Human immunodeficiency virus (HIV)-specific CD8+-T-cell responses are critical in restricting viral replication and altering the course of HIV infection (38, 48). Ex vivo cytotoxic T lymphocytes (CTL) are much less frequent in vertically infected children than in adults (9, 10, 41, 42, 46, 60). In contrast, after in vitro culture, memory CTL can be readily detected in HIV-infected children, with magnitude, breadth, and specificity similar to those observed in adults (9, 10, 41, 46). These memory CTL can be detected during the first weeks or months of life (10, 41, 56, 75), and the presence of HIV-specific CTL is associated with slower evolution toward disease (10). Complete suppression of viral replication following combined therapy is less frequent in children than in adults (40, 51, 52), and the kinetics of viral decrease is slower (49). However, children have a more active thymus function than adults (19, 44), which allows better lymphocyte regeneration following combined therapy (13, 69). In children treated with combined therapy before 3 months of age, and with undetectable viral load, neither HIV-specific antibodies nor HIV-specific CD4 or CD8 cells were detected (43). This contrasts with the beneficial effect of early treatment on preservation of HIV-specific T-cell immunity in adults (8, 50). New immunotherapeutic interventions are being developed for adults (8, 50). These treatments are of great interest for children due to observance problems and because complete viral suppression is more difficult to obtain in children than in adults. Therefore, it is crucial to characterize the dynamics of HIV-specific T cells in pediatric HIV infection. On encountering viral antigen, naive CD8+ T cells proliferate and differentiate into effector cells able to lyse infected cells and to secrete cytokines. As virus is cleared, most activated effector cells undergo apoptosis, but some survive and enter the memory pool that persists for long periods (76). Most memory cells are in a resting state, unable to secrete cytokine or to lyse infected cells, until reactivation on reexposure to viral antigen (61). During persistent infection, continuous stimulation of T cells may lead to dysfunction, anergy, or clonal exhaustion. In the absence of CD4+ T cells, exhaustion and anergization of CD8+ T cells is more rapid (78). The complex interplay between the antigen load, virus-specific CD8+-T-cell dynamics and function, and the immune status of the infected host has been illustrated by a number of studies of HIV-infected adults. In untreated chronically HIV-infected patients, CTL numbers are inversely correlated with the plasma viral load (5, 24, 25, 34, 54, 58). On the other hand, reduction of HIV replication by potent antiretroviral therapy is associated with a decline in HIV-specific CD8+ T cells (11, 55). Evidence that a fraction of HIV-specific CD8+ T cells are not functional was recently obtained (22, 31). Furthermore, loss of gamma-interferon (IFN-γ)-producing cells with persistence of tetramer binding cells was observed in subjects progressing to AIDS (32). CTL derived from the memory pool after in vitro expansion are easily detected by the standard 51Cr release assay in most HIV-infected children (9). In contrast, ex vivo-activated effector cytolytic cells are infrequently detected immediately after isolation (9). Cytokine synthesis following short-term ex vivo stimulation with the antigen can be measured using assays that are more sensitive than the chromium release assay and could be used to quantify the effector subset of HIV-specific CD8+ T cells (3, 18). Cytokine production can be measured at the single-cell level using the enzyme-linked immunospot (ELISPOT) technique, allowing direct calculation of T-cell frequencies (15). This technique is very sensitive and has been found to reliably detect CD8+ T cells in various human diseases (35, 63, 64, 74), including HIV infection (16, 36). The aim of the present work was to evaluate the use of the ELISPOT assay for the ex vivo study of HIV-specific CD8+ T cells from HIV-infected children. We chose to focus our initial effort on two immunodominant HLA-A∗0201-restricted HIV epitopes that are frequently recognized by infected children, as we showed previously using tetramers (66). In this cross-sectional study, HLA-A∗0201-positive HIV-infected children were systematically tested for the presence of HIV-specific CD8+ T cells using the ELISPOT assay. The frequencies of HIV-specific IFN-γ-producing CD8+ T cells were compared to the frequencies of HIV-specific CD8+ T cells measured by tetramer labeling, and relationships with biological parameters of HIV infection were investigated. The results from the ELISPOT and the tetramer assays were well correlated, but a comparison of the results from both assays suggests that a significant fraction of CD8+ T cells were unable to produce IFN-γ. Most importantly, the frequencies of ex vivo-activated HIV-specific CD8+-T-cell-mediated IFN-γ production were positively correlated with plasma HIV RNA, showing that this subset of antiviral CD8+ T cells is dependent upon continuous antigenic stimulation.