Limits...
Rapid Computational Prediction of Thermostabilizing Mutations for G Protein-Coupled Receptors.

Bhattacharya S, Lee S, Grisshammer R, Tate CG, Vaidehi N - J Chem Theory Comput (2014)

Bottom Line: Deriving thermostable mutants has been a successful strategy to stabilize GPCRs in detergents, but this process is experimentally tedious.The prediction using the stability score improves when using an ensemble of receptor conformations compared to a single structure, showing that receptor flexibility is important.We improved the thermostability prediction by including other properties such as residue-based stress and the extent of allosteric communication by each residue in the stability score.

View Article: PubMed Central - PubMed

Affiliation: Division of Immunology, Beckman Research Institute of the City of Hope , 1500 East Duarte Rd, Duarte, California 91010, United States.

ABSTRACT
G protein-coupled receptors (GPCRs) are highly dynamic and often denature when extracted in detergents. Deriving thermostable mutants has been a successful strategy to stabilize GPCRs in detergents, but this process is experimentally tedious. We have developed a computational method to predict the position of the thermostabilizing mutations for a given GPCR sequence. We have validated the method against experimentally measured thermostability data for single mutants of the β1-adrenergic receptor (β1AR), adenosine A2A receptor (A2AR) and neurotensin receptor 1 (NTSR1). To make these predictions we started from homology models of these receptors of varying accuracies and generated an ensemble of conformations by sampling the rigid body degrees of freedom of transmembrane helices. Then, an all-atom force field function was used to calculate the enthalpy gain, known as the "stability score" upon mutation of every residue, in these receptor structures, to alanine. For all three receptors, β1AR, A2AR, and NTSR1, we observed that mutations of hydrophobic residues in the transmembrane domain to alanine that have high stability scores correlate with high experimental thermostability. The prediction using the stability score improves when using an ensemble of receptor conformations compared to a single structure, showing that receptor flexibility is important. We also find that our previously developed LITiCon method for generating conformation ensembles is similar in performance to predictions using ensembles obtained from microseconds of molecular dynamics simulations (which is computationally hundred times slower than LITiCon). We improved the thermostability prediction by including other properties such as residue-based stress and the extent of allosteric communication by each residue in the stability score. Our method is the first step toward a computational method for rapid prediction of thermostable mutants of GPCRs.

No MeSH data available.


Related in: MedlinePlus

Comparison of enrichments for different receptor models. The nomenclaturesare as follows: β1AR homo-D3DR impliesβ1AR homology model using D3DR as template;A2AR homo-β1AR implies A2AR homology modelbased on β1AR as template; the other models are namedaccordingly. The RMSD of the homology models from their respectivecrystal structures are tabulated in SupportingInformation Table S3.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4230369&req=5

fig4: Comparison of enrichments for different receptor models. The nomenclaturesare as follows: β1AR homo-D3DR impliesβ1AR homology model using D3DR as template;A2AR homo-β1AR implies A2AR homology modelbased on β1AR as template; the other models are namedaccordingly. The RMSD of the homology models from their respectivecrystal structures are tabulated in SupportingInformation Table S3.

Mentions: In the last section, we showed that theLITiConDesign method showed good predictability for thermostable mutationsstarting from GPCR crystal structures. However, the real case scenariois that the crystal structure of the GPCR to be thermostabilized willnot be known. In this section, we discuss the thermostability predictionresults derived using homology models for β1AR, A2AR, and NTSR1 without using any crystal structure informationon these receptors. Figure 3 shows a comparisonof the enrichment as a function of number of mutants for several homologymodels of β1AR, A2AR, and NTSR1. To observehow the accuracy of the homology model affects the predictabilityof thermostability, we selected several template structures of varyingsequence similarity to β1AR, namely β2AR, dopamine receptor D3DR (67% and 40% sequence similarity),A2AR (33% sequence similarity), and CXCR4 (24% sequencesimilarity). For A2AR and NTSR1, we evaluated one homologymodel each, using β1AR and β2ARas templates, respectively. Unlike β1AR, A2AR and NTSR1 have no close template crystal structures. Therefore,we wanted to test how generic templates such as β1AR or β2AR perform in thermostability prediction.Since the NTSR1 homology model was based on the inactive state ofβ2AR, we evaluated its predictability using the −NTdata. In contrast, for evaluating thermostability prediction usingthe NTSR1 crystal structure, we had used the +NT data, since the crystalstructure shows active state characteristics. We also plotted theenrichments corresponding to 50% cutoff for each model against theCα RMSD in the coordinates of the homology modelsfrom the respective crystal structures in Figure 4. While the β1AR models based on β2AR (RMSD of the β1AR to its crystal structureis 1.6 Å) and D3DR (RMSD 2.5 Å) performed thebest among homology models, the CXCR4 based model (RMSD of the β1AR model to its crystal structure is 3.1 Å) performedthe worst because of its distant sequence and structural homologywith β1AR. For A2AR, the performance ofthe β1AR based homology model is worse compared tothe A2AR crystal structure (65% vs 70%). Unlike β1AR, there are no other templates that are closer in sequencesimilarity to A2AR, for which crystal structures are available(e.g., adenosine receptors other than A2AR). Therefore,we could not test the effect of template closeness on prediction performancefor A2AR. The RMSD in coordinates of the various homologymodels from their respective crystal structures and their percentagerecovery of the thermostable mutants are tabulated in Table S3 ofthe Supporting Information. Overall, wefind that for homology models that are within 2.5 Å (RMSD ofthe Cα atoms in the TM region) to their respectivecrystal structures, there is little change in performance of the thermostabilitypredictions. The performance of the thermostability prediction isworse if the homology model has RMSD above 3 Å to its crystalstructure.


