HTLV-1-Associated Myelopathy (HAM/TSP) is a progressive neuroinflammatory disorder for which no disease-modifying treatment exists. Modest clinical benefit from type I interferons (IFN-α/β) in HAM/TSP contrasts with its recently identified IFN-inducible gene signature. In addition, IFN-α treatment in vivo decreases proviral load and immune activation in HAM/TSP, whereas IFN-β therapy decreases tax mRNA and lymphoproliferation. We hypothesize this "IFN paradox" in HAM/TSP might be explained by both cell type- and gene-specific effects of type I IFN in HTLV-1-associated pathogenesis. Therefore, we analyzed ex vivo transcriptomes of CD4+ T cells, PBMCs and whole blood in healthy controls, HTLV-1-infected individuals, and HAM/TSP patients. First, we used a targeted approach, simultaneously quantifying HTLV-1 mRNA (HBZ, Tax), proviral load and 42 host genes with known antiretroviral (anti-HIV) activity in purified CD4+ T cells. This revealed two major clusters ("antiviral/protective" vs. "proviral/deleterious"), as evidenced by significant negative (TRIM5/TRIM22/BST2) vs. positive correlation (ISG15/PAF1/CDKN1A) with HTLV-1 viral markers and clinical status. Surprisingly, we found a significant inversion of antiretroviral activity of host restriction factors, as evidenced by opposite correlation to in vivo HIV-1 vs. HTLV-1 RNA levels. The anti-HTLV-1 effect of antiviral cluster genes was significantly correlated to their adaptive chimp/human evolution score, for both Tax mRNA and PVL. Six genes of the proposed antiviral cluster underwent lentivirus-driven purifying selection during primate evolution (TRIM5/TRIM22/BST2/APOBEC3F-G-H), underscoring the cross-retroviral evolutionary imprint. Secondly, we examined the genome-wide type I IFN response in HAM/TSP patients, following short-term ex vivo culture of PBMCs with either IFN-α or IFN-β. Microarray analysis evidenced 12 antiretroviral genes (including TRIM5α/TRIM22/BST2) were significantly up-regulated by IFN-β, but not IFN-α, in HAM/TSP. This was paralleled by a significant decrease in lymphoproliferation by IFN-β, but not IFN-α treatment. Finally, using published ex vivo whole blood transcriptomic data of independent cohorts, we validated the significant positive correlation between TRIM5, TRIM22, and BST2 in HTLV-1-infected individuals and HAM/TSP patients, which was independent of the HAM/TSP disease signature. In conclusion, our results provide ex vivo mechanistic evidence for the observed immunovirological effect of in vivo IFN-β treatment in HAM/TSP, reconcile an apparent IFN paradox in HTLV-1 research and identify biomarkers/targets for a precision medicine approach.