CD47-blocking immunotherapies stimulate macrophage-mediated destruction of small-cell lung cancer

Abstract

Small-cell lung cancer (SCLC) is a highly aggressive subtype of lung cancer with limited treatment options. CD47 is a cell-surface molecule that promotes immune evasion by engaging signal-regulatory protein alpha (SIRPα), which serves as an inhibitory receptor on macrophages. Here, we found that CD47 is highly expressed on the surface of human SCLC cells; therefore, we investigated CD47-blocking immunotherapies as a potential approach for SCLC treatment. Disruption of the interaction of CD47 with SIRPα using anti-CD47 antibodies induced macrophage-mediated phagocytosis of human SCLC patient cells in culture. In a murine model, administration of CD47-blocking antibodies or targeted inactivation of the Cd47 gene markedly inhibited SCLC tumor growth. Furthermore, using comprehensive antibody arrays, we identified several possible therapeutic targets on the surface of SCLC cells. Antibodies to these targets, including CD56/neural cell adhesion molecule (NCAM), promoted phagocytosis in human SCLC cell lines that was enhanced when combined with CD47-blocking therapies. In light of recent clinical trials for CD47-blocking therapies in cancer treatment, these findings identify disruption of the CD47/SIRPα axis as a potential immunotherapeutic strategy for SCLC. This approach could enable personalized immunotherapeutic regimens in patients with SCLC and other cancers.

Figure 1. CD47 is a therapeutic target for SCLC.

(A) Histological analysis of macrophage infiltration in SCLC patient samples (n = 79). Specimens were stained for the macrophage markers CD68 and CD163. Samples were scored from 1 to 3 (1, low; 2, moderate; 3, intense) based on macrophage infiltration, with representative images of each score depicted (left). One sample exhibited no evidence of macrophage infiltration (not shown). Chart depicts a summary of macrophage infiltration scores as varied by tumor stage (right) (correlation coefficient r = 0.2721; P = 0.0153). Scale bar: 300 μm. (B) CD47 expression on human SCLC cell lines (n = 6) by flow cytometry. Dotted black line, unstained NCI-H82 cells. (C) CD47 expression on SCLC patient sample PDX NJH29 (left) and quantification of CD47 on PDX samples from chemonaive patients (n = 3) and patients with recurrent tumors after chemotherapy (treated, n = 4). (D) Gating strategy used for flow cytometry analysis of phagocytosis assays performed with human macrophages (CD45⁺) and calcein AM–labeled SCLC cells. Percentages of calcein AM⁺ macrophages out of total CD45⁺ macrophage population are indicated. (E) Representative images of cell populations after sorting. The double-positive population contains macrophages with engulfed tumor cells. Scale bar: 20 μm. Experiment performed twice with similar results. (F) Summary of phagocytosis assays using human macrophages and calcein AM–labeled SCLC cell lines (left) or primary NJH29 SCLC cells as analyzed by flow cytometry. SCLC cells were treated with vehicle control (PBS) or anti-CD47 antibodies (Hu5F9-G4). Assays performed with macrophages from independent donors (n = 4) and depicted as the percentages of calcein AM⁺ macrophages (right) or normalized to the maximal response by each donor (left). Data represent mean ± SD. **P < 0.01; ****P < 0.0001, 2-way ANOVA with Šidák correction (left) or 2-tailed t test (right).

Figure 2. The crystal structure of Hu5F9-G4 diabody in complex with CD47 demonstrates competitive antagonism.

(A) Crystal structure of CD47-ECD in complex with Hu5F9-G4 diabody (PDB 5IWL). For clarity, only 1 domain from each diabody chain is shown. The CDR1, CDR2, and CDR3 loops from the V_H domain (cyan) and the CDR1 and CDR2 loops from the V_L domain (pink) contribute to binding the CD47-ECD epitope. Contact residues are indicated in the inset. (B) Hu5F9-G4/CD47-ECD structure superimposed on SIRPα/CD47-ECD (PDB ID 2JJS), demonstrating a shared binding interface. (C) CD47-binding interface showing residues interacting with only SIRPα (red), only Hu5F9-G4 (blue), or both ligands (purple).

Figure 3. CD47-blocking antibodies inhibit the growth of human SCLC tumors in vivo.

