TIPS-1132; No. of Pages 7
Opinion
Modeling G protein-coupled receptors in complex with biased agonists Stefano Costanzi Department of Chemistry, American University, Washington, DC 20016, USA Center for Behavioral Neuroscience, American University, Washington, DC 20016, USA
The biological response to the activation of G proteincoupled receptors (GPCRs) typically originates from the simultaneous modulation of various signaling pathways that lead to distinct biological consequences. Hence, ‘biased agonists’ (i.e., compounds that selectively activate one of the pathways while blocking the others) are highly sought-after molecules to provide fine-tuned pharmacological interventions. This review describes strategies that can be deployed to model the conformation of GPCRs in complex with ligands endowed with specific signaling profiles useful for the generation of hypotheses on the structural requirements for the activation of different signaling pathways or for rational computer-aided ligand discovery campaigns. In particular, it focuses on strategies potentially applicable to model the global or local conformational states of GPCRs stabilized by specific ligands. Introduction: GPCRs and biased signaling GPCRs are integral membrane proteins that function as cellular receivers for stimuli that, in most cases, are given by extracellular molecules known as receptor agonists [1,2]. These can be endogenous compounds, for instance neurotransmitters or hormones, but also exogenous natural or artificial compounds, for instance odorants or drugs. On binding to the receptors, agonists trigger the activation of receptor-mediated signaling pathways that initiate with the interaction of intracellular signaling proteins with the domains of the receptor exposed to the cytosol. Other ligands, known as antagonists, impede the agonist-mediated activation of the receptor. Finally, ligands known as inverse agonists, besides interfering with agonists as do antagonists, suppress the constitutive activity of the receptor [3,4]. Several studies pioneered by Robert Lefkowitz demonstrated that the biological response to the activation of GPCRs typically originates from the simultaneous modulation of various signaling pathways [1–3,5–9]. Some of these are mediated by the activation of G proteins, whereas others are modulated by proteins known as arrestins. Because the activation of different pathways leads to distinct biological consequences, there is significant Corresponding author: Costanzi, S. ([email protected]). Keywords: flexible docking; global conformational changes; local conformational changes; molecular dynamics; normal mode analysis; virtual screening. 0165-6147/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tips.2014.04.004
interest in the identification of ‘biased agonists’ of GPCRs, which are agonists that selectively activate only one signaling response. As a result, biased agonists have been identified for several GPCRs [2], some examples of which include the b-adrenoceptors [8], the parathyroid hormone receptor [10], a nicotinic acid receptor (GPR109A) [11], the angiotensin receptors [12–18], the opioid receptors [19–21], and the dopamine receptors [22]. The molecular bases of biased agonisms rest on the fact that GPCRs are essentially allosteric systems: the binding of agonists to a first site stabilizes a conformation of the receptor that promotes the activation of specific intracellular signaling partners that bind to a second site. In particular, the receptors are formed by a single polypeptide chain that spans the membrane seven times, with seven ahelical transmembrane domains connected by three intracellular and three extracellular loops [23]. The structure folds to form a membrane-embedded helical bundle, with the N terminus of the chain in the extracellular milieu and the C terminus in the cytosol. For most GPCRs, the binding of agonists occurs in a cavity formed within the helical bundle and exposed to the extracellular milieu [23]. Conversely, intracellular partners bind to regions of the receptor exposed to the cytosol [24,25]. As research studies have shown, GPCRs can assume a range of conformations stabilized by different ligands and associated with different signaling states; some conformations are signaling silent, whereas others trigger the activation of signaling cascades [4,9,26–30]. Specifically, some conformations prevent the interaction of the receptor with all of its signaling partners. Conversely, other conformations promote the association with one or more intracellular signaling proteins, some of which require prior phosphorylation of specific amino acid residues of the receptor exposed to the cytosol [25]. The fact that biased agonists have been found for several systems suggests that the receptor conformations associated with the activation of different pathways are distinct; hence, one conformation can be silent with respect to a given pathway but lead to the activation of another pathway [9]. Thanks to several technical and methodological advancements, the field of GPCR structural biology is currently in full blossom and is yielding a steady output of structures solved at atomic resolution [31–36]. Each experimental structure reflects a specific conformational state of the receptor stabilized by the ligand and all of the other conditions employed for the structural determination. Hence, to gain insights into the conformational Trends in Pharmacological Sciences xx (2014) 1–7
1
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
requirements for the activation of the various signaling pathways, efforts are being made toward the solution of structures of GPCRs in complex with ligands endowed with different signaling profiles [37–39]. In particular, there is significant interest in the solution of GPCR structures in complex with biased agonists through X-ray crystallography [37,38] and toward the generation of NMR-supported hypotheses on the conformations linked to the various signaling states [39]. One of the challenges related to the characterization of the molecular mechanisms that underlie biased agonism is that some of the structures solved in complex with biased agonists, although showing receptor–ligand interactions not found with non-biased ligands, captured the receptor in its inactive conformation and do not show the conformational changes responsible for biased signaling [37]. In the absence of the experimental structure of a GPCR in the conformational state stabilized by a given ligand, a structure of the same receptor solved in a different conformation has the potential to serve as a starting point for computational experiments intended to model the conformational state of interest. In particular, efforts to model the conformation of a GPCR in complex with a specific ligand can have two distinct scopes: a broader one, intended to forecast the global conformational state of the receptor–ligand complex [40–45]; or a more circumscribed one, intended to forecast the local conformational state of the ligand-binding cavity (Figure 1) [46–61]. In this review, I discuss ways in which experimental structures of GPCRs cocrystallized with non-biased ligands could be used as platforms to generate models of the receptors in complex with biased agonists. I begin with an illustration of the published strategies that attempt to model the overall structural features of GPCRs in complex with biased agonists [40,41]. I then provide an overview of the strategy that my research group has employed to: (i) model the local conformation of the ligand-binding cavity of GPCRs stabilized by specific compounds [46]; and (ii) use such models to steer virtual screening campaigns toward the identification of ligands with desired signaling profiles [62]. We reported the application of such a procedure to the Global modeling
Receptor
Biased agonist
+
Local modeling TRENDS in Pharmacological Sciences
Figure 1. Modeling the global and local conformational state of G protein-coupled receptors (GPCRs) in complex with biased agonists. Given the experimental structure of a GPCR and a model of a given biased agonist, different modeling strategies can be applied to the generation of hypotheses on the global conformation of the resulting complex or its local conformation with respect to the ligand-binding cavity.
