IntroductionData are sparse about the potential health risks of chronic low-dose contamination of humans by uranium (natural or anthropogenic) in drinking water. Previous studies report some molecular imbalances but no clinical signs due to uranium intake.ObjectivesIn a proof-of-principle study, we reported that metabolomics is an appropriate method for addressing this chronic low-dose exposure in a rat model (uranium dose: 40 mg L−1; duration: 9 months, n = 10). In the present study, our aim was to investigate the dose–effect pattern and identify additional potential biomarkers in urine samples.MethodsCompared to our previous protocol, we doubled the number of rats per group (n = 20), added additional sampling time points (3 and 6 months) and included several lower doses of natural uranium (doses used: 40, 1.5, 0.15 and 0.015 mg L−1). LC–MS metabolomics was performed on urine samples and statistical analyses were made with SIMCA-P+ and R packages.ResultsThe data confirmed our previous results and showed that discrimination was both dose and time related. Uranium exposure was revealed in rats contaminated for 9 months at a dose as low as 0.15 mg L−1. Eleven features, including the confidently identified N1-methylnicotinamide, N1-methyl-2-pyridone-5-carboxamide and 4-hydroxyphenylacetylglycine, discriminated control from contaminated rats with a specificity and a sensitivity ranging from 83 to 96 %, when combined into a composite score.ConclusionThese findings show promise for the elucidation of underlying radiotoxicologic mechanisms and the design of a diagnostic test to assess exposure in urine, in a dose range experimentally estimated to be above a threshold between 0.015 and 0.15 mg L−1.Electronic supplementary materialThe online version of this article (doi:10.1007/s11306-016-1092-8) contains supplementary material, which is available to authorized users.