© 2007 Science Publications In silico Modeling of α1A-Adrenoceptor: Interaction of its Normal and Mutated Active Sites with Noradrenaline as well as its Agonist and Antagonist

Noradrenaline, like most other neurotransmitters, acts through various adrenoceptor subtypes. The structure and active site of adrenoceptors for the binding of noradrenaline were unknown, however, such information are crucial for understanding the molecular mechanism of action of neurotransmitters, including noradrenaline, in health and disease as well as for drug designing. In this in silico study, we modeled the �1A-adrenoceptor; a G protein coupled receptor and defined its active site. Further, molecular docking and interaction of noradrenaline and its agonist as well as antagonist with the so defined active site of the receptor was studied before and after in silico site directed mutation of several amino acid residues forming the active site. Our results indicate that the ARG166 is the most crucial residue for binding of noradrenaline and methoxamine to �1A- adrenoceptor and ILE178 is the most important residue for binding of prazosin to it. Thus, the observations provide new insights into the structure function relationship of �1A-adrenoceptor. A significant finding of this study is that the same residue of the active site may not be necessary for binding of a receptor with its natural ligand and its pharmacologically active known agonist and antagonist.


INTRODUCTION
Rapid eye movement (REM) sleep is an integral component of sleep-waking rhythm, is present in mammals including humans and throughout ones life span, although its quantum varies through age and in different species [1] . Its importance may be gauged by the fact that REM sleep loss tends to be compensated by its rebound increase and if the deprivation is prolonged, several patho-physio-psycho-behavioral disorders set in [2][3][4][5][6][7][8][9] . As a mechanism of action it has been shown that cessation of activity of the noradrenaline (NA) containing neurons in the locus coeruleus is a prerequisite for REM sleep generation [10] and upon REM sleep deprivation they do not cease activity [11] resulting in increased levels of NA in the brain [12][13][14] . This increased NA is the primary cause for REM sleep deprivation associated changes including increased Na-K ATPase activity [15] neuronal cytomorphometry [16] and apoptosis [17] and all these changes were mediated by NA acting upon 1adrenoceptor. Additionally we showed that at least the increase in Na-K ATPase activity was mediated by 1A-adrenoceptor [18] . For further understanding of the mechanism of action at the molecular level, the next step was to study the interactions between NA and 1A-adrenoceptor, however, although the structure of NA was known, the structure and active site of the 1A-adrenoceptor was unknown. Hence, we took the help of bioinformatics to first model the 1Aadrenoceptor in silico and deposited it in the PDB as 2F75. Thereafter, in this study we modeled the active site of the 1A-adrenoceptor and carried out in silico mutation of different amino acids within the modeled active site to determine the component of the adrenoceptor that is essential for binding with NA. In addition to NA, we also studied the interactions of the so modeled active site of the receptor with methoxamine, prazosin and WB4101 the known pharmacologically active agonist and antagonist of 1A-adrenoceptor.

