**4. Role of homology modeling in unraveling the structures of GPCRs: a success story**

Protein based virtual screening requires knowledge of three dimensional structure of targets. Researchers will have to face an overwhelming number of potential targets like GPCRs for which no or very few experimental 3-D information is available. Therefore, it is crucial in the near future to be able to use not only X-ray or NMR structures, but also GPCR models for protein-based virtual screening of chemical libraries. There are a lot of difficulties in obtaining significant amounts of pure and active recombinant GPCRs and this has been a huge problem in generating a lot of high resolution three dimensional structures of GPCRs. Low resolution GPCR structures of either bacteriorhodopsin or Bovine rhodopsin have paved way for many GPCR models. These models proved to be ineffective as they were not reliable enough for structure-based ligand design. The solving of crystal structure of the inactive dark-state rhodopsin back in the year 2001 was a huge mile stone in the structural study of GPCRs as a number of homology models of other class A GPCRs have been reported since then based on this structure.

Generally the importance of crystal structures is that they are useful to map sequence differences and to help analyze if the ortholog variant may affect the ligand binding and signaling of that particular GPCR. The first prerequisite for experimentally solving a protein structure is obtaining large amounts of stable, purified, homogeneous protein which can be used as templates to build a homology model. By means of in silico methods like homology modeling, crystal structures can be used to predict the effect of such ortholog variants. First step in developing homology models is the alignment of fingerprint motifs that are common among the family which are then are extrapolated to assign coordinates for the entire helical bundle. On the basis of databases of loop conformations and based on the specific application loop regions are either ignored or modeled accordingly [35]. As the template and query sequences used in homology modeling both belong to the GPCR family, the seven transmembrane (TM) helixes were properly transformed in the models according to that of the template structure. The RMSD between the model and the template structure must always preferably be less than range of 3 Å. Further the models were validated with the help of ERRAT plot [36], PROCHECK [37] and VERIFY3D [38].

One of the test case wherein homology modeling proved to be effective with the structural studies of GPCRs is the work done by Bissantz et al. where 3-D models of the D3, β2, and δ-opioid receptors were generated for future agonist screening as already several full agonists were known for each of these GPCRs [39]. Many GPCR models were set up to speculate if the "activated state" of GPCRs was conformationally more flexible than the antagonist-bound ground state. Apomorphine and pergolide (D3 receptor), epinephrine, and nylidrine (2 receptor), SNC-80, and TAN67 (−opioid receptor) were the agonists used for the refinement and two agonist-bound models were built for each receptor. An alternative activated-state model was also generated by substituting the single ligand-biased receptor to do comparative studies. When the amino acid sequences of the target receptors were aligned to the sequence of the Bovine rhodopsin template, the alignment coincided with the known structural features of GPCRs. It was observed that despite the low sequence identity when taking the whole TM sequence in account, the structurally

**61**

identity of 28%) [40].

*Importance of Homology Modeling for Predicting the Structures of GPCRs*

and functionally important amino acids were highly conserved or compensated by amino acids of high similarity. These GPCR models are static though proteins are in reality more or less flexible which gave rise to more problems associated with docking. GPCR models based on a template with an identity of 20–30% can be expected to be of higher accuracy than when modeling other type of proteins based on a template with low-sequence identity and this was proven to be rue in this test case where in the antagonist-bound state models of three human GPCRs were proven to be suitable for virtual screening of GPCR antagonists. Although single template based models were seen to be less reliable. This was because all GPCR models that were used as templates have been derived from the inactive state of Bovine rhodospin, which was closer to an "antagonist-bound state" than to an "agonist-bound state" of the target GPCR and though their active site can be expanded the following conformational changes occurring in the receptor activation process could not be stimulated. A similar unreliability with the single template model was observed with all the GPCR homology models developed based on β2AR. Many models exist for β2AR, some of which have been improved upon with supporting biochemical data. All of these models were more similar to rhodopsin than β2AR. This was mainly because they were all homology models generated from single structural templates. The addition of multiple structural templates and conformational states to the pool of information on GPCRs later paved way to a new generation of more potent therapeutics targeting GPCR family. It is also not conclusive to come to a judgment where this unreliability of single template modeling stands strong as these modeling were conducted at a time when there were only few templates available. Judith Varady et al. used Bovine rhodopsin template to build the model of dopamine 3 (D3) subtype receptor which is a promising lead in treating drug addictions. The transmembrane helical region of the D3 receptor was modeled using Bovine rhodopsin template includes the ligand-binding site and showed sequence identity in the twilight region during homology modeling (sequence

