1. Biochemistry and Chemical Biology
Download icon

Concerted conformational dynamics and water movements in the ghrelin G protein-coupled receptor

  1. Maxime Louet
  2. Marina Casiraghi
  3. Marjorie Damian
  4. Mauricio GS Costa
  5. Pedro Renault
  6. Antoniel AS Gomes
  7. Paulo R Batista
  8. Céline M'Kadmi
  9. Sophie Mary
  10. Sonia Cantel
  11. Severine Denoyelle
  12. Khoubaib Ben Haj Salah
  13. David Perahia
  14. Paulo M Bisch
  15. Jean-Alain Fehrentz
  16. Laurent J Catoire
  17. Nicolas Floquet
  18. Jean-Louis Banères  Is a corresponding author
  1. IBMM, Univ Montpellier, CNRS, ENSCM, France
  2. Laboratoire de Biologie Physico-Chimique des Protéines Membranaires, UMR 7099, CNRS, Université de Paris, Institut de Biologie Physico-Chimique (FRC 550), France
  3. Laboratoire de Biologie et Pharmacologie Appliquées, UMR 8113 CNRS, Ecole Normale Supérieure Paris-Saclay, France
  4. Programa de Computação Científica, Fundação Oswaldo Cruz, Brazil
  5. Laboratório de Física Biológica, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Brazil
Research Article
  • Cited 0
  • Views 370
  • Annotations
Cite this article as: eLife 2021;10:e63201 doi: 10.7554/eLife.63201

Abstract

There is increasing support for water molecules playing a role in signal propagation through G protein-coupled receptors (GPCRs). However, exploration of the hydration features of GPCRs is still in its infancy. Here, we combined site-specific labeling with unnatural amino acids to molecular dynamics to delineate how local hydration of the ghrelin receptor growth hormone secretagogue receptor (GHSR) is rearranged upon activation. We found that GHSR is characterized by a specific hydration pattern that is selectively remodeled by pharmacologically distinct ligands and by the lipid environment. This process is directly related to the concerted movements of the transmembrane domains of the receptor. These results demonstrate that the conformational dynamics of GHSR are tightly coupled to the movements of internal water molecules, further enhancing our understanding of the molecular bases of GPCR-mediated signaling.

Introduction

G protein-coupled receptors (GPCRs) are major players in many central biological processes (Lagerström and Schiöth, 2008). The diversity in the signaling properties of GPCRs indicates that this process cannot be fully described by the limited number of conformational states captured by X-ray crystallography and cryoelectron microscopy (cryo-EM). Indeed, GPCRs likely explore complex conformational landscapes, characterized by several meta-stable structural states. The relative distribution of these states is controlled by ligands, signaling proteins, and the environment, ultimately dictating the signaling output (Casiraghi et al., 2019; Hilger et al., 2018; Wingler and Lefkowitz, 2020). As a consequence, the conformational dynamics of GPCRs and its modulation by the receptor’s environment are under intense scrutiny, as this should illuminate how signal transduction occurs.

Among all the components in the receptor environment that control receptor dynamics and signaling behavior, the solvent is to play an important but yet unexplored role. Many GPCR experimental structures indicate the occurrence of water molecules within their transmembrane (TM) regions. Some are located in the ligand-binding pocket and directly contribute to the energetics of ligand binding (Deflorian et al., 2020). Others lie in different cavities and have been proposed to be central to the allosteric propagation of the conformational rearrangements required for receptor activation (Lesca et al., 2018; Huang et al., 2015; Zhao et al., 2020). Yet, mechanistic models describing the dynamics and role of water molecules in GPCR functioning remain speculative, as only limited experimental information is available. Indeed, the relationship between GPCR conformational dynamics and local hydration has been mostly inferred from molecular dynamics (MD) simulations with rhodopsin (Grossfield et al., 2008) and other receptors as well (Yuan et al., 2014; Venkatakrishnan et al., 2019). Although these simulations provide invaluable information on the arrangement and movements of water molecules, they nevertheless require further experimental support.

