative breast cancer (TNBC) xenograft models. In both models, CD47 was abundantly expressed on the cell surface (Supplementary Figs. S2A and B). NOD-SCID mice with tumors of 150–200 mm3 were administered once a week for 3 weeks with CC-90002 at doses ranging between 1 and 10 mg/kg. At the termination of the study, > 50% of the animals remained tumor free with lowest dose (1 mg/kg) tested in MDA-MB-231 xenograft model. Similarly in an AA1126 patient-derived xenograft model, a dose-dependent tumor growth inhibition was observed with tumor-free animals at 3 mg/kg (Supplementary Figs. S2C and D).

To determine the potential mechanism of action of CC-90002, NOD-SCID mice with RPMI 8226 tumors were treated with either a single or multiple doses of 10 mg/kg CC-90002 and plasma and tumor samples were collected at different time points for cytokine and chemokine analysis. Tumor samples were also used to detect the expression of CD47, binding of CC-90002 to tumor cells and CC-90002-induced macrophage infiltration. RPMI 8226 tumors showed an abundant expression of CD47 on the tumor cell membranes (Supplementary Fig. S3). Following single-dose administration, there was a progressive time-dependent increase in the binding of CC-90002 to tumor cells in vivo from 6 to 72 h (Supplementary Fig. S4). Single-dose administration of CC-90002 caused a significant, time-dependent increase in the influx of F4-80-positive macrophages into the tumors from 3 to 96 h following CC-90002 treatment (Fig. 5A, Supplementary Fig. S5). This statistically significant influx was sustained until 21-day post-single dose (Supplementary Fig. S5) and greatly increased in tumors from the animals treated with multiple doses of CC-90002.

Fig. 5
figure 5

CC-90002 mechanism of action in RPMI8226 human multiple myeloma xenograft model (representation of 3 independent experiments). NOD-SCID mice with RPMI 8226 tumors were treated with a single doses of 10 mg/kg CC-90002. Plasma and tumor samples were collected for cytokine and chemokine analysis. Tumor samples were processed for immunohistochemistry using F4-80 antibodies to detect macrophages. A. Immunohistochemistry images showing the macrophages (brown staining) in the tumors from the animals treated with vehicle, hIgG4 or CC-90002. B. Cytokine and chemokine induction in the plasma and tumors of mice treated with hIgG4 or CC-90002. Tumor and plasma samples were harvested at time points indicated. Concentration of mouse MCP-1, MIP-1α and IL-1β in tumor lysates and plasma was determined using enzyme-linked immunosorbent assays (ELISA). Data were calculated as the increase over the hIgG4 control. Results are presented as the mean (n = 5) and error bars represent standard error of the mean. Statistical analysis was with one-way ANOVA with Dunnett’s post hoc test

Full size image

Among the panels of host (murine) and tumor (human)-derived chemokines and cytokines evaluated, murine-derived chemokines (MCP-1and MIP-1α) and cytokine (IL-1β) showed an increase following treatment with CC-90002 (Fig. 5B). The induction of MCP-1 in tumor and plasma of CC-90002-treated animals at 1 h appeared to precede the infiltration of macrophages into the tumor. The presence of MCP-1 and MIP-1α chemokines known to recruit inflammatory cells to tumor sites [36] suggested that macrophages in CC-90002-treated tumors were responding to these signals. Infiltrating macrophages accompanied by host-derived chemokines (MCP-1 and MIP-1α) and cytokine (IL 1β) induction may contribute to the observed antitumor activity of CC-90002.

