Human retroviruses are derived from simian ones through cross-species transmission. These retroviruses are associated with little pathogenicity in their natural hosts, but in humans, HIV causes AIDS, and human T-cell leukemia virus type 1 (HTLV-1) induces adult T-cell leukemia–lymphoma (ATL). We analyzed the proviral sequences of HTLV-1, HTLV-2, and simian T-cell leukemia virus type 1 (STLV-1) from Japanese macaques ( Macaca fuscata ) and found that APOBEC3G (A3G) frequently generates G-to-A mutations in the HTLV-1 provirus, whereas such mutations are rare in the HTLV-2 and STLV-1 proviruses. Therefore, we investigated the mechanism of how HTLV-2 is resistant to human A3G (hA3G). HTLV-1, HTLV-2, and STLV-1 encode the so-called antisense proteins, HTLV-1 bZIP factor (HBZ), Antisense protein of HTLV-2 (APH-2), and STLV-1 bZIP factor (SBZ), respectively. APH-2 efficiently inhibits the deaminase activity of both hA3G and simian A3G (sA3G). HBZ and SBZ strongly suppress sA3G activity but only weakly inhibit hA3G, suggesting that HTLV-1 is incompletely adapted to humans. Unexpectedly, hA3G augments the activation of the transforming growth factor (TGF)-β/Smad pathway by HBZ, and this activation is associated with ATL cell proliferation by up-regulating BATF3/IRF4 and MYC . In contrast, the combination of APH-2 and hA3G, or the combination of SBZ and sA3G, does not enhance the TGF-β/Smad pathway. Thus, HTLV-1 is vulnerable to hA3G but utilizes it to promote the proliferation of infected cells via the activation of the TGF-β/Smad pathway. Antisense factors in each virus, differently adapted to control host cellular functions through A3G, seem to dictate the pathogenesis.