An antibody-toxin conjugate targeting CD47 linked to the bacterial toxin listeriolysin O for cancer immunotherapy

Abstract

Antigen-presenting cells phagocytose tumor cells and subsequently cross-present tumor-derived antigens. However, these processes are impeded by phagocytosis checkpoints and inefficient cytosolic transport of antigenic peptides from phagolysosomes. Here, using a microbial-inspired strategy, we engineered an antibody-toxin conjugate (ATC) that targets the 'don't eat me' signal CD47 linked to the bacterial toxin listeriolysin O from the intracellular bacterium Listeria monocytogenes via a cleavable linker (CD47-LLO). CD47-LLO promotes cancer cell phagocytosis by macrophages followed by LLO release and activation to form pores on phagolysosomal membranes that enhance antigen cross-presentation of tumor-derived peptides and activate cytosolic immune sensors. CD47-LLO treatment in vivo significantly inhibited the growth of both localized and metastatic breast and melanoma tumors and improved animal survival as a monotherapy or in combination with checkpoint blockade. Together, these results demonstrate that designing ATCs to promote immune recognition of tumor cells represents a promising therapeutic strategy for treating multiple cancers.

Extended Data Fig. 1 ∣. In vitro toxicity assays of CD47-LLO with cancer cell lines.

a-h, Flow cytometry analysis and quantification of apoptosis and necrosis in 4T1Br4 (a-b), EO771 (c-d), KPC (e-f), and D4M.3A (g-h) cells after treatment with 2 μg/mL CD47-LLO for 0, 6, or 24 hours. Data shown represent mean ± s.d. (b, d, f, h) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test. Data show at least n = 3 biologically independent experiments.

Extended Data Fig. 2 ∣. CD47-LLO promotes dendritic cell phagocytosis, lysosomal permeabilization, antigen presentation, and cGAS-STING activation in vitro.

a, Flow cytometry analysis and b, quantification of phagocytic activity of bone marrow-derived dendritic cells (BMDCs) as evaluated by flow cytometry. BMDCs were collected from n = 3 C57BL6 mice. c, Flow cytometry analysis and d, quantification of BMDCs stained with acridine orange. BMDCs were collected from n = 3 C57BL6 mice. e, Flow cytometry analysis and f, quantification of cross-presentation of SIINFEKL–H2Kb peptides on the surfaces of BMDCs (n = 3). g, Flow cytometry analysis and h, quantification of pSTING levels in BMDCs isolated from co-cultures with EO771 cells. n = 4 C57BL6 mice. Data shown represent mean ± s.d. (b, d, f, h) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test.

Extended Data Fig. 3 ∣. Antitumour effect of intratumoural CD47-LLO in 4T1Br4 and EO771 models.

a, 4T1Br4 tumour volumes after intratumoural injection of anti-CD47 or CD47-LLO. Mice were treated as described in Supplementary Fig. 2. b-c, Flow cytometry analysis of CD4⁺ (b) and CD8⁺ (c) tumour-associated lymphocytes in EO771 tumours at day 16 after tumour inoculation in each group. d, Flow cytometry analysis of SIINFEKL–H2Kb tetramer⁺CD8⁺ T cells within the tumour microenvironment. Data shown represent mean ± s.e.m. (a) analyzed by two-way analysis of variance with Tukey’s multiple comparisons test.

Extended Data Fig. 4 ∣. Biodistribution of CD47-LLO after intratumoural and intraperitoneal administration in vivo.

a, Representative fluorescence images and b, quantification of EO771 tumour-bearing mice taken at predetermined times after intratumoural injection of IR800CW-tagged CD47-LLO (25 μg). n = 2 biologically independent mice. Bilateral tumours enclosed in circles. c, Ex vivo fluorescence images and d, quantification of tumour and major organs collected at 24 h after intratumoural administration. e, Quantification of EO771 tumours taken at predetermined times after intraperitoneal injection of IR800CW-tagged CD47-LLO (100 μg). n = 4 biologically independent mice. f, Ex vivo fluorescence images and g, quantification of tumour and major organs collected at 36 h after intraperitoneal administration. Data shown represent mean ± s.d. (b, d, e, g) analyzed by two-sided unpaired Student’s t test (g).