(A) Growth of NCI-H82 cells in the subcutaneous tissue of NSG mice. Mice were randomized into groups treated with vehicle control (PBS) or anti-CD47 antibodies (Hu5F9-G4). Growth was evaluated by tumor volume measurements. Points, measurements from independent animals; bars, median. (B) Growth of GFP-luciferase⁺ PDX NJH29 tumors in the subcutaneous tissue of NSG mice as evaluated by bioluminescence imaging. Mice were randomized into groups treated with vehicle control (PBS) or anti-CD47 antibodies (Hu5F9-G4). (C) Representative bioluminescence images of NJH29 tumors on day 85 after engraftment. (D) Growth of NJH29 tumors as evaluated by tumor volume measurements. (E) Survival of mice bearing PDX NJH29 tumors treated with the indicated therapies. P = 0.0004 by Mantel-Cox test. (A–E) Black arrows indicate the start of treatment; points indicate measurements from independent animals. Volume measurements at each time point are staggered for clarity. Bars indicate median values. n = 7–8 mice per treatment cohort. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Mann-Whitney U test unless otherwise indicated.

Figure 4. CD47-blocking antibodies are effective in an orthotopic model of SCLC.

(A) Representative images of mice injected with SCLC tumor cells showing tumor growth in the thoracic area (left, arrowhead), with some mice developing metastases to the liver (right, arrowhead). (B) Tumor growth monitored postmortem under normal light (top) and under fluorescent light (for GFP expression, bottom). Scale bar: 2 mm. H, heart. (C) Representative bioluminescence images of NCI-H82 tumors on day 46 after engraftment in control mice (PBS) and mice treated with CD47-blocking antibodies. (D) Bioluminescence measurements over time. Points represent values from individual mice; bars represent median. Note dropout of mice in PBS cohort over time due to mortality. n = 8–9 mice per treatment cohort. *P < 0.05; **P < 0.01, Mann-Whitney U test. (E) Survival of mice bearing NCI-H82 tumors treated with the indicated therapies. P < 0.0001 by Mantel-Cox test. (D and E) Arrows indicate start of treatment.

Figure 5. Genetic ablation of Cd47 inhibits SCLC growth in vivo.

(A) Growth of Cas9 control or CD47 knockout NCI-H82 cells after subcutaneous engraftment in NSG mice. (B–E) Generation of a mouse Cd47 knockout SCLC cell line. KP1 cells were subjected to CRISPR/Cas9 genome editing to create a KP1 Cd47 knockout variant. (B) FACS analysis showing CD47 expression on the surface of KP1 cells (red) compared with KP1 Cd47 knockout cells after 2 rounds of sorting (blue). Dotted black line represents unstained KP1 cells. (C) In vitro phagocytosis assay performed with RFP⁺ mouse macrophages and KP1 or KP1 Cd47 knockout cells. Data represent mean ± SD from 3 replicates. ****P < 0.0001, 2-tailed t test. (D and E) Growth of control or Cd47 knockout mouse KP1 cells in immunocompetent (D) or NSG (E) mice. (A, D, and E) Points indicate measurements from independent animals. Volume measurements at each time point are staggered for clarity. Bars indicate median values. Cohorts consisted of n = 9–10 (A) or n = 15 (D and E) mice. *P < 0.05; **P < 0.01; ****P < 0.0001, Mann-Whitney U test (A, D, and E).

Figure 6. Comprehensive antibody screening identifies targets on SCLC for combination therapy with high-affinity SIRPα variants.

Antigen expression on the surface of 4 SCLC cell lines and patient sample NJH29 was assessed by flow cytometry using a collection of 332 antibodies targeting surface antigens. (A) Histogram depicting geometric mean fluorescence intensity (MFI) of all antibodies screened for binding. Median values for each antibody across all 5 samples were fit to Gaussian distribution (black line). Negative antigens (gray), low antigens (red), and high antigens (blue) defined based on MFI thresholds as described in Supplemental Methods (see Supplemental Table 2 for full antigen list). The number of antigens in each category is indicated in parentheses. Geo, geometric. (B) Ranked list of the 39 antigens identified as high based on median MFI across all 5 SCLC samples. (C–D) Phagocytosis of NCI-H82 cells (C) and NCI-H524 cells (D) in response to tumor-binding antibodies alone (red) or in combination with high-affinity SIRPα variant CV1 monomer (blue). Points, measurements from individual donors; bars, median values. Three clones of anti-CD56 (NCAM) antibodies were tested, as well as antibodies to CD24, CD29, CD99, and CD47 (Hu5F9-G4). (E) Phagocytosis of NCI-H82 SCLC cells in response to varying concentrations of the anti-CD56 antibody lorvotuzumab alone (red) or in combination with the high-affinity SIRPα variant CV1 monomer (blue). Data represent mean ± SD. (C–E) Phagocytosis was normalized to the maximal response by each donor (n = 4–8 donors). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, 2-way ANOVA with Šidák correction.