To analyze the hydration pattern of GPCRs, we used here an original strategy initially described with a model soluble protein (Amaro et al., 2015). This strategy combines site-specific labeling with unnatural amino acids (UAAs), fluorescence spectroscopy, and MD. This approach was applied to the growth hormone secretagogue receptor (GHSR). In addition to being a model for class A GPCRs, GHSR is a major target in pharmacology. Indeed, this receptor and its natural peptide agonist, ghrelin, are involved in most important biological processes such as the control of food intake, glucose metabolism, or reward and stress behaviors (Müller et al., 2015). Using the emission properties of a particular UAA whose fluorescence emission properties are related to its hydration (Amaro et al., 2015; Choudhury et al., 2008), we found here that the ghrelin receptor hydration pattern is likely remodeled by orthosteric ligands and the lipid environment. In parallel, MD simulations provided a structural framework to the fluorescence observations and demonstrated that such a remodeling may be associated with collective movements of GHSR TM domains. Taken together, our data illuminate GPCR signaling with a mechanism where specific changes in the local hydration of the receptor could occur in a concerted manner with its conformational dynamics, in direct relationship to the activation process.

Results

GHSR labeling

We used L-(7-hydroxycoumarin-4-yl)-ethylglycine as a reporter of receptor local hydration. This UAA contains the L-(7-hydroxycoumarin-4-yl) (7H4MC) moiety whose emission properties are correlated to the presence of water molecules in its vicinity (Amaro et al., 2015). 7H4MC-ethylglycine was synthesized as described in the Materials and methods section and introduced in GHSR using codon suppression technology (Wang et al., 2006). To analyze receptor activation in a relevant membrane-like environment, the labeled receptor was inserted into lipid nanodiscs formed by the scaffolding MSP1E3D1 protein and a POPC:POPG mixture (see Materials and methods) (Damian et al., 2012). Under such conditions, homogeneous nanodisc populations of functional receptors were obtained (Figure 1—figure supplement 1). Of importance, the active receptor was purified through a ligand affinity chromatography step to ensure all the receptors in our preparations were competent with regard to ligand binding (Ferré et al., 2019). Accordingly, we repeatedly demonstrated that the receptor obtained under such conditions is totally functional with regard to ligand binding, and that its pharmacological profile is closely related to that of GHSR expressed in HEK cell membranes (Ferré et al., 2019).

7H4MC-ethylglycine was introduced at several, single positions within the TM domains of GHSR, namely Y812.42, W1042.65, Y1062.67, F1193.28, F1213.30, I1343.43, F1794.61, W2155.41, Y2325.58, V2686.40, F2726.44, Y3037.33, or S3157.45 (superscript numbers follow Ballesteros-Weinstein numbering [Ballesteros and Weinstein, 1995; Figure 1A]). In all the cases, protein expression yields markedly decreased but, with the exception of the F1193.28 mutant that was not expressed at a detectable level, the amounts of purified receptor obtained were still compatible with the fluorescence experiments, that is, in the range of a hundred of µg per liter of bacterial culture. However, the modified receptors bearing 7H4MC-ethylglycine at position Y812.42, W1042.65, Y1062.67, F1193.28, F1213.30, F1794.61, and Y3037.43 could not be purified through the ligand affinity chromatography step, indicating that replacing the native residue with 7H4MC-ethylglycine affected their three-dimensional fold and/or their ability to bind ligands. In addition, replacing W2155.41 with 7H4MC-ethylglycine markedly decreased the basal activity of the receptor, although this mutant could still bind its ligands and be activated by ghrelin. In contrast, for the other positions, that is, I1343.43, Y2325.58, V2686.40, F2726.44, and S3157.45, replacing the naturally occurring residue with 7H4MC-ethylglycine affected neither ghrelin binding nor the receptor-catalyzed Gq activation in a relevant manner (Figure 1B, C, Figure 1—figure supplement 2, 3, 4, 5, 6). Hence, only these mutants were considered in our analyses. As shown in Figure 1D, the modified proteins displayed an emission spectrum characteristic of the 7H4MC moiety, while the wild-type receptor had no significant emission signal when excited at the same wavelength. This indicates an efficient incorporation of the labeled UAA into the receptor.

