1,2-Dibenzamidobenzene Inhibitors of Human Factor Xa David K. Herron,* Theodore Goodson, Jr., Michael R. Wiley, Leonard C. Weir, ‡ Jeffrey A. Kyle, Ying K. Yee, Ann Louise Tebbe, Jennifer M. Tinsley, David Mendel, John J. Masters, Jeffry B. Franciskovich, J. Scott Sawyer, Douglas W. Beight, Andrew M. Ratz, Guy Milot, † Steven E. Hall, ‡ Valentine J. Klimkowski, James H. Wikel, Brian J. Eastwood, Richard D. Towner, Donetta S. Gifford-Moore, Trelia J. Craft, and Gerald F. Smith Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, Indiana 46285, Sphinx Pharmaceuticals, Eli Lilly & Company, Durham, North Carolina 27707, and Sphinx Pharmaceuticals, Eli Lilly & Company, Cambridge, Massachusetts 02139 Received June 24, 1999 High-throughput screening of a combinatorial library of diamidophenols yielded lead compounds with the ability to inhibit human factor Xa (fXa) at micromolar concentrations (e.g. compound 4, fXa apparent K ass ) 0.64 × 10 6 L/mol). SAR studies in this novel structural series of fXa inhibitors showed that the phenolic hydroxyl group was not essential for activity. The best activity was found in substituted 1,2-dibenzamidobenzenes in which the phenyl group of one benzoyl group (A-ring) was substituted in the 4-position with relatively small lipophilic or polarizable groups such as methoxy, vinyl, or chloro and the phenyl group of the other benzoyl group (B-ring) was substituted in the 4-position with larger lipophilic groups such as tert- butyl or dimethylamino. The central phenyl ring (C-ring) tolerated a wide variety of substituents, but methoxy, methanesulfonamido, hydroxyl, and carboxyl substitution produced slightly higher levels of activity than other substituents when present in combination with favorable B-ring substitution. Methylation of the amide nitrogen atoms was found to greatly decrease activity. Compound 12 is the highest affinity fXa inhibitor in this group of compounds, having fXa apparent K ass ) 25.5 × 10 6 L/mol, about 40× more active than the original lead. This lead series does not show potent inhibition of human thrombin. A model for the binding of these ligands to the fXa active site is proposed. The model is consistent with the observed SAR and can serve to guide future SAR studies. Introduction Thromboembolic diseases continue as a leading cause of morbidity and mortality in developed nations. A primary medical strategy to treat and to prevent such diseases has been the use of anticoagulants: heparin and low-molecular-weight heparins for parenteral short- term treatments and vitamin K antagonists such as warfarin for chronic oral therapy. However, there are serious and well-documented liabilities associated with the chronic use of warfarin, 1 and there has been a consequent worldwide pharmaceutical discovery effort to discover safer and more effective anticoagulant compounds which could be used as chronic oral therapy. Initially, a large effort focused on the target thrombin on the basis that direct inhibitors of the fibrinogen- clotting/platelet-activating protease could produce an- tithrombotic effects in all types of thrombotic disease. Thus far, however, an orally effective thrombin inhibitor has not been successfully developed 2 and the discovery focus to resolve this large unmet medical need has shifted to the serine protease factor Xa. Factor Xa (fXa) produces thrombin by activating prothrombin during blood clotting whether coagulation is triggered by tissue factor or by blood-contact mechan- isms. 3,2a Accordingly, fXa inhibitors have the potential to be effective in all types of coagulation-based throm- boembolic processes by means of repression of thrombin generation. Potential advantages over thrombin inhibi- tors include the indirect repression of many moles of generated thrombin per fXa inhibitor molecule, rather than the 1:1 stoichiometry necessary for direct thrombin inhibition, and the avoidance of direct interference with multiple thrombin-mediated activities (platelet aggrega- tion, protein C activation, factor XIII activation, factors V and VIII feedback activation). 4 Natural and recom- binant fXa inhibitor proteins and small molecular weight reversible fXa inhibitors have been shown to be anticoagulant and antithrombotic in animal models. 5 The overall development of inhibitors for fXa has to some extent built upon the effort and success for R-thrombin. A comparison of their active site binding regions 12,14 reveals that they both have the relatively deep S1 or “specificity” pocket. The interior of the S1 pocket is fairly hydrophobic except for the presence of the acidic carboxyl side chain of Asp189 at the base. It is this acidic group that is usually found forming a strong salt bridge with a basic amino acid like arginine or lysine or with other positively charged moieties such as benzamidine. Both enzymes possess the S3 binding site made up of Gly216. Additionally, both have a hydrophobic “distal” S4 region; however the structure and character of this area differ somewhat between the two enzymes. In R-thrombin the S4 area is a broad solvent-exposed hydrophobic region, made up of the side * To whom correspondence should be addressed. Tel: (317) 276- 4670. Fax: (317) 276-1417. E-mail: dkherron@lilly.com. ‡ Sphinx Pharmaceuticals, North Carolina. † Sphinx Pharmaceuticals, Massachusetts. 859 J. Med. Chem. 2000, 43, 859-872 10.1021/jm990326m CCC: $19.00 © 2000 American Chemical Society Published on Web 02/23/2000