The role of the HLA class I-restricted, CD8+, herpes simplex virus (HSV)-specific cytotoxic T lymphocytes (CTL) in the control of human HSV infections is controversial because previous reports suggest that a substantial portion of the antigen-specific lytic response is mediated by CD4+ cells. To address this question directly, we isolated HSV-specific CD8+ CTL clones from a patient with recurrent genital herpes. These CTL were cloned by coculturing responder peripheral blood mononuclear cells (PBMC) with phytohemagglutinin-stimulated PBMC that had been infected with live HSV-2 and then irradiated prior to the addition of responder cells. After 1 week, CTL were cloned by limiting dilution using phytohemagglutinin stimulation and allogeneic feeder PBMC. Seven clones were isolated; all seven clones were CD8+ CD4- CD3+ DRbright, six lysed only HSV-2-infected targets, and one lysed both HSV-1- and HSV-2-infected targets. Antigen presentation was restricted by two to three different HLA class I loci. To determine the antigens recognized by these HSV-specific CTL, target cells were infected with HSV in the presence of acyclovir, 5,6-dichloro-1-beta-D-ribofuranosylbenzimidazole, or cycloheximide in a series of drug block/release protocols to limit the repertoire of viral gene expression to select transcriptional classes. Five of the clones exhibited a different pattern of cytotoxicity, suggesting that each recognized a distinct HSV antigen. One of the clones appears to be directed against an immediate-early antigen; six of the clones recognize virion proteins. Five of these clones recognized internal virion proteins that could be introduced into target cells by HSV infection in the absence of virus gene expression. Antigen specificity was further tested by using vaccinia virus vectors that express glycoproteins gD2 and gB2 or the tegument protein VP16. One clone lysed vaccinia virus/gD2-infected target cells; the remaining clones did not recognize any of these gene products. The diversity of the CD8+ response from a single individual indicated that several different antigens are recognized when presented in the context of a variety of class I HLA alleles, a pattern that markedly differs from that described for another human herpesvirus, cytomegalovirus.