The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors

We evaluated CD47 expression on dissociated patient ovarian, breast, colon, bladder, glioblastoma, hepatocellular carcinoma, and prostate tumor cells by flow cytometry. Viability dyes and antibodies targeted to CD45, CD31, and H-2K^b/d were used to exclude dead, nontumor (lineage), and mouse cells. CD47 expression was detected on nearly all cancer cells from every primary and xenograft patient tumor sample evaluated (Fig. 1A; for representative analyses, see Fig. S1A). CD47 was also highly expressed on tumor samples when analyzed by immunofluorescence (Fig. S1B). As previously reported, CD47 was also detected on dissociated normal (noncancer) cells (Fig. 1A) (13–15). To evaluate the relative expression level of CD47 by tumor cells, we performed quantitative flow cytometry on dissociated cells from separate tumor or matched adjacent normal (nontumor) tissue specimens diagnosed by pathologists at Stanford University. On average, tumor cells expressed ∼3.3-fold more CD47 than corresponding normal cells (Fig. 1B).

We previously demonstrated that increased CD47 mRNA expression levels were correlated with poor clinical outcomes in patients with acute myeloid leukemia and non-Hodgkin's lymphoma (16, 17). To determine if CD47 mRNA expression levels were also a prognostic factor in human solid tumors, we analyzed gene-expression data from several previously described cohorts of patients with ovarian cancers, gliomas, and glioblastomas (Table S1) (18–25). In a univariate analysis, stratification of patients into “CD47 high” and “CD47 low” groups based on an optimum threshold revealed that high CD47 mRNA expression levels were associated with a decreased probability of progression-free (Fig. 2 A and B) and overall (Fig. 2 C–F) survival. These results suggest that CD47 expression levels may be a clinically relevant prognostic factor in some solid tumors.

We previously demonstrated that blockade of CD47-mediated SIRPα signaling using targeted monoclonal antibodies (mAbs) induced the phagocytosis of leukemia, lymphoma, and bladder cancer cells by human and mouse macrophages (5, 16, 17, 26). We expanded our evaluation of human solid tumors and tested the ability of purified blocking (clones B6H12.2 and Bric126) and nonblocking (clone 2D3) anti-CD47 mAbs to induce the phagocytosis of cells from dissociated patient ovarian, breast, and colon tumors, patient-derived glioblastoma neurospheres, and colon cancer cell lines. In contrast to cells treated with isotype-matched mouse IgG, anti-HLA, or nonblocking anti-CD47 (2D3) control mAbs, solid tumor cells treated with the blocking anti-human CD47 (hCD47) mAbs B6H12.2 and Bric126 were efficiently phagocytosed by human and mouse macrophages (Fig. 3 B and C; see Fig. 3A for representative images and Movies S1 and S2 for videos demonstrating phagocytosis of human tumor cells following anti-hCD47 mAb treatment). To ensure anti-hCD47 therapy can also effectively target and eliminate cancer stem cells (CSCs), FACS-purified colorectal CSCs (lineage⁻ESA⁺CD44⁺) were isolated from freshly dissociated xenograft tumors (27). Blocking anti-hCD47 mAbs induced the phagocytosis of purified colorectal CSCs by both human and mouse macrophages (Fig. 3D). These results suggest that CD47 is a legitimate therapy target, and that blocking anti-hCD47 mAbs may be effective therapeutic agents to inhibit solid tumor growth by enabling macrophages to eliminate both CSCs and their differentiated progeny.

Tumor-associated macrophages (TAMs) often aid tumor growth and may be functionally impaired and potentially ineffective at phagocytosis of tumor cells (28–30). To determine if TAMs retain the ability to phagocytose tumor cells upon anti-hCD47 mAb treatment, we isolated TAMs (F4/80⁺) and human tumor cells (GFP⁺) by FACS from single-cell suspensions prepared from large (>1 cm³) subcutaneous xenograft tumors established in NOD/SCID/γ (NSG) mice. These tumors were established with patient breast, bladder, or liver cancer cells previously transduced with a GFP encoding lentivirus. Purified tumor cells and TAMs were mixed in the presence of control IgG1 or anti-hCD47 mAbs and phagocytosis was evaluated by flow cytometry detection of GFP⁺ TAMs. Blockade of CD47 increased TAM phagocytosis of human tumor cells (Fig. 3E; see Fig. S2 for representative analyses), demonstrating that TAMs can be converted into antitumor effectors by anti-hCD47 antibody therapy.

