The design of cell-targeted protein therapeutics can be informed by natural protein-protein interactions that use cooperative physical contacts to achieve cell type specificity. Here we applied this approach in vivo to the anemia drug erythropoietin (EPO), to direct its activity to EPO receptors (EPO-Rs) on red blood cell (RBC) precursors and prevent interaction with EPO-Rs on nonerythroid cells, such as platelets. Our engineered EPO molecule was mutated to weaken its affinity for EPO-R, but its avidity for RBC precursors was rescued via tethering to an antibody fragment that specifically binds the human RBC marker glycophorin A (huGYPA). We systematically tested the impact of these engineering steps on in vivo markers of efficacy, side effects, and pharmacokinetics. huGYPA transgenic mice dosed with targeted EPO exhibited elevated RBC levels, with only minimal platelet effects. This in vivo selectivity depended on the weakening EPO mutation, fusion to the RBC-specific antibody, and expression of huGYPA. The terminal plasma half-life of targeted EPO was ∼28.3 h in transgenic mice vs. ∼15.5 h in nontransgenic mice, indicating that huGYPA on mature RBCs acted as a significant drug sink but did not inhibit efficacy. In a therapeutic context, our targeting approach may allow higher restorative doses of EPO without platelet-mediated side effects, and also may improve drug pharmacokinetics. These results demonstrate how rational drug design can improve in vivo specificity, with potential application to diverse protein therapeutics.protein engineering | drug targeting | platelet | scFv