Sialylated CD43 is a glyco-immune checkpoint for macrophage phagocytosis

Abstract

Macrophages in the tumor microenvironment exert potent anti-tumorigenic activity through phagocytosis. Yet therapeutics that enhance macrophage phagocytosis have not improved outcomes in clinical trials for patients with acute myeloid leukemia (AML) or myelodysplastic syndrome (MDS). To systematically identify regulators of phagocytosis, we performed genome-scale CRISPR knockout screens in human leukemia cells co-cultured with human monocyte-derived macrophages. Surprisingly, we found that whereas the classic "don't eat me" signal CD47 inhibited mouse macrophages, it did not inhibit phagocytosis by human macrophages. In contrast, the O-linked glycosylation and sialylation pathways were strong negative regulators of phagocytosis. In AML, the cell surface O-linked glycoprotein CD43 was the major effector of the O-linked glycosylation and sialylation pathways. Genetic deletion or antibody blockade of CD43 enhanced macrophage phagocytosis. This work highlights the importance of using human platforms to identify immune checkpoints, and nominates CD43 as a glyco-immune regulator of human macrophage phagocytosis.

Fig. 1:. Genome-wide CRISPR screens identify key pathways that regulate antibody-dependent cellular phagocytosis and antibody-independent phagocytosis of human leukemia.

a, Schematic of antibody-independent cellular phagocytosis screen design. b, Scatter plot of log fold change of enrichment or depletion of genes in the leukemia fraction of MV411 (x-axis) or MOLM13 (y-axis) co-cultured with macrophages versus leukemia alone. Circle size indicates average -log10 (p-value) across both cell lines. c, Scatter plot of log fold change of enrichment or depletion of genes in the macrophage fraction of MV411 (x-axis) or MOLM13 (y-axis) co-cultured with macrophages versus leukemia alone. Circle size indicates average -log10 (p-value) across both cell lines. d, Schematic of competitive co-culture experiment of human leukemia cells with human or mouse macrophages. e, Flow cytometry plot of leukemia cells with and without CD47 knockout cultured alone or with human macrophages stimulated with LPS. Barplots show quantification of the percent of total cells phagocytosed and log2 normalized ratio of knockout cells versus control. Data show the mean ± s.e.m. of four technical replicates and are representative of two different experiments. f, Flow cytometry plot of human leukemia cells co-cultured with LPS stimulated mouse macrophages. Barplots shown quantify the percent of phagocytosed cells and ratio of knockout to control remaining after macrophage co-culture. Data show the mean ± s.e.m. of four technical replicates and are representative of two different experiments. g, Schematic of co-culture experiment of human leukemia treated with anti-CD47 (MIAP410) antibody prior to co-culture with human macrophages pre-treated with or without Fc receptor blocking antibodies. Barplot shows the percent of phagocytosed cells in the presence or absence of macrophage Fc receptor blockade. Data show the mean ± s.e.m. of four technical replicates and are representative of two different experiments. h, Schematic of antibody-dependent cellular phagocytosis screen design. i, Scatter plot of log fold change of enrichment or depletion of genes in the leukemia fraction of anti-CD47 treated MV411 (x-axis) or MOLM13 (y-axis) co-cultured with macrophages versus leukemia alone. Circle size indicates average log10 (p-value) across both cell lines. j, Competitive co-culture experiment of control or knockout leukemia cells co-cultured with human macrophages. Log2 normalized ratio of the knockout to control and percent of phagocytosed leukemia are plotted. Data in e-g, j were analyzed by unpaired, two-sided Student’s t-test, * p<0.05, ** p<0.01, *** p<0.001.

Fig. 2:. Integrated analysis of genome-scale antibody-dependent and antibody-independent cellular phagocytosis screens reveals O-glycan pathway genes as key regulators of macrophage phagocytosis.

a, Genes ranked by the average of the difference between the log fold change of enrichment or depletion of the leukemia fraction and macrophage fraction. Circle size is scaled -log10 (p-value). Top 15 enriched genes (blue) and top 5 depleted genes (red) are listed and arranged by statistical significance. b, Schematic of competitive co-culture experiment of PTPN6 knockout leukemia. Percent of phagocytosed cells and log2 ratio of knockout to control are shown. Data represent mean ± s.e.m. of four technical replicates and are representative of two independent experiments. c, Scatter plot of the difference between the log fold change of enrichment or depletion in the leukemia fraction and macrophage fraction in the ADCP screen (x-axis) and AICP screen (y-axis). Circle size is the average -log10(p-value). Colors denote pathway annotation. d, Schematic of the O-glycosylation pathway highlighting top gene knockouts that enhance AICP and ADCP (blue text). Data in b were analyzed by unpaired, two-sided Student’s t-test, ** p< 0.01

Fig. 3:. The O-linked glycosylation pathway inhibits human macrophage phagocytosis of leukemia cells through terminal sialic acid residues

