Computational Chemistry Successes and Limitations

Despite the uncertainties in aligning sequences of remotely related GPCRs to that of bovine rhodopsin, successful outcomes based on the resulting homology models have been cited (see, e.g., a recent review [6]). For example, using an X-ray structure (PDB' s 1F88) of bovine rhodopsin as a structural template, Bissantz et al. [5] created antagonist-bound homology models of three human GPCRs (dopamine D3, muscarinic M1, and vasopressin V1a) and agonist -bound homology models of three human GPCRs (dopamine D3, p2-adrenergic, and 8-opioid). Using three docking algorithms and seven scoring functions, the authors screened six databases, each containing 3D structures of 990 random analogs plus 10 known antagonists or agonists for each target. As the authors concluded, the homology models that were based on bovine rhodopsin were effective in identifying known antagonists that they seeded in the database but were not sufficiently accurate for identifying known agonists. To develop improved agonist models, the authors also utilized their own knowledge- and pharmacophore-based modeling protocol. Successful construction and validation of models have relied heavily on experimental results (e.g., mutational data, structure-activity relationships [SAR], etc.) [6]. In one example, Xie et al.

[56] developed a homology model, which agreed well with known biochemical and structural data, of the human CB2 receptor, again using the X-ray structure of bovine rhodopsin as a template. Methodology that avoids the use of bovine rhodopsin as a template for constructing GPCR models has also been reported. Vaidehi et al., using the ab initio structure prediction algorithms, MembStruk and HierDock, reproduced [57] the bovine rhodopsin X-ray crystal structure within a root-mean-square (RMS) difference of 3.1 A spanning the transmembrane region. In addition, the authors predicted protein structure and function without knowledge of the experimental structures for four classes of GPCR targets (p1AR, EDG6, human sweet receptor, and mouse/rat I7 olfactory receptors). Shacham et al. developed [58] PREDICT, an algorithm that combines knowledge of the amino acid sequence with properties of the membrane environment without utilizing information about the 3D structure of rhodopsin. Using this methodology, the authors reproduced the X-ray geometry of rhodopsin within an RMS difference of 3.87 A spanning the transmembrane region. The methodology also showed potential in creating other GPCR homology models for structure-based drug discovery, including the screening of virtual libraries. In other studies, hybrid approaches, including the use of receptor-ligand pharmacophores [59] and ligand-based homology modeling [60], were developed to enhance [43] the success of GPCR homology modeling efforts when a bovine rhodopsin X-ray structure is used as the template. For example, the use of structure-activity relationships, site-directed mutagenesis data, and affinity labeling studies have been integrated with various ligand-based approaches that resulted in a few encouraging examples [12,39-41].The lack of broad implementation demonstrates that GPCR modeling has not yet reached the stage of other successful drug-discovery approaches, due to the unavailability of high-resolution ligand-mediated GPCR X-ray structures.

The potential impact of using X-ray structures of ligand-activated GPCR proteins directly has recently been assessed with the high -resolution p2AR/ carazolol X-ray structure as a prototype for drug discovery. High-throughput docking alone efficiently extracted [61] low-nanomolar compounds from very large databases, demonstrating that GPCR X-ray structures could indeed play as critical a role in a drug-discovery setting as X-ray structures of other targets. Furthermore, the same X-ray structure was used for the docking of a number of known p2AR ligands [61], including timolol. The accuracy of the predicted binding modes was validated with a later publication that reported a p2AR X-ray structure complexed with the antagonist timolol [27, 62]. The predicted binding mode [61, 62] of timolol in the p2AR/carazolol X-ray structure and the p2AR/timolol X-ray structure are in very good agreement.