Rapid Computational Prediction of Thermostabilizing Mutations for G Protein-Coupled Receptors.

Bhattacharya S, Lee S, Grisshammer R, Tate CG, Vaidehi N - J Chem Theory Comput (2014)

Comparison of enrichments for different receptor models. The nomenclaturesare as follows: β1AR homo-D3DR impliesβ1AR homology model using D3DR as template;A2AR homo-β1AR implies A2AR homology modelbased on β1AR as template; the other models are namedaccordingly. The RMSD of the homology models from their respectivecrystal structures are tabulated in SupportingInformation Table S3.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4230369&req=5

fig4: Comparison of enrichments for different receptor models. The nomenclaturesare as follows: β1AR homo-D3DR impliesβ1AR homology model using D3DR as template;A2AR homo-β1AR implies A2AR homology modelbased on β1AR as template; the other models are namedaccordingly. The RMSD of the homology models from their respectivecrystal structures are tabulated in SupportingInformation Table S3.
Mentions: In the last section, we showed that theLITiConDesign method showed good predictability for thermostable mutationsstarting from GPCR crystal structures. However, the real case scenariois that the crystal structure of the GPCR to be thermostabilized willnot be known. In this section, we discuss the thermostability predictionresults derived using homology models for β1AR, A2AR, and NTSR1 without using any crystal structure informationon these receptors. Figure 3 shows a comparisonof the enrichment as a function of number of mutants for several homologymodels of β1AR, A2AR, and NTSR1. To observehow the accuracy of the homology model affects the predictabilityof thermostability, we selected several template structures of varyingsequence similarity to β1AR, namely β2AR, dopamine receptor D3DR (67% and 40% sequence similarity),A2AR (33% sequence similarity), and CXCR4 (24% sequencesimilarity). For A2AR and NTSR1, we evaluated one homologymodel each, using β1AR and β2ARas templates, respectively. Unlike β1AR, A2AR and NTSR1 have no close template crystal structures. Therefore,we wanted to test how generic templates such as β1AR or β2AR perform in thermostability prediction.Since the NTSR1 homology model was based on the inactive state ofβ2AR, we evaluated its predictability using the −NTdata. In contrast, for evaluating thermostability prediction usingthe NTSR1 crystal structure, we had used the +NT data, since the crystalstructure shows active state characteristics. We also plotted theenrichments corresponding to 50% cutoff for each model against theCα RMSD in the coordinates of the homology modelsfrom the respective crystal structures in Figure 4. While the β1AR models based on β2AR (RMSD of the β1AR to its crystal structureis 1.6 Å) and D3DR (RMSD 2.5 Å) performed thebest among homology models, the CXCR4 based model (RMSD of the β1AR model to its crystal structure is 3.1 Å) performedthe worst because of its distant sequence and structural homologywith β1AR. For A2AR, the performance ofthe β1AR based homology model is worse compared tothe A2AR crystal structure (65% vs 70%). Unlike β1AR, there are no other templates that are closer in sequencesimilarity to A2AR, for which crystal structures are available(e.g., adenosine receptors other than A2AR). Therefore,we could not test the effect of template closeness on prediction performancefor A2AR. The RMSD in coordinates of the various homologymodels from their respective crystal structures and their percentagerecovery of the thermostable mutants are tabulated in Table S3 ofthe Supporting Information. Overall, wefind that for homology models that are within 2.5 Å (RMSD ofthe Cα atoms in the TM region) to their respectivecrystal structures, there is little change in performance of the thermostabilitypredictions. The performance of the thermostability prediction isworse if the homology model has RMSD above 3 Å to its crystalstructure.

Bottom Line: Deriving thermostable mutants has been a successful strategy to stabilize GPCRs in detergents, but this process is experimentally tedious.The prediction using the stability score improves when using an ensemble of receptor conformations compared to a single structure, showing that receptor flexibility is important.We improved the thermostability prediction by including other properties such as residue-based stress and the extent of allosteric communication by each residue in the stability score.

View Article: PubMed Central - PubMed

Affiliation: Division of Immunology, Beckman Research Institute of the City of Hope , 1500 East Duarte Rd, Duarte, California 91010, United States.

ABSTRACT
G protein-coupled receptors (GPCRs) are highly dynamic and often denature when extracted in detergents. Deriving thermostable mutants has been a successful strategy to stabilize GPCRs in detergents, but this process is experimentally tedious. We have developed a computational method to predict the position of the thermostabilizing mutations for a given GPCR sequence. We have validated the method against experimentally measured thermostability data for single mutants of the β1-adrenergic receptor (β1AR), adenosine A2A receptor (A2AR) and neurotensin receptor 1 (NTSR1). To make these predictions we started from homology models of these receptors of varying accuracies and generated an ensemble of conformations by sampling the rigid body degrees of freedom of transmembrane helices. Then, an all-atom force field function was used to calculate the enthalpy gain, known as the "stability score" upon mutation of every residue, in these receptor structures, to alanine. For all three receptors, β1AR, A2AR, and NTSR1, we observed that mutations of hydrophobic residues in the transmembrane domain to alanine that have high stability scores correlate with high experimental thermostability. The prediction using the stability score improves when using an ensemble of receptor conformations compared to a single structure, showing that receptor flexibility is important. We also find that our previously developed LITiCon method for generating conformation ensembles is similar in performance to predictions using ensembles obtained from microseconds of molecular dynamics simulations (which is computationally hundred times slower than LITiCon). We improved the thermostability prediction by including other properties such as residue-based stress and the extent of allosteric communication by each residue in the stability score. Our method is the first step toward a computational method for rapid prediction of thermostable mutants of GPCRs.

No MeSH data available.


Related in: MedlinePlus