CD43 functions as a CD47 independent “don't eat me” signal and novel immune checkpoint in AML Free

CD47 has gained significant attention as a critical anti-phagocytic molecule and therapeutic target for blood cancer immunotherapy. However, the CD47-targeting monoclonal antibody magrolimab recently failed to meet its primary endpoint in Phase III ENHANCE trial in myeloid malignancies. This outcome highlights an urgent need to identify novel regulatory mechanisms for CD47 or to explore entirely new “Don't eat me signal” to overcome existing therapeutic limitations and optimize macrophage-mediated tumor clearance.

To search for potential regulators of CD47, we performed genome-wide CRISPR knock-out (KO) screens in AML cell lines (MOLM13 and U937). We identified 53 positive and 25 negative hits based on P_neg/pos<0.05 that consistently influenced CD47 levels in both cell lines. Notably, these genes were predominantly enriched in O-glycosylation, fucosylation and N-glycosylation pathways. To further investigate the role of O-glycosylation and fucosylation in regulating CD47, we carried out CRISPR-mediated genetic deletion of 13 genes. Deletion of Core 1 O-glycosylation-modifying genes C1GALT1 and C1GALT1C1, or the heavily O-glycosylated protein CD43, increased cell surface CD47, and binding to SIRPa-Fc recombinant protein, establishing these targets as negatively regulators of CD47. In contrast, deletion of SLC35C1, regulator of fucosylation, decreased cell surface CD47 and reduced binding to SIRPa-Fc recombinant protein, implying as a positive regulator of CD47. To further validate, we utilized O-glycosylation inhibitors that resulted in upregulation of CD47 levels while the SLC35C1 inhibitor downregulated CD47 in a dosage and time-dependent manner.

Glycan analysis revealed that CD47 lacks O-glycosylation sites, hence we hypothesized that O-glycosylation indirectly regulates CD47 expression via CD43. Supportive to this, despite higher CD47 expression, disruption of O-glycosylation (via C1GALT1/C1GAT1C1 KO) or CD43 KO increased macrophage phagocytosis (~3-fold on average, P_adj < 0.0001). Further investigation revealed that the CD43-blocking antibody MEM59 and O-glycosylation inhibitors enhanced phagocytosis in AML (MOLM13 and MV4-11), and lymphoma cells (Raji) as well as AML primary tumors. These findings strongly highlight highly O-glycosylated CD43 as a NOVEL “don't eat me” immune checkpoint in hematological tumors.

We next investigated the underlying mechanism of CD43-dependent “don't eat me” signaling. Classic activation of ADAM10/17 using PMA induced CD43 shedding in MOLM13 cells, with more pronounced effects in O-glycosylation-deficient cells. ADAM10 inhibition, but not ADAM17 inhibition, effectively prevented CD43 shedding and restored CD43 levels in O-glycosylation-deficient cells. This suggested that CD43 in AML is susceptible to ADAM10-mediated proteolytic cleavage, particularly under O-glycosylation-deficient conditions. We engineered SIGLEC1/7 knockout THP1 cells to test these lectins as potential CD43 receptors on macrophages. SIGLEC1/7 deficiency failed to abolish CD43-mediated anti-phagocytic activity, indicating that CD43 signals through alternative macrophage receptors. Negatively charged sialic acids have been proposed to prevent phagocytosis through electrostatic repulsion. However, CD43 deletion or blockade had minimal impact on AML cell zeta potential, ruling out electrostatic repulsion as the primary mechanism.

We finally evaluated the validity of CD43 as a putative immune checkpoint for AML patients. Transcriptomic data from BeatAML and TCGA cohorts revealed significantly elevated CD43 expression on leukemic blasts compared to normal counterpart. CD43 overexpression positively correlated with immunosuppressive M2 macrophage signatures while negatively correlating with anti-tumor immune cell signatures. To isolate tumor-intrinsic CD43 effects from immune cell contribution, we performed Kaplan-Meier survival analysis on BeatAML patients with high tumor purity (>60%), demonstrating that elevated CD43 expression was associated with significantly worse overall survival. These findings for the first time suggest that CD43 on AML blasts promotes an immunosuppressive tumor microenvironment and contributes to poor patient outcomes.

Collectively, our work identifies O-glycosylated CD43 as a novel “don't eat me” immune checkpoint that operates independently of CD47, providing a promising therapeutic target to enhance macrophage-mediated tumor clearance in hematologic malignancies.