2
generation of models that well discriminate between agonists and blockers; although theoretically conceivable, its application to the construction of models that can distinguish between agonists endowed with different signaling profiles (e.g., agonists biased toward the exclusive activation of a single signaling pathway) remains to be demonstrated. I conclude with a view of the possible evolution of the field. The case studies that we discuss are all related to the b2-adrenoceptor (b2AR), a prototypical GPCR naturally activated by epinephrine and targeted by drugs for various indications including lung diseases [63] and hypertension [64]. Modeling the overall conformation of GPCRs in complex with biased ligands Broad modeling endeavors aspire to forecast the global conformational state of the receptor in complex with ligands endowed with different signaling profiles. Reaching this goal would provide powerful mechanistic insights into the requirements for the activation of the different signaling pathways. Although modeling global conformational states is challenging, the results of several studies indicate that the field is moving closer to its attainment [40–45]. In the following paragraphs we describe two case studies that illustrate two modeling strategies that have been applied or are potentially applicable to the global conformation of GPCRs in complex with biased ligands [40,41]. The first is a modeling report of the effect of ligands endowed with different signaling profiles in the stabilization of different conformations of the b2AR, described in 2013 by Tikhonova and coworkers [40]. Specifically, the authors investigated how receptors bound to ligands with different signaling profiles transition from the active to the inactive state through a modeling approach based on accelerated molecular dynamics. This is a modification of conventional molecular dynamics that aims at achieving a more thorough conformational sampling by ‘boosting’ dihedral potentials, thus allowing rotations of the dihedral angles defined by atomic bonds in the backbone and the side chains of proteins to overcome energetic barriers. The authors subjected to the computational study models of the receptor in complex with salbutamol, a non-biased agonist also known as albuterol that is widely used for the treatment of asthma, and the biased agonist N-cyclopentylbutanepherine, an agonist that is more efficacious toward the b-arrestin than the Gs-mediated signaling pathway [5]. Both models were constructed on the basis of the activated crystal structure of the receptor solved in complex with a nanobody [65]. To study the transition of the receptor from the activated to the inactive state, the authors conducted the molecular dynamics simulations in the absence of the nanobody, which is fundamental for the stabilization of the activated conformation [65]. Hence, they observed that, in the course of the simulations, the receptor moved toward an inactive-like conformation in presence of both the nonbiased and the biased agonist, as well as in the absence of a ligand. However, interestingly, they also observed that the simulations conducted with the two different ligands caused distinct patterns of motion to the seventh transmembrane domain (TM7) of the receptor. Because it drives the receptor toward its inactive conformation, the
TIPS-1132; No. of Pages 7
Opinion approach taken by the authors is not suitable for the prediction of the conformation of the receptor stabilized by ligands with specific signaling profiles. Most likely, such a prediction would require the inclusion of the correct intracellular partner in the modeled system rather than the deletion of the one found in the experimental structure of the receptor used as the modeling platform. Moreover, it would probably require modeling the appropriate phosphorylation state of the cytosolic residues of the receptor. Nevertheless, studies like the one described by the authors can provide insights on the different patterns of interresidue interactions promoted by ligands endowed with different signaling profiles. The second case study relates to the development of a modeling approach intended to generate hypotheses on the coupling between the motions occurring in different functional sites of a same GPCR. The study, reported in 2009 by Balabin and coworkers [41], was based on a modification of coarse-grained normal mode analysis known as ‘allosteric coupling analysis’, which entails modeling the local structural perturbations occurring at each site and then studying the collective structural motions in the receptor through which the local perturbations are coupled. Specifically, the authors analyzed the allosteric relationships between the ligand-binding site and sites located in the cytosolic region of the receptor, which, as mentioned, is the region through which GPCRs interact with their cognate signaling proteins. Analyzing with this method rhodopsin and the b2AR – specifically, two crystal structures of the two receptors solved in the inactive state – the authors concluded that functional sites, unlike other sites, show strong couplings dominated by only a few normal modes. Although the study analyzed only the receptors in complex with the cocrystallized inverse agonists – 11-cis-retinal for rhodopsin and carazolol for the b2AR – one can envision doing the same for various ligands. In particular, by analyzing agonists biased for specific pathways, one could possibly generate modeling hypotheses regarding which allosteric sites of the receptor are most affected by compounds endowed with different signaling profiles. Modeling the binding cavity of GPCR–ligand complexes More circumscribed modeling attempts are intended to forecast the binding mode of ligands endowed with different signaling profiles, such as the orientation and conformation that the ligands assume when bound to the receptor, and model the local conformational state of the ligand-binding cavity [46–61]. Focusing the modeling on the conformation of the ligand-binding cavity, a task that is more readily achievable with the current computational techniques, does not shed much light on the activation mechanisms. However, it can provide a significant level of structural accuracy related to this crucial region, thus yielding valuable theoretical models applicable to ligand discovery campaigns [62]. As explained in the next section, a receptor structure solved in complex with a given ligand, when used for molecular recognition campaigns, tends to favor the identification of ligands endowed with similar signaling properties [66,67]. This is unsurprising, especially considering
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
a major limitation of molecular docking procedures, which are the computational procedures intended to place a ligand within a receptor-binding cavity in the right conformation and the right orientation and to score the interactions of the ligand with the receptor. The limitation is that, in most cases, during molecular docking the receptor is held rigid, with flexibility granted only to the ligand. Most structures have been solved in complex with unbiased antagonists or inverse agonists, which would seem to restrict the applicability of the rational ligand discovery efforts to this class of compounds [31,32,68]. However, these structures can be employed as a platform to construct models of the binding cavity of the same receptor in complex with molecules endowed with different signaling profiles, potentially including biased agonists. As an example of the feasibility of such a task, I illustrate how, in my research group, we successfully modeled the agonist-bound conformation of the binding cavity of the b2AR [46]. Using as the basis for our modeling study the crystal structure of the receptor solved in complex with the inverse agonist carazolol, we generated models of the binding cavity of the receptor in complex with 60 ligands: 30 agonists and 30 blockers. A traditional docking procedure, due to the rigid representation of the receptor that we discussed, would not yield any prediction of the conformation adopted by the binding cavity of a receptor when in complex with a docked ligand. Hence, we constructed the models through one of the available procedures that grants flexibility to both the ligand and the receptor (Box 1) [69]. Two elements were key to the success of our modeling effort: (i) the fact that we granted flexibility to large contiguous portions of the receptor, which allowed us to capture subtle local rearrangements of the Box 1. Modeling local conformational changes For our study of the local conformational changes associated with the binding of agonists and blockers of the b2AR, we employed the ‘induced fit’ protocol implemented in the Schro¨dinger molecular modeling package (Schro¨dinger, LLC; http://www.schrodinger.com). In this approach, which is one of the many that have been devised to overcome the receptor rigidity associated with conventional docking, the flexibility issue is addressed by performing three modeling procedures in a sequence: (i) the docking of a flexible ligand to a rigid representation of the receptor, to yield a first receptor–ligand complex; (ii) the sampling, around the docked ligand, of the rotameric state of the side chains of several selected residues that surround the dock ligand, while also allowing minor backbone movements through energy minimizations; and (iii) the redocking of the ligand to the modified conformation of the receptor to yield a final model of the receptor–ligand complex. Key to the success of our modeling effort was the peculiar way in which we selected the receptor residues to which flexibility was to be granted. Specifically, although traditionally flexible residues are defined as those that have at least one atom within a sphere of a specified radius centered on the docked ligand, in our study we granted flexibility to all of the residues within three uninterrupted sequence segments that define the entire exofacial portion of the receptor ligand-banding cavity. Each segment comprised the exofacial half of two adjacent transmembrane domains (TMs) and the entire interconnecting extracellular loop (EL): TM2–EL1–TM3; TM4–EL2–TM5; and TM6–EL3–TM7. The N terminus and the exofacial half of TM1 were not granted flexibility because they are too distantly located from the cavity. The definition of three contiguous large segments of flexible residues allowed the generation of hypotheses on the subtle conformational changes of the backbone of the receptor within the ligand-binding cavity. 3
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
backbone within the ligand-binding cavity (Box 1); and (ii) the use of statistical analyses that assisted the identification of the conformational features that consistently distinguished the complexes of the receptor with agonists from those with blockers. The resulting models well captured the conformational differences in the binding cavity of the receptor when in complex with agonists or blockers, which, as explained in the next section, have a fundamental impact on the results of virtual screening campaigns. In particular, we accurately predicted that the most pronounced divergence between agonist-bound and blocker-bound structures would be a subtle displacement of a specific residue located in the fifth transmembrane domain; namely, Ser207. As shown in Figure 2, this modeling hypothesis was entirely confirmed by the crystal structure of the receptor in complex with a full agonist solved by Kobilka and coworkers [65]. In agreement with our models, these authors noted that
(A)
(C)
(B)
(D)
TRENDS in Pharmacological Sciences
Figure 2. Modeling the local conformational state of the binding cavity of G protein-coupled receptor (GPCR)–ligand complexes. Certain modeling techniques are intended to model only local conformational states. The figure shows a model of the b2-adrenoceptor (b2AR) in complex with the agonist epinephrine constructed on the basis of the crystal structure of the same receptor with the blocker carazolol [46]. (A) The model, in yellow, is compared with its parent blocker-bound structure of the receptor, in red [Protein Data Bank (PDB) ID: 2RH1].(B) The model, again in yellow, is compared with the experimentally solved agonist-bound structure of the receptor, in green (PDB ID: 3P0G). (C,D) Blow-out views of the areas included within the white rectangles in (A,B), respectively. Although the global conformation of the receptor is close to the blocker-bound parent structure, the models reflect accurately a crucial local conformational difference of the ligandbinding cavity in complex with agonists and blockers. Specifically, the arrows and the insets on the lower right side of each model illustrate how the model captured the conformation that residue Ser207 assumes when in complex with agonists. As revealed by X-ray crystallography, this is the most prominent difference found between the ligand-binding cavity of the blocker-bound and agonist-bound structures of the b2AR [65]. The ribbons and tubes schematically illustrate the backbone of the receptor. The bound ligands are shown as space-filling models; Ser207 is shown as a ball-and-stick model.
4
‘The greatest difference between inactive and active structures in the ligand-binding site is an inward bulge of TM5 centered around Ser207. . .whose Ca position shifts by ˚ ’. 2.1 A Beyond the procedure described in the previous paragraph, numerous alternative modeling strategies have been applied to the generation of hypotheses on the local conformation of the binding cavity of GPCRs in complex with specific ligands. These include normal mode analyses and Monte Carlo conformational searches [47–56], systematic rotation of transmembrane domains [57–59], methods based on alanine scanning [60], and even unbiased simulations in which the ligands spontaneously associate with their receptors [61]. One can envision that, given a set of ligands endowed with different biased signaling profiles, these strategies could also be applied to modeling the local conformations stabilized by these ligands. However, a caveat that needs to be considered is that this task may be complicated by the fact that the structural differences between biased and non-biased agonists are often subtle. Steering structure-based drug discovery campaigns As Shoichet and coworkers demonstrated, when crystal structures of GPCRs are targeted by virtual screening based on molecular docking, the campaigns tend to yield active compounds with signaling properties similar to those of the cocrystallized ligands [66,67]. In other words, screenings that target structures solved in complex with blockers tend to prioritize blockers, whereas those that target structures solved in complex with agonists tend to prioritize agonists. However, as detailed above, the structure of the receptor can be modified through the flexible docking of a compound known to have signaling properties different from those of the cocrystallized ligand following various procedures. Importantly, we found that modifying the crystal structures in such a fashion before virtual screening steers the campaign toward the retrieval of ligands endowed with signaling properties in line with those of the compound around which the structure was modified [62]. This observation arises from controlled computational experiments that we performed by docking agonists and blockers of the b2AR to different crystal structures and models of the receptor [62]. Specifically, our controlled screenings targeted crystal structures of the receptor solved in different signaling states as well as models obtained by flexibly docking agonists within the binding cavity of a blocker-bound crystal structure obtained in the study that we illustrated in the previous section [46]. As our results indicate, the campaigns that targeted an activated structure prioritized agonists over blockers, whereas those that targeted inactive structures prioritized blockers over agonists. Notably, our results also indicate that the campaigns that targeted the blocker-bound structures modified through the flexible docking of agonists before the screening were effectively steered toward the prioritization of agonists (Figure 3). These data suggest that receptor models in which the binding cavity has been modified through the flexible docking of a biased agonist might be used to steer virtual screening campaigns toward the retrieval of additional biased agonists.
