Background: An efficient gene transfer can be achieved in mouse liver by a rapid tail vein injection of a large volume of plasmid DNA solution (hydrodynamics-based transfection). The mechanism of gene transfer by this procedure is not known. It must be related to the uptake and intracellular fate of DNA. Methods: We have investigated the problem by following the uptake by mouse liver and the intracellular distribution of DNA after a rapid tail vein injection of a large (2.0 ml) or a small (0.2 ml) volume of 35S-DNA solution. Total and acid-soluble radioactivity were measured in liver homogenates at increasing times after injection, and their subcellular distributions were established by centrifugation methods and compared with the distributions of marker enzymes of the membrane compartments involved in endocytosis: alkaline phosphodiesterase (plasma membrane) and cathepsin C (lysosomes). Results: 35S-DNA uptake by the liver is similar when a small or a large volume of injection is used but its degradation is markedly slower after a 2.0 ml injection. When a small volume of injection is given, distribution of radioactivity after differential centrifugation indicates that the plasmid DNA is endocytosed and reaches lysosomes where it is hydrolysed. After a large volume injection, part of 35S-DNA has the same fate, another part remains acid-precipitable for at least 1 h and is associated with structures sedimenting at low centrifugation speed in the nuclear fraction N. Analysis of that fraction by gradient centrifugation suggests that these structures are plasma membrane fragments that could originate from the apical domain of hepatocytes. The proportion of 35S-DNA associated with hepatocytes is about doubled after a large volume injection. Fractionation of isolated hepatocytes by centrifugation confirms results obtained on the whole liver. Treatment of the N fraction or isolated hepatocytes with pancreatic DNAse illustrates that 35S-DNA that remains bound to plasma membrane after a large volume injection is located on the outer face. Conclusions: The fact that after an hydrodynamic injection 35S-DNA remains bound to the outside face of the plasma membrane for at least 1 h indicates that it is not, or very slowly, internalised during that period. The relatively small difference in the amount of DNA picked up by hepatocytes depending on the type of injection could not explain the absence of expression after a conventional injection and the strong expression after a hydrodynamic injection. If DNA enters the cells by endocytosis, even after an hydrodynamic injection, its persistence at the outside face of the plasma membrane could favour transfection by allowing hepatocytes to dispose for a relatively long time of a reservoir of intact DNA.