1. Relaxation of the methoxamine-precontracted rat small mesenteric artery by endothelium-derived hyperpolarizing factor (EDHF) was compared with relaxation to the cannabinoid, anandamide (arachidonylethanolamide). EDHF was produced in a concentration- and endothelium-dependent fashion in the presence of NG-nitro-L-arginine methyl ester (L-NAME, 100 microM) by either carbachol (pEC50 [negative logarithm of the EC50] = 6.19 +/- 0.01, Rmax [maximum response] = 93.2 +/- 0.4%; n = 14) or calcium ionophore A23187 (pEC50 = 6.46 +/- 0.02, Rmax = 83.6 +/- 3.6%; n = 8). Anandamide responses were independent of the presence of endothelium or L-NAME (control with endothelium: pEC50 = 6.31 +/- 0.06, Rmax = 94.7 +/- 4.6%; n = 10; with L-NAME: pEC50 = 6.33 +/- 0.04, Rmax = 93.4 +/- 6.0%; n = 4). 2. The selective cannabinoid receptor antagonist, SR 141716A (1 microM) caused rightward shifts of the concentration-response curves to both carbachol (2.5 fold) and A23187 (3.3 fold). It also antagonized anandamide relaxations in the presence or absence of endothelium giving a 2 fold shift in each case. SR 141716A (10 microM) greatly reduced the Rmax values for EDHF-mediated relaxations to carbachol (control, 93.2 +/- 0.4%; SR 141716A, 10.7 +/- 2.5%; n = 5; P < 0.001) and A23187 (control, 84.8 +/- 2.1%; SR 141716A, 3.5 +/- 2.3%; n = 6; P < 0.001) but caused a 10 fold parallel shift in the concentration-relaxation curve for anandamide without affecting Rmax. 3. Precontraction with 60 mM KCl significantly reduced (P < 0.01; n = 4 for all) relaxations to 1 microM carbachol (control 68.8 +/- 5.6% versus 17.8 +/- 7.1%), A23187 (control 71.4 +/- 6.1% versus 3.9 +/- 0.45%) and anandamide (control 71.1 +/- 7.0% versus 5.2 +/- 3.6%). Similar effects were seen in the presence of 25 mM K+. Incubation of vessels with pertussis toxin (PTX; 400 ng ml-1, 2 h) also reduced (P < 0.01; n = 4 for all) relaxations to 1 microM carbachol (control 63.5 +/- 7.5% versus 9.0 +/- 3.2%), A23187 (control 77.0 +/- 5.8% versus 16.2 +/- 7.1%) and anandamide (control 89.8 +/- 2.2% versus 17.6 +/- 8.7%). 4. Incubation of vessels with the protease inhibitor phenylmethylsulphonyl fluoride (PMSF; 200 microM) significantly potentiated (P < 0.01), to a similar extent (approximately 2 fold), relaxation to A23187 (pEC50: control, 6.45 +/- 0.04; PMSF, 6.74 +/- 0.10; n = 4) and anandamide (pEC50: control, 6.31 +/- 0.02; PMSF, 6.61 +/- 0.08; n = 8). PMSF also potentiated carbachol responses both in the presence (pEC50: control, 6.25 +/- 0.01; PMSF, 7.00 +/- 0.01; n = 4; P < 0.01) and absence (pEC50: control, 6.41 +/- 0.04; PMSF, 6.88 +/- 0.04; n = 4; P < 0.001) of L-NAME. Responses to the nitric oxide donor S-nitroso-N-acetylpenicillamine (SNAP) were also potentiated by PMSF (pEC50: control, 7.51 +/- 0.06; PMSF, 8.00 +/- 0.05, n = 4, P < 0.001). 5. EDHF-mediated relaxation to carbachol was significantly attenuated by the K+ channel blocker tetraethylammonium (TEA; 1 mM) (pEC50: control, 6.19 +/- 0.01; TEA, 5.61 +/- 0.01; n = 6; P < 0.01). In contrast, TEA (1 mM) had no effect on EDHF-mediated relaxation to A23187 (pEC50: control, 6.47 +/- 0.04; TEA, 6.41 +/- 0.02, n = 4) or on anandamide (pEC50: control, 6.28 +/- 0.06; TEA, 6.09 +/- 0.02; n = 5). TEA (10 mM) significantly (P < 0.01) reduced the Rmax for anandamide (control, 94.3 +/- 4.0%; 10 mM TEA, 60.7 +/- 4.4%; n = 5) but had no effect on the Rmax to carbachol or A23187. 6. BaCl2 (100 microM), considered to be selective for blockade of inward rectifier K+ channels, had no significant effect on relaxations to carbachol or A23187, but caused a small shift in the anandamide concentration-response curve (pEC50: control, 6.39 +/- 0.01; Ba2+, 6.20 +/- 0.01; n = 4; P < 0.01). BaCl2 (1 mM; which causes non-selective block of K+ channels) significantly (P < 0.01) attenuated relaxations to all three agents (pEC50 values: carbachol, 5.65 +/- 0.02; A23187, 5.84 +/- 0.04; anandamide, 5.95 +/- 0.02; n = 4 for each). 7. Apamin (1mu M), a selective blocker of small conductance, Ca2+-activated, K+ channels (SKCa), 4-aminopyridine (1mM), a blocker of delayed rectifier, voltage-dependent, K+ channels (Kv), and ciclazindol (10mu M), an inhibitor of Kv and adenosine 5'-triphosphate (ATP)-sensitive K+ channels (KATP), significantly reduced EDHF-mediated relaxations to carbachol, but had no significant effects on A23187 or anandamide responses. 8. Glibenclamide (10mu M), a KATP inhibitor and charybdotoxin (100 or 300nM), a blocker of several K+ channel subtypes, had no significant effect on relaxations to any of the agents. Iberiotoxin (50nM), an inhibitor of large conductance, Ca2+-activated, K+ channels (BKCa), had no significant effect on the relaxation responses, either alone or in combination with apamin (1muM). Also, a combination of apamin (1muM) with either glibenclamide (10muM) or 4-aminopyridine (1mM) did not inhibit relaxation to carbachol significantly more than apamin alone. Neither combination had any significant effect on relaxation to A23187 or anandamide. 9. A combination of apamin (1muM) with charybdotoxin (100nM) abolished EDHF-mediated relaxation to carbachol, but had no significant effect on that to A23187. Apamin (1muM) and charybdotoxin (300nM) together consistently inhibited the response to A23187, while apamin (1muM) and ciclazindol (10muM) together inhibited relaxations to both carbachol and A23187. None of these toxin combinations had any significant effect on relaxation to anandamide. 10. It was concluded that the differential sensitivity to K+ channel blockers of EDHF-mediated responses to carbachol and A23187 might be due to actions on endothelial generation of EDHF, as well as its actions on the vascular smooth muscle, and suggests care must be taken in choosing the means of generating EDHF when making comparative studies. Also, the relaxations to EDHF and anandamide may involve activation of cannabinoid receptors, coupled via PTX-sensitive G-proteins to activation of K+ conductances. The results support the hypothesis that EDHF is an endocannabinoid but relaxations to EDHF and anandamide show differential sensitivity to K+ channel blockers, therefore it is likely that anandamide is not identical to EDHF in the small rat mesenteric artery.