Файл: Extracellular and Intracellular Signaling (книга).pdf

C

Figure 11.2 A Definition of the helical axis; B Table showing relative orientation parameters for the GPCRs with crystal structures; C Correlation of deviation in orientation parameters with sequence identity/similarity.

positioned on the hydrophobic plane. Two angles, y and f, specify the tilt angles of the helix and the angle Z corresponds to the helix rotation angle about its axis. The two tilt angles (y,f) and the rotation angle (Z) require a definition of the helical axis, which needs to account for the reality of bent helices as prolines are commonly found in the TM helices. We use a helical axis that corresponds to the lowest moment of inertia vector for the helix obtained by diagonalizing the moment of inertia matrix for the helix using only heavy backbone atoms.

We rotate the membrane-aligned GPCRs from the OPM database in the x–y plane such that the helical axis of TM helix (TMH) 3 goes through the origin, and that of TMH 2 intersects the x-axis. Figure 11.2B shows the relative six orientation parameters for all seven helices for the crystallized GPCRs relative to b2 adrenergic receptor and of bovine rhodopsin (cis-retinal bound form) relative to bovine opsin (the retinal free form) that is considered a conformation along the activation pathway of rhodopsin.

In order to correlate the sequence variability of these GPCRs with their helix geometries, we calculated all-to-all (across these systems) RMS (root-mean- squared) deviations in position R (x,y position in the z ¼ 0 plane), and angles y, f, Z averaged over all helices and plotted them against the corresponding sequence identity and similarity. The equations used for the deviations between

amino acids (that make up the TM helices) into the lipid bilayer.36 An implicit membrane potential of mean force has also been obtained recently for each amino acid as a function of the membrane normal using experimental structures of a-helical membrane proteins.37 The 3D-structure of these a- helical membrane proteins, to which GPCRs belong, is strongly a ected by interhelical interactions (mainly H-bonds and salt-bridges).38 An accurate structure-prediction methodology needs to be able to sample and describe these interhelical interactions very thoroughly.

Availability of a good structure (from experiment or modeling) allows for the molecular dynamics (MD) simulations to be performed on these proteins in their native lipid environment under ambient conditions. MD simulations of biomolecular systems have come of age and are contributing enormously to the understanding of their dynamical behavior. With more a ordable and more powerful computers, the dynamics of these membrane proteins can be followed in their explicit lipid environment for hundreds of nanoseconds or more. This situation is only going to improve with time, allowing for dynamics over even longer timescales of the order of microseconds. However, the conformational changes that accompany GPCR activation are known to occur on the millisecond or higher timescales, suggesting that explicit all-atom dynamics will not be able to describe these large conformational changes for some years; coarse grained simulations may do it sooner.

Many methods have been used to obtain model structures for membrane

proteins due to their pharmacological importance. These methods have been reviewed elsewhere.34,39 For GPCRs the main approach has been homology

modeling (using the X-ray structure of bovine rhodopsin as a reference until 2007, and others more recently). Because of their low homology to other GPCRs of pharmacological interest, most studies have used constraints based on mutation and binding experiments coupled to the homologous rhodopsin structure to guide additional mutation experiments. These structures have not generally been su ciently accurate for predicting binding sites of ligands. Methods are also available for predicting structures of membrane proteins in general.40

Our group has been developing de novo computational approaches (not based on homology) such as MembStruk and HierDock,41 for predicting the 3D structure of a GPCR, and its ligand binding sites. MembStruk method involved prediction of the TM regions, helix optimization based on TM regions, placement of optimized helices in a template (rhodopsin), followed by a local optimization of the helix rotations. These methods have been applied successfully to dopamine,42 adrenaline,43 muscarine,44 chemokine,45 prostaglandin DP46 and serotonin47 receptors. In all these cases, MembStruk was used to generate an ensemble of GPCR structures, out of which only one structure was carried forward for docking. After predicting the best structure for the GPCR, we used the HierDock procedure to locate the binding region and to predict the binding configuration in this region. Then we compared the predicted structures for the ligand-GPCR complex with experimental binding and mutation data and in some cases with experimental Structure-Activity- Relationship (SAR) data.

C

Figure 11.3 The predicted structure of human CCR1 bound to BX471 (from ref. 45). A Side view. B Top view. C Detailed binding site view.

As reviewed recently,48 this led to excellent results for modest-sized ligands such as dopamine and epinephrine and even for ligands such as haloperidol and domperidone. For example, Figure 11.3 shows the predicted structure for human CCR1 along with the binding site for its antagonist BX471.45 In this GPCR system, the mutagenesis studies were performed after the predictions, providing confidence in these prediction methods.

We have recently replaced the MembStruk method with the GEnSeMBLE (GPCR Ensemble of Structures in Membrane BiLayer Environment) method, which besides implementing an improved TM prediction step (PredicTM) contains a helix rotation optimization step (BiHelix) based on pairwise helix interactions that performs complete and thorough sampling of B1 billion conformations in a highly e cient way (see Figure 11.4). This method was recently applied to predict the structure of human adenosine A2A receptor and its ligand binding site.49 The structural comparison of predicted and crystal structure is shown in Figure 11.5. The method was able to predict the ligand to

˚

within 2.8 A of the crystal pose and also identified 9 out of 12 protein residues in the binding site.

11.3.2GPCR Activation: Conformation Driven Functional Selectivity

GPCRs undergo activation in response to an extracellular signal (e.g. an agonist ligand) and that information is relayed inside the cell through the coupling of the activated GPCR with cytoplasmic G proteins, which is normally followed by its coupling to b-arrestins that eventually ends in the internalization of the GPCR embedded into endosomes inside the cell. In that respect they can be called nature’s allosteric robots that enable transduction of extracellular signals across the plasma membrane.

Proteins by their very nature are dynamic macromolecules and under physiological conditions exist in an ensemble of conformations. This dynamic motion can be considered molecular ‘‘breathing’’50 and is supported by NMR data as well as computer simulations. This motion can be visualized as that of a golf ball (protein) rolling on a golf course (the potential energy surface) being randomly kicked around (due to thermal fluctuations) by a force (the temperature, solvent, etc.). The bottom of the protein’s multi-dimensional potential energy surface (corresponding to troughs on the golf course), called an energy well, can accommodate a micro-ensemble of isoenergetic and similar conformations. The greater the depth of an energy well, the more time the protein will spend in the conformations corresponding to that well. Favorable interaction with another molecule (e.g. a ligand or another protein) will change the character of the protein’s potential energy surface and the energy wells resulting in a di erent ensemble of protein conformations that would prefer binding to that other molecule. Now we will see how these ideas can be applied to GPCRs.

11.3.2.1Multi-Conformational View of GPCRs

One of the fundamental challenges in structure determination of GPCRs is their conformational flexibility. The Kobilka group has shown evidence for multiple conformational states in b2 adrenergic receptor even in the presence of a single ligand,51 consistent with the energy landscape idea presented earlier. Debra Kendall’s group has shown convincingly that a single mutation in the CB1 receptor can change the constitutively active receptor into the inactive form or the active form.52 This strongly suggests that even in the absence of any ligand, GPCRs are capable of major conformational changes with di erent functional outcomes.

These observations present a special problem in understanding structure and function of GPCRs: conformational changes in the protein structure are an essential aspect of its function. Thus, to understand GPCR activation we need to consider the multiple conformational states that the receptor can have under physiological conditions and we must consider the changes in these populations as the ligand interacts with them. This presents an enormous challenge to purely experimental structure determinations since the crystal must have all ligand-protein complexes identical, as evidenced by the fact that all currently available GPCR structures are in their inactive form (except opsin).