New Allosteric Modulators of Metabotropic Glutamate Receptor 5 (mGluR5) Found by Ligand‐Based Virtual Screening

Abstract

Glutamate is the major excitatory neurotransmitter in the mammalian central nervous system. Its long-term modulating affect on synaptic transmission is mediated by the metabotropic glutamate receptors (mGluR), a family of class III G protein-coupled receptor (GPCR) proteins. This class of GPCRs is characterized by the typical seven transmembrane domain and an additional large extracellular ligand-binding domain. Allosteric modulators that bind to the receptor within the seven transmembrane helix region—that is, the region where the ligand-binding pocket is located in type I and type II GPCRs— represent an attractive class of molecules for addressing several neural disorders associated with mGluR activity. The first selective allosteric antagonists for mGluR5, SIB-1751 (1) and SIB-1893 (2), were published in 1999 (Scheme 1). SIB-1751 was identified by high-throughput screening (HTS), and SIB1893 resulted from a UNITY search for analogues. In phosphoinositol (PI) hydrolysis assays, the two molecules revealed IC50 values of 3.1 and 2.3 mm, respectively. Chemical variation of SIB-1893 resulted in the much more potent, highly selective mGluR5 antagonist 2-methyl-6-(phenylethynyl)pyridine (MPEP, 3), which had an IC50 of 36 nm in PI hydrolysis assays. [4] Several MPEP analogues (4–9) with reported low nanomolar activity have been published in the scientific and patent literature since then. Nonetheless, the mode of action of these ligands is not completely understood. Recent publications of MPEP and MPEP derivatives also reported off-target activity and a short plasma half life. In particular, the latter could be attributed to potential metabolic instability of the ethynyl linker. Pharmacophore-based similarity searching has been proven to be suited to finding new ligands that exhibit similar biological activity but are based on a different chemical scaffold. Using a set of known specific allosteric antagonists of mGluR5 (3–9), 5] which were compiled from scientific and patent literature, we applied a hierarchical, ligand-based virtual-screening approach to identify novel compounds that modulate mGluR5. First, a “drug-likeness” estimation by an artificial neural-network system was employed to prescreen just for molecules with a predicted “drug-like” structure. For subsequent similarity searching, we used the CATS3D descriptor, a 3D extension of the original topology-based CATS approach, both of which were designed for the sake of “scaffold-hopping”. 10] CATS3D encodes the conformation of a molecule in the form of a histogram that contains the normalized frequencies of all pairs of atoms of a molecule. Atom pairs were subdivided into groups that were characterized by atom–atom distance ranges in Euclidean space and six different pharmacophore types. 20 equal-distance bins from 0–20 were used. One of the pharmacophore types—cation, anion, hydrogenbond acceptor, hydrogen-bond donor, polar (hydrogen-bond acceptor and hydrogen-bond donor) or hydrophobic—was assigned to each atom by using the ph4_aType function of MOE (Chemical Computing Group Inc. , Montr al, Canada; URL: http://www.chemcomp.com). The CATS3D representation is rotationand translation-invariant, and therefore enables rapid pair-wise comparisons of molecules without the need to explicitly align the molecules. To form a hypothesis about receptor-bound 3D conformations of 3–9, we used the flexible-alignment tool of MOE. Ligands were successively aligned from 3 to 9 (Figure 1), and conformations were chosen based on existing knowledge