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
AUC (agonists vs. blockers)
1
0.8
0.6
0.4
0.2
0 2RH1 3P0G 2RH1-epinephrine TRENDS in Pharmacological Sciences
Figure 3. Prioritizing agonists versus blockers. When ligands of the b2-adrenoceptor (b2AR) are computationally docked at an inactive structure of the receptor, such as the structure cocrystallized with the blocker carazolol [Protein Data Bank (PDB) ID: 2RH1], blockers receive better scores that agonists. Conversely, when they are docked at an activated structure of the receptor, such as the structure cocrystallized with the agonist BI-167017 (PDB ID: 3P0G), agonists receive better scores than blockers. However, if an agonist, such as epinephrine, is flexibly docked into the carazolol-bound structure of the receptor, the resulting model favors agonists over blockers. The plot illustrates the results of our study analyzed through a receiver operating characteristic (ROC) analysis. An area under the ROC curve (AUC) of 1 indicates complete prioritization of agonists versus blockers, an AUC of 0 indicates complete prioritization of blockers versus agonists, and an AUC of 0.5 indicates absence of prioritization of one ligand versus the other [62].
Concluding remarks I have illustrated strategies that have been or could be employed to model the conformational state of a GPCR in complex with biased ligands, either globally [40,41] or just related to the ligand-binding cavity [46]. One must stress that, currently, this is a remarkably challenging task. As discussed, predicting the overall conformational state of a receptor in a given signaling state probably requires, besides a biased ligand, correct modeling of the interactions of the receptors with the appropriate intracellular signaling pathways and the correct phosphorylation state of the receptor residues exposed to the cytosol. Moreover, concerning local modeling efforts, the subtle structural differences between biased and non-biased agonists may make it particularly difficult to distinguish the different conformational states of the binding cavity stabilized by different ligands. For many receptor systems, modeling attempts are hindered by the absence of known ligands endowed with a significant signaling bias. Despite these difficulties, it is likely that the task may become more feasible in the near future. First, the generation of more knowledge about biased signaling and the identification of more biased agonists will probably foster the identification of further biased ligands. Hence, if a few partially biased agonists are identified through the models that it is possible to build on current knowledge, these ligands could then be utilized to generate more accurate models subsequently applicable to the identification of agonists endowed with a stronger bias, in a virtuous cycle. Moreover, it is probable that better modeling techniques will become available for the construction of more accurate
models of both the overall and the local conformation of GPCRs in complex with specific ligands. As a general direction, the key to the generation of accurate models may be found in the expert combination of experimental and computational techniques. The availability of crystal structures solved in complex with biased ligands will no doubt assist the modeling process [37,38]. Further insights are likely to come from NMR, which is a powerful tool for conformational studies and is being applied with increasing success to the analysis of GPCRs in complex with different ligands, including biased agonists [26,39,70–73]. The structural constraints derived from NMR experiments may represent the most efficacious way to construct experimentally supported computational 3D models of GPCRs in complex with biased ligands applicable to drug discovery campaigns. Moreover, the modeling process, in particular with respect to global conformational studies, may be further assisted by the experimental identification of phosphorylation sites and the solution of the structure of the intracellular signaling partners of GPCRs [25,74]. In conclusion, as biological knowledge increases and computational techniques improve, it is likely that molecular modeling will play an increasingly relevant role in the understanding of the structural bases of biased signaling, fostering the rational discovery of GPCR ligands endowed with highly specific signaling profiles. References 1 Pierce, K. et al. (2002) Seven-transmembrane receptors. Nat. Rev. Mol. Cell Biol. 3, 639–650
5
TIPS-1132; No. of Pages 7
Opinion 2 Reiter, E. et al. (2012) Molecular mechanism of b-arrestin-biased agonism at seven-transmembrane receptors. Annu. Rev. Pharmacol. Toxicol. 52, 179–197 3 Rajagopal, S. et al. (2011) Quantifying ligand bias at 7-transmembrane receptors. Mol. Pharmacol. 80, 367–377 4 Urban, J.D. et al. (2007) Functional selectivity and classical concepts of quantitative pharmacology. J. Pharmacol. Exp. Ther. 320, 1–13 5 Drake, M.T. et al. (2008) Beta-arrestin-biased agonism at the b2adrenergic receptor. J. Biol. Chem. 283, 5669–5676 6 Martin, N.P. et al. (2004) PKA-mediated phosphorylation of the b1adrenergic receptor promotes Gs/Gi switching. Cell. Signal. 16, 1397– 1403 7 Lefkowitz, R. and Shenoy, S. (2005) Transduction of receptor signals by b-arrestins. Science 308, 512–517 8 Wisler, J.W. et al. (2007) A unique mechanism of b-blocker action: carvedilol stimulates b-arrestin signaling. Proc. Natl. Acad. Sci. U.S.A. 104, 16657–16662 9 Kahsai, A.W. et al. (2011) Multiple ligand-specific conformations of the b2-adrenergic receptor. Nat. Chem. Biol. 7, 692–700 10 Gesty-Palmer, D. et al. (2006) Distinct b-arrestin- and G proteindependent pathways for parathyroid hormone receptor-stimulated ERK1/2 activation. J. Biol. Chem. 281, 10856–10864 11 Walters, R.W. et al. (2009) b-arrestin1 mediates nicotinic acid-induced flushing, but not its antilipolytic effect, in mice. J. Clin. Invest. 119, 1312–1321 12 Monasky, M.M. et al. (2013) The b-arrestin-biased ligand TRV120023 inhibits angiotensin II-induced cardiac hypertrophy while preserving enhanced myofilament response to calcium. Am. J. Physiol. Heart Circ. Physiol. 305, H856–H866 13 Violin, J.D. et al. (2013) GPCR biased ligands as novel heart failure therapeutics. Trends Cardiovasc. Med. 23, 242–249 14 Boerrigter, G. et al. (2012) TRV120027, a novel b-arrestin biased ligand at the angiotensin II type I receptor, unloads the heart and maintains renal function when added to furosemide in experimental heart failure. Circ. Heart Fail. 5, 627–634 15 Kim, K.S. et al. (2012) b-arrestin-biased AT1R stimulation promotes cell survival during acute cardiac injury. Am. J. Physiol. Heart Circ. Physiol. 303, H1001–H1010 16 Boerrigter, G. et al. (2011) Cardiorenal actions of TRV120027, a novel b-arrestin-biased ligand at the angiotensin II type I receptor, in healthy and heart failure canines: a novel therapeutic strategy for acute heart failure. Circ. Heart Fail. 4, 770–778 17 DeWire, S.M. and Violin, J.D. (2011) Biased ligands for better cardiovascular drugs: dissecting G-protein-coupled receptor pharmacology. Circ. Res. 109, 205–216 18 Shukla, A.K. et al. (2008) Distinct conformational changes in b-arrestin report biased agonism at seven-transmembrane receptors. Proc. Natl. Acad. Sci. U.S.A. 105, 9988–9993 19 Zhou, L. et al. (2013) Development of functionally selective, small molecule agonists at kappa opioid receptors. J. Biol. Chem. 288, 36703–36716 20 White, K.L. et al. (2013) Identification of novel functionally selective kappa opioid receptor scaffolds. Mol. Pharmacol. 85, 83–90 21 Chen, X.T. et al. (2013) Structure–activity relationships and discovery of a G protein biased mu opioid receptor ligand, [(3-methoxythiophen2-yl)methyl]({2-[(9R)-9-(pyridin-2-yl)-6-oxaspiro-[4.5]decan-9yl]ethyl})amine (TRV130), for the treatment of acute severe pain. J. Med. Chem. 56, 8019–8031 22 Lane, J.R. et al. (2013) Structure-based ligand discovery targeting orthosteric and allosteric pockets of dopamine receptors. Mol. Pharmacol. 84, 794–807 23 Palczewski, K. and Orban, T. (2013) From atomic structures to neuronal functions of G protein-coupled receptors. Annu. Rev. Neurosci. 36, 139–164 24 Rasmussen, S.G. et al. (2011) Crystal structure of the b2 adrenergic receptor–Gs protein complex. Nature 477, 549–555 25 Nobles, K.N. et al. (2011) Distinct phosphorylation sites on the b2adrenergic receptor establish a barcode that encodes differential functions of b-arrestin. Sci. Signal. 4, ra51 26 Nygaard, R. et al. (2013) The dynamic process of b2-adrenergic receptor activation. Cell 152, 532–542 27 Kobilka, B. (2007) G protein coupled receptor structure and activation. Biochim. Biophys. Acta 1768, 794–807 6