MATERIALS AND METHODS
Homology Modeling: The sequence of rat 1Aadrenoceptor was obtained from the Swiss-Prot Database [19] ; the protein contains 466 amino acid (Accession number: P43140). BLAST [20] search among the proteins of known 3D structure revealed that the bovine rhodopsin showed the highest (21%) sequence identity score with 1A-adrenoceptor; the sequence similarity was 31%. 3D-JURY [21] also picked bovine rhodopsin 1GZM at 2.65 Å as the template. Hence, in this study X-ray structure of bovine rhodopsin from Bos taurus (PDB code: 1l9h) [22] was selected as a template and based on that the 3D structure of 1A-adrenoceptor was predicted. However, the initial N-terminal residues 1-35 and final C-terminal residues 342-466 were not modeled because these were absent in the template. The 3DJIGSAW, automated homology modeling tool [23] , was used to model the 1A-adrenoceptor framework for residues 36-341. STRIDE [24] , which uses hydrogen bond energy and main chain dihedral angles to recognize helix, coils and strands was used to predict the secondary structure of the so modeled 1Aadrenoceptor. Hydropathy of 1A-adrenoceptor was analyzed according to the algorithm of Kyte and Doolittle [25] . The weighted root mean square deviation (RMSD) of the modeled protein was calculated using combinatorial extension (CE) algorithm [26] . Solvent accessibility was measured using the program GEPOL [27] . The computed structure of the 1Aadrenoceptor obtained was refined by energy minimization with CHARMM force field until the energy showed stability in sequential repetition [28] . The stability of the so predicted theoretical model was evaluated sterically with procheck [29] .
Gold docking: The chemical structures of 1Aadrenoceptor agonists and antagonists were extracted from pubchem database (http://pubchem.ncbi.nlm.nih.gov/). Structures of all the five ligands (adrenaline, NA, methoxamine, prazosin and WB4101) were retrieved into two-dimensional MDL/SDF format. Three -dimensional coordinates were generated using the CORINA program [30] . The chemical structures of NA and the known pharmacological agonist (methoxamine) and antagonist (prazosin and WB4101) of 1A-adrenoceptor are shown in Fig. 1. Docking was performed using GOLD Software (Genetic Optimization Ligand Docking). GOLD uses a genetic algorithm to explore the full range of ligand conformational flexibility with partial flexibility of the protein [31] . Docking procedure consisted of three interrelated components; a) identification of binding site b) a search algorithm to effectively sample the search space (the set of possible ligand positions and conformations on the protein surface) and c) a scoring function. The GOLD fitness function consisted of four components: a) proteinligand hydrogen bond energy (external H-bond); b) protein-ligand van der Waals (vdw) energy (external vdw); c) ligand internal vdw energy (internal vdw); d) ligand torsional strain energy (internal torsion). Standard default settings, consisting of population size-100, number of islands-5, selection pressure-1.1, niche size-2, migrate-10, cross over-95, number of operations-100,000, number of dockings-10, were adopted for GOLD docking. For each ligand-1Aadrenoceptor binding, 10 docking conformations (poses) were tested and the best GOLD score was selected for studies. The consensus scoring function XSCORE was used to estimate the binding affinity of the 1A-adrenoceptor with its ligands NA, its agonist methoxamine as well as antagonist prazosin and WB4101 [32] . SILVER was used to predict the interactions of 1A-adrenoceptor and ligand complex [31] .  In silico Site-directed Mutagenesis: In silico sitedirected mutagenesis has been widely used to identify the critical residue(s) for binding of a ligand; NA with 1A-adrenoceptor, in this case. The following in silico mutations were performed using Swiss-Pdb Viewer Software [33] and minimized the structure with CHARMM force field [28] : a) polar amino acid(s) was replaced with polar amino acid(s); b) non-polar group replaced with non-polar group; c) polar group was replaced with non-polar amino acid(s); d) non-polar group replaced with polar amino acid(s); either one amino acid was replaced at a time or a set of amino acids were mutated simultaneously.

RESULTS AND DISCUSSION
Procheck evaluation: The sequence information of rat 1A-adrenoceptor was described in the section materials and methods. Secondary structure assignment using STRIDE provided the physical features of the modeled structure (protein), such as helix, coil as well as extended strands (Fig. 2). We found a proline residue, known to introduce kink, in the middle of the 5 th , 6 th and 7 th transmembrane helices that might alter the helicity of the bundle. To avoid error we used appropriate parameters [34,35] so that helix bundle geometry was properly represented. The output of hydropathy analysis of 1A-adrenoceptor was comparable to the results of secondary structure assignment. The 3D structural model of 1Aadrenoceptor was obtained as described in the materials and methods section. The stereo-chemical quality of the modeled 1A-adrenoceptor was estimated by Procheck. The phi/psi angles of 81.7% residues fell in the most favored regions, 13.2% residues lied in the additional allowed regions and 2.9% fell in the generously allowed regions; only 2.2% of residues lied in the disallowed conformations (Fig. 3). Thus, statistical analysis suggests that the backbone conformation of our predicted model of 1A-adrenoceptor was almost as good as that of the template; the 3D conformation of the predicted model of 1A-adrenoceptor has been shown in Fig. 4. The weighted RMSD of C α trace between the template (1L9H) and final refined model of 1Aadrenoceptor was 1.5Å with a significant Z-score of 6.5.