We hypothesized that anti-CD47 mAbs would enable the phagocytosis of xenotransplanted solid tumor cells, resulting in an inhibition or elimination of the developing tumor. Tumor cells from dissociated patient xenograft tumors (see Table S2 for tumor specimen details) were injected into immunodeficient NSG mice (see Table S3 for full experimental details). NSG mice lack B, T, and NK cells, but retain macrophages with phagocytic potential (Fig. 3). Unfortunately, the nonblocking 2D3 clone anti-hCD47 hybridoma is unavailable and we were therefore unable include this antibody treatment as a control. Ovarian cancer xenotransplantation models were established in the peritoneal cavity (Fig. 4A). Prior transduction of these cells with a lentivirus designed to express GFP and Luciferase enabled the use of bioluminescent imaging to monitor tumor growth. Treatment with anti-hCD47 mAbs inhibited tumor growth as evidenced by bioluminescence imaging (Fig. 4A, see Fig. 4B for representative images). Importantly, mice treated with anti-hCD47 mAbs demonstrated a dramatic increase in survival (Fig. 4C). Next, patient breast cancer cells were xenotransplanted into the mammary fat pad. After 8 wk, all (five of five) control IgG-treated mice developed large tumors but no tumors were palpable in the anti-hCD47 mAb-treated mice (Fig. 4D). Antibody treatments were discontinued and the mice were monitored for an additional 4 mo. No tumors developed in the mammary fat pad of anti-hCD47–treated mice, indicating that the mAb therapy had fully eliminated the breast cancer cells, including the CSCs (Fig. 4D). Treatment of a second breast cancer sample resulted in a significant inhibition of tumor growth (Fig. 4E). No therapeutic benefit was observed following anti-CD47 mAb treatment of a third patient breast tumor (Fig. S3B). Patient colon cancer cells transduced with a GFP and luciferase-encoding lentivirus were injected subcutanteously onto the back of NSG mice. Several weeks after engraftment, anti-hCD47 antibody infusions resulted in a significant inhibition of tumor growth, as evaluated by bioluminescence (Fig. S3A) and tumor mass (Fig. 4F). Glioblastoma cells transduced with the GFP Luciferase–encoding lentivirus were implanted into the left hemisphere of mouse brains. Subsequent intraperitoneal treatment with anti-hCD47 mAbs resulted in a significant reduction in tumor growth and possible elimination of the tumor in some mice (Fig. 4 G and H). In two additional xenotransplantation models of bladder and ovarian cancer, a substantial inhibition of tumor growth was observed in the majority of mice; anti-CD47 mAb treatment failed to halt tumor growth in rare cases (Fig. S3 C and D). These results indicate that anti-CD47 antibodies can dramatically inhibit the growth of human solid tumors by blocking the ability of CD47 to transmit the “don't-eat-me” signal to macrophages.

We hypothesized that inhibitory antibodies against CD47 would induce the phagocytosis of circulating tumor cells, or be therapeutically effective against newly seeded tumors, thereby preventing tumor metastasis. We identified a bladder cancer sample that spontaneously metastasized to the lymph nodes and lungs when engrafted subcutaneously onto the back of NSG mice. In contrast to previous experiments (Fig. 4), anti-hCD47 mAb therapy was not initiated until a large tumor mass was detected. Evaluated by gross examination, 24 lymph-node metastases were detected in control IgG-treated mice (Fig. 5 A and B). A single, substantially smaller, secondary lymph node was observed in one anti-hCD47 mAb-treated mouse (Fig. 5 A and B). The presence of micrometastases in the lungs was also evaluated by microscopic analysis. Thirty-seven total micrometastases were detected in eight IgG-treated mice (Fig. 5 C and D). In contrast, solitary lung micrometastases were observed in three anti-hCD47–treated mice (Fig. 5 C and D). Anti-hCD47 mAb therapy also resulted in a significant inhibition of primary site tumor growth over the duration of the experiment (Fig. S4). The inhibitory effect of anti-CD47 mAbs on tumor metastasis was confirmed using a metastatic head and neck squamous cell carcinoma patient sample. When injected subcutaneously onto the back of NSG mice, this tumor rapidly metastasizes to mouse lymph nodes. Because of the aggressive metastasis of this sample, antibody therapy was initiated before detection of a palpable primary tumor. Evaluated by gross examination, large secondary lymph nodes were observed in 80% (four of five) of the control IgG treated mice, but zero (zero of five) metastatic lesions were detected in mice following anti-hCD47 mAb therapy (Fig. 5E). These results indicate that CD47-targeted antibodies will not only inhibit primary tumor growth, but may also be effective in preventing tumor metastasis.

