We demonstrate that human T lymphocytes proliferate in vitro to highly purified human heat-shock protein 60 (Hu. hsp60). The response to this self Ag was confined to the CD45RA+ RO− T cell subset, with minimal responses by adult CD45RA− RO+ T cells. Experiments using keyhole limpet hemocyanin as a prototypic novel Ag, or tetanus toxoid as a recall Ag, were consistent with the notion that CD45RA+ RO− and CD45RA− RO+ T cell subsets can be designated as naive and memory cells, respectively; thus, responses to Hu. hsp60 were confined to the putative naive subset. In contrast, both CD45RA+ RO− and CD45RA− RO+ T cell populations proliferated to bacterial hsp60 from Mycobacterium leprae, Escherichia coli, or Chlamydia trachomatis. However, only CD45RA− RO+(memory) T cells responded to a mycobacterial hsp60-derived peptide previously defined as a major bacteria-specific epitope. Experiments with cord blood T cells, which are CD45RA+ RO− and can be considered truly naive, showed that the peptide could elicit responses from naive T cells in vitro; cord blood cells also responded to Hu. hsp60. Since bacterial hsp60 Ags contain both conserved and nonconserved epitopes, we speculate that in vivo challenge with bacterial hsp60 will activate T cells capable of seeing either type of epitope, but only those that see nonconserved epitopes maintain the CD45RA− RO+ memory phenotype. However, T cells recognizing conserved epitopes, while not apparently being recruited to the memory pool, may nevertheless play a role in immunoregulation, particularly in the context of inflammation, when expression of Hu. hsp60 is increased.