CC-90002 antitumor activity is mediated by macrophages

As CC-90002 treatment dramatically enhances macrophage infiltration in xenograft tumors, this increased macrophage infiltration was proposed to mediate antitumor activity in RPMI8226 tumor model. To test this hypothesis, we used liposome-clodronate for systemic depletion of macrophages in RPMI-8226 xenograft model. In initial single-dose studies, liposome-clodronate at 50 mg/kg caused a near complete (~ 98%) depletion of F4/80+ CD115+ CD11b+ population from bone marrow and peripheral blood by 24 h and the depleted cells were replenished to normal level by 2 and 6 days, respectively (Supplementary Fig. S6). To investigate the effect of macrophage ablation on CC-90002-mediated antitumor activity, mice with established subcutaneous RPMI8226 tumors of approximately 400 mm3 were administered with liposome-clodronate and CC-90002. Liposome-clodronate treatment was initiated 24 h prior to the administration of CC-90002. As shown in Fig. 6A, co-administration of liposome-clodronate and CC-90002 abrogated the CC-90002-elicited antitumor activity in RPMI8226 tumors. Analysis of samples collected at the termination of the study indicated that treatment with liposome-clodronate not only depleted monocyte/ macrophages from blood of RPMI8226 tumor-bearing mice, but also reduced macrophage infiltration in the tumor that was mediated by CC-90002 (Fig. 6B, C), suggesting macrophages play an important role in CC-90002-elicitied antitumor activity.

Fig. 6
figure 6

Macrophage depletion abrogates antitumor activity of CC-90002 in RPMI8226 tumors. A. NOD-SCID mice were transplanted subcutaneously with RPMI8226 cells. Mice with tumors of approximately 400 mm3 were randomized into 4 groups and treated with liposome, hIgG4 (10 mg/kg, QW), liposome-clodronate (50 mg/kg, Q3D) or CC-90002 (10 mg/kg, QW). In combination group, CC-90002 treatment was initiated 1 day after liposome-clodronate administration. Tumor volumes were assessed twice/week. The data represent mean ± SEM. Statistical analysis was with one-way ANOVA with Dunnett’s post hoc comparing data with liposome plus hIgG4 treatment group. B & C. Lip-clodronate treatment reduced macrophage infiltration in tumors treated by CC-90002. Tumor samples collected at the end of the study were processed for immunohistochemistry using F4-80 antibodies and quantitated. H-score quantification of F4-80 + macrophages in tumors (B). The data represent mean ± SEM. Immunohistochemistry images showing Liposome-clodronate treatment reduced macrophage infiltration in tumors treated with CC-90002 (C)

Full size image

CC-90002 has an acceptable PK and favorable safety profile in non-human primates

Pharmacokinetic (PK) and toxicity assessments of CC-90002 were performed in cynomolgus monkeys following single or repeated doses (weekly administration for one month). In the single-dose toxicity study, female monkeys were administered CC-90002 at dose levels of 0, 10, 30, and 100 mg/kg via IV injection; doses in the repeated-dose study were 0, 20, 60, and 150 mg/kg/dose. CC-90002 exhibited approximately 2.8-fold higher clearance at 10 mg/kg compared to 30 and 100 mg/kg after one week (Table 1). Similarly, in the repeat dose study, the clearance at 20 mg/kg in Week 1 was greater than at higher doses. At doses of 30 mg/kg and higher, the target is likely saturated, leading to lower clearance. Exposures increased in a greater than dose proportional manner from 10 to 30 mg/kg and in an approximately dose proportional manner from 30 to 150 mg/kg. The volume of distribution was close to plasma volume after target saturation, as is expected for mAb therapeutics with IV administration. The half-life of CC-90002 in monkeys was approximately 4 to 5 days at doses of 30 mg/kg or higher (after target saturation) and was shorter when target saturation was not complete. A single dose of CC 90002 at these dose levels were well tolerated; the no-observed-effect level (NOEL) was 100 mg/kg. In the repeated-dose toxicity study, systemic exposure (AUC0-168 h) of CC-90002 increased in a greater than dose proportional manner in the dose range of 20 to 60 mg/kg/week and increased in an approximately dose proportional in the dose range of 60 to 150 mg/kg/week. Target-mediated disposition was responsible for higher clearance and nonlinear kinetics in monkeys up to 20 mg/kg. Maximum plasma concentration (Cmax) increased in an approximately dose proportional manner. The volume of distribution was close to plasma volume. Systemic clearance of CC-90002 increased and systemic exposures decreased between Weeks 1 and 4 and corresponded with the presence of anti-drug antibodies (ADA) for Week 4 (Supplementary Table S2). Despite the generation and presence of ADA, exposure to circulating free CC-90002 was achieved and maintained throughout the dosing phase at all dose levels (Supplementary Table S2), enabling appropriate evaluation of potential toxicities. No sex difference in exposure was observed and no accumulation of CC-90002 was detected in the repeat dose toxicity study (Supplementary Table S2). All animals survived to their scheduled necropsy and there were no CC-90002-related clinical signs or effects on body weight, food consumption, ophthalmology, electrocardiography, heart rate, blood pressure, respiratory rate, hematology, bone marrow cytology, coagulation, clinical chemistry or urinalysis parameters, and no CC-90002-related macroscopic or histopathologic findings. Based on the absence of adverse test article-related findings, the no-observed-adverse-effect level (NOAEL) was 150 mg/kg dosed once weekly, corresponding to a mean, combined gender, Week 4 AUC0-168 h of 335,000 µg∙hr/mL. CC-90002-related findings were limited to higher spleen weights and minimal to slight lymphoid hyperplasia of the spleen and mesenteric and axillary lymph nodes. These changes are consistent with an immunological response to the administration of a human protein to monkeys and were therefore not considered adverse. Following the 5-week recovery period, lymphoid hyperplasia was still evident; however, the severity of hyperplasia was minimal only, indicating partial recovery.

