Cationic amphiphiles have been shown to mediate gene transfer to eukaryotic cells, although the nature and fate of the lipid-DNA complexes is still a matter of debate. Negative staining transmission electron microscopy (TEM) of the complexes in physiological medium, as well as thin-section TEM of transfected cells has been used to visualize the particles and the possible pathways leading to transgene expression. Lipopolyamines form a network of tubular micelles into which plasmid DNA is intertwined and condensed; the cationic particles contain hundreds of plasmid molecules and are heterogeneous with respect to size (0.1-0.5 microgram) and shape. Adherent cells (293M, 3T3, MRC5, primary leptomeningeal cells) take them up readily within minutes by spontaneous endocytosis. Among suspension cells, lymphocytes only incidentally show cytoplasmic inclusions and monocytes degrade the particles by phagocytosis. The marked decrease in transfection efficiency generally observed between adherent and nonadherent cells is thus due to reduce cell binding. This suggests that cationic particles bind to membrane components responsible for Ca2+-mediated cell anchoring to the extracellular matrix. Cation/anion-mediated endocytosis leads to endosomes that are entirely filled with the particles. Consequently, two escape mechanisms may operate: disruption of the lamellar envelope in close contact with tubular micelles, and endosome buffering by the lipopolyamine in response to proton entry, leading to osmotic swelling and endosome rupture. Even for moderately transfected MRC5 cells, 10(2)-10(3) particles are found either free or in cytoplasmic vacuoles 24 h after transfection, highlighting a very inefficient nuclear translocation process. Such high numbers are also the clue to the small concentration window between transfection and cytotoxicity that is often observed with nonviral vectors. Nuclear particle inclusions are sometimes seen, yet it is unclear whether plasmid uncoating (before expression) takes place by anion exchange in the cytoplasm or in the nucleus. The still lower efficiency of free plasmid translocation to the nucleus suggests an active role for the cationic lipid during this step. Although the last stages of the transfection mechanism remain unclear, the present work shows that the major barrier which hampers in vitro gene delivery with cationic vectors is nuclear translocation (and cell entry for nonadherent cells), providing precise targets for the design of improved nonviral vectors.