The infectivity and replication of human (HIV-1), feline (FIV), and murine (LP-BM5) immunodeficiency viruses are all inhibited by several nucleoside analogues after intracellular conversion to their triphosphorylated derivatives. At the cellular level, the main problems in the use of these drugs concern their limited phosphorylation in some cells (e.g., macrophages) and the cytotoxic side effects of nucleoside analogue triphosphates. To overcome these limitations a new nucleoside analogue homodinucleotide, di(thymidine-3'-azido-2',3'-dideoxy-D-riboside)-5'-5'-p(1)-p(2)-pyrophosphate phosphate (AZTp(2)AZT), was designed and synthesized. AZTp(2)AZT was a poor in vitro inhibitor of HIV reverse transcriptase, although it showed antiviral and cytotoxic activities comparable to those of the parent AZT when added to cultures of a HTLV-1 transformed cell line. AZTp(2)AZT encapsulated into erythrocytes was remarkably stable. Induction of erythrocyte-membrane protein clusterization and subsequent phagocytosis of AZTp(2)AZT-loaded cells allowed the targeted delivery of this impermeant drug to macrophages where its metabolic activation occurs. The addition of AZTp(2)AZT-loaded erythrocytes to human, feline, and murine macrophages afforded almost complete in vitro protection of these cells from infection by HIVBa-L, FIV, and LP-BM5, respectively. Therefore, AZTp(2)AZT, unlike the membrane-diffusing azidothymidine, acts as a very efficient antiretroviral prodrug following selective targeting to macrophages by means of loaded erythrocytes.