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
28 Kobilka, B.K. and Deupi, X. (2007) Conformational complexity of Gprotein-coupled receptors. Trends Pharmacol. Sci. 28, 397–406 29 Yao, X. et al. (2006) Coupling ligand structure to specific conformational switches in the b2-adrenoceptor. Nat. Chem. Biol. 2, 417–422 30 Warne, T. and Tate, C.G. (2013) The importance of interactions with helix 5 in determining the efficacy of b-adrenoceptor ligands. Biochem. Soc. Trans. 41, 159–165 31 Stevens, R.C. et al. (2013) The GPCR Network: a large-scale collaboration to determine human GPCR structure and function. Nat. Rev. Drug Discov. 12, 25–34 32 Katritch, V. et al. (2013) Structure–function of the G protein-coupled receptor superfamily. Annu. Rev. Pharmacol. Toxicol. 53, 531–556 33 Venkatakrishnan, A.J. et al. (2013) Molecular signatures of G-proteincoupled receptors. Nature 494, 185–194 34 Tate, C.G. and Schertler, G.F. (2009) Engineering G protein-coupled receptors to facilitate their structure determination. Curr. Opin. Struct. Biol. 19, 386–395 35 Kruse, A.C. et al. (2013) Applications of molecular replacement to G protein-coupled receptors. Acta Crystallogr. D: Biol. Crystallogr. 69, 2287–2292 36 Steyaert, J. and Kobilka, B.K. (2011) Nanobody stabilization of G protein-coupled receptor conformational states. Curr. Opin. Struct. Biol. 21, 567–572 37 Warne, T. et al. (2012) Crystal structures of a stabilized b1adrenoceptor bound to the biased agonists bucindolol and carvedilol. Structure 20, 841–849 38 Wacker, D. et al. (2013) Structural features for functional selectivity at serotonin receptors. Science 340, 615–619 39 Liu, J.J. et al. (2012) Biased signaling pathways in b2-adrenergic receptor characterized by 19F-NMR. Science 335, 1106–1110 40 Tikhonova, I.G. et al. (2013) Simulations of biased agonists in the b2 adrenergic receptor with accelerated molecular dynamics. Biochemistry 52, 5593–5603 41 Balabin, I.A. et al. (2009) Coarse-grained modeling of allosteric regulation in protein receptors. Proc. Natl. Acad. Sci. U.S.A. 106, 14253–14258 42 Provasi, D. et al. (2011) Ligand-induced modulation of the free-energy landscape of G protein-coupled receptors explored by adaptive biasing techniques. PLoS Comput. Biol. 7, e1002193 43 Dror, R.O. et al. (2011) Activation mechanism of the b2-adrenergic receptor. Proc. Natl. Acad. Sci. U.S.A. 108, 18684–18689 44 Dror, R.O. et al. (2011) Pathway and mechanism of drug binding to Gprotein-coupled receptors. Proc. Natl. Acad. Sci. U.S.A. 108, 13118– 13123 45 Katritch, V. et al. (2009) Analysis of full and partial agonists binding to b2-adrenergic receptor suggests a role of transmembrane helix V in agonist-specific conformational changes. J. Mol. Recognit. 22, 307–318 46 Vilar, S. et al. (2011) In silico analysis of the binding of agonists and blockers to the b2-adrenergic receptor. J. Mol. Graph. Model. 29, 809– 817 47 Vilar, S. et al. (2011) Docking-based virtual screening for ligands of G protein-coupled receptors: not only crystal structures but also in silico models. J. Mol. Graph. Model. 29, 614–623 48 Cavasotto, C.N. (2012) Normal mode-based approaches in receptor ensemble docking. Methods Mol. Biol. 819, 157–168 49 Cavasotto, C.N. and Abagyan, R.A. (2004) Protein flexibility in ligand docking and virtual screening to protein kinases. J. Mol. Biol. 337, 209– 225 50 Cavasotto, C.N. et al. (2005) Representing receptor flexibility in ligand docking through relevant normal modes. J. Am. Chem. Soc. 127, 9632– 9640 51 Rueda, M. et al. (2009) Consistent improvement of cross-docking results using binding site ensembles generated with elastic network normal modes. J. Chem. Inf. Model. 49, 716–725 52 Reynolds, K. et al. (2009) Identifying conformational changes of the b2 adrenoceptor that enable accurate prediction of ligand/receptor interactions and screening for GPCR modulators. J. Comput. Aided Mol. Des. 23, 273–288 53 Phatak, S.S. et al. (2010) Ligand-steered modeling and docking: a benchmarking study in class A G-protein-coupled receptors. J. Chem. Inf. Model. 50, 2119–2128
TIPS-1132; No. of Pages 7
Opinion 54 Katritch, V. and Abagyan, R. (2011) GPCR agonist binding revealed by modeling and crystallography. Trends Pharmacol. Sci. 32, 637–643 55 Rueda, M. et al. (2012) ALiBERO: evolving a team of complementary pocket conformations rather than a single leader. J. Chem. Inf. Model. 52, 2705–2714 56 Katritch, V. et al. (2012) Ligand-guided receptor optimization. Methods Mol. Biol. 857, 189–205 57 Bhattacharya, S. et al. (2008) Ligand-stabilized conformational states of human b2 adrenergic receptor: insight into G-protein-coupled receptor activation. Biophys. J. 94, 2027–2042 58 Bhattacharya, S. and Vaidehi, N. (2010) Computational mapping of the conformational transitions in agonist selective pathways of a G-protein coupled receptor. J. Am. Chem. Soc. 132, 5205–5214 59 Bhattacharya, S. and Vaidehi, N. (2012) LITiCon: a discrete conformational sampling computational method for mapping various functionally selective conformational states of transmembrane helical proteins. Methods Mol. Biol. 914, 167–178 60 Bottegoni, G. et al. (2008) A new method for ligand docking to flexible receptors by dual alanine scanning and refinement (SCARE). J. Comput. Aided Mol. Des. 22, 311–325 61 Dror, R.O. et al. (2013) Structural basis for modulation of a G-proteincoupled receptor by allosteric drugs. Nature 503, 295–299 62 Costanzi, S. and Vilar, S. (2012) In silico screening for agonists and blockers of the b2 adrenergic receptor: implications of inactive and activated state structures. J. Comput. Chem. 33, 561–572 63 Cazzola, M. et al. (2013) b2-agonist therapy in lung disease. Am. J. Respir. Crit. Care Med. 187, 690–696
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
J.A. and Liggett, S.B. (2011) Cardiovascular 64 Johnson, pharmacogenomics of adrenergic receptor signaling: clinical implications and future directions. Clin. Pharmacol. Ther. 89, 366–378 65 Rasmussen, S.G. et al. (2011) Structure of a nanobody-stabilized active state of the b2 adrenoceptor. Nature 469, 175–180 66 Weiss, D.R. et al. (2013) Conformation guides molecular efficacy in docking screens of activated b2 adrenergic G protein coupled receptor. ACS Chem. Biol. 8, 1018–1026 67 Kolb, P. et al. (2009) Structure-based discovery of b2-adrenergic receptor ligands. Proc. Natl. Acad. Sci. U.S.A. 106, 6843–6848 68 Jacobson, K.A. and Costanzi, S. (2012) New insights for drug design from the X-ray crystallographic structures of G-protein-coupled receptors. Mol. Pharmacol. 82, 361–371 69 Sherman, W. et al. (2006) Novel procedure for modeling ligand/receptor induced fit effects. J. Med. Chem. 49, 534–553 70 Kim, T.H. et al. (2013) The role of ligands on the equilibria between functional states of a G protein-coupled receptor. J. Am. Chem. Soc. 135, 9465–9474 71 Bokoch, M. et al. (2010) Ligand-specific regulation of the extracellular surface of a G-protein-coupled receptor. Nature 463, 108–112 72 Horst, R. et al. (2013) b2-Adrenergic receptor activation by agonists studied with 19F NMR spectroscopy. Angew. Chem. Int. Ed. Engl. 52, 10762–10765 73 Horst, R. et al. (2013) b2-Adrenergic receptor solutions for structural biology analyzed with microscale NMR diffusion measurements. Angew. Chem. Int. Ed. Engl. 52, 331–335 74 Shukla, A.K. et al. (2013) Structure of active b-arrestin-1 bound to a Gprotein-coupled receptor phosphopeptide. Nature 497, 137–141
7
Opinion
Modeling G protein-coupled receptors in complex with biased agonists Stefano Costanzi Department of Chemistry, American University, Washington, DC 20016, USA Center for Behavioral Neuroscience, American University, Washington, DC 20016, USA
