Memory lymphocytes are important mediators of the immune response. These cells are long-lived and undergo clonal expansion upon reexposure to specific antigen, differentiating into effector cells that secrete Ig or cytokines while maintaining a residual pool of memory T and B lymphocytes. Here, the ability of antigen-specific lymphocytes to undergo repeated cycles of antigen-driven clonal expansion and contraction is exploited in a therapeutic protocol aimed at regulating protein delivery. The principle of this strategy is to introduce genes encoding proteins of therapeutic interest into a small number of antigen-specific B lymphocytes. Output of therapeutic protein can then be regulated in vivo by manipulating the size of the responder population by antigen challenge. To evaluate whether such an approach is feasible, we developed a mouse model system in which Emu- and Iglambda-based vectors were used to express human erythropoietin (hEPO) gene in B lymphocytes. These mice were then immunized with the model antigen phycoerythrin (PE), and immune splenocytes (or purified PE-specific B lymphocytes) were adoptively transferred to normal or mutant (EPO-deficient) hosts. High levels of hEPO were detected in the serum of adoptively transferred normal mice after PE administration, and this responsiveness was maintained for several months. Similarly, in EPO-deficient anemic recipients, antigen-driven hEPO expression was shown to restore hematocrit levels to normal. These results show that antigen-mediated regulation of memory lymphocytes can be used as a strategy for delivering therapeutic proteins in vivo.