Résumé : Bovine leukemia virus (BLV) Tax protein, a transcriptional activator of viral expression, is essential for viral replication in vivo. Tax is believed to be involved in leukemogenesis because of its second function, immortalization of primary cells in vitro. These activities of Tax can be dissociated on the basis of point mutations within specific regions of the protein. For example, mutation of the phosphorylation sites at serines 106 and 293 abrogates immortalization potential in vitro but maintains transcriptional activity. This type of mutant is thus particularly useful for unraveling the role of Tax immortalization activity during leukemogenesis independently of viral replication. In this report, we describe the biological properties of BLV recombinant proviruses mutated in the Tax phosphorylation sites (BLVTax106+293). Titration of the proviral loads by semiquantitative PCR revealed that the BLV mutants propagated at wild-type levels in vivo. Furthermore, two animals (sheep 480 and 296) infected with BLVTax106+293 developed leukemia or lymphosarcoma after 16 and 36 months, respectively. These periods of time are within the normal range of latencies preceding the onset of pathogenesis induced by wild-type viruses. The phenotype of the mutant-infected cells was characteristic of a B lymphocyte (immunoglobulin M positive) expressing CD11b and CD5 (except at the final stage for the latter marker), a pattern that is typical of wild-type virus-infected target cells. Interestingly, the transformed B lymphocytes from sheep 480 also coexpressed the CD8 marker, a phenotype rarely observed in tumor biopsies from chronic lymphocytic leukemia patients. Finally, direct sequencing of the tax gene demonstrated that the leukemic cells did not harbor revertant proviruses. We conclude that viruses expressing a Tax mutant unable to transform primary cells in culture are still pathogenic in the sheep animal model. Our data thus provide a clear example of the discordant conclusions that can be drawn from in vitro immortalization assays and in vivo experiments. These observations could be of interest for other systems, such as the related human T-cell leukemia virus type 1, which currently lack animal models allowing the study of the leukemogenic process.