Extended Data Fig. 5 ∣. Tumour-associated CD45+ cells detected by single-cell RNA sequencing.

a, Uniform manifold approximation and projection (UMAP) of all single cells from 3 treatment groups color-coded by cluster. b, Heatmap of 20 differentially expressed genes in clusters, ranked by false discovery rate (FDR). c, Dot plot showing marker expression for different clusters. Dot size indicates the percentage of cells in each cluster expressing the gene, and colors indicate the average expression levels.

Extended Data Fig. 6 ∣. Tumour-associated neutrophils are enriched in CD47-LLO tumours.

a, Representative images show levels of Ly-6g+ (top) and Ly-6g+/Cd11b+ (bottom) cells detected by immunostaining in 4T1Br4 breast tumour frozen tissue sections. b, Quantification of Ly-6g⁺ cells per DAPI⁺ cells per field of view (n = 3 biologically independent tumours per condition, with >2000 DAPI⁺ cells counted per tumour). c, Uniform manifold approximation and projection (UMAP) of only CD45⁺ granulocytes color-coded by sample. d, UMAP projection of only CD45⁺ granulocytes color-coded by cluster. e, Numbers of cells (y-axis) from each cluster (x-axis) color-coded by sample. f, Percentage of cells (y-axis) from each cluster (x-axis) color-coded by sample. g, Dot plot depicting the top 5 differentially expressed genes per macrophage cluster. The dot size indicates the percentage of cells in each cluster expressing the gene, and colors indicate the average expression levels. Data shown represent mean ± s.d. (b) analyzed by two-sided unpaired Student’s t test.

Extended Data Fig. 7 ∣. Differentially expressed genes define tumour-associated macrophages clusters.

a, Dot plot depicting the top 10 differentially expressed genes per macrophage cluster. Dot size indicates the percentage of cells in each cluster expressing the gene, and colors indicate the average expression levels. b, Dot plot depicting the differential expression of relevant genes for cluster assignments per macrophage cluster.

Extended Data Fig. 8 ∣. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology Biological Process (GOBP) analysis of tumour-associated macrophage clusters.

a, Gene set enrichment analysis using the KEGG gene set for tumour-associated macrophages, with heatmap displaying the 10 most upregulated and downregulated pathways in each cluster ranked by their normalized enrichment scores (NES). b, Gene set enrichment analysis using the GOBP gene set for tumour-associated macrophages, with heatmap displaying the 10 most upregulated and downregulated pathways in each cluster ranked by their NES.

Extended Data Fig. 9 ∣. In vivo toxicity of CD47-LLO.

a, Lymphocyte counts at day 2 and day 9 after intraperitoneal injection of drug (50 μg IgG, 50 μg anti-CD47, 100 μg CD47-LLO, or 100 μg CD47-LLO + 200 μg anti-PD1). b, Red blood cell (RBC) counts at day 2 and day 9 after drug injection. c, Serum blood urea nitrogen (BUN) levels at day 9 after drug injection. d, Serum aspartate transaminase / alanine transaminase (AST/ALT) levels at day 9 after drug injection. e, Body weight changes in mice at day 2 and day 9 after drug injection. Data show at least n = 3 C57BL6 mice per treatment group. f, Serum IL-6 levels at 4 hours after drug injection measured by enzyme-linked immunosorbent assay (ELISA). Data show n = 4 C57BL6 mice per treatment group. g, Serum IL-1β levels at 4 hours after drug injection measured by ELISA. Data show n = 4 C57BL6 mice per treatment group. h, Serum concentrations of anti-LLO antibody at 6 weeks after drug injection by sandwich immunogenicity assay. Data show n = 3 mice per IgG treatment group and n = 5 mice per CD47-LLO treatment group. Data shown represent mean ± s.d. (a, b, c, d, e, f, g, h) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test (a, b, c, d, e) or two-sided unpaired Student’s t test (f, g, h).

