Examining Radiation-Induced In Vivo and In Vitro Gene Expression Changes of the Peripheral Blood in Different Laboratories for Biodosimetry Purposes: First RENEB Gene Expression Study

The risk of a large-scale event leading to acute radiation exposure necessitates the development of high-throughput methods for providing rapid individual dose estimates. Our work addresses three goals, which align with the directive of the European Union's Realizing the European Network of Biodosimetry project (EU-RENB): 1. To examine the suitability of different gene expression platforms for biodosimetry purposes; 2. To perform this examination using blood samples collected from prostate cancer patients (in vivo) and from healthy donors (in vitro); and 3. To compare radiation-induced gene expression changes of the in vivo with in vitro blood samples. For the in vitro part of this study, EDTA-treated whole blood was irradiated immediately after venipuncture using single X-ray doses (1 Gy/min(-1) dose rate, 100 keV). Blood samples used to generate calibration curves as well as 10 coded (blinded) samples (0-4 Gy dose range) were incubated for 24 h in vitro, lysed and shipped on wet ice. For the in vivo part of the study PAXgene tubes were used and peripheral blood (2.5 ml) was collected from prostate cancer patients before and 24 h after the first fractionated 2 Gy dose of localized radiotherapy to the pelvis [linear accelerator (LINAC), 580 MU/min, exposure 1-1.5 min]. Assays were run in each laboratory according to locally established protocols using either microarray platforms (2 laboratories) or qRT-PCR (2 laboratories). Report times on dose estimates were documented. The mean absolute difference of estimated doses relative to the true doses (Gy) were calculated. Doses were also merged into binary categories reflecting aspects of clinical/diagnostic relevance. For the in vitro part of the study, the earliest report time on dose estimates was 7 h for qRT-PCR and 35 h for microarrays. Methodological variance of gene expression measurements (CV ≤10% for technical replicates) and interindividual variance (≤twofold for all genes) were low. Dose estimates based on one gene, ferredoxin reductase (FDXR), using qRT-PCR were as precise as dose estimates based on multiple genes using microarrays, but the precision decreased at doses ≥2 Gy. Binary dose categories comprising, for example, unexposed compared with exposed samples, could be completely discriminated with most of our methods. Exposed prostate cancer blood samples (n = 4) could be completely discriminated from unexposed blood samples (n = 4, P < 0.03, two-sided Fisher's exact test) without individual controls. This could be performed by introducing an in vitro-to-in vivo correction factor of FDXR, which varied among the laboratories. After that the in vitro-constructed calibration curves could be used for dose estimation of the in vivo exposed prostate cancer blood samples within an accuracy window of ±0.5 Gy in both contributing qRT-PCR laboratories. In conclusion, early and precise dose estimates can be performed, in particular at doses ≤2 Gy in vitro. Blood samples of prostate cancer patients exposed to 0.09-0.017 Gy could be completely discriminated from pre-exposure blood samples with the doses successfully estimated using adjusted in vitro-constructed calibration curves.