Thrombin inhibitors are potentially useful in medicine for their anticoagulant and antithrombotic effects. We synthesized and evaluated diverse heterocycle-activated ketones based on the d-Phe-Pro-Arg, and related thrombin active-site recognition motifs, as candidate inhibitors. The peptide-based alpha-ketoheterocycles were typically prepared by either an imidate or a Weinreb amide route (Schemes 1 and 2), the latter of which proved to be more general. Test compounds were generally assayed for inhibition of human alpha-thrombin and bovine trypsin. From a structure-based design standpoint, the heterocycle allows one to explore and adjust interactions within the S1' subsite of thrombin. The preferred alpha-ketoheterocycle is a pi-rich 2-substituted azole with at least two heteroatoms proximal to the carbon bearing the keto group, and a preferred thrombin inhibitor is 2-ketobenzothiazole 3, with a potent K(i) value of 0.2 nM and ca. 15-fold selectivity over trypsin. 2-Ketobenzothiazole 13 exhibited exceedingly potent thrombin inhibition (K(i) = 0.000 65 nM; slow tight binding). Several alpha-ketoheterocycles had thrombin K(i) values in the range 0.1-400 nM. The "Arg" unit in the alpha-ketoheterocycles can be sensitive to stereomutation under mildy basic conditions. For example, 2-ketothiazoles 4 and 59 readily epimerize at pH 7.4, although they are fairly stable stereochemically at pH 3-4; thus, suitable conditions had to be selected for the enzymatic assays. Lead d-Phe-Pro-Arg 2-benzothiazoles 3, 4, and 68 displayed good selectivity for thrombin over other key coagulation enzymes (e.g., factor Xa, plasmin, protein Ca, uPA, tPA, and streptokinase); however, their selectivity for thrombin over trypsin was modest (<25-fold). Compounds 3, 4, and 68 exhibited potent in vitro antithrombotic activity as measured by inhibition of gel-filtered platelet aggregation induced by alpha-thrombin (IC(50) = 30-40 nM). They also proved to be potent anticoagulant/antithrombotic agents in vivo on intravenous administration, as determined in the canine arteriovenous shunt (ED(50) = 0.45-0.65 mg/kg) and the rabbit deep vein thrombosis (ED(50) = 0.1-0.4 mg/kg) models. Intravenous administration of 3, and several analogues, to guinea pigs caused hypotension and electrocardiogram abnormalities. Such cardiovascular side effects were also observed with some nonguanidine inhibitors and inhibitors having recognition motifs other than d-Phe-Pro-Arg. 2-Benzothiazolecarboxylates 4 and 68 exhibited significantly diminished cardiovascular side effects, and benzothiazolecarboxylic acid 4 had the best profile with respect to therapeutic index. The X-ray crystal structures of the ternary complexes 3-thrombin-hirugen and 4-thrombin-hirugen depict novel interactions in the S(1)' region, with the benzothiazole ring forming a hydrogen bond with His-57 and an aromatic stacking interaction with Trp-60D of thrombin's insertion loop. The benzothiazole ring of 3 displaces the Lys-60F side chain into a U-shaped gauche conformation, whereas the benzothiazole carboxylate of 4 forms a salt bridge with the side chain of Lys-60F such that it adopts an extended anti conformation. Since 3 has a 10-fold greater affinity for thrombin than does 4, any increase in binding energy resulting from this salt bridge is apparently offset by perturbations across the enzyme (viz. Figure 4). The increased affinity and selectivity of 2-ketobenzothiazole inhibitors, such as 3, may be primarily due to the aromatic stacking interaction with Trp-60D. However, energy contour calculations with the computer program GRID also indicate a favorable interaction between the benzothiazole sulfur atom and a hydrophobic patch on the surface of thrombin.