ytotoxic T lymphocytes (CTLs) are part of the major host defence against trespassing intracellular pathogens. It has been shown in vitro that human immunodeficiency virus type 1 (HIV-1)specific CTLs can eliminate virus-infected cells via MHC class I-restricted killing 1,2 and can interfere with HIV-1 replication via secretion of various antiviral cytokines (reviewed in Ref. 3), thus it is widely held that antiviral CTL responses to HIV-1 are 'salutary' to the infected host. However, after a decade of intense research on HIV-1-specific CTLs (Refs 1, 2) their precise role during the clinical course of HIV-1 infection is still not fully resolved. The majority of studies to date seem to support the concept that HIV-1-specific CTLs contribute to controlling viral replication and thus to delaying the onset of disease. However, several observations have suggested that HIV-1-specific CTLs are 'pathogenic' to the patient. HIV-1-specific CTLs have been detected in bronchoalveolar lavage fluids of some HIV-1-infected patients suffering from lymphocytic alveolitis or interstitial pneumonitis 2 , and in cerebrospinal fluid of some infected patients suffering from neurological disorders 4 . Zinkernagel has even suggested that severe depletion of CD4 ϩ T cells and progression to AIDS result from the killing of huge numbers of infected cells by HIV-1-specific CTLs (Refs 5, 6). If this were true, it could greatly affect the current rationale of developing AIDS vaccines and treatment of HIV-1-infected patients.
In order to start composing a coherent model for the role of HIV-1-specific CTLs in the pathogenesis of AIDS, we need to (1) fully characterize HIV-1-specific CTL responses during the different clinical stages of disease, with respect to dynamics and epitope specificity; (2) understand why the virus persists in the face of vigorous and ongoing HIV-1-specific CTL responses; and (3) reveal why antiviral CTLs fail to provide lifelong protection from progression to AIDS in the vast majority of patients.
HIV-1-specific CTL responses during the natural history of AIDS CTLs during primary HIV-1 infection
Upon sexual or parenteral transmission of HIV-1 it usually takes ~2-4 weeks before clinical symptoms of acute infection ensue. In about 50-70% of infected individuals an acute retroviral syndrome develops with mild influenza-like manifestations from which most patients usually fully recover. To date, HIV-1-specific CTL responses have only been documented in a limited number of patients with an acute retroviral syndrome [7][8][9][10][11][12] . HIV-1-specific CTLs have been observed as early as a few days following the onset of acute symptoms and in general before (neutralizing) antibody responses could be detected (see Fig. 1). The appearance of HIV-1-specific CTLs usually parallels a striking diminution of the viremia in infected patients. Likewise, elegant studies in rhesus monkeys have shown that upon deliberate infection with simian immunodeficiency virus (SIV), CTLs appear as early as 4-7 days post virus inoculation, and coincide with viral clearance from blood and lymph nodes 13,14 . Collectively these studies indicate that CTLs are recruited very early during the encounter with HIV-1, and that in part they may be responsible for the initial control of HIV-1 replication. Interestingly, some individuals were observed who remained HIV-1 seronegative despite frequent exposure to the virus, and seem to harbour HIV-1-specific CTL responses 15 . These CTL responses may be coincidental, demonstrating that some degree of viral replication has occurred, but could also suggest that the virus has been 'repelled' because of efficient HIV-1specific CTL responses. Furthermore, in a group of women from Gambia who seemed to have escaped HIV-1 infection despite several years of high-risk sexual behaviour, three out of six patients had CTLs that recognized HIV-1 and HIV-2 crossreactive epitopes. This suggests that cellular immunity to HIV-2 protects against infection with HIV-1 (Ref. 16). Similarly, recent animal studies demonstrated that a live attenuated HIV-2 vaccine could protect cynomolgus macaques against development of AIDS for more than five years after infection with a pathogenic strain of SIV (Ref. 17). These results are quite encouraging, but the exact features for control and clearance of HIV-1 infection remain to be elucidated further. A critical note has been put forward by Phillips 18 who produced a mathematical model of termination of the primary viremia in the absence of HIV-1-specific immune responses. His theory seems to be supported by the lack of detectable HIV-1-specific CTL responses reported in some individuals. Clearly more systematic studies of patients with acute HIV-1 infection are needed with respect to other effector