Three-dimensional model of the human CCR5 receptor was developed by Fano A. et al. using a homology-based approach starting from the X-ray structure of the bovine rhodopsin receptor [41]. The reliability of these models was accessed using molecular docking and molecular dynamics studies. During this work there was no experimentally solved three dimensional chemokine receptor structures available and hence became a major hurdle in the deeper researches on the structural properties of these receptors. Therefore main ways to investigate the properties of CCR5 were homology modeling studies along with site-directed mutagenesis (SDM). Therefore a new model of CCR5 was built after consolidating all the information from the previously built models and also incorporating extensive molecular dynamics simulations (MD). Furthermore, flexible docking of a synthetic antagonist TAK779 and a novel docking protocol for natural agonists RANTES and MIP-1β was employed to develop the CCR5 models. The first crystal structure of bovine rhodopsin by Palczewski et al. served as the perfect template to build this model as the sequence identity increased to ∼30% from previously being less than 20% when considering only the transmembrane helices (TMHs), and several of the amino acid residues essential for maintaining CCR5's architecture and receptor function were highly conserved. Pair wise alignment between the template and human CCR5 was carried out in CLUSTAL W and it was found that the anti-parallel β sheet loop of the second extracellular loop (ECL2) had higher sequence homology to the template. Out of the four cysteines which form two disulfide links in CCR5, Cys101-Cys178 had the anti-parallel β sheet loop of ECL2 and thus this loop was constructed by homology from the template structure using MODELER 6.2 [42]. The Cα Cartesian coordinates of the seven transmembrane helices and ECL2 were

*DOI: http://dx.doi.org/10.5772/intechopen.94402*

#### *Importance of Homology Modeling for Predicting the Structures of GPCRs DOI: http://dx.doi.org/10.5772/intechopen.94402*

*Homology Molecular Modeling - Perspectives and Applications*

unturned page in the global research of GPCRs.

reported since then based on this structure.

[37] and VERIFY3D [38].

**a success story**

structures of GPCRs solved every year as recorded by GPCR-EXP database. There are still many more GPCR structures that are yet to be solved and these remain as an

**4. Role of homology modeling in unraveling the structures of GPCRs:** 

Protein based virtual screening requires knowledge of three dimensional structure of targets. Researchers will have to face an overwhelming number of potential targets like GPCRs for which no or very few experimental 3-D information is available. Therefore, it is crucial in the near future to be able to use not only X-ray or NMR structures, but also GPCR models for protein-based virtual screening of chemical libraries. There are a lot of difficulties in obtaining significant amounts of pure and active recombinant GPCRs and this has been a huge problem in generating a lot of high resolution three dimensional structures of GPCRs. Low resolution GPCR structures of either bacteriorhodopsin or Bovine rhodopsin have paved way for many GPCR models. These models proved to be ineffective as they were not reliable enough for structure-based ligand design. The solving of crystal structure of the inactive dark-state rhodopsin back in the year 2001 was a huge mile stone in the structural study of GPCRs as a number of homology models of other class A GPCRs have been

Generally the importance of crystal structures is that they are useful to map sequence differences and to help analyze if the ortholog variant may affect the ligand binding and signaling of that particular GPCR. The first prerequisite for experimentally solving a protein structure is obtaining large amounts of stable, purified, homogeneous protein which can be used as templates to build a homology model. By means of in silico methods like homology modeling, crystal structures can be used to predict the effect of such ortholog variants. First step in developing homology models is the alignment of fingerprint motifs that are common among the family which are then are extrapolated to assign coordinates for the entire helical bundle. On the basis of databases of loop conformations and based on the specific application loop regions are either ignored or modeled accordingly [35]. As the template and query sequences used in homology modeling both belong to the GPCR family, the seven transmembrane (TM) helixes were properly transformed in the models according to that of the template structure. The RMSD between the model and the template structure must always preferably be less than range of 3 Å. Further the models were validated with the help of ERRAT plot [36], PROCHECK

One of the test case wherein homology modeling proved to be effective with the structural studies of GPCRs is the work done by Bissantz et al. where 3-D models of the D3, β2, and δ-opioid receptors were generated for future agonist screening as already several full agonists were known for each of these GPCRs [39]. Many GPCR models were set up to speculate if the "activated state" of GPCRs was conformationally more flexible than the antagonist-bound ground state. Apomorphine and pergolide (D3 receptor), epinephrine, and nylidrine (2 receptor), SNC-80, and TAN67 (−opioid receptor) were the agonists used for the refinement and two agonist-bound models were built for each receptor. An alternative activated-state model was also generated by substituting the single ligand-biased receptor to do comparative studies. When the amino acid sequences of the target receptors were aligned to the sequence of the Bovine rhodopsin template, the alignment coincided with the known structural features of GPCRs. It was observed that despite the low sequence identity when taking the whole TM sequence in account, the structurally