Figure 1 with 6 supplements see all
Growth hormone secretagogue receptor (GHSR) labeling.

(A) Position of the labeled residues in GHSR sequence. Red labeling indicates positions that were deleterious to GHSR expression and/or function. Green labeling indicates positions that did not markedly affect the pharmacological properties of the isolated receptor and were considered in the present work. (B) FRET-monitored competition assays of ghrelin for binding to GHSR assembled into nanodiscs. (C) GTP turnover for Gq catalyzed by GHSR and its labeled counterparts in the absence of ligand (apo) or in the presence of 10 µM of JMV3011, ghrelin, JMV3002, or SPA (substance-P analog). (D) Normalized emission spectrum of the apo wild-type and labeled GHSR with λexc set at 320 nm. Data in (B) and (C) is the mean value ± SD of three experiments. Statistical analyses for the data in (C) are provided in Figure 1—figure supplement 3.

GHSR local hydration

We then investigated whether the fluorescence properties of 7H4MC-ethylglycine could report on the local hydration features of GHSR. To this end, we analyzed the 7H4MC emission profile for each of the positions considered. An excitation wavelength of 320 nm was systematically used to excite the neutral form of the fluorophore (Amaro et al., 2015). The emission spectra were deconvoluted into their separate components using the procedure initially described (Amaro et al., 2015). A hydration parameter H was then determined that corresponded to the sum of the contributions of the anionic and tautomer forms. This parameter is an indicator of the extent of hydration at the position considered, as the higher the H parameter the higher local hydration (Amaro et al., 2015). A difference in the H parameter inferred for the 7H4MC probe at the different positions of the apo GHSR was observed depending on the position considered (Figure 2—figure supplement 1, 2). Indeed, all positions were hydrated to some extent, but some displayed a high H value characteristic of high hydration (3.43, 6.40) whereas others displayed a low H value suggestive of a lower local hydration (5.58, 6.44, 7.45). This indicates that 7H4MC fluorescence is a good indicator to discriminate between different local hydration states in the receptor structure. Besides, these data show that local hydration, as reported by 7H4MC fluorescence, depends on the region of the TM domain considered, with some regions more accessible to the solvent than others, even for closely related positions in the receptor structure (e.g., V2686.40 and F2726.44).

Impact of ligands on GHSR local hydration

We then used 7H4MC fluorescence to monitor the impact of ligand binding on the hydration pattern of GHSR. To this end, the H parameter was measured in the presence of saturating concentrations in ligands from different pharmacological classes, that is, the natural full agonist (ghrelin), a neutral antagonist (JMV3011), a Gq-biased partial agonist (JMV3002), and an inverse agonist (substance-P analog [SPA]) (M'Kadmi et al., 2015Figure 2—figure supplement 3). Binding of JMV3011 to labeled GHSR was not accompanied by a measurable change in the hydration parameter for any of the positions considered (Figure 2, Figure 2—figure supplement 4). This is to be related to our previous observations demonstrating that binding of this compound was not associated with any change in the conformation of isolated GHSR (Mary et al., 2012; Damian et al., 2015). In contrast, changes in the H parameter were observed at some specific positions upon binding of either the full agonist ghrelin, the Gq-biased agonist JMV3002, or the inverse agonist SPA. Specifically, ghrelin binding was associated with an increase in the hydration parameter at position 5.58 while local hydration at position 6.44 decreased (Figure 2—figure supplement 4, 5, 6). Besides 5.58 and 6.44, no relevant change in the hydration parameter was observed for the other positions (Figure 2—figure supplement 4). This indicates that agonist-induced GHSR activation is accompanied by a concerted parallel increase and decrease of the local hydration in specific regions of the receptor, namely here TM5 and TM6. Interestingly, no change was observed for V2686.40 whereas a decrease in the H parameter was measured for the probe at F2726.44, although both positions are close in GHSR structure. This suggests that 7H4MC fluorescence is well adapted to monitor hydration changes in a very local environment and that changes in local hydration likely occur at specific, spatially restricted sites. Perhaps not surprisingly, binding of the inverse agonist SPA was accompanied by a change in the hydration parameter opposite to that observed with ghrelin, that is, the H parameter decreased and increased for positions 5.58 and 6.44, respectively (Figure 2, Figure 2—figure supplement 4), consistent with the opposite effect of ghrelin and SPA on GHSR activation and conformational landscape (Mary et al., 2012). Finally, the hydration pattern in the presence of JMV3002 was different from that observed in the presence of ghrelin. Indeed, while the binding of this compound was still accompanied by a decrease in the hydration parameter at position 6.44, as in the case of ghrelin binding, no change was observed for the probe at position 5.58 (Figure 2, Figure 2—figure supplement 4). This could be related to the differences in the pharmacological profile of the two compounds, as ghrelin is a full agonist whereas JMV3002 is a Gq partial agonist that triggers neither Gi activation nor arrestin recruitment (M'Kadmi et al., 2015).

