Modeling of ex vivo immune response reveals a central role for CD8+CD25+ cells in Adult T-cell Leukemia survival: a long-term prospective cohort study

Abstract

Adult T-cell leukemia (ATL) is a rare but aggressive CD4+CD25+ leukemia triggered by HTLV-1 infection. We quantified ex vivo levels of CD4+ and CD8+ (sub)populations and their antiproliferative, pro-apoptotic, antiviral and immunomodulatory interplay in short-term culture of primary cells from ATL patients, in a long-term prospective study (819 person-years of follow-up). We integrated clinical, cellular and molecular data into a data mining approach combining linear (consensus HIerarchical Tree clustering) and non-linear (BAYesian network) models. This HIT-BAY approach revealed an association between CD8+CD25+ cells and survival, independent of CD4+CD25+ cells. Moreover, CD8+CD25+ levels at diagnosis significantly predicted 5-year survival in ATL patients (p = 0.037), which was confirmed in multivariable Cox regression models correcting for clinical forms. We provide in vivo support for our ex vivo model in a unique patient on AZT monotherapy, for whom adding IFN-α resulted in a rapid (<2 weeks) and pronounced (>70%) decline in leukemic/non-leukemic cell ratio, accompanied by a ten-fold increase CD8+CD25+ levels, and followed by long-term survival. Finally, transcriptomic analysis of a CD8+CD25+ gene module and replication in an independent ATL cohort revealed shared cytotoxic activity of both CD8+CD25+ and Tax-specific CD8+ cells as an underlying mechanism for prolonged survival. In conclusion, our integrated data mining approach allowed us to integrate ex vivo clinical and immunological data, revealing a central role for CD8+CD25+ cells in ATL survival. HIT-BAY might serve as a prototype to model immune response and facilitate biomarker and therapeutic target discovery in real-world cohorts, including rare malignancies.