Extended Data Fig. 10 ∣. Hematoxylin and eosin staining of paraffin sections of major organs 2 days after intraperitoneal injection of CD47-LLO or CD47-LLO and anti-PD1.

Scale bar, 200 μm.

Fig. 1 ∣. Conjugation and purification of anti-CD47 antibody with Listeriolysin O.

a, Schematic of antibody-toxin conjugation. The CD47 antibody and Listeriolysin O (LLO) were modified by using the click chemistry labelling reagent DBCO-PEG4-NHS ester and the crosslinking reagent SPDP-PEG11-azide, respectively. The modified LLO protein was mixed with the antibody and allowed to react overnight at room temperature. Unconjugated antibody and Listeriolysin O proteins were removed by affinity and size exclusion chromatographic methods, respectively. b, SDS-PAGE results of azide-labelled LLO, DBCO-labelled anti-CD47, conjugation reaction, affinity column passthrough, and size exclusion input sample. Experiment was repeated independently three times with similar results; a representative result is shown. c, Size exclusion chromatogram of conjugate (solid line) and anti-CD47 alone (dashed line). Experiment was repeated independently three times with similar results; a representative result is shown. d, Representative SDS-PAGE results of conjugate in reductive and non-reductive loading buffers from n = 3 independent experiments. e, Haemolysis assay of LLO, CD47-LLO, and CD47-LLO treated with reducing agent (5 mM DTT, 30 min) from n = 3 independent experiments. Haemolysis percentage was normalized to 0.1% Triton X-100. f, Apoptosis assays by flow cytometry of mouse bone marrow-derived macrophages (BMDMs) treated with CD47-LLO or LLO for 24 hours from n = 3 independent experiments. g, Proposed mechanism of action of CD47-LLO. Tumour cells sidestep phagocytosis by amplifying their expression of the “don’t eat me” signal, CD47 (red boxes, left). The anti-CD47 antibody promotes tumour cell phagocytosis by antigen-presenting cells (APCs; macrophages or dendritic cells [DCs]). Inside phagolysosomes, LLO monomers dissociate from anti-CD47 and form membrane permeations that allow tumour DNA and antigenic peptides to escape into the cytosol (inset, right). This in turn activates cGAS-STING to produce type I interferons (IFNs) and enhance cross-presentation of neoantigens. SIRPalpha, Signal regulatory protein alpha. Data shown represent mean ± s.d. (e, f) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test. n = 3 biologically independent experiments.

Fig. 2 ∣. CD47-LLO promotes phagocytosis, lysosomal permeabilization, and antigen presentation in vitro.

a, Representative images of BMDM phagocytosis of EO771 cells (white arrows). Scale bars, 20 μm. b, Quantification of BMDMs containing EO771 cells per field of view (n=20 for IgG, n=18 for IgG-LLO, n=27 for anti-CD47, n=26 for CD47-LLO from n=3 independent experiments). c, Flow cytometry analysis and d, quantification of BMDM phagocytosis. n = 4 independent experiments for IgG, IgG-LLO, CD47-LLO (4°C), and CD47-LLO (NaN₃). n=5 for anti-CD47 and n=6 for CD47-LLO. e, Flow cytometry analysis and f, quantification of CD86 and CD206 expression in BMDMs. n = 4 independent experiments. g, TEM images of BMDMs with gold nanoparticles inside phagolysosomes (green arrows) or within the cytoplasm (red arrows). Yellow arrow shows wall perforation. h, Quantification of intact vacuoles; n=5 independent experiments. i, Confocal images of BMDMs (white arrows) stained with LysoTracker Red. Scale bars, 20 μm. j, Quantification of LysoTracker Red puncta per BMDM (n=114 for anti-CD47, n=119 for CD47-LLO. n=3 independent experiments). k, Flow cytometry analysis of BMDMs stained with acridine orange. l, Quantification of acridine red and m, acridine green from n=4 independent experiments. n, Flow cytometry analysis and o, quantification of SIINFEKL–H2Kb peptides cross-presentation on BMDMs (n=4). p, Flow cytometry analysis and q, quantification of OT-I CD8⁺ T cell proliferation(n=5). Data shown represent mean ± s.d. (b, d, f, j, l, m, o, q) or mean ± s.e.m. (h) analyzed by two-sided unpaired Student’s t tests (h, j) or one-way analysis of variance with Tukey’s multiple comparisons test (b, d, f, l, m, o, q).