**60**

and functionally important amino acids were highly conserved or compensated by amino acids of high similarity. These GPCR models are static though proteins are in reality more or less flexible which gave rise to more problems associated with docking. GPCR models based on a template with an identity of 20–30% can be expected to be of higher accuracy than when modeling other type of proteins based on a template with low-sequence identity and this was proven to be rue in this test case where in the antagonist-bound state models of three human GPCRs were proven to be suitable for virtual screening of GPCR antagonists. Although single template based models were seen to be less reliable. This was because all GPCR models that were used as templates have been derived from the inactive state of Bovine rhodospin, which was closer to an "antagonist-bound state" than to an "agonist-bound state" of the target GPCR and though their active site can be expanded the following conformational changes occurring in the receptor activation process could not be stimulated. A similar unreliability with the single template model was observed with all the GPCR homology models developed based on β2AR. Many models exist for β2AR, some of which have been improved upon with supporting biochemical data. All of these models were more similar to rhodopsin than β2AR. This was mainly because they were all homology models generated from single structural templates. The addition of multiple structural templates and conformational states to the pool of information on GPCRs later paved way to a new generation of more potent therapeutics targeting GPCR family. It is also not conclusive to come to a judgment where this unreliability of single template modeling stands strong as these modeling were conducted at a time when there were only few templates available. Judith Varady et al. used Bovine rhodopsin template to build the model of dopamine 3 (D3) subtype receptor which is a promising lead in treating drug addictions. The transmembrane helical region of the D3 receptor was modeled using Bovine rhodopsin template includes the ligand-binding site and showed sequence identity in the twilight region during homology modeling (sequence identity of 28%) [40].

Three-dimensional model of the human CCR5 receptor was developed by Fano A. et al. using a homology-based approach starting from the X-ray structure of the bovine rhodopsin receptor [41]. The reliability of these models was accessed using molecular docking and molecular dynamics studies. During this work there was no experimentally solved three dimensional chemokine receptor structures available and hence became a major hurdle in the deeper researches on the structural properties of these receptors. Therefore main ways to investigate the properties of CCR5 were homology modeling studies along with site-directed mutagenesis (SDM). Therefore a new model of CCR5 was built after consolidating all the information from the previously built models and also incorporating extensive molecular dynamics simulations (MD). Furthermore, flexible docking of a synthetic antagonist TAK779 and a novel docking protocol for natural agonists RANTES and MIP-1β was employed to develop the CCR5 models. The first crystal structure of bovine rhodopsin by Palczewski et al. served as the perfect template to build this model as the sequence identity increased to ∼30% from previously being less than 20% when considering only the transmembrane helices (TMHs), and several of the amino acid residues essential for maintaining CCR5's architecture and receptor function were highly conserved. Pair wise alignment between the template and human CCR5 was carried out in CLUSTAL W and it was found that the anti-parallel β sheet loop of the second extracellular loop (ECL2) had higher sequence homology to the template. Out of the four cysteines which form two disulfide links in CCR5, Cys101-Cys178 had the anti-parallel β sheet loop of ECL2 and thus this loop was constructed by homology from the template structure using MODELER 6.2 [42]. The Cα Cartesian coordinates of the seven transmembrane helices and ECL2 were

copied from the corresponding template (PDB: 1F88) and the N-terminal domain and the remaining loops were built de novo using MODELER 6.2. Confirmations of the models were done using PROCHECK and were selected as the input structure for MD Loop Refinement. The resulting model consisting of the TMHs and all ECLs and ICLs, was validated by MD conformational analysis, which showed it to be consistent with the then currently available SDM data and was used to gain insights into the molecular basis of the initiation and development of HIV-1 infection. This information could be useful in the rational design of HIV-1 entry blockers.

Chronologically, the time when the structural information about chemokine receptors was unavailable Gugan et al. in 2012 carried out the investigations on the binding site of CCR2 [43]. A comparative model was generated using the template structure of CXCR4 (PDB ID: 3ODU [44]). The structure of CXCR4 (PDB ID: 3ODU) was elucidated in 2010. One of the key findings along with the binding site residues is that the disulfide bridge was produced between Cys113-Cys190 of the selected CCR2 model and was also later observed in the crystal structure which was elucidated in 2016 (PDB ID: 5T1A [45]).

