The search for synthetic analogues of somatostatin (SRIF) which exhibit selective affinities for the five known receptor subtypes (sst1-5) has generated a large number of potent agonist analogues. Many of these agonists display good subtype selectivities and affinities for the subtypes 2, 3, and 5, with very few selective for sst1 or sst4. Until the recent report by Bass and co-workers (Mol. Pharmacol. 1996, 50, 709-715; erratum Mol. Pharmacol. 1997, 51, 170), no true antagonists of somatostatin had been discovered, let alone any displaying differential receptor subtype selectivity. In this present study, we further explore the effect of this putative L,5D6 antagonist motif on somatostatin octapeptide analogues with a cyclic hexapeptide core. The most potent antagonist found to date is H-Cpa-cyclo[DCys-Tyr-DTrp-Lys-Thr-Cys]-Nal-NH2, PRL-2970 (21), which has an IC50 of 1.1 nM in a rat pituitary growth hormone in vitro antagonist assay versus SRIF (1 nM). This analogue bound to cloned human somatostatin subtype 2 receptors with a Ki of 26 nM. The highest hsst2 affinity analogue was H-Cpa-cyclo[DCys-Pal-DTrp-Lys-Tle-Cys]-Nal-NH2, PRL-2915 (15), with a Ki of 12 nM (IC50 = 1.8 nM). This analogue was also selective for hsst2 over hsst3 and hsst5 by factors of 8 and 40, respectively, and had no agonist activity when tested alone at concentrations up to 10 microM. Regression analysis of the binding affinities versus the observed antagonist potencies revealed high correlations for hsst2 (r = 0.65) and hsst3 (r = 0.52) with a less significant correlation to hsst5 (r = 0.40). This is quite different from the somatostatin agonist analogues which show a highly significant correlation to hsst2 (r > 0.9). Receptor-selective somatostatin antagonists should provide valuable tools for characterizing the many important physiological functions of this neuropeptide.