Infection with EBV provides a unique opportunity to follow the human CD8(+) T cell response to a persistent, genetically stable agent from the primary phase, as seen in infectious mononucleosis (IM) patients, into long-term memory. This study focuses on the response to an immunodominant HLA-A2.01-restricted epitope, GLCTLVAML, from the EBV-lytic cycle Ag BMLF1. TCR analysis of the highly amplified primary response to this epitope revealed markedly oligoclonal receptor usage among in vitro-derived clones, with similar clonotypes dominant in all three IM patients studied. Direct staining of IM T cell preparations with the A2.01/GLCTLVAML tetramer linked this oligoclonal epitope-specific response with appropriate Vbeta subset expansions in the patients' blood. These patients were studied again >2 years later, at which time TCR analysis of in vitro-reactivated clones suggested that rare clonotypes within the primary response had now come to dominate memory. Five additional A2. 01-positive IM patients were studied prospectively for Vbeta subset representation within primary and memory epitope-specific populations as identified by tetramer staining. In each case, the primary response contained large Vbeta2, Vbeta16, or Vbeta22 components, and in three of five cases the originally dominant Vbeta was represented very poorly, if at all, in memory. We conclude 1) that an EBV epitope-specific primary response large enough to account for up to 10% CD8(+) T cells in IM blood may nevertheless be dominated by just a few highly expanded clonotypes, and 2) that with persistent viral challenge such dominant T cell clonotypes may be lost and replaced by others in memory.