Table 1 Pharmacokinetic parameters of CC-90002 in Cynomolgus Monkeys following a single intravenous administration
Full size table

Discussion

Macrophage receptors discriminate altered-self molecules on aged or dying cells from normal self-markers on healthy cells, resulting in either activation or inhibition of the phagocytosis response (2). The CD47-SIRPα axis is a critical checkpoint regulator of phagocytosis inhibiting macrophage-mediated phagocytosis [4]. Several studies have demonstrated that CD47 blocking antibodies enable the phagocyte-mediated elimination of cancer cells and several molecules blocking this axis have entered clinical development [24]. Our results demonstrated that CC-90002 is a unique anti-CD47 antibody that does not promote hemagglutination while maintaining high affinity to CD47 and inhibition of CD47-SIRPα interaction. We showed that CC-90002 killed a panel of hematological cancer cell lines and primary cells from patients in vitro, and exhibited antitumor activity in subcutaneous and disseminated xenograft models in a macrophage-dependent manner.

The data presented here demonstrate that CC-90002 potently blocks the CD47-SIRPα interaction, thereby enabling macrophage-mediated phagocytosis of tumor cells. Although, the in vitro phagocytic activity of CD47.IgG4PE was similar to that of CD47.IgG1 and CD47.IgG4P, the safety profile of CD47.G4PE was clearly differentiated from CD47. IgG1 and CD47.IgG4P in cynomologus monkey toxicity study. Compared to NOEL at 100 mg/kg for CC-90002, a NOEL for CD47.IgG1 was not identified due to mortality, alterations in clinical pathology, and gross and histopathology findings observed at 10 mg/kg and a NOEL also was not identified for CD47.IgG4P due to decreased platelets at 10 mg/kg (data not shown). The antibody recognizes a unique conformational epitope on CD47 without promoting hemagglutination. A plausible explanation for this unique selectivity is due to the differential conformation and varying densities of epitopes in different cells [37]. In addition, CC-90002 was designed to have reduced Fc effector function to minimize the risk of ADCC and CDC. Although mechanisms behind CC-90002 binding affinities to tumor cells and RBCs may differ and require additional studies, similar to our data, AO-176, a humanized CD47 mAb in IgG2 Fc format [18], and SRF231, an IgG4 [19], were shown to have tumor cell killing activity without inducing hemagglutination. Moreover, AO-176 has been shown to have cell autonomous killing activity of unknown mechanism. Differential attributes of antibodies such as binding kinetics, epitope recognition, internalization, and downstream signaling contribute to the distinct behaviors of different anti-CD47 antibodies [38, 39].

CC-90002 enabled antibody-mediated phagocytosis of hematological cancer cell lines and primary AML patient samples in vitro. The effect was both cell line- and donor monocyte-dependent. In the MM xenograft models, weekly dosing of CC-90002 yielded tumor regression and tumor-free animals. Interestingly, CC-90002 treatment significantly enabled the phagocytosis of two lenalidomide-resistant MM cell lines in vitro and showed significant antitumor activity in vivo. This activity indicates that CC-90002 may be active in settings of IMiDS® relapse or insensitivity as potential cause of lenalidomide-resistance in cell lines including H929/R1 was reported to be due to lower cereblon expression, point mutations and chromosomal deletion [30, 40] that might not alter the CD47-SIPRα axis.

