The monitoring of isotopic fractionations in in vitro cultured human cell samples is a very promising and under-exploited tool to help identify the metabolic processes leading to disease-induced isotopic fractionations or decipher metabolic pathways of toxic metals in these samples. One of the limitations is that the analytes are often present at small amounts, ranging from tens to hundreds of ng, thus making challenging low-uncertainty isotope ratio determinations. Here we present a new procedure for U, Cu and Zn purification and isotope ratio determinations in cultured human neuron-like cells exposed to natural U. A thorough study of the influence of the limiting factors impacting the uncertainty of δ238U, δ66Zn and δ65Cu is also carried out. These factors include the signal intensity, which determines the within-day measurement reproducibility, the procedural blank correction and the matrix effects, which determine the accuracy of the mass bias correction models. Given the small Cu and U amounts in the cell samples, 15–30 and 20 ng respectively, a highly efficient sample introduction system was employed in order to improve the analyte transport to the plasma and, hence, the signal intensity. With this device, the procedural blanks became the main uncertainty source of δ238U and δ65Cu values, accounting over 65% of the overall uncertainty. The matrix effects gave rise to inaccuracies in the mass bias correction models for samples finally dissolved in the minimal volumes required for the analysis, 100–150 µL, leading to biases for U and Cu. We will show how these biases can be cancelled out by dissolving the samples in volumes of at least 300 µL for Cu and 450 µL for U. Using our procedure, expanded uncertainties (k = 2) of around 0.35‰ for δ238U and 0.15‰ for δ66Zn and δ65Cu could be obtained. The analytical approach presented in this work is also applicable to other biological microsamples and can be extended to other elements and applications.< Réduire