Fig. 3 ∣. CD47-LLO activates STING signalling in vitro and primary breast cancer in vivo.

a, Wild-type (WT) mice were inoculated with 4T1Br4 breast tumours and treated with intratumoural anti-CD47 or CD47-LLO as described in Extended Data Fig. 4a. Representative images show levels of phosphorylated STING and macrophages (F4/80⁺). Arrowheads show F4/80⁺ and pSTING⁺ co-labelled cells. Scale bar, 20 μm. b, Quantification of F4/80⁺ cells per DAPI⁺ cells per field of view (n=16 for anti-CD47, n=12 for CD47-LLO from n=3 biologically independent tumours per condition). c, Distribution of pSTING fluorescence intensity within F4/80⁺ cells. n=3 biologically independent tumors per condition. d, Western blotting of proteins in the cGAS-STING pathway from BMDMs. A representative result is shown from n=3 independent experiments. e, Flow cytometry analysis and f, quantification of pSTING levels in BMDMs isolated from EO771 co-cultures. n=4 independent experiments. g, RT-PCR results of interferon (IFN)-α expression levels in BMDMs. n=3 independent experiments. h, RT-PCR results of interferon (IFN)-β expression levels in BMDMs n=3 independent experiments. i, Tumour necrosis factor (TNF)-α levels in cell culture supernatants (n=3). j,k WT or STING^−/− mice were inoculated with EO771 breast tumour cells and treated with intratumoural IgG or CD47-LLO. Growth curves for EO771 breast tumours (j) and comparison of tumour volumes at day 12 (k) after intratumoural injection are shown (j). n=7 for CD47-LLO-treated and n=6 for IgG-treated WT mice; n=5 for CD47-LLO-treated and n=4 for IgG-treated STING^−/− mice. Data shown represent mean ± s.e.m. (b, j) or mean ± s.d. (f, g, h, i, k) analyzed by two-sided unpaired Student’s t tests (b), one-way analysis of variance with Tukey’s multiple comparisons test (f, g, h, i, k), two-way analysis of variance with Tukey’s multiple comparisons test (j), or unpaired two-tailed Mann–Whitney U test of the median value of pSTING intensity within F4/80⁺ cells (c).

Fig. 4 ∣. CD47-LLO drives tumour antigen-driven T cell responses in vivo.

