1. The human EP3 prostaglandin receptor is a seven transmembrane, G protein-coupled receptor that couples to inhibition of adenylyl cyclase. The receptor occurs as at least six isoforms which result from alternative splicing. The isoforms are identical over the first 359 amino acids, comprising the seven transmembrane helices, but differ in the carboxyl terminal tail which ranges in length from 6 to 65 amino acids beyond the common region. 2. We have stably expressed in CHO-K1 cells four of the isoforms (EP3I-EP3IV) and a form of the EP3 receptor (T-359) truncated at the carboxyl-terminal region defined by the alternative splicing site at amino acid number 359. 3. Isoforms EP3I and EP3II showed concentration-dependent inhibition of forskolin-stimulated adenylyl cyclase in CHO-K1 cells by the EP3 receptor agonist, sulprostone. The IC50 calculated for sulprostone inhibition was 0.2 nM for EP3I and 0.15 nM for EP3II. The maximum extent of inhibition was 80% for both isoforms. 4. Isoforms EP3III and EP3IV showed marked constitutive activity, inhibiting forskolin-stimulated adenylyl cyclase in the absence of agonist. EP3IV also displayed some agonist-dependent inhibition whereas EP3III was fully constitutively active. 5. The truncated receptor T-359 was fully constitutively active, inhibiting forskolin-stimulated adenylyl cyclase by about 70% in the absence of agonist, and showed no agonist-dependent inhibition, in agreement with a similar truncation of the mouse EP3 receptor. 6. To confirm that differences in cyclic AMP level between isoforms represent constitutive activity, we treated cells with pertussis toxin for 6 h to abolish Gi function. Pertussis toxin reversed sulprostone-mediated inhibition of cyclic AMP formation in EP3I and EP3II and abolished constitutive activity of EP3III, EP3IV and T-359 so that the level of forskolin-stimulated cyclic AMP produced was the same in all cells and similar to that obtained in mock-transfected cells. In mock-transfected cells, sulprostone had no effect on forskolin-stimulated cyclic AMP formation. 7. For these experiments we chose clones that showed similar expression levels of each isoform, as determined by binding of [3H]-prostaglandin E2 (PGE2) (EP3I, 0.71; EP3II, 1.47; EP3IV, 1.59 pmol mg-1 protein). Mock-transfected cells showed no detectable binding of [3H]-PGE2. In addition, we performed a detailed study of the effects of expression level on constitutive activity. Over a six fold range of expression there was no change in the properties of each isoform with regard to whether it was constitutively active or not. 8. The degree of constitutive activity correlated with the inverse of the length of the C-terminal tail of the isoforms. However, no correlation was found between isoforms from human and mouse: whereas EP3II shows no constitutive activity, its mouse homologue, EP3 gamma, shows almost complete constitutive activity, even though the C-terminal domains of the receptors following the splice site differ in only 7 of 29 amino acids.