1. The nature and cellular mechanisms that are responsible for endothelium-dependent relaxations resistant to indomethacin and NG-nitro-L-arginine methyl ester (L-NAME) were investigated in phenylephrine (PE) precontracted isolated carotid arteries from the rabbit. 2. In the presence of the cyclo-oxygenase inhibitor, indomethacin (10 microM), acetylcholine (ACh) induced a concentration- and endothelium-dependent relaxation of PE-induced tone which was more potent than the calcium ionophore A23187 with pD2 values of 7.03 +/- 0.12 (n = 8) and 6.37 +/- 0.12 (n = 6), respectively. The ACh-induced response was abolished by removal of the endothelium, but was not altered when indomethacin was omitted (pD2 value 7.00 +/- 0.10 and maximal relaxation 99 +/- 3%, n = 6). Bradykinin and histamine (0.01-100 microM) had no effect either upon resting or PE-induced tone (n = 5). 3. In the presence of indomethacin plus the NO synthase inhibitor, L-NAME (30 microM), the response to A23187 was abolished. However, the response to ACh was not abolished, although it was significantly inhibited with the pD2 value and the maximal relaxation decreasing to 6.48 +/- 0.10 and 67 +/- 3%, respectively (for both P < 0.01, n = 8). The L-NAME/indomethacin insensitive vasorelaxation to ACh was completely abolished by preconstriction of the tissues with potassium chloride (40 mM, n = 8). 4. The Ca(2+)-activated K+ (KCa) channel blockers, tetrabutylammonium (TBA, 1 mM, n = 5) and charybdotoxin (CTX, 0.1 microM, n = 5), completely inhibited the nitric oxide (NO) and prostacyclin (PGI2)-independent relaxation response to ACh. However, iberiotoxin (ITX, 0.1 microM, n = 8) or apamin (1-3 microM, n = 6) only partially inhibited the relaxation. 5. Inhibitors of the cytochrome P450 mono-oxygenase, SKF-525A (1-10 microM, n = 6), clotrimazole (1 microM, n = 5) and 17-octadecynoic acid (17-ODYA, 3 microM, n = 7) also reduced the NO/PGI2-independent relaxation response to ACh. 6. In endothelium-denuded rings of rabbit carotid arteries, the relaxation response to exogenous NO was not altered by either KCa channel blockade with apamin (1 microM, n = 5) or CTX (0.1 microM, n = 5), or by the cytochrome P450 mono-oxygenase blockers SKF-525A (10 microM, n = 4) and clotrimazole (10 microM, n = 5). However, the NO-induced response was shifted to the right by LY83583 (10 microM, n = 4), a guanylyl cyclase inhibitor, with the pD2 value decreasing from 6.95 +/- 0.14 to 6.04 +/- 0.09 (P < 0.01). 7. ACh (0.01-100 microM) induced a concentration-dependent relaxation of PE-induced tone in endothelium-denuded arterial segments sandwiched with endothelium-intact donor segments. This relaxation to ACh was largely unaffected by indomathacin (10 microM) plus L-NAME (30 microM), but abolished by the combination of indomethacin, L-NAME and TBA (1 mM, n = 5). 8. These data suggest that in the rabbit carotid artery: (a) ACh can induce the release of both NO and EDHF, whereas A23187 only evokes the release of NO from the endothelium, (b) the diffusible EDHF released by ACh may be a cytochrome P450-derived arachidonic acid metabolite, and (c) EDHF-induced relaxation involves the opening of at least two types of KCa channels, whereas NO mediates vasorelaxation via a guanosine 3': 5'-cyclic monophosphate (cyclic GMP)-mediated pathway, in which a cytochrome P450 pathway and KCa channels do not seem to be involved.