This paper represented the interface between two long and very productive periods of research. The first of these started in 1976 with our discovery of thromboxane synthase (Needleman eta]., 1976) and prostacyclin (Moncada et a1 1976a) and the second began with this paper and led to the discovery the following year of the identity of EDRF as nitric oxide (NO) (Palmer et ul., 1987) and the subsequent elucidation of its many biological roles.Endothelium-dependent relaxation and EDRF had been identified by Furchgott and Zawadzki (1980) using fresh vascular tissues and many people had confirmed their findings using a variety of tissues (see Furchgott, 1984; Moncada et al., 1987); Vanhouttc, 1989). A number of experiments were carried out attempting to transfer biologically active EDRF from one piece of vascular tissue to another. Some of these were sufficiently successful for a half-life for EDRF to be calculated (Griffith el ul., 1984; Cocks et al., 1985; Forstermann et al., 1985). Many other experiments were carried out attempting to identify the chemical structure of EDRF and to inhibit its activity (Singler et al., 1984; Pinto et al., 1985; MacDonald et al., 1986).We tried to approach the problem in a different way. The difficulties associated with the research on EDRF at that time were related to the indirect nature of the majority of the experiments in which endothelium-dependent relaxation was measured or, when working with endothelium-derived relaxing factor, to the very small amounts of material that were being used. We thought that if proper pharmacological studies were to be carried out and the chemical structure was to he identified then we needed a larger quantity of material and a reliable and, as much as possible, a quantitative bioassay system. We decided to culture endothelial cells on microcarrier beads and to perfuse them in a modified chromatography column. We used the perfusate to superfuse a series of bioassay tissues, both vascular and non-vascular, which when combined appropriately allowed the differential bioassay of prostacyclin and EDRF.The system was useful in several ways since, first, it allowed the culture of many millions of endothelial cells which opened the door to further important studies. Secondly, although the cells were not as responsive as vascular tissue, they did respond to substances such as bradykinin and the Ca2+ ionophore A23 187 by releasing both prostacyclin and EDRF and this allowed the study of some of the biological characteristics of EDRF itself. Thirdly, the bioassay superfusion technique, as many times before (see Vane, 1964), proved to be immensely informative and reliable.We were particularly well placed to do this work since, during the mid-l970s, we had gained a great deal of experience doing bioassay of unstable arachidonic acid metabolites such as prostacyclin and thromboxane A, (TXA,), which had a half life of approximately 30 s. With Stuart Bunting we had struggled and succeeded in bioassaying TXA, generated from different sources (Moncada et al., I...