The biological response to the activation of G proteincoupled receptors (GPCRs) typically originates from the simultaneous modulation of various signaling pathways that lead to distinct biological consequences. Hence, ‘biased agonists’ (i.e., compounds that selectively activate one of the pathways while blocking the others) are highly sought-after molecules to provide fine-tuned pharmacological interventions. This review describes strategies that can be deployed to model the conformation of GPCRs in complex with ligands endowed with specific signaling profiles useful for the generation of hypotheses on the structural requirements for the activation of different signaling pathways or for rational computer-aided ligand discovery campaigns. In particular, it focuses on strategies potentially applicable to model the global or local conformational states of GPCRs stabilized by specific ligands. Introduction: GPCRs and biased signaling GPCRs are integral membrane proteins that function as cellular receivers for stimuli that, in most cases, are given by extracellular molecules known as receptor agonists [1,2]. These can be endogenous compounds, for instance neurotransmitters or hormones, but also exogenous natural or artificial compounds, for instance odorants or drugs. On binding to the receptors, agonists trigger the activation of receptor-mediated signaling pathways that initiate with the interaction of intracellular signaling proteins with the domains of the receptor exposed to the cytosol. Other ligands, known as antagonists, impede the agonist-mediated activation of the receptor. Finally, ligands known as inverse agonists, besides interfering with agonists as do antagonists, suppress the constitutive activity of the receptor [3,4]. Several studies pioneered by Robert Lefkowitz demonstrated that the biological response to the activation of GPCRs typically originates from the simultaneous modulation of various signaling pathways [1–3,5–9]. Some of these are mediated by the activation of G proteins, whereas others are modulated by proteins known as arrestins. Because the activation of different pathways leads to distinct biological consequences, there is significant Corresponding author: Costanzi, S. ([email protected]). Keywords: flexible docking; global conformational changes; local conformational changes; molecular dynamics; normal mode analysis; virtual screening. 0165-6147/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tips.2014.04.004
interest in the identification of ‘biased agonists’ of GPCRs, which are agonists that selectively activate only one signaling response. As a result, biased agonists have been identified for several GPCRs [2], some examples of which include the b-adrenoceptors [8], the parathyroid hormone receptor [10], a nicotinic acid receptor (GPR109A) [11], the angiotensin receptors [12–18], the opioid receptors [19–21], and the dopamine receptors [22]. The molecular bases of biased agonisms rest on the fact that GPCRs are essentially allosteric systems: the binding of agonists to a first site stabilizes a conformation of the receptor that promotes the activation of specific intracellular signaling partners that bind to a second site. In particular, the receptors are formed by a single polypeptide chain that spans the membrane seven times, with seven ahelical transmembrane domains connected by three intracellular and three extracellular loops [23]. The structure folds to form a membrane-embedded helical bundle, with the N terminus of the chain in the extracellular milieu and the C terminus in the cytosol. For most GPCRs, the binding of agonists occurs in a cavity formed within the helical bundle and exposed to the extracellular milieu [23]. Conversely, intracellular partners bind to regions of the receptor exposed to the cytosol [24,25]. As research studies have shown, GPCRs can assume a range of conformations stabilized by different ligands and associated with different signaling states; some conformations are signaling silent, whereas others trigger the activation of signaling cascades [4,9,26–30]. Specifically, some conformations prevent the interaction of the receptor with all of its signaling partners. Conversely, other conformations promote the association with one or more intracellular signaling proteins, some of which require prior phosphorylation of specific amino acid residues of the receptor exposed to the cytosol [25]. The fact that biased agonists have been found for several systems suggests that the receptor conformations associated with the activation of different pathways are distinct; hence, one conformation can be silent with respect to a given pathway but lead to the activation of another pathway [9]. Thanks to several technical and methodological advancements, the field of GPCR structural biology is currently in full blossom and is yielding a steady output of structures solved at atomic resolution [31–36]. Each experimental structure reflects a specific conformational state of the receptor stabilized by the ligand and all of the other conditions employed for the structural determination. Hence, to gain insights into the conformational Trends in Pharmacological Sciences xx (2014) 1–7
1
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
requirements for the activation of the various signaling pathways, efforts are being made toward the solution of structures of GPCRs in complex with ligands endowed with different signaling profiles [37–39]. In particular, there is significant interest in the solution of GPCR structures in complex with biased agonists through X-ray crystallography [37,38] and toward the generation of NMR-supported hypotheses on the conformations linked to the various signaling states [39]. One of the challenges related to the characterization of the molecular mechanisms that underlie biased agonism is that some of the structures solved in complex with biased agonists, although showing receptor–ligand interactions not found with non-biased ligands, captured the receptor in its inactive conformation and do not show the conformational changes responsible for biased signaling [37]. In the absence of the experimental structure of a GPCR in the conformational state stabilized by a given ligand, a structure of the same receptor solved in a different conformation has the potential to serve as a starting point for computational experiments intended to model the conformational state of interest. In particular, efforts to model the conformation of a GPCR in complex with a specific ligand can have two distinct scopes: a broader one, intended to forecast the global conformational state of the receptor–ligand complex [40–45]; or a more circumscribed one, intended to forecast the local conformational state of the ligand-binding cavity (Figure 1) [46–61]. In this review, I discuss ways in which experimental structures of GPCRs cocrystallized with non-biased ligands could be used as platforms to generate models of the receptors in complex with biased agonists. I begin with an illustration of the published strategies that attempt to model the overall structural features of GPCRs in complex with biased agonists [40,41]. I then provide an overview of the strategy that my research group has employed to: (i) model the local conformation of the ligand-binding cavity of GPCRs stabilized by specific compounds [46]; and (ii) use such models to steer virtual screening campaigns toward the identification of ligands with desired signaling profiles [62]. We reported the application of such a procedure to the Global modeling
Receptor
Biased agonist
+
Local modeling TRENDS in Pharmacological Sciences
Figure 1. Modeling the global and local conformational state of G protein-coupled receptors (GPCRs) in complex with biased agonists. Given the experimental structure of a GPCR and a model of a given biased agonist, different modeling strategies can be applied to the generation of hypotheses on the global conformation of the resulting complex or its local conformation with respect to the ligand-binding cavity.