a,b WT female mice were inoculated with EO771 breast tumours and treated with intratumoural anti-CD47, CD47-LLO, IgG-LLO, or a non-cleavable CD47-LLO (nc) conjugate (Supplementary Fig. 3a). Tumour volumes were monitored and analyzed for the indicated periods (a) and quantified at day 12 after tumour inoculation (b). n=6 for IgG-LLO, n=6 CD47-LLO-nc, and n=7 for all remaining groups. c, D4M.3A tumour volumes (Supplementary Fig. 3b). n = 6 for all groups. d, Survival curves for each treatment group. n = 6 for CD47-LLO, and n = 7 for anti-CD47. e, Flow cytometry analysis and f, quantification of CD86⁺ and CD206⁺ tumour-associated macrophages in 4T1Br4 tumours. n = 4 biologically independent experiments. g, Quantification of CD4⁺ and h, CD8⁺ cells in EO771 tumours. n = 4 biologically independent tumours per condition. i, Schematic illustrating the mechanism of SIINFEKL–H2Kb tetramer⁺CD8⁺ T-cell expansion. j, Flow cytometry analysis and k, quantification of SIINFEKL–H2Kb tetramer⁺CD8⁺ T cells (n=4 biologically independent tumours per condition). l, Flow cytometry analysis and m, quantification of SIINFEKL–H2Kb tetramer⁺CD8⁺ T cells within the spleen (n=4 biological replicates). n, Schematic illustrating bilateral breast tumour model. o, Treated and untreated tumour growth curves. n=6 for all groups. p, Flow cytometry analysis and q, quantification of SIINFEKL–H2Kb tetramer⁺CD8⁺ T cells within the untreated tumour (n=4 biological replicates). Data shown represent mean ± s.e.m. (a, c, o) or mean ± s.d. (b, f, g, h, k, m, q) analyzed by two-sided unpaired Student’s t test (d, f, m, q), one-way analysis of variance with Tukey’s multiple comparisons test (b, g, h, k), two-way analysis of variance with Tukey’s multiple comparisons test (a, c, o), or two-sided log-rank (Mantel–Cox) test (d).

Fig. 5 ∣. Transcriptome analysis reveals CD47-LLO TAM proinflammatory signatures.

a-k, EO771 tumours were harvested on day 3 after intratumoural treatment with IgG (n = 3), anti-CD47 (n = 4), or CD47-LLO (n = 4), and subjected to fluorescence-activated cell sorting on CD45⁺ cells. Cells were fixed, and 100,000 CD45⁺ cells/tumour were pooled into the respective treatment groups and subjected to scRNA-seq. a, Flow cytometry analysis and b, quantification of CD45⁺ cells isolated from EO771 tumours treated with intratumoural IgG, anti-CD47, or CD47-LLO. n = 3 biological replicates. c, Uniform Manifold Approximation and Projection (UMAP) mapping of 13,891 single cells from 3 treatment groups show the composition of the different cell types. The UMAP projections are shown by assignment and d, by sample. e, UMAPs of only macrophages color-coded by cluster. f, Top 20 differentially expressed genes in 10 macrophage clusters ranked by FDR. Gene expression values were centered, scaled, and transformed to a −2 to 2 scale. g, UMAP projection of macrophage populations showing the distribution of transcriptional signatures of pre-defined macrophage and monocyte subsets. Inset shows a dot plot depicting the average expression and percent expression of signature genes per macrophage cluster. h, Heatmap depicting expression of cGAS-STING pathway related-genes in macrophage clusters by treatment group and cluster. i, Heatmap depicting expression of phagocytosis related-genes in macrophage clusters by treatment group and cluster. Data shown represent mean ± s.d (b) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test.

Fig. 6 ∣. CD47-LLO facilitates innate and adaptive cell clustering and signalling in vivo.

a, Uniform Manifold Approximation and Projection (UMAP) of macrophage populations isolated from EO771 tumours after intratumoural treatment with IgG (n = 3), anti-CD47 (n = 4), or CD47-LLO (n = 4). b, Fraction of cells (y-axis) from each cluster (x-axis) color-coded by sample. c, Heatmap displaying select upregulated and downregulated pathways from KEGG gene set enrichment analysis. d, CD11c⁺ cells (arrowheads), CD4⁺ cells (arrows), and CD8⁺ cells (asterisks) in 4T1Br4 tumours. Scale bar, 10 μm. e-g, Percentage of CD11c⁺ cells (e), CD4+ cells (f), and CD8+ cells (g) per field of view from n=3 biologically independent tumours per condition. h, Quantification of percentages of cells that colocalized in triad clusters from n=3 independent experiments. i, UMAP projection of lymphocyte populations showing the distribution of pre-defined T cell subsets from the three treatment groups. j, Cell-to-cell communication networks inferred with CellChat software. Strength of cell-to-cell interactions is represented in the edge width. Data shown represent mean ± s.d. (e, f, g, h) analyzed by two-sided unpaired Student’s t test (e, f, g, h).

