Erythrocyte-mediated mRNA delivery enables in vivo generation of CAR-myeloid cells for cancer immunotherapy

Background: Engineering myeloid cells with chimeric antigen receptors (CARs) holds great therapeutic promise, yet their generation in vivo remains challenging. Despite the success of COVID-19 vaccines, targeted mRNA delivery to immune cells remains inefficient. Conventional lipid nanoparticles (LNPs) primarily target the liver upon systemic administration, limiting their utility for lymphoid organs. Moreover, LNPs suffer from low endosomal escape rate and may induce off-target toxicities. Here, we developed an erythrocyte-mediated mRNA delivery system in which mRNA-loaded LNPs are covalently conjugated to erythrocyte membrane proteins in a “plug-and-play” manner, termed mRNA-LNP-Ery. Leveraging erythrocytes' natural splenic homing, mRNA-LNP-Ery enables selective spleen-specific targeting and efficient mRNA delivery to immune cells for cancer immunotherapy. Method: We characterized the biodistribution of mRNA-LNP-Ery across multiple mRNA cargoes. To elucidate internalization mechanisms, time-course confocal imaging was performed. Anti-HER2 and anti-CD19 CAR mRNAs were delivered via erythrocytes, and the therapeutic efficacy was assessed in corresponding syngeneic tumor models. Result: Exploiting erythrocytes' intrinsic splenic homing capacity, mRNA-LNP-Ery achieved highly selective and efficient mRNA delivery to the spleen while minimizing hepatic uptake. Cellular characterization showed preferential targeting of splenic CD11b⁺ myeloid cells. Mechanistically, mRNA-LNP-Ery was internalized through phagocytosis while escaping lysosomal degradation, resulting in markedly enhanced mRNA translation and protein expression. To evaluate the therapeutic potential of this platform, we delivered mRNAs encoding CARs targeting HER2 or CD19, which generated functional CAR-myeloid cells in vivo , which adopted a pro-inflammatory, antigen-presenting phenotype. These cells migrated to tumors, eliminated cancer cells, and remodeled the tumor microenvironment, leading to increased infiltration of effector T and NK cells. Functionally, mRNA-LNP-Ery induced stronger and more durable tumor regression than conventional mRNA-LNPs across multiple tumor models, including immune-cold tumors, despite using only one-tenth the mRNA dose. The antitumor effect was abolished in splenectomized mice and partially diminished in nude mice, indicating dependence on both CAR-myeloid formation in the spleen and crosstalk with adaptive immunity. Moreover, mRNA-LNP-Ery induced minimal systemic toxicity, underscoring the safety and translational potential of this spleen-targeted delivery approach. Conclusion: Together, our findings establish a clinically translatable erythrocyte-based mRNA platform that integrates with existing LNP technology to enable direct in vivo immune cell programming and advance CAR-immune cell therapies for solid tumors.