2
generation of models that well discriminate between agonists and blockers; although theoretically conceivable, its application to the construction of models that can distinguish between agonists endowed with different signaling profiles (e.g., agonists biased toward the exclusive activation of a single signaling pathway) remains to be demonstrated. I conclude with a view of the possible evolution of the field. The case studies that we discuss are all related to the b2-adrenoceptor (b2AR), a prototypical GPCR naturally activated by epinephrine and targeted by drugs for various indications including lung diseases [63] and hypertension [64]. Modeling the overall conformation of GPCRs in complex with biased ligands Broad modeling endeavors aspire to forecast the global conformational state of the receptor in complex with ligands endowed with different signaling profiles. Reaching this goal would provide powerful mechanistic insights into the requirements for the activation of the different signaling pathways. Although modeling global conformational states is challenging, the results of several studies indicate that the field is moving closer to its attainment [40–45]. In the following paragraphs we describe two case studies that illustrate two modeling strategies that have been applied or are potentially applicable to the global conformation of GPCRs in complex with biased ligands [40,41]. The first is a modeling report of the effect of ligands endowed with different signaling profiles in the stabilization of different conformations of the b2AR, described in 2013 by Tikhonova and coworkers [40]. Specifically, the authors investigated how receptors bound to ligands with different signaling profiles transition from the active to the inactive state through a modeling approach based on accelerated molecular dynamics. This is a modification of conventional molecular dynamics that aims at achieving a more thorough conformational sampling by ‘boosting’ dihedral potentials, thus allowing rotations of the dihedral angles defined by atomic bonds in the backbone and the side chains of proteins to overcome energetic barriers. The authors subjected to the computational study models of the receptor in complex with salbutamol, a non-biased agonist also known as albuterol that is widely used for the treatment of asthma, and the biased agonist N-cyclopentylbutanepherine, an agonist that is more efficacious toward the b-arrestin than the Gs-mediated signaling pathway [5]. Both models were constructed on the basis of the activated crystal structure of the receptor solved in complex with a nanobody [65]. To study the transition of the receptor from the activated to the inactive state, the authors conducted the molecular dynamics simulations in the absence of the nanobody, which is fundamental for the stabilization of the activated conformation [65]. Hence, they observed that, in the course of the simulations, the receptor moved toward an inactive-like conformation in presence of both the nonbiased and the biased agonist, as well as in the absence of a ligand. However, interestingly, they also observed that the simulations conducted with the two different ligands caused distinct patterns of motion to the seventh transmembrane domain (TM7) of the receptor. Because it drives the receptor toward its inactive conformation, the
TIPS-1132; No. of Pages 7
Opinion approach taken by the authors is not suitable for the prediction of the conformation of the receptor stabilized by ligands with specific signaling profiles. Most likely, such a prediction would require the inclusion of the correct intracellular partner in the modeled system rather than the deletion of the one found in the experimental structure of the receptor used as the modeling platform. Moreover, it would probably require modeling the appropriate phosphorylation state of the cytosolic residues of the receptor. Nevertheless, studies like the one described by the authors can provide insights on the different patterns of interresidue interactions promoted by ligands endowed with different signaling profiles. The second case study relates to the development of a modeling approach intended to generate hypotheses on the coupling between the motions occurring in different functional sites of a same GPCR. The study, reported in 2009 by Balabin and coworkers [41], was based on a modification of coarse-grained normal mode analysis known as ‘allosteric coupling analysis’, which entails modeling the local structural perturbations occurring at each site and then studying the collective structural motions in the receptor through which the local perturbations are coupled. Specifically, the authors analyzed the allosteric relationships between the ligand-binding site and sites located in the cytosolic region of the receptor, which, as mentioned, is the region through which GPCRs interact with their cognate signaling proteins. Analyzing with this method rhodopsin and the b2AR – specifically, two crystal structures of the two receptors solved in the inactive state – the authors concluded that functional sites, unlike other sites, show strong couplings dominated by only a few normal modes. Although the study analyzed only the receptors in complex with the cocrystallized inverse agonists – 11-cis-retinal for rhodopsin and carazolol for the b2AR – one can envision doing the same for various ligands. In particular, by analyzing agonists biased for specific pathways, one could possibly generate modeling hypotheses regarding which allosteric sites of the receptor are most affected by compounds endowed with different signaling profiles. Modeling the binding cavity of GPCR–ligand complexes More circumscribed modeling attempts are intended to forecast the binding mode of ligands endowed with different signaling profiles, such as the orientation and conformation that the ligands assume when bound to the receptor, and model the local conformational state of the ligand-binding cavity [46–61]. Focusing the modeling on the conformation of the ligand-binding cavity, a task that is more readily achievable with the current computational techniques, does not shed much light on the activation mechanisms. However, it can provide a significant level of structural accuracy related to this crucial region, thus yielding valuable theoretical models applicable to ligand discovery campaigns [62]. As explained in the next section, a receptor structure solved in complex with a given ligand, when used for molecular recognition campaigns, tends to favor the identification of ligands endowed with similar signaling properties [66,67]. This is unsurprising, especially considering
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
a major limitation of molecular docking procedures, which are the computational procedures intended to place a ligand within a receptor-binding cavity in the right conformation and the right orientation and to score the interactions of the ligand with the receptor. The limitation is that, in most cases, during molecular docking the receptor is held rigid, with flexibility granted only to the ligand. Most structures have been solved in complex with unbiased antagonists or inverse agonists, which would seem to restrict the applicability of the rational ligand discovery efforts to this class of compounds [31,32,68]. However, these structures can be employed as a platform to construct models of the binding cavity of the same receptor in complex with molecules endowed with different signaling profiles, potentially including biased agonists. As an example of the feasibility of such a task, I illustrate how, in my research group, we successfully modeled the agonist-bound conformation of the binding cavity of the b2AR [46]. Using as the basis for our modeling study the crystal structure of the receptor solved in complex with the inverse agonist carazolol, we generated models of the binding cavity of the receptor in complex with 60 ligands: 30 agonists and 30 blockers. A traditional docking procedure, due to the rigid representation of the receptor that we discussed, would not yield any prediction of the conformation adopted by the binding cavity of a receptor when in complex with a docked ligand. Hence, we constructed the models through one of the available procedures that grants flexibility to both the ligand and the receptor (Box 1) [69]. Two elements were key to the success of our modeling effort: (i) the fact that we granted flexibility to large contiguous portions of the receptor, which allowed us to capture subtle local rearrangements of the Box 1. Modeling local conformational changes For our study of the local conformational changes associated with the binding of agonists and blockers of the b2AR, we employed the ‘induced fit’ protocol implemented in the Schro¨dinger molecular modeling package (Schro¨dinger, LLC; http://www.schrodinger.com). In this approach, which is one of the many that have been devised to overcome the receptor rigidity associated with conventional docking, the flexibility issue is addressed by performing three modeling procedures in a sequence: (i) the docking of a flexible ligand to a rigid representation of the receptor, to yield a first receptor–ligand complex; (ii) the sampling, around the docked ligand, of the rotameric state of the side chains of several selected residues that surround the dock ligand, while also allowing minor backbone movements through energy minimizations; and (iii) the redocking of the ligand to the modified conformation of the receptor to yield a final model of the receptor–ligand complex. Key to the success of our modeling effort was the peculiar way in which we selected the receptor residues to which flexibility was to be granted. Specifically, although traditionally flexible residues are defined as those that have at least one atom within a sphere of a specified radius centered on the docked ligand, in our study we granted flexibility to all of the residues within three uninterrupted sequence segments that define the entire exofacial portion of the receptor ligand-banding cavity. Each segment comprised the exofacial half of two adjacent transmembrane domains (TMs) and the entire interconnecting extracellular loop (EL): TM2–EL1–TM3; TM4–EL2–TM5; and TM6–EL3–TM7. The N terminus and the exofacial half of TM1 were not granted flexibility because they are too distantly located from the cavity. The definition of three contiguous large segments of flexible residues allowed the generation of hypotheses on the subtle conformational changes of the backbone of the receptor within the ligand-binding cavity. 3
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
backbone within the ligand-binding cavity (Box 1); and (ii) the use of statistical analyses that assisted the identification of the conformational features that consistently distinguished the complexes of the receptor with agonists from those with blockers. The resulting models well captured the conformational differences in the binding cavity of the receptor when in complex with agonists or blockers, which, as explained in the next section, have a fundamental impact on the results of virtual screening campaigns. In particular, we accurately predicted that the most pronounced divergence between agonist-bound and blocker-bound structures would be a subtle displacement of a specific residue located in the fifth transmembrane domain; namely, Ser207. As shown in Figure 2, this modeling hypothesis was entirely confirmed by the crystal structure of the receptor in complex with a full agonist solved by Kobilka and coworkers [65]. In agreement with our models, these authors noted that
(A)
(C)
(B)
(D)
TRENDS in Pharmacological Sciences
Figure 2. Modeling the local conformational state of the binding cavity of G protein-coupled receptor (GPCR)–ligand complexes. Certain modeling techniques are intended to model only local conformational states. The figure shows a model of the b2-adrenoceptor (b2AR) in complex with the agonist epinephrine constructed on the basis of the crystal structure of the same receptor with the blocker carazolol [46]. (A) The model, in yellow, is compared with its parent blocker-bound structure of the receptor, in red [Protein Data Bank (PDB) ID: 2RH1].(B) The model, again in yellow, is compared with the experimentally solved agonist-bound structure of the receptor, in green (PDB ID: 3P0G). (C,D) Blow-out views of the areas included within the white rectangles in (A,B), respectively. Although the global conformation of the receptor is close to the blocker-bound parent structure, the models reflect accurately a crucial local conformational difference of the ligandbinding cavity in complex with agonists and blockers. Specifically, the arrows and the insets on the lower right side of each model illustrate how the model captured the conformation that residue Ser207 assumes when in complex with agonists. As revealed by X-ray crystallography, this is the most prominent difference found between the ligand-binding cavity of the blocker-bound and agonist-bound structures of the b2AR [65]. The ribbons and tubes schematically illustrate the backbone of the receptor. The bound ligands are shown as space-filling models; Ser207 is shown as a ball-and-stick model.
4
‘The greatest difference between inactive and active structures in the ligand-binding site is an inward bulge of TM5 centered around Ser207. . .whose Ca position shifts by ˚ ’. 2.1 A Beyond the procedure described in the previous paragraph, numerous alternative modeling strategies have been applied to the generation of hypotheses on the local conformation of the binding cavity of GPCRs in complex with specific ligands. These include normal mode analyses and Monte Carlo conformational searches [47–56], systematic rotation of transmembrane domains [57–59], methods based on alanine scanning [60], and even unbiased simulations in which the ligands spontaneously associate with their receptors [61]. One can envision that, given a set of ligands endowed with different biased signaling profiles, these strategies could also be applied to modeling the local conformations stabilized by these ligands. However, a caveat that needs to be considered is that this task may be complicated by the fact that the structural differences between biased and non-biased agonists are often subtle. Steering structure-based drug discovery campaigns As Shoichet and coworkers demonstrated, when crystal structures of GPCRs are targeted by virtual screening based on molecular docking, the campaigns tend to yield active compounds with signaling properties similar to those of the cocrystallized ligands [66,67]. In other words, screenings that target structures solved in complex with blockers tend to prioritize blockers, whereas those that target structures solved in complex with agonists tend to prioritize agonists. However, as detailed above, the structure of the receptor can be modified through the flexible docking of a compound known to have signaling properties different from those of the cocrystallized ligand following various procedures. Importantly, we found that modifying the crystal structures in such a fashion before virtual screening steers the campaign toward the retrieval of ligands endowed with signaling properties in line with those of the compound around which the structure was modified [62]. This observation arises from controlled computational experiments that we performed by docking agonists and blockers of the b2AR to different crystal structures and models of the receptor [62]. Specifically, our controlled screenings targeted crystal structures of the receptor solved in different signaling states as well as models obtained by flexibly docking agonists within the binding cavity of a blocker-bound crystal structure obtained in the study that we illustrated in the previous section [46]. As our results indicate, the campaigns that targeted an activated structure prioritized agonists over blockers, whereas those that targeted inactive structures prioritized blockers over agonists. Notably, our results also indicate that the campaigns that targeted the blocker-bound structures modified through the flexible docking of agonists before the screening were effectively steered toward the prioritization of agonists (Figure 3). These data suggest that receptor models in which the binding cavity has been modified through the flexible docking of a biased agonist might be used to steer virtual screening campaigns toward the retrieval of additional biased agonists.