To validate the safety and efficacy of targeting CD47 in fully immune competent hosts, we engrafted 50,000 MT1A2 mouse breast cancer cells into the mammary fat pad of syngeneic FVB mice (31). Upon detection of palpable tumors, 400 μg of control IgG or anti-mouse CD47 (mCD47) mAbs were injected into the mammary fat pad every other day (Fig. 6A). This antibody dose was chosen to reveal any potential toxicity and is not necessarily indicative of the minimal effective dose. After 30 d, tumors were resected and weighed. The anti-mCD47 mAb clone MIAP410 (mouse IgG1) significantly inhibited tumor growth (Fig. 6B) (32). A second anti-mCD47 antibody, clone MIAP301 (rat IgG2A) produced a modest, yet insignificant, inhibition of tumor growth (Fig. 6B) (33). Whereas extensive tumor burden with sparse lymphocytic infiltrate can be viewed in control IgG-treated tumors, Stanford University pathologists observed only isolated tumor cell nests surrounded by substantial inflammatory cell infiltrates anti-mCD47 treated tumors (Fig. 6C). Importantly, the therapeutic anti-mCD47 mAbs produced no unacceptable toxicity over the course of the experiment, despite having spread systemically (Fig. S5). These experiments were analyzed at the end of the therapeutic regimen; acute infusion of these antibodies led to a short-term anemia (Fig. S6). A similar lack of toxicity was observed 1 wk after administering anti-mCD47 antibodies via intraperitoneal injection into tumor free BALB/c mice (Table S4). No increase in cell death was observed when MT1A2 cells were treated overnight with anti-CD47 antibodies, suggesting that direct toxicity is unlikely to be the primary mechanism by which the anti-CD47 antibodies produce a therapeutic effect (Fig. 6D).

Tissue specimens were obtained from consented patients as approved by Stanford University Institutional Review Board protocols. Colorectal cancer samples were obtained as previously described (27). Stanford University pathologists defined tumor and matched adjacent normal (nontumor) tissue specimens.

CD47 expression was dichotomized in large (n ≥ 50 samples) solid tumor datasets for which clinical annotations were available. The Maxstat package (version 0.7–13) (47) of the R programming language (version 2.11) was used to define high- and low-CD47 groups. Log-rank P values and hazard ratios with 95% confidence intervals (CI) were derived by Kaplan–Meier analyses. For Affymetrix arrays, data were normalized from raw CEL files using MAS5 with a custom chip definition file-mapping array oligonucleotides to Entrez gene identifiers (48).

For in vitro phagocytosis assay, 5 × 10⁴ macrophages were plated per well in a 24-well tissue-culture plate. Tumor cells were labeled with 2.5 μM carboxyfluorescein succinimidyl ester (CFSE) according to the manufacturer's protocol (Invitrogen). Macrophages were incubated in serum-free medium for 2 h before adding 2 × 10⁵ CFSE-labeled live tumor cells. The indicated antibodies (10 μg/mL) were added and incubated for 2 h at 37°. Macrophages were repeatedly washed and subsequently imaged with an inverted microscope (Leica DMI6000B). The phagocytic index was calculated as the number of phagocytosed CFSE⁺ cells per 100 macrophages. For ex vivo assays, TAMs (F4/80⁺; eBioscience) and human tumor cells (GFP⁺) were isolated by FACS from single-cell suspensions prepared from large (>1 cm³) subcutaneous xenograft tumors established in NSG mice. Purified TAMs and tumor cells and were mixed in the presence of control IgG1 or anti-CD47 (B6H12) antibody (20 μg/mL) and incubated for 2–4 h. Phagocytosis was then determined by flow cytometry detection of GFP⁺ TAMs.

The anti-hCD47 (B6H12) hybridoma was obtained from the ATCC. Hybridoma cells were cultured under standard conditions and antibodies were purified by Protein G. For quantification of CD47 expression, cells were labeled with a saturating concentration of a 1:1 phycoerythrin- (PE) conjugated anti-CD47 antibody (BD Pharmingen) and analyzed using a BD LSR Fortessa Analyzer. BD QuantiBRITE PE beads (BD Pharmingen) were analyzed at the same settings and conditions as the patient samples. Median absolute CD47 antibody binding for each sample was determined from a calibration curve constructed from the QuantiBRITE bead data using the FlowJo Data Analysis software calibration tool.