In the similar manner, Changdev et al. in 2013 developed the 3-D model for CCR5 using the template CXCR4 (PDB ID: 3ODU; resolution 2.5 Å) modeled by MODELER 9.2 [46] to explore the biding site of the receptor [47]. Significantly, the modeled structure coincides with the crystal structure of the CCR5 (PDB ID:4MBS [20]) whose structural information was determined by Tan et al. in 2013 [48].

The research by Anand et al. in 2011 on the accuracy of homology modeling revealed the comparison study between the reported models along with the crystal structure of CCR5 (PDB ID:4MBS)[49]. The findings have identified the importance of multi-template model in determining the insights of structural information of the receptor possessing its own merits and demerits. The inhibitor Maraviroc was docked to the single template and multi-template models of bovine rhodopsin (PDB ID: 1F88), β2 adrenergic receptor (PDB ID: 2RH1 [50]) and CXCR4 (PDB ID: 3ODU). The critical salt-bridge interaction established by Maraviroc with Glu283 of the receptor was genuinely observed in modeled structure and crystal structure.

In the process of building model of a particular GPCR usually many models are constructed with varying side chains and almost identical backbone. This is done to check which model among all the constructed models shows maximum affinity towards various ligands. So the model showing consisting binding mode is selected for further analysis. An example of this is the study done by Mateusz N et al. where 400 homology models of serotonin 5-HT1A receptor, one of the most documented monoamine GPCR, was modeled using Modeler 7v7 [51] with the crystal structure of bovine rhodopsin (PDB:1F88) as template [52]. These models varied considerably in their side chain but the polypeptide backbone varied only marginally from the template. Arylpiperazines test ligands were docked to all the 400 models with default parameters without any constraints. A detailed analysis of the docking poses revealed intrinsic information about crucial ionic bonds that were formed d almost exclusively in the case of receptors with the gauche(−) conformation of the Asp3.32 ø1 angle. Such insights led to the development of 200 new homology models with all the changes incorporated. Molecular docking was once again done on all the 200 new models and the complexes were scored using various scoring functions to choose the best models.

The past few years have seen remarkable advances in the structural biology of G-protein coupled receptors (GPCRs) and separate databases exist to study GPCRs. The applications of structural studies of GPCRs have various goals and these goals trigger myriad scientific investigations. For the GPCRs whose structures have now been solved, the homology models developed earlier based on rhodopsin, have been the first step in discovering the versatility of their structural studies. Due to the

**63**

**Author details**

**5. Conclusion**

Ananthasri Sailapathi1

**Conflict of interest**

University of Madras, Chennai, India

provided the original work is properly cited.

Gugan Kothandan1

, Seshan Gunalan1

2 Department of Microbiology, Assam University, Assam, India

\*Address all correspondence to: drgugank@gmail.com

1 Biopolymer Modeling Laboratory, CAS in Crystallography and Biophysics,

© 2020 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,

\* and Diwakar Kumar2

The authors declare no conflict of interest.

, Kanagasabai Somarathinam1

,

*Importance of Homology Modeling for Predicting the Structures of GPCRs*

being constructed with the upcoming elucidated structures of GPCRs.

increase in the available GPCR structures, the templates used to build the structure for homology modeled GPCRs show a drastic increase in similarity and query coverage in the recent years. This enhances the structure of the models which are

The research in GPCRs is a global phenomenon and this is possible only if we have structural insights based on structural studies of GPCRs. Owing to the difficulty in crystallizing the GPCRs, it was once construed that structural studies of GPCRs were impossible. But with the technological advancements in the computational techniques, building a model structure based on the homology of a particular receptor with a template structure became possible. Thus homology modeling and models generated via tools like MODELER unraveled the unexplored arenas in the research of GPCRs. These models served a greater purpose to the pharmaceutical industries wherein GPCRs became famous drug discovery targets. The many experimental structures constructed using previously solved structures as templates were further scrutinized based on their efficiency in showing a consisting binding mode with various ligands. Recent times have seen use of Cryo EM techniques in solving structures of GPCRs. But still contribution made by techniques like homology modeling in the structural studies of GPCRs will always remain as a mile stone.

*DOI: http://dx.doi.org/10.5772/intechopen.94402*

increase in the available GPCR structures, the templates used to build the structure for homology modeled GPCRs show a drastic increase in similarity and query coverage in the recent years. This enhances the structure of the models which are being constructed with the upcoming elucidated structures of GPCRs.