The limitations of using a remotely related GPCR as a template for homol-ogy modeling have been demonstrated in recent reports. Costanzi has shown -63] that docking carazolol into a rhodopsin-based p2AR homology model produced poorer results than those obtained by the direct use of the p2AR X-ray structure. Similarly, when epinephrine was docked to another rhodop-sin - based p2AR homology model, the hydroxy alkylamine moiety of the predicted binding mode [ 64] was involved in interactions different from those observed for the identical structural component within carazolol. The difference in interactions may be an artifact of the model or, for example, a reflection of different binding modes adopted by an inverse agonist (carazolol) and an agonist (epinephrine). In other efforts [44, 65] , a ligand-mediated GPCR template, the p2AR X- ray structure, was used for predicting other Class A GPCR structures. The binding of ZM241385 is very different in mode and location in the A2 a adenosine receptor X- ray structure [ 29] and the p2AR -based homology model [44] of the A2a adenosine receptor. Similarly, the p2AR X-ray structure is now being used for agonist drug-discovery efforts [66-68]. For example, improved efficiency in database mining for agonists is observed when the p2AR X-ray structure is modified to contain a binding pocket that more closely represents a closed (active) form that is expected for agonist binding [66] .

16.4. CLASS C GPCRs

While the 7TM domain is the common component of GPCRs and is directly involved in the intracellular interaction with G proteins for all classes of GPCRs, large extracellular (N-terminal) extensions play key roles in Class B and Class C GPCRs. For Class C GPCRs, with regard to the extracellular regions, recent X- ray structures also have provided new opportunities that parallel those described above for the 7TM region. Indeed, conceptually analogous insights and questions are emerging.

16.4.1. Global Architecture

Class C GPCRs include the eight different mGluRs, which are subdivided into three subclasses (mGluR1 and 5 in subclass I; mGluR2 and 3 in subclass II; and mGluR4, 6, 7, and 8 in subclass III), as well as the sweet taste receptors, umami receptors, and the GABAB receptors. We focus on the mGluR receptors. The extracellular, N-terminal portion of these receptors contains the so called Venus flytrap (VFT) domain and the cysteine-rich (C-rich) domain (see Fig. 16.4). Unlike the Class A receptors whose ligand-binding site is the 7TM

Orthosteric (Glutamate) Binding Site

Upper Lobe Lower Lobe

Orthosteric (Glutamate) Binding Site

Upper Lobe Lower Lobe

Transmembrane Domain

Figure 16.4 Schematic representation of the Class C GPCRs.

VFT Domain

C-Rich Domain

Transmembrane Domain

Extracellular Domain

Figure 16.4 Schematic representation of the Class C GPCRs.

domain, as described above, the mGluR binding site of the endogenous ligand, glutamate, is the VFT domain. The C-rich domain lies between the N-terminal VFT and C-terminal 7TM domains. It is now well established that these proteins exist as dimers in the cell membrane.

From a drug-discovery perspective, by analogy to the binding sites of other receptors or enzymes, the VFT region would be the site to target for new ligands (agonists or antagonists) as it represents the endogenous ligand's binding site. Drug-discovery efforts, particularly before the availability of the VFT X-ray structures described below, tended to be built around the glutamate template. The belief that the VFT binding site required highly polar ligands, coupled with the argument that evolutionary pressure preserves common recognition features in the VFT to bind the common glutamate ligand, resulting in less subtype selectivity, together with pharmacological arguments, has led to the search for ligands that bind in the 7TM domain. These ligands act as "allosteric" modulators on the activity of the endogenous ligand, glutamate. Examples of positive and negative allosteric modulators are now common [69-71].

Fully understanding the mechanism of action of these multi-domain dimers is daunting, especially when considered in light of the challenges of just the 7TM domain discussed earlier for Class A GPCRs. While the site for endogenous ligand binding may be different for Class C GPCRs, ultimately, they too couple to G proteins and would be expected to have common coupling features. It is therefore intriguing, and arguably understandable, that a 7TM -acting positive allosteric modulator of mGluR5 was found to function as a full agonist in an N- t erminal truncated (no extracellular [EC] region) receptor [72] - In spite of the complexity, here too, recent X-ray structures are leading to valuable developments.

0 0

Post a comment