Fig. 7 ∣. CD47-LLO requires macrophages and CD8+ T cells for tumour cell elimination.

a, Schematic for generating F4/80- or CD8-depleted mice and subsequent treatment with intratumoural CD47-LLO or anti-CD47. b, Efficiency of F4/80+ cell depletion was evaluated ex vivo in mouse tumors after treatment with anti-CD47 (n=3), CD47-LLO (n=4), anti-CD47 and anti-CSF-IR antibody (n=4) or CD47-LLO and anti-CSF-IR antibody (n=6) or. c) Tumour volumes were monitored and analyzed for the indicated periods (d) and quantified at 10 days after treatment initiation. For anti-CSF1R-treated animals, n=6 for anti-CD47 and n=8 for CD47-LLO. For IgG-treated animals, n=4 for anti-CD47 and n=7 for CD47-LLO. e, Efficiency of CD8 T-cell depletion was evaluated ex vivo in mouse spleens after treatment with CD47-LLO with and without anti-CD8 antibody. Data show n=3 C57BL6 mice per treatment group. f, Tumour volumes were monitored and analyzed for the indicated periods and quantified at 10 days after treatment initiation (g). n = 5 for all treatment groups. Data shown represent mean ± s.e.m. (c, f) or mean ± s.d. (b, d, e, g) analyzed by one-way analysis of variance with Tukey’s multiple comparisons test (b, d, e, g), two-way analysis of variance with Tukey’s multiple comparisons test (c, f).

Fig. 8 ∣. CD47-LLO drives systemic antitumour immunity to inhibit breast cancer metastasis.

a, Representative IVIS spectrum images and b, quantified signal intensity of Luc-4T1Br4 breast tumours before tumour debulking (i.e., on day 12 after tumour inoculation) and after tumour debulking (i.e., days 16–36). n=7 for CD47-LLO and n=7 for anti-CD47. c, Survival curves for each treatment group. d, Arrowheads show CD8⁺ cells colabeled with anti-PD1. Scale bar, 20 μm. e, Quantification of CD8⁺ cells that co-labeled with anti-PD1 (n=3 biologically independent experiments). f, PD-L1 protein expression levels in tumour frozen tissue sections. Scale bar, 20 μm. g, Quantification of DAPI⁺ cells that co-labeled with anti-PD-L1 (n=3 biologically independent experiments). h, Flow cytometry analysis and i, quantification of CD8⁺ T cells isolated from 4T1Br4 tumours treated with intraperitoneal CD47-LLO (n=5), anti-CD47 (n=5), anti-CD47 plus anti-PD1 (n=6), or CD47-LLO + anti-PD1 (n=8). j, Representative IVIS spectrum images and k, quantified signal intensity of Luc-4T1Br4 breast tumours. n=4 for intraperitoneal CD47-LLO, n=7 for anti-CD47. For mice treated with anti-PD1, n=6 for anti-PD1 alone, n=8 for anti-CD47 and n=6 for CD47-LLO. l, Survival curves for each treatment group. m, Tumour volumes were monitored for the indicated periods. n=4 tumour-naïve animals, n=3 intratumoural CD47-LLO animals, n=4 intraperitoneal CD47-LLO animals. n, Schematic illustrating the establishment of splenic CD8⁺ T cells tumour co-cultures. o, Interferon (IFN)-gamma ELISPOT assay, scale bar 1mm, and p, quantification. n=3 biologically independent samples. Data shown represent mean ± s.e.m. (b, e, g, k, m) or mean ± s.d. (i, p) analyzed by two-sided unpaired Student’s t test (e, g, p), Welch analysis of variance with Dunnet’s multiple comparisons test (i), two-way analysis of variance with Benjamini, Krieger, and Yekutieli multiple comparisons test (b, k), or two-sided log-rank (Mantel–Cox) test (c, l).