1. We have used a cascade bioassay system and isolated arterial ring preparations to investigate the contribution of Ca2+ release from endothelial intracellular stores to nitric oxide (NO) production evoked by increases in shear stress and by acetylcholine in rabbit aorta. 2. Experiments were performed before and following incubation with either the endoplasmic reticulum Ca(2+)-ATPase inhibitors cyclopiazonic acid (CPA, 10 microM) and thapsigargin (TSG, 1 microM) or ryanodine (30, 100 microM) which binds to a specific endoplasmic reticulum Ca(2+)-release channel. 3. In cascade bioassay all three agents induced relaxations of the recipient ring (CPA, 24.4 +/- 3.8%; TSG, 51.5 +/- 10.6%; ryanodine, 17.4 +/- 1.6%) which were significantly attenuated by preincubation of the donor with 100 microM NG-nitro-L-arginine methyl ester (L-NAME). However, in isolated rings, only CPA and TSG induced L-NAME-sensitive relaxations (CPA 52.7 +/- 6.5%; TSG 61.3 +/- 7%). 4. Addition of superoxide dismutase (SOD) to the donor perfusate evoked relaxations of the recipient ring in cascade bioassay (13.3 +/- 1.4%, n = 22). Prior administration of SOD attenuated relaxations to TSG (23.2 +/- 3.8% n = 4) and ryanodine (1.7 +/- 0.8%, n = 4), and pre-incubation with TSG and ryanodine blunted SOD-induced responses (4 +/- 1.5%, n = 4 and 8.9 +/- 1.1%, n = 4, respectively). By contrast, no interaction was observed between the relaxations evoked by SOD and CPA. In isolated rings, SOD exerted no direct relaxant and did not modulate relaxations to CPA, TSG or ryanodine. 5. In cascade bioassay studies time-averaged shear stress was manipulated with dextran (1-4% w/v, 800000 MW) to increase perfusate viscosity. NO-dependent relaxation of the recipient ring induced by increased perfusate viscosity was significantly attenuated by CPA (P < 0.01; n = 6) and TSG (P < 0.05; n = 7), but not by ryanodine (n = 6). 6. Endothelium-dependent relaxations to acetylcholine (0.1-30 microM) in cascade bioassay and in isolated aortic ring preparations were markedly attenuated by pretreatment with CPA and TSG, but were unaffected by ryanodine. Ryanodine and CPA caused only a small attenuation of endothelium-independent relaxations to sodium nitroprusside (0.001-10 microM), whereas TSG had no effect. 7. We conclude that release of Ca2+ from CPA- and TSG-sensitive endothelial stores is necessary for NO release evoked by acute flow changes and agonists in rabbit abdominal aorta. Ca(2+)-induced Ca2+ release via the ryanodine-sensitive release channel plays no direct role in these responses. Free radical interactions may complicate the interpretation of findings in cascade bioassay compared with isolated ring preparations.