Several nonviral vectors including linear polyethylenimine (L-PEI) confer a pronounced lung tropism to plasmid DNA when injected into the mouse tail vein in a nonionic solution. and results We have optimized this route by injecting 50 microg DNA with excess L-PEI (PEI nitrogen/DNA phosphate = 10) in a large volume of 5% glucose (0.4 ml). In these conditions, 1-5% of lung cells were transfected (corresponding to 2 ng luciferase/mg protein), the other organs remaining essentially refractory to transfection (1-10 pg luciferase/mg protein). beta-Galactosidase histochemistry confirmed alveolar cells, including pneumocytes, to be the main target, thus leading to the puzzling observation that the lung microvasculature must be permeable to cationic L-PEI/DNA particles of ca 60 nm. A smaller injected volume, premixing of the complexes with autologous mouse serum, as well as removal of excess free L-PEI, all severely decreased transgene expression in the lung. Arterial or portal vein delivery did not increase transgene expression in other organs. These observations suggest that effective lung transfection primarily depends on the injection conditions: the large nonionic glucose bolus prevents aggregation as well as mixing of the cationic complexes and excess free L-PEI with blood. This may favour vascular leakage in the region where the vasculature is dense and fragile, i.e. around the lung alveoli. Cationic particles can thus reach the epithelium from the basolateral side where their receptors (heparan sulphate proteoglycans) are abundant.