TIPS-1132; No. of Pages 7
Opinion
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
AUC (agonists vs. blockers)
1
0.8
0.6
0.4
0.2
0 2RH1 3P0G 2RH1-epinephrine TRENDS in Pharmacological Sciences
Figure 3. Prioritizing agonists versus blockers. When ligands of the b2-adrenoceptor (b2AR) are computationally docked at an inactive structure of the receptor, such as the structure cocrystallized with the blocker carazolol [Protein Data Bank (PDB) ID: 2RH1], blockers receive better scores that agonists. Conversely, when they are docked at an activated structure of the receptor, such as the structure cocrystallized with the agonist BI-167017 (PDB ID: 3P0G), agonists receive better scores than blockers. However, if an agonist, such as epinephrine, is flexibly docked into the carazolol-bound structure of the receptor, the resulting model favors agonists over blockers. The plot illustrates the results of our study analyzed through a receiver operating characteristic (ROC) analysis. An area under the ROC curve (AUC) of 1 indicates complete prioritization of agonists versus blockers, an AUC of 0 indicates complete prioritization of blockers versus agonists, and an AUC of 0.5 indicates absence of prioritization of one ligand versus the other [62].
Concluding remarks I have illustrated strategies that have been or could be employed to model the conformational state of a GPCR in complex with biased ligands, either globally [40,41] or just related to the ligand-binding cavity [46]. One must stress that, currently, this is a remarkably challenging task. As discussed, predicting the overall conformational state of a receptor in a given signaling state probably requires, besides a biased ligand, correct modeling of the interactions of the receptors with the appropriate intracellular signaling pathways and the correct phosphorylation state of the receptor residues exposed to the cytosol. Moreover, concerning local modeling efforts, the subtle structural differences between biased and non-biased agonists may make it particularly difficult to distinguish the different conformational states of the binding cavity stabilized by different ligands. For many receptor systems, modeling attempts are hindered by the absence of known ligands endowed with a significant signaling bias. Despite these difficulties, it is likely that the task may become more feasible in the near future. First, the generation of more knowledge about biased signaling and the identification of more biased agonists will probably foster the identification of further biased ligands. Hence, if a few partially biased agonists are identified through the models that it is possible to build on current knowledge, these ligands could then be utilized to generate more accurate models subsequently applicable to the identification of agonists endowed with a stronger bias, in a virtuous cycle. Moreover, it is probable that better modeling techniques will become available for the construction of more accurate
models of both the overall and the local conformation of GPCRs in complex with specific ligands. As a general direction, the key to the generation of accurate models may be found in the expert combination of experimental and computational techniques. The availability of crystal structures solved in complex with biased ligands will no doubt assist the modeling process [37,38]. Further insights are likely to come from NMR, which is a powerful tool for conformational studies and is being applied with increasing success to the analysis of GPCRs in complex with different ligands, including biased agonists [26,39,70–73]. The structural constraints derived from NMR experiments may represent the most efficacious way to construct experimentally supported computational 3D models of GPCRs in complex with biased ligands applicable to drug discovery campaigns. Moreover, the modeling process, in particular with respect to global conformational studies, may be further assisted by the experimental identification of phosphorylation sites and the solution of the structure of the intracellular signaling partners of GPCRs [25,74]. In conclusion, as biological knowledge increases and computational techniques improve, it is likely that molecular modeling will play an increasingly relevant role in the understanding of the structural bases of biased signaling, fostering the rational discovery of GPCR ligands endowed with highly specific signaling profiles. References 1 Pierce, K. et al. (2002) Seven-transmembrane receptors. Nat. Rev. Mol. Cell Biol. 3, 639–650
5
TIPS-1132; No. of Pages 7
Opinion 2 Reiter, E. et al. (2012) Molecular mechanism of b-arrestin-biased agonism at seven-transmembrane receptors. Annu. Rev. Pharmacol. Toxicol. 52, 179–197 3 Rajagopal, S. et al. (2011) Quantifying ligand bias at 7-transmembrane receptors. Mol. Pharmacol. 80, 367–377 4 Urban, J.D. et al. (2007) Functional selectivity and classical concepts of quantitative pharmacology. J. Pharmacol. Exp. Ther. 320, 1–13 5 Drake, M.T. et al. (2008) Beta-arrestin-biased agonism at the b2adrenergic receptor. J. Biol. Chem. 283, 5669–5676 6 Martin, N.P. et al. (2004) PKA-mediated phosphorylation of the b1adrenergic receptor promotes Gs/Gi switching. Cell. Signal. 16, 1397– 1403 7 Lefkowitz, R. and Shenoy, S. (2005) Transduction of receptor signals by b-arrestins. Science 308, 512–517 8 Wisler, J.W. et al. (2007) A unique mechanism of b-blocker action: carvedilol stimulates b-arrestin signaling. Proc. Natl. Acad. Sci. U.S.A. 104, 16657–16662 9 Kahsai, A.W. et al. (2011) Multiple ligand-specific conformations of the b2-adrenergic receptor. Nat. Chem. Biol. 7, 692–700 10 Gesty-Palmer, D. et al. (2006) Distinct b-arrestin- and G proteindependent pathways for parathyroid hormone receptor-stimulated ERK1/2 activation. J. Biol. Chem. 281, 10856–10864 11 Walters, R.W. et al. (2009) b-arrestin1 mediates nicotinic acid-induced flushing, but not its antilipolytic effect, in mice. J. Clin. Invest. 119, 1312–1321 12 Monasky, M.M. et al. (2013) The b-arrestin-biased ligand TRV120023 inhibits angiotensin II-induced cardiac hypertrophy while preserving enhanced myofilament response to calcium. Am. J. Physiol. Heart Circ. Physiol. 305, H856–H866 13 Violin, J.D. et al. (2013) GPCR biased ligands as novel heart failure therapeutics. Trends Cardiovasc. Med. 23, 242–249 14 Boerrigter, G. et al. (2012) TRV120027, a novel b-arrestin biased ligand at the angiotensin II type I receptor, unloads the heart and maintains renal function when added to furosemide in experimental heart failure. Circ. Heart Fail. 5, 627–634 15 Kim, K.S. et al. (2012) b-arrestin-biased AT1R stimulation promotes cell survival during acute cardiac injury. Am. J. Physiol. Heart Circ. Physiol. 303, H1001–H1010 16 Boerrigter, G. et al. (2011) Cardiorenal actions of TRV120027, a novel b-arrestin-biased ligand at the angiotensin II type I receptor, in healthy and heart failure canines: a novel therapeutic strategy for acute heart failure. Circ. Heart Fail. 4, 770–778 17 DeWire, S.M. and Violin, J.D. (2011) Biased ligands for better cardiovascular drugs: dissecting G-protein-coupled receptor pharmacology. Circ. Res. 109, 205–216 18 Shukla, A.K. et al. (2008) Distinct conformational changes in b-arrestin report biased agonism at seven-transmembrane receptors. Proc. Natl. Acad. Sci. U.S.A. 105, 9988–9993 19 Zhou, L. et al. (2013) Development of functionally selective, small molecule agonists at kappa opioid receptors. J. Biol. Chem. 288, 36703–36716 20 White, K.L. et al. (2013) Identification of novel functionally selective kappa opioid receptor scaffolds. Mol. Pharmacol. 85, 83–90 21 Chen, X.T. et al. (2013) Structure–activity relationships and discovery of a G protein biased mu opioid receptor ligand, [(3-methoxythiophen2-yl)methyl]({2-[(9R)-9-(pyridin-2-yl)-6-oxaspiro-[4.5]decan-9yl]ethyl})amine (TRV130), for the treatment of acute severe pain. J. Med. Chem. 56, 8019–8031 22 Lane, J.R. et al. (2013) Structure-based ligand discovery targeting orthosteric and allosteric pockets of dopamine receptors. Mol. Pharmacol. 84, 794–807 23 Palczewski, K. and Orban, T. (2013) From atomic structures to neuronal functions of G protein-coupled receptors. Annu. Rev. Neurosci. 36, 139–164 24 Rasmussen, S.G. et al. (2011) Crystal structure of the b2 adrenergic receptor–Gs protein complex. Nature 477, 549–555 25 Nobles, K.N. et al. (2011) Distinct phosphorylation sites on the b2adrenergic receptor establish a barcode that encodes differential functions of b-arrestin. Sci. Signal. 4, ra51 26 Nygaard, R. et al. (2013) The dynamic process of b2-adrenergic receptor activation. Cell 152, 532–542 27 Kobilka, B. (2007) G protein coupled receptor structure and activation. Biochim. Biophys. Acta 1768, 794–807 6