Protein model deposition:
The atomic coordinates of the theoretical model of 1A-adrenoceptor have been deposited with RCSB Protein Data Bank [36] , which can be accessed with the code 2F75.
Gold docking: One of the most appropriate methods to predict structural features of active site of a molecule is through docking studies with selected ligands. Therefore, once a theoretical model of the 1Aadrenoceptor was obtained, its active site was predicted and characterized in silico by docking pharmacologically confirmed analogues/agonists (adrenaline, noradrenaline and methoxamine) and antagonists (Prazosin and WB4101) with the receptor.
For such prediction we considered a cavity size of 10 Å around SER188, because it (SER188) has been reported to be the key residue for binding with adrenaline, a methylated product of NA that also binds to 1A-adrenoceptor with comparable affinity [37] . Such interactions enabled us to predict the following interactive residues at the active site in our modeled 1A-adrenoceptor, ASP106, ARG166, ILE178, ASN179, TRY184, SER188 and SER192 and the details are explained below. Interactions with adrenaline: The following 4 hydrogen bond interactions were observed between adrenaline and 1A-adrenoceptor (Fig. 5a)   Interactions with noradrenaline: There are the following 6 hydrogen bond interactions between NA and 1A-adrenoceptor (Fig. 5b) (Fig. 5c) (Table 1).

