Sabadinine: A Potential Non-Peptide Anti-Severe Acute-Respiratory-Syndrome Agent Identified Using Structure-Aided Design Jeffrey H. Toney,* ,‡ Sonia Navas-Martı ´n, # Susan R. Weiss, # and Andreas Koeller § Department of Chemistry and Biochemistry and Department of Computer Science, Montclair State University, 1 Normal Avenue, Upper Montclair, New Jersey 07043, and Department of Microbiology, University of Pennsylvania Medical School, Johnson Pavilion 202F, 36th Street and Hamilton Walk, Philadelphia, Pennsylvania 19104 Received July 3, 2003 Abstract: A novel human coronavirus has been reported to be the causative agent of severe acute respiratory syndrome (SARS). Since replication of HcoVs depends on extensive proteolytic processing, the main proteinase, 3CL pro , is an attractive drug target for anti-SARS agents. We have employed molecular docking of a chemical database into the active site of 3CL pro to search for non-peptidyl inhibitors. One compound was identified to be the natural product sabadinine, isolated from a historical herbal remedy. The recent spread of severe acute respiratory syn- drome (SARS) has caused concern because of its rela- tively high fatality rate, particularly among the elderly. 1 A novel human coronavirus (HCoV) is reported to be the causative agent of SARS. 2 The viral main proteinase (3CL pro or M pro ) encoded by HCoV has been identified as an attractive drug target, and the X-ray crystal structure has been determined recently. 3 The structural relationships between proteinases HCoV 3CL pro , porcine coronavirus (transmissible gastroenteritis virus (TGEV)) M pro or 3CL pro , and human rhinovirus (HRV) serotype 2 3C pro have been used to identify AG7088 as a starting point for structure-aided design of anti-SARS drugs. 3 AG7088 is an HRV 3C pro inhibitor currently in clinical trials for the common cold. Since AG7088 is a peptidyl molecule, it was considered to be of great interest to search for non-peptidyl inhibitors of 3CL pro , a cysteine protease. We have employed the 3CL pro crystal structure in an automated docking computer software of flexible ligands to macromolecules called AutoDock 4 to screen a large public database of small molecules as a first step toward the selection of compounds for testing in antiviral assays. Results. We chose the National Cancer Institute (NCI) diversity set that is representative of a collection of approximately 140 000 chemicals as a potential source for biologically relevant chemicals predicted to be 3CL pro inhibitors. This set of 1853 compounds was selected by the NCI as three-point pharmacophores and includes both synthetic compounds and natural prod- ucts. AutoDock performs ligand docking on precomputed grid maps representing different atom types in the receptor. The docking grid was centered on Cys 144 ,a member of the catalytic dyad of 3CL pro . The algorithm calculates the conformation of each compound with the lowest energy, and the chemical database was rank- ordered. These energies were then used as a guide to predict potential binding to the enzyme active site. The algorithm performed 10 docking optimizations per ligand, each starting from a different random initial location of the compound. Calculations for the same ligand with a root-mean-square deviation of <0.5 Å from each other were considered to be a cluster of solutions, since they represent reproducible results. The top 5% of the compounds ranked according to docking energy were then sorted by the size of the cluster. The resulting statistics for all compounds are provided in Figure 1. Within the lowest energy dockings, 10 compounds (∼0.5%) showed clustering of 5 or greater out of the 10 independent docking studies. The 10 compounds selected using the criteria de- scribed above were then evaluated for desirable chemi- cal properties as therapeutics. While most compounds within this group did not have desirable physicochem- ical properties, sabadinine (diversity 1043) was excep- tional (Scheme 1). Figure 2 shows sabadinine docked into the active site of 3CL pro having a docking energy of -11.6 kcal/mol and a clustering of 9 out of 10 conformers. The lowest energy conformer of sabadinine is characterized by both steric factors and hydrogen bonding within the active site of 3CL pro . Sabadinine is * To whom correspondence should be addressed. Telephone: 973- 655-6864. Fax: 973-655-7772. E-mail: toneyje@mail.montclair.edu. ‡ Department of Chemistry and Biochemistry, Montclair State University. # University of Pennsylvania Medical School. § Department of Computer Science, Montclair State University. Figure 1. Histogram of the lowest docking energy for each conformer using the National Cancer Institute diversity set. Scheme 1 1079 J. Med. Chem. 2004, 47, 1079-1080 10.1021/jm034137m CCC: $27.50 © 2004 American Chemical Society Published on Web 01/24/2004