Human pathogenic retroviruses do not have common loci of integration. However, many factors, such as chromatin structure, transcriptional activity, DNA-protein interaction, CpG methylation, and nucleotide composition of the target sequence, may influence integration site selection. These features have been investigated by in vitro integration reactions or by infection of cell lines with recombinant retroviruses. Less is known about target choice for integration in vivo. The present study was conducted in order to assess the characteristics of cellular sequences targeted for human T-cell leukemia virus type 1 (HTLV-1) integration in vivo. Sequencing integration sites from >/=200 proviruses (19 kb of sequence) isolated from 29 infected individuals revealed that HTLV-1 integration is not random at the level of the nucleotide sequence. The virus was found to integrate in A/T-rich regions with a weak consensus sequence at positions within and without of the hexameric repeat generated during integration. These features were not associated with a preference for integration near active regions or repeat elements of the host chromosomes. Most or all of the regions of the genome appear to be accessible to HTLV-1 integration. As with integration in vitro, integration specificity in vivo seems to be determined by local features rather than by the accessibility of specific regions.