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
28 Kobilka, B.K. and Deupi, X. (2007) Conformational complexity of Gprotein-coupled receptors. Trends Pharmacol. Sci. 28, 397–406 29 Yao, X. et al. (2006) Coupling ligand structure to specific conformational switches in the b2-adrenoceptor. Nat. Chem. Biol. 2, 417–422 30 Warne, T. and Tate, C.G. (2013) The importance of interactions with helix 5 in determining the efficacy of b-adrenoceptor ligands. Biochem. Soc. Trans. 41, 159–165 31 Stevens, R.C. et al. (2013) The GPCR Network: a large-scale collaboration to determine human GPCR structure and function. Nat. Rev. Drug Discov. 12, 25–34 32 Katritch, V. et al. (2013) Structure–function of the G protein-coupled receptor superfamily. Annu. Rev. Pharmacol. Toxicol. 53, 531–556 33 Venkatakrishnan, A.J. et al. (2013) Molecular signatures of G-proteincoupled receptors. Nature 494, 185–194 34 Tate, C.G. and Schertler, G.F. (2009) Engineering G protein-coupled receptors to facilitate their structure determination. Curr. Opin. Struct. Biol. 19, 386–395 35 Kruse, A.C. et al. (2013) Applications of molecular replacement to G protein-coupled receptors. Acta Crystallogr. D: Biol. Crystallogr. 69, 2287–2292 36 Steyaert, J. and Kobilka, B.K. (2011) Nanobody stabilization of G protein-coupled receptor conformational states. Curr. Opin. Struct. Biol. 21, 567–572 37 Warne, T. et al. (2012) Crystal structures of a stabilized b1adrenoceptor bound to the biased agonists bucindolol and carvedilol. Structure 20, 841–849 38 Wacker, D. et al. (2013) Structural features for functional selectivity at serotonin receptors. Science 340, 615–619 39 Liu, J.J. et al. (2012) Biased signaling pathways in b2-adrenergic receptor characterized by 19F-NMR. Science 335, 1106–1110 40 Tikhonova, I.G. et al. (2013) Simulations of biased agonists in the b2 adrenergic receptor with accelerated molecular dynamics. Biochemistry 52, 5593–5603 41 Balabin, I.A. et al. (2009) Coarse-grained modeling of allosteric regulation in protein receptors. Proc. Natl. Acad. Sci. U.S.A. 106, 14253–14258 42 Provasi, D. et al. (2011) Ligand-induced modulation of the free-energy landscape of G protein-coupled receptors explored by adaptive biasing techniques. PLoS Comput. Biol. 7, e1002193 43 Dror, R.O. et al. (2011) Activation mechanism of the b2-adrenergic receptor. Proc. Natl. Acad. Sci. U.S.A. 108, 18684–18689 44 Dror, R.O. et al. (2011) Pathway and mechanism of drug binding to Gprotein-coupled receptors. Proc. Natl. Acad. Sci. U.S.A. 108, 13118– 13123 45 Katritch, V. et al. (2009) Analysis of full and partial agonists binding to b2-adrenergic receptor suggests a role of transmembrane helix V in agonist-specific conformational changes. J. Mol. Recognit. 22, 307–318 46 Vilar, S. et al. (2011) In silico analysis of the binding of agonists and blockers to the b2-adrenergic receptor. J. Mol. Graph. Model. 29, 809– 817 47 Vilar, S. et al. (2011) Docking-based virtual screening for ligands of G protein-coupled receptors: not only crystal structures but also in silico models. J. Mol. Graph. Model. 29, 614–623 48 Cavasotto, C.N. (2012) Normal mode-based approaches in receptor ensemble docking. Methods Mol. Biol. 819, 157–168 49 Cavasotto, C.N. and Abagyan, R.A. (2004) Protein flexibility in ligand docking and virtual screening to protein kinases. J. Mol. Biol. 337, 209– 225 50 Cavasotto, C.N. et al. (2005) Representing receptor flexibility in ligand docking through relevant normal modes. J. Am. Chem. Soc. 127, 9632– 9640 51 Rueda, M. et al. (2009) Consistent improvement of cross-docking results using binding site ensembles generated with elastic network normal modes. J. Chem. Inf. Model. 49, 716–725 52 Reynolds, K. et al. (2009) Identifying conformational changes of the b2 adrenoceptor that enable accurate prediction of ligand/receptor interactions and screening for GPCR modulators. J. Comput. Aided Mol. Des. 23, 273–288 53 Phatak, S.S. et al. (2010) Ligand-steered modeling and docking: a benchmarking study in class A G-protein-coupled receptors. J. Chem. Inf. Model. 50, 2119–2128
TIPS-1132; No. of Pages 7
Opinion 54 Katritch, V. and Abagyan, R. (2011) GPCR agonist binding revealed by modeling and crystallography. Trends Pharmacol. Sci. 32, 637–643 55 Rueda, M. et al. (2012) ALiBERO: evolving a team of complementary pocket conformations rather than a single leader. J. Chem. Inf. Model. 52, 2705–2714 56 Katritch, V. et al. (2012) Ligand-guided receptor optimization. Methods Mol. Biol. 857, 189–205 57 Bhattacharya, S. et al. (2008) Ligand-stabilized conformational states of human b2 adrenergic receptor: insight into G-protein-coupled receptor activation. Biophys. J. 94, 2027–2042 58 Bhattacharya, S. and Vaidehi, N. (2010) Computational mapping of the conformational transitions in agonist selective pathways of a G-protein coupled receptor. J. Am. Chem. Soc. 132, 5205–5214 59 Bhattacharya, S. and Vaidehi, N. (2012) LITiCon: a discrete conformational sampling computational method for mapping various functionally selective conformational states of transmembrane helical proteins. Methods Mol. Biol. 914, 167–178 60 Bottegoni, G. et al. (2008) A new method for ligand docking to flexible receptors by dual alanine scanning and refinement (SCARE). J. Comput. Aided Mol. Des. 22, 311–325 61 Dror, R.O. et al. (2013) Structural basis for modulation of a G-proteincoupled receptor by allosteric drugs. Nature 503, 295–299 62 Costanzi, S. and Vilar, S. (2012) In silico screening for agonists and blockers of the b2 adrenergic receptor: implications of inactive and activated state structures. J. Comput. Chem. 33, 561–572 63 Cazzola, M. et al. (2013) b2-agonist therapy in lung disease. Am. J. Respir. Crit. Care Med. 187, 690–696
Trends in Pharmacological Sciences xxx xxxx, Vol. xxx, No. x
J.A. and Liggett, S.B. (2011) Cardiovascular 64 Johnson, pharmacogenomics of adrenergic receptor signaling: clinical implications and future directions. Clin. Pharmacol. Ther. 89, 366–378 65 Rasmussen, S.G. et al. (2011) Structure of a nanobody-stabilized active state of the b2 adrenoceptor. Nature 469, 175–180 66 Weiss, D.R. et al. (2013) Conformation guides molecular efficacy in docking screens of activated b2 adrenergic G protein coupled receptor. ACS Chem. Biol. 8, 1018–1026 67 Kolb, P. et al. (2009) Structure-based discovery of b2-adrenergic receptor ligands. Proc. Natl. Acad. Sci. U.S.A. 106, 6843–6848 68 Jacobson, K.A. and Costanzi, S. (2012) New insights for drug design from the X-ray crystallographic structures of G-protein-coupled receptors. Mol. Pharmacol. 82, 361–371 69 Sherman, W. et al. (2006) Novel procedure for modeling ligand/receptor induced fit effects. J. Med. Chem. 49, 534–553 70 Kim, T.H. et al. (2013) The role of ligands on the equilibria between functional states of a G protein-coupled receptor. J. Am. Chem. Soc. 135, 9465–9474 71 Bokoch, M. et al. (2010) Ligand-specific regulation of the extracellular surface of a G-protein-coupled receptor. Nature 463, 108–112 72 Horst, R. et al. (2013) b2-Adrenergic receptor activation by agonists studied with 19F NMR spectroscopy. Angew. Chem. Int. Ed. Engl. 52, 10762–10765 73 Horst, R. et al. (2013) b2-Adrenergic receptor solutions for structural biology analyzed with microscale NMR diffusion measurements. Angew. Chem. Int. Ed. Engl. 52, 331–335 74 Shukla, A.K. et al. (2013) Structure of active b-arrestin-1 bound to a Gprotein-coupled receptor phosphopeptide. Nature 497, 137–141
7
