Pharmacophore Modeling as an Efficient Tool in the Discovery of Novel Noncompetitive AMPA Receptor Antagonists Maria Letizia Barreca,* ,² Rosaria Gitto, ² Silvana Quartarone, ² Laura De Luca, ² Giovambattista De Sarro, ‡ and Alba Chimirri ² Dipartimento Farmaco-Chimico, Universita ` di Messina, Viale Annunziata, 98168 Messina, Italy, and Dipartimento di Medicina Sperimentale e Clinica, Universita ` Magna Græcia,Via T. Campanella, 88100 Catanzaro, Italy Received October 24, 2002 A three-dimensional pharmacophore model for the binding of noncompetitive AMPA receptor antagonists was developed in order to map common structural features of highly active compounds. This hypothesis, which consists of two hydrophobic regions, one hydrogen bond acceptor and one aromatic region, was successfully used as framework for the design of a new class of allosteric modulators containing a tetrahydroisoquinoline skeleton and for in silico screening. The promising biological results suggested that the identified molecules might be useful “lead compounds” for future drug development. INTRODUCTION Glutamate (L-Glu) is considered the main excitatory neurotransmitter in the mammalian central nervous system (CNS), activating ionotropic (iGluRs) as well as metabotropic receptors (mGluRs). iGluRs are thought to play an important physiological role in learning and memory, whereas their excessive activation seems to be related to the neuronal death associated with stroke, global ischemia, and epilepsy. 1-3 iGluRs are divided into three classes on the basis of sequence identity and pharmacological response to the agonists 2-amino-3-(3-hydroxy-5-methylisoxazol-4-yl)pro- pionic acid (AMPA), kainic acid (KA), and N-methyl-D- aspartic acid (NMDA). 4 Considering the pivotal role of AMPA receptor (AMPAR) in physiological and pathological CNS phenomena, signifi- cant effort has been focused toward the synthesis of specific ligands as a source of potential anticonvulsant and neuro- protective agents. 5,6 The first AMPAR antagonists reported in the literature were the competitive inhibitors such as NBQX, heterocyclic- fused quinoxalinones, isatinoximes, etc. Later, the search for compounds able to interact with AMPAR resulted in the identification of several classes of modulators which proved to act via a noncompetitive mechanism of action. For a long time 2,3-benzodiazepine derivatives were the only noncompetitive AMPAR antagonists available as phar- macological tools; within this class of compounds GYKI 52466 (1) was the prototypic anticonvulsant and neuropro- tective agent in rodents. 7 Subsequently, talampanel (2) was identified, and clinical trials in patients with severe epilepsy not responsive to other drugs are now under way. 8 Moreover, taking the 2,3-benzodiazepines as template, highly potent arylphthalazine derivatives (e.g. 3, SYM-2207) have been synthesized. 9 In the field of 2,3-benzodiazepine analogues, we reported chemical and biological studies of a large series of 7,8- dimethoxy-2,3-benzodiazepines (e.g. 4, CFM-2) which have been shown to be specific noncompetitive AMPAR antago- nists and potent anticonvulsant agents in various seizure models. 10-14 Recently a new class of noncompetitive AMPAR antago- nists has been identified by Pfizer scientists, who discovered CP-4650223 (5) as lead compound of this series of quinazo- lines. 15,16 Despite the insightful interest for allosteric modulators of AMPAR and their crucial role in specific diseases, their mechanism has not been completely clarified to date, and a comprehensive study of the structural features affecting potency and selectivity is still lacking. Furthermore, no information is currently available about the location and composition of the AMPAR negative allosteric ligand binding region. In the absence of such three-dimensional structure-based information, we attempted to identify the hypothetical 3-D ligand-based pharmacophore model by using the common features hypothesis generation approach (HipHop) imple- mented in the program Catalyst. 17 In particular, HipHop algorithm finds common feature pharmacophore models among a set of highly active compounds thus carrying out a “qualitative model” (without the use of activity data), which represents the essential 3D arrangement of functional groups common to a set of molecules for interacting with a specific biological target, i.e., AMPAR in the current study. This approach is the most appropriate for our ligands; in fact, the “quantitative” hypothesis generation method is not suitable considering that the available in vitro and in vivo biological data were evaluated through different experimental protocols and therefore these values are not homogeneous. Thus, the main goal of this work was the deciphering of main three-dimensional structural requirements that are relevant in a molecule in order to noncompetitively interact with AMPAR. * Corresponding author phone: (++39) 090-6766464; fax: (++39) 090- 355613; e-mail: barreca@pharma.unime.it. ² Universita ` di Messina. ‡ Universita ` Magna Græcia. 10.1021/ci025625q CCC: $25.00 © xxxx American Chemical Society PAGE EST: 5 Published on Web 00/00/0000