PURPOSE: In a nuclear or radiological event, an early diagnostic or prognostic tool is needed to distinguish the worried well from low-exposed and those individuals who may later develop life-threatening hematologic acute radiation syndrome (H-ARS). In previous studies, we identified and validated genes in peripheral blood for this purpose. To gain a deeper understanding and to make methodological improvements, we examined the contribution of the peripheral blood´s cell populations on radiation-induced gene expression changes. MATERIALS AND METHODS: EDTA whole blood from six healthy donors was X-irradiated with 0 and 4 Gy, cultured in vitro for 24 h and cell populations of T-lymphocytes (CD3), B-lymphocytes (CD19), NK-cells (CD56) and granulocytes (CD15) were separated using immunomagnetic methods. Whole blood was used as a positive control to ensure the expected radiation-induced gene expression response based on previous examinations. Purity of cell separation and cell counts was validated using immunofluorescence imaging flow cytometry. After RNA isolation, gene expressions of FDXR, DDB2, POU2AF1 and WNT3 were examined in the cell populations and whole blood. RESULTS: The cell populations, on average, contributed to the total RNA amount with a ratio of 11.6 for T-lymphocytes: 1.2 for B-cells: 1.2 for NK-cells: 1.0 for granulocytes. In order to estimate the contribution of gene expression per cell population, the baseline (0 Gy) as well as the radiation-induced fold-change in gene expression relative to unexposed was considered for each gene. After considering all three parameters, the T-lymphocytes (74.8%/80.5%) contributed predominantly to the radiation-induced up-regulation observed for FDXR/DDB2 and the B-lymphocytes (97.1%/83.8%) for down-regulated POU2AF1/WNT3 with a similar effect on whole blood gene expression measurements reflecting a corresponding order of magnitude. CONCLUSIONS: T-lymphocytes and B-lymphocytes contributed predominantly to the radiation-induced up-regulation of FDXR/DDB2 and down-regulation of POU2AF1/WNT3. Further separation of cell populations will not increase the diagnostic sensitivity, but complicate an efficient workflow. Also, this study identifies undesired limitations of widely used whole blood in vitro models, but it still underlines the use of FDXR and DDB2 for biodosimetry purposes and POU2AF1 and WNT3 for effect prediction of acute health effects.