A new in vitro model has been developed for investigating extravascular diffusion of therapeutic agents in tumour tissue. V79-171b or EMT6/Ak cells are grown on porous Teflon support membranes and submerged in a large reservoir of medium, to give diffusion-limited 'multicellular membranes' (MMs) c. 200 microm in thickness. MMs are histologically similar to multicellular spheroids, but their planar rather than spherical geometry facilitates direct measurement of the flux of radiolabelled agents through the multicellular structure. For [14C]urea, flux kinetics through V79-171b MMs was modelled as simple diffusion, yielding a diffusion coefficient in the MM (DMM) of 1.45 x 10(-6) cm2 s(-1), 11-fold lower than in culture medium. Flux of the 3H-labelled DNA intercalator 9-[3-(N,N-dimethylamino)propylamino]acridine (DAPA) was dramatically slower than urea. Modelling this over the first 5 h gave a DMM of 1.3 x 10(-8) cm2 s(-1), but over longer times the kinetics was not consistent with simple diffusion. Flux of DAPA was markedly increased in the presence of 50 mM ammonium chloride, indicating that sequestration in acidic endosomes is a major impediment to flux. Accumulation in cytoplasmic vesicles was confirmed by fluorescence microscopy. The DAPA flux kinetics, with and without ammonium chloride, was well fitted by a reaction-diffusion model with reversible cellular uptake (modelled as binding), using uptake parameters determined in separate experiments with V79-171b single-cell suspensions. This study demonstrates the utility of the MM model for determining extravascular transport parameters, and indicates that much of the impediment to diffusion of basic DNA intercalators in tumour tissue may arise from lysosomal sequestration rather than DNA binding.