Interactions with prazosin:
There was only one interaction between non-polar residue O atom of ILE178 that forms hydrogen bond with O26 of prazosin (Fig. 5d) (Table 2). This scoring function included terms accounting for van der Waals interaction, hydrogen bonding, deformation penalty and hydrophobic effect. The XSCORE performs better in identifying the correct bound conformations used for docking analysis. We evaluated top three docking conformations using GOLD score and XSCORE for all the ligands with the modeled structure (Table 3), however, as a normal practice, we have taken the best gold score for further studies.
In silico mutagenesis: The hydrogen bond residues mentioned above were present in the cavity that formed active site of the 1A-adrenoceptor. Hence, to further explore that which of those residues was critical for binding of NA and its agonist/antagonist, we performed in silico mutation of the active site residues and then studied the binding affinities of the mutated active site with NA and its agonist as well as antagonist mentioned above (Table 4).
In an earlier in situ mutation study [37] , SER 188 and SER192 of 1A-adrenoceptor were mutated and their interactions with adrenaline, a methylated product of NA, was studied. Such study identified SER188 as the key residue for binding with adrenaline. To start with we evaluated whether the same residue of the receptor was playing a crucial role for binding with NA and known specific pharmacological agonist methoxamine and known specific antagonist prazosin as well as WB4101, we mutated either SER188 with ALA 188 or, SER192 with ALA192 or both simultaneously (SER188 and SER192 with ALA188 and ALA192). It was observed that there was no change in the cavity and binding affinity of NA, methoxamine and prazosin.
Thereafter, we replaced all four polar amino acid residues of 1A-adrenoceptor (ASP106, ARG166, ASN179, TYR184) with four non-polar amino acid residues (VAL106, ALA166, LEU179, GLY184); all simultaneously or one at a time. Solvent Accessibility Surface area (SAS) was calculated for natural 1Aadrenoceptor (without mutation) and that after each of such in silico mutation mentioned above ( Table 5). The results showed that the size of the cavity and binding affinity with NA, methoxamine and prazosin were significantly decreased. Subsequently, we studied the effects of replacing one amino acid at a time of the active site of the modeled receptor on SAS and binding affinities with NA, methoxamine, prazosin and WB4101. Replacement of polar residue ASP106 with a non-polar residue VAL106 did not significantly change the cavity size and binding affinity of 1Aadrenoceptor with NA, methoxamine and prazosin. Similarly, upon replacement of the polar residue ASN179 with a non-polar residue LEU179, there was no change in the cavity size and binding affinity of NA, methoxamine and prazosin with 1A-adrenoceptor.
Substitution of the non-polar residue ILE178 with a polar residue THR178 increased the cavity size but there was no change in the binding affinity of NA and methoxamine, however, the prazosin binding affinity with 1A-adrenoceptor was decreased. Mutation of the polar residue ARG166 with a non-polar residue ALA166 increased the cavity size and the binding affinities of NA, methoxamine and prazosin were significantly decreased with 1A-adrenoceptor. Substitution of the polar residue TYR184 with a nonpolar residue GLY184 increased the cavity size but the binding affinities of NA and methoxamine were not significantly altered, however, the prazosin binding affinity with the receptor was decreased. These data suggest that ARG166 is the most crucial residue for binding of NA and methoxamine to 1A-adrenoceptor; however, not for prazosin because ARG166 is not its interacting residue. Our prediction may be supported by the fact that the ARG166 is conserved in all the four mammalian species 1A-adrenoceptor whose sequence we compared (Fig. 6). The sequence alignment was derived using CLUSTAL W package [38] . Further, ILE178 is the most important residue for binding of prazosin to 1A-adrenoceptor. However, at present we cannot comment on why in all the mutations the binding affinities of WB4101 to 1A-adrenoceptor increased. Finally, a word of caution that our model does not account for the portions of the sequence of the 1A-adrenoceptor, which was not present in the template bovine rhodopsin. Nevertheless, as an indirect evidence, we calculated the distance from the active site (ARG166) to N-(29.78 Å) and C-(46.00 Å) terminals of the modeled receptor and the length of the NA Table 4: In silico mutation of 1A-adrenoceptor and their interactions with NA, its agonist methoxamine and antagonist prazosin and WB4101.
GOLD score and XSCORE were used to calculate the binding affinities of normal and mutated 1A adrenoceptor Mutation has been shown under substitution as follows: SER188 to ALA 188=S188A; SER192 to ALA192=S192A; SER188 to ALA188/SER192 to ALA192= S188A/S192A; ASP 106 to VAL106=D106V; ARG166 to ALA166=R166A; ASN179 to LEU179=N179L; TYR184 to GLY184=Y184G; ILE178 to THR178=I178T  4. Since to our knowledge there is no biochemical data available to confirm if the non-modeled portion of the receptor is not necessary for NA binding, based on the distance, it is expected that indeed those regions are unlikely to be directly participating for NA binding to 1A-adrenoceptor. Also, we are aware that while mutation, we have considered the polar and non-polar properties of the amino acids; their biological potency and whether they can really be synthesized are unknown hence, our prediction are subject to in vivo confirmation.

CONCLUSION
The structure of 1A-adrenoceptor was modeled in silico based on X-ray crystallography structure of bovine rhodopsin (1l9h) taken as a template.
The strong correlation between the GOLD score and XSCORE values add confidence to the accuracy of the model. The validity of the refined modeled structure was evaluated by studying its molecular docking affinities with known pharmacological agonist and antagonists before and after in silico site-directed mutagenesis of the active site of the model. The findings provide new information about the intra-and inter-molecular interactions characterizing the 1Aadrenoceptor-NA complex. Our analysis provides strong evidence that ARG166 is the most crucial residue for binding of NA and methoxamine to 1Aadrenoceptor and ILE178 is the most important residue for binding of prazosin to 1A-adrenoceptor. These findings advance our knowledge on specific interactions on NA binding with 1A-adrenoceptor, which subject to confirmation in vivo may facilitate designing of related therapeutics.