Figure 2 with 6 supplements see all
Local hydration of growth hormone secretagogue receptor (GHSR) as a function of ligands.

H parameter for the 7H4MC-labeled GHSR in the absence of ligand (apo) and in the presence of JMV3011, ghrelin, JMV3002, or SPA (substance-P analog). All ligands were used at a 10 µM concentration. In all cases, the data represents the mean value ± SD of three experiments. Statistical analyses are provided in Figure 1—figure supplement 2 and 4.

Structural bases of the changes in water accessibility

To provide a structural framework to our experimental observations and observe possible differences between the hydration pattern of inactive and active-like conformers of wild-type GHSR, we then ran five MD simulations of 5 µs for each conformational state of the receptor, totalizing 50 µs of all-atoms simulation. The crystal structure of the inactive, antagonist-loaded state of the receptor has been solved (Shiimura et al., 2020) and was used as a starting point for our MD studies. Besides, two cryo-EM structures of the ghrelin receptor in complex with ghrelin or a synthetic agonist and a Gq mimetic have been posted on the BioRxiv preprint server (https://doi.org/10.1101/2021.06.09.447478). However, since the coordinates of these structures are not yet available, we had to model an active-like state of GHSR in the absence of its cognate G protein (see Materials and methods). A projection of all conformers explored during our simulations confirmed their compatibility with experimental structures, describing mainly inactive and intermediate states (based on the classification in the GPCRdb; Pándy-Szekeres et al., 2018), the latter corresponding to an activated receptor without the G protein (Figure 3—figure supplement 1).

Interestingly, analysis of the statistical water distribution in GHSR confirmed that differences in the hydration pattern could exist depending on its inactive/active states (Figure 3). More importantly, these differences effectively occurred in the regions where the 7H4MC-ethylglycine residue had been inserted in our experiments. Four out of five simulations starting from the X-ray (inactive) structure of GHSR converged toward a same hydration pattern (Figure 3A-D). In the last simulation (Figure 3E), the water statistically occupied a larger volume on the intracellular side of the receptor. Of interest, this distribution of water molecules in the inactive state was in agreement with the distribution described recently for other receptors of the same family using similar methods (Venkatakrishnan et al., 2019; Bertalan et al., 2020). In agreement with our experimental data, the water distribution in the inactive conformation of the receptor showed the presence of water molecules around F2726.44 whereas Y2325.58 was not solvated. Accordingly, I1343.43 was also in contact with water molecules. However, and in contradiction with our experiments, V2686.40 was not in contact with water molecule in the inactive state whereas S3157.45 was. V2686.40 occupies a central position in the receptor whereas S3157.45 is close to the interface between TM6 and TM7 (Figure 4). If insertion of 7H4MC-ethylglycine at positions 1343.43, 2325.58, and 2726.44 are more conservative in terms of residue size, the insertion of 7H4MC in place of a valine (V2686.40) or a serine (S3157.45) suggests a stronger adaptation of the receptor fold to these mutations. To clarify the possible orientations of 7H4MC-ethylglycine into the receptor, we thus computed adiabatic maps for positions 2686.40 and 3157.45 (Figure 4—figure supplement 1). In the case of 7H4MC-ethylglycine at position 2686.40, the adiabatic map confirmed that this large residue, in comparison to a valine, allowed interaction with solvent molecules (Figure 4—figure supplement 1). For the 3157.45 position, even the adiabatic map suggested that this residue could conserve its initial orientation toward the interior of the receptor and should be highly solvated.