Primary AML patient samples displayed increased expression of CD47 on the cell surface compared to normal cell counterparts [9]. Furthermore, higher levels of CD47 mRNA expression were an independent poor prognostic factor in AML patients. Recently, positive clinical efficacy was observed with combination therapy of azacitidine and magrolimab, a humanized anti-CD47 antibody, in untreated AML, particularly TP53-mutant AML patients [41]. Our data demonstrated CC-90002-induced robust phagocytosis of primary AML patient leukemic cells in vitro by macrophages. Furthermore, in vivo, CC-90002 treatment resulted in prolonged survival in HL-60 disseminated xenograft model, compared to IgG4 isotype control-treated mice.

The in vivo activity of CC-90002 in inhibiting the tumor growth in various xenograft models in immunocompromised mice is generally consistent with previously reported anti-CD47 antibodies such as AO-176 [18] and SRF231[19], subtle differences exist. In our studies with TNBC models, 3 weekly doses of 1 mg/kg produced tumor-free animals in MDA-MB-231 xenograft model, while AO-176 produced an 83% tumor growth inhibition after 4 weeks of dosing at 15 mg/kg, 5 times/week. Binding affinities, PK properties, and Fc-mediated clearance may influence the frequency and dose levels in animal studies. It should be cautioned, however, that the active dose level of CC-90002 in patients cannot be fully elucidated in immunodeficient mouse models due to the lack of CC-90002 binding to mouse CD47 and the absence of a competent immune system.

From the current study, multiple lines of evidence points to the macrophages as being central to the CC-90002 elicited antitumor activity in MM and AML tumor models. The massive time-dependent infiltration of macrophages into the tumor following CC-90002 treatment preceding the induction of myeloid-derived chemokines MCP-1, MIP-1α and IL-1β in the plasma and tumor indicates that myeloid cells are the principal component of CC-90002-induced antitumor activity. The biphasic response of host-derived chemokines and cytokines suggest that initial induction of chemokines as early as 1–3 h following dosing by resident myeloid cells further recruit additional cells resulting in massive infiltration of macrophages into the tumor leading to further elevation of MCP-1 and MIP-1α followed by tumor cell death and necrosis culminating in tumor regression. Additional evidence of the role of macrophages in the CC-90002-induced antitumor activity came from liposome-encapsulated clodronate treatment. While liposome-clodronate, which depletes macrophages and other phagocytic cells, has previously been shown to have profound effect on tumor progression attributing to macrophage-derived insulin-like growth factor 1 in MM [42], liposome-clodronate as a single agent showed marginal effect on the tumor growth. Macrophage depletion using liposome-clodronate led to near complete abrogation of CC-90002-elicited antitumor activity in a RPMI8226 MM xenograft model. These data suggest reduced but residual FcR function of CC-90002 is sufficient to mediate antitumor activity in human tumor cell-NOD-SCID macrophage setting. Relative affinity differences between effector cell and tumor cell SIRPα-CD47 axis in various species influencing phagocytosis have been documented [43] and blocking human CD47-human SIRPα in man may require addition perturbations. Additionally in immunocompetent host, CC-90002 may have additional macrophage-independent antitumor activities. Indeed, activated T cells and NK cells were recently reported to also express SIRPα [44, 45].

In conclusion, CC-90002, an investigational anti-CD47 antibody with IgG4-PE isotype, showed substantial antitumor activity in multiple models of MM, AML and solid tumors in vitro and in vivo without having hemagglutination properties. While further understanding of how CD47-SIRPα axis interruption and subsequent Fc-FcR engagement contributes to CC-90002 activity might be needed, minimal residual Fc-function in IgG4PE and macrophage engagement appears to be central to the mechanism of CC-90002-elicited antitumor activity. In non-human primates, CC-90002 was well tolerated and allowed for the calculation of a clinical start dose and schedule in oncology patients. The cumulative data from preclinical pharmacology studies have provided rationale to support initiation of the clinical studies CC-90002-ST-001 (NCT02367196) and CC-90002-AML-001 (NCT02641002) with CC-90002 as a single agent and in combination with other cancer therapeutic modalities in patients with hematologic and solid tumor malignancies.