a, Schematic of competitive phagocytosis assay, representative flow cytometry plots, and bar graphs of relative phagocytosis of C1GALT1-deficient, C1GALT1C1-deficient, or control versus control leukemia cells (MV411 and MOLM13). b, Schematic of competitive phagocytosis assay, representative flow cytometry plots, and bar graphs of relative phagocytosis of SLC39A9-deficient, SLC35A2-deficient, or control versus control sgRNA leukemia cells (MV411 and MOLM13) c. Antibody-dependent cellular phagocytosis (ADCP) strategy: control or C1GALT1-deficient MV411 cells were engineered to ectopically express mouse CD8𝛂 d. In vivo ADCP model: control or C1GALT1-deficient were intravenously engrafted into sublethally irradiated NOD SCID IL2rg^−/− mice. Mice bearing C1GALT1-deficient or control MV411 cells were intraperitoneally injected with systemic anti-CD8𝛂 antibodies starting on day +5. e. Survival of mice challenged with C1GALT1-deficient or control MV411 leukemia cells. f, Schematic of competitive phagocytosis assay experimental design and bar graphs of relative phagocytosis of C1GALT1, C1GALT1C1, or control sgRNA versus control sgRNA leukemia cells pre-treated with varying doses of V. cholerae sialidase prior to phagocytosis assays. Both knockout and control cells were pre-treated with sialidase prior to co-culture. g, Schematic of competitive phagocytosis assay and bar graphs of relative phagocytosis of C1GALT1-deficient, C1GALT1C1-deficient, or control MV411 leukemia cells versus control cells pre-treated with V. cholerae sialidase prior to phagocytosis assays. Only control cells (but not knockout cells) were pre-treated with sialidase prior to co-culture. For a-b and f-g, genetically modified or sialidase-treated leukemia cells were co-cultured with or without IFNγ-stimulated macrophages for 18–24 hours. Bar graphs indicate log (fold change) of the ratio of knockout cells relative to control after co-culture with macrophages. Data represent mean ± s.d. of four technical replicates and are representative of 3–5 independent experiments For e, data represent three independent survival experiments.

Fig. 4:. Sialylated CD43 is the major downstream effector of the O-linked glycosylation pathway and is sufficient to inhibit macrophage phagocytosis of human leukemia

a, Scatter plot of functional impact in CRISPR AICP screens (x-axis) vs. relative gene expression (y-axis) of all known human cell surface proteins. Size of each dot is scaled to reflect the % of potential O-linked glycosylation sites. b, Schematic of competitive phagocytosis assay, representative flow cytometry plots, and bar graphs of relative phagocytosis of SPN (CD43)-deficient or control sgRNA versus control leukemia cells (MV411 and MOLM13) in coculture assays with IFNγ-stimulated macrophages. c. Representative flow cytometry plots and bar graphs of relative phagocytosis of SPN (CD43)-deficient or control leukemias overexpressing CD8𝛂 versus control MV411 leukemia cells overexpressing CD8𝛂 in coculture assays with varying doses of anti-CD8𝛂 opsonizing antibodies. d. Representative flow cytometry plots and bar graphs of relative phagocytosis of control leukemias overexpressing luciferase, C1GALT1, or CD43 versus control MV411 leukemia cells overexpressing luciferase in coculture assays with IFNγ-stimulated macrophages. e. Survival of mice challenged with SPN (CD43)-deficient or control MV411 leukemia cells overexpressing CD8𝛂. Mice were intraperitoneally injected with systemic anti-CD8𝛂 antibodies starting on day +5. f. Representative flow plots for surface expression of sialylated CD43 in control, C1GALT1-deficient, C1GALT1C1-deficient, or CD43-deficient MV411 leukemias. g. Representative western blots for sialylated CD43 levels in MV411 leukemia cells treated without or with V. cholerae sialidase. h. Representative western blots for total CD43 levels in MV411 leukemia cells treated without or with V. cholerae sialidase i. Representative flow cytometry plots and bar graphs of relative phagocytosis of C1GALT1-deficient leukemias overexpressing luciferase, C1GALT1, or CD43 versus control MV411 leukemia cells overexpressing luciferase in coculture assays with IFNγ-stimulated macrophages. j. Relative phagocytosis of CD43-deficient or dual C1GALT1- and CD43-deficient leukemias after co-culture with IFNγ-stimulated macrophages. b,c,d,h Bar graphs indicate log (fold change) of the ratio of knockout cells relative to control after co-culture with IFNγ-stimulated macrophages. Data represent mean ± s.d. of four technical replicates and are representative of 3–5 independent experiments.

Fig. 5:. anti-CD43 antibodies enhance human macrophage phagocytosis of leukemia cell lines and primary patient samples

a-b, Uniform manifold approximation and projection (UMAP) of single cell RNA-sequencing (scRNA-seq) profiles of bone marrow aspirate cells taken from healthy patients or AML patients. Expression of SPN/CD43 is shown on the far right panel. c, Dot plot summary of z-score scaled expression of CD43 across normal and malignant cell subtypes. Size of each dot is scaled to reflect the absolute % of cells in which SPN transcripts were detected. d, Representative western blots for total CD43 expression in human leukemia cell lines (MV411, MOLM13), primary AML blasts (AML_01 - AML07), and normal immune cell subsets (bulk PBMCs, monocytes). e, Flow cytometry plots for sialylated CD43 expression on fresh blasts from AML patients (AML01 - AML06) f, Schematic of phagocytosis experiments with anti-CD43 antibodies and bar graphs of absolute phagocytosis of human leukemia cell lines (MV411, MOLM13, HEL) after addition of various doses of anti-CD43 antibodies to macrophage co-culture assays g, Schematic of phagocytosis assays of patient-derived AML blasts and bar graphs of absolute phagocytosis of primary AML cells after addition of varying doses of anti-CD43 antibodies to macrophage co-culture assays h, Schematic of phagocytosis assays with FcR blockade prior to addition of either anti-CD47 or anti-CD43 antibodies and bar graphs of absolute phagocytosis of MV411 leukemias with or without Fc blockade prior to addition of anti-CD47 or anti-CD43 antibodies. i, Schematic of phagocytosis assays with AML patient-derived macrophages and bar graphs of relative phagocytosis of control or CD43-deficient leukemias versus control MV411 leukemias j, Schematic of phagocytosis assays with AML patient-derived macrophages and bar plots of absolute phagocytosis of MV411 leukemias with or without anti-CD43 antibodies. f,g,h,i,j Data represent mean ± s.d. of four technical replicates and are representative of 3–5 independent experiments.