Following the identification of the C-C chemokines RANTES, MIP-1alpha and MIP-1beta as major human immunodeficiency virus (HIV)-suppressive factors produced by CD8+ T cells, several chemokine receptors were found to serve as membrane co-receptors for primate immunodeficiency lentiretroviruses. The two most widely used co-receptors thus far recognized, CCR5 and CXCR4, are expressed by both activated T lymphocytes and mononuclear phagocytes. CCR5, a specific RANTES, MIP-1alpha and MIP-1 receptor, is used preferentially by non-MT2-tropic HIV-1 and HIV-2 strains and by simian immunodeficiency virus (SIV), whereas CXCR4, a receptor for the C-X-C chemokine SDF-1, is used by MT2-tropic HIV-1 and HIV-2, but not by SIV. Other receptors with a more restricted cellular distribution, such as CCR2b, CCR3 and STRL33, can also function as co-receptors for selected viral isolates. The third variable region (V3) of the gp120 envelope glycoprotein of HIV-1 has been fingered as a critical determinant of the co-receptor choice. Here, we document a consistent pattern of evolution of viral co-receptor usage and sensitivity to chemokine-mediated suppression in a longitudinal follow-up of children with progressive HIV-1 infection. Viral isolates obtained during the asymptomatic stages generally used only CCR5 as a co-receptor and were inhibited by RANTES, MIP-1alpha and MIP-1beta, but not by SDF-1. By contrast, the majority of the isolates derived after the progression of the disease were resistant to C-C chemokines, having acquired the ability to use CXCR4 and, in some cases, CCR3, while gradually losing CCR5 usage. Surprisingly, most of these isolates were also insensitive to SDF-1, even when used in combination with RANTES. An early acquisition of CXCR4 usage predicted a poor prognosis. In children who progressed to AIDS without a shift to CXCR4 usage, all the sequential isolates were CCR5-dependent but showed a reduced sensitivity to C-C chemokines. Discrete changes in the V3 domain of gp120 were associated with the loss of sensitivity to C-C chemokines and the shift in co-receptor usage. These results suggest an adaptive evolution of HIV-1 in vivo, leading to escape from the control of the antiviral C-C chemokines.