Structure Prediction of the Prostaglandin E2 Receptor Stephen Dutz Courtney Edwards August 23, 2007 Advisor: William A Goddard III California Institute of Technology

22 Outline Biological Significance Project Objectives Methodology and Progress Future Work

3 Biological Significance G protein-coupled receptors (GPCRs) allow cells to sense their surroundings 3 Image courtesy of

44 Biological Significance Cont’d Bovine rhodopsin – only GPCR structure determined by x-ray techniques. Drug targets – Currently approximately 50% of marketed drugs target GPCRs – If structure is known drugs can be more efficiently designed Images courtesy:

55 Project Objectives Predict the 3-D structure of Prostaglandin E2 receptors, subtypes EP3 and EP4 Computationally test the binding between the predicted structure and known agonists/ antagonists (with their respective binding constants) Compare to Prostaglandin E2 subtype DP GPCR (conserved residues, H-bond interactions, etc.)

6 EP3 & EP4 Receptors 6 EP3 Fever generation Allergies Angiogenesis suppression EP4 Bone formation Inflammatory bowel disease Joint inflammation

7 Steps to Obtain Sequences of 7 Helices Performed a BLAST search with EP4 sequence – Resulted in 935 mammalian hits with E-value = 0.1 as a cutoff Aligned the sequences using ClustalW Used aligned sequences with “PredicTM” – Generates a prediction for each helix – Predicts hydrophobic centers based on residues

8 Hydrophobic Profile

99 Prediction of 7 helices * TM GSVSVAFPITMLLTGFVGNALAMLLVSR * TM 2f KSLYVVIVVPTTLLQGVLDTLALWGICLLFSK * TM RLCTFFGLTMTVFGLSSLFIASAMAVE * TM 4f GVGLVPLLAFALVALWVGLLVAR * TM GNLFFASAFAFLGLLALTVTFSCNLATIK * TM 6f TQNFIMKLMMILLPSWCVSLVCMIGMLQIATE * TM ECNFFLIAVRLASLNQILDPWVYLLLR * CENTER * ***Hydrophobic Centers are in bold

10 Sequences of the 7 helices are compared to a template (frog rhodopsin), resulting in a structure file Helical dynamics: – Cartesian / Neimo dynamics – Charged / neutral FFs EP4 Continued

11 Rotation Helices were individually rotated 360° in 30° increments – After each rotation side chains were adjusted to remove bad clashes – Various energies were calculated Conserved residues, and calculated energies will be analyzed

12 EP4 Structure 12

13 EP4 Future Work Combinatorial analysis Molecular Dynamics on entire system Finish build Dock ligands

14 EP3 Prediction BLAST, ClustalW, PredicTM, MD The initial structure was compared to that of the previously determined DP GPCR – A D-type receptor that preferentially binds prostaglandin PGD2 – Key conserved residues and geometric positions of helices are examined Helices were individually rotated 360° in 30° increments

15

16 TM2 Problem TM2 had an unfavorable kink that needed to be resolved. Maestro was used to manipulate φ & ψ angles to correct kink in helix – Before correction Tyr residue was sticking out of bundle, after rotation position is improved

18 Combinatorial Analysis A set of optimal rotations was selected – Comprised of the two to three most energetically favorable rotations for each helix Structure files were produced for all possible combinations of these rotations Side chains were adjusted to minimize bad clashes Energies were calculated for each structure file – Same method as in rotational analysis H-bond interactions were analyzed

19 Results of Combinatorial Analysis H1H2H3H4H5H6H7 Tot Ep + Es (kcal/mol) Es + Tot_fm (kcal/mol)

20 EP3 Final Remarks and Next Steps An optimal structure was selected based on H-bond interactions and total Energy Side chain adjustment is unfavorable – Continue with current structure Run MD on total structure Finish build Dock known agonists and antagonists – In parallel, attempt to produce a new structure in which side chain adjustment is not necessary

21 Conclusions The method presented for GPCR structure prediction is based on first principles – Has been proven most recently on the Protaglandin DP subtype receptor – Have worked in collaboration with Pharma to predict optimal agonists and antagonists for a particular GPCR Currently in clinical trials Goddard group continuing to optimize the prediction methodology Goal: To nearly automate GPCR structure prediction with the hopes of determining optimal ligands to bind the GPCRs of interest

22 Thank You! Bill Goddard, Caltech Youyong Li, Caltech Ravi Abrol, Caltech SoCalBSI faculty & students Funding: – NIH – NSF – LA/Orange County Biotechnology Center