Common metabolomics platforms require about 10(6) cells, which has a limited throughput due to the time-consuming steps of cell culture and preparation. There is a demand for metabolic profiling methods to improve analytical throughput and detection sensitivity based on small amount of cells. In this study, we proposed a high-throughput scheme, integrating 96-well plate cell cultivation, in-situ cell pretreatment, and sensitive dansylation labeling coupled with LC-MS analysis of metabolites inside HepG2 cells (of the order of magnitude of 10(3) cells in each well). A simple and rapid cell pretreatment was performed showing good extraction efficiency and good precision (the RSDs smaller than 5%) for polar metabolites. The recovery in metabolite extraction evaluated with three isotope-labeled amino acids was from 89.7 to 106.3% at low, medium, and high concentrations. The suitability of the method was illustrated by exploring influences of different fatty acids on HepG2 cells.