Figure 3 with 1 supplement see all
Water distribution in growth hormone secretagogue receptor (GHSR) as observed along the five independent 5 µs molecular dynamics (MD) simulations starting from either the inactive (A to E) or the active (F to J) states of the receptor.

(E and J) panels show simulations where GHSR transited from inactive to active (E) or from active to inactive (J) states, respectively. The backbone of the protein is represented as a transparent-white cartoon, while the five positions at which the L-(7-hydroxycoumarin-4-yl) (7H4MC)-ethylglycine residue was inserted are shown in licorice. Blue or orange surfaces respectively describe the hydration of the receptor using a probability of 0.3. Volumetric maps were computed using the volmap tool of visual molecular dynamics (VMD).

Figure 4 with 5 supplements see all
Amino acid positions and hydration patterns of inactive growth hormone secretagogue receptor (GHSR) (A) and active GHSR (B) explored by molecular dynamics (MD) simulations.

GHSR is represented in white ribbons. Volumetric maps in solid surface represent the water distribution with a probability of presence of 0.3. Meshes represent the most probable (probability of 0.3) positions of residues I1343.43, Y2325.58, V2686.40, F2726.44, and S3157.45 in both states. Snapshots representing the residues in their respective shapes are drawn in licorice for visualization. Volumetric maps were computed using the volmap tool of visual molecular dynamics (VMD).

Interestingly, a different hydration pattern was found in the fifth simulation. This profile was explained by a spreading of TM6 during the simulation, thus leading to conformers close to those observed when starting from the active-like state (Figure 4—figure supplement 2). Accordingly, the resulting hydration pattern was very close to that obtained in the simulations starting from the active-like state (Figure 3F-I). In this pattern, water molecules were more uniformly distributed in the receptor including its lower, intracellular part. Indeed, the main structural difference between both states was the spreading of TM6 (Figure 4—figure supplement 2), which contributed to a large water influx into the intracellular moiety of the receptor. In this pattern, and in agreement with the fluorescence experiments, I1343.43, Y2326.44, and V2686.40 were all in close contact with water molecules whereas F2726.44 was flipped toward TM5, contributing to reduce its interactions with surrounding water (Figure 4—figure supplement 3). This structural feature was found to be conserved in all GPCRs (Figure 4).

We also obtained a simulation starting from the active state that showed a different behavior from other simulations, that is, a closure and a loss of hydration in the intracellular part of the receptor due to a motion of TM7 inside the receptor (Figure 3J). Such a motion was compatible with the direction coded by the experimental structures and shown by principal component analysis (PCA), where the first two eigenvectors displayed this inward motion of TM7 concomitant to TM6 spreading (Figure 4—figure supplement 4). Accordingly, in this simulation, the resulting hydration pattern was very close to those observed in the simulations starting from the inactive state (Figure 3A-D).

Impact of lipids on GHSR local hydration

In their native environment, receptors are surrounded not only by the solvent but also by the lipid bilayer. To provide an illustration of the impact of the environment of the ghrelin receptor on its local hydration, we finally analyzed the effect of the lipid composition of the nanodiscs on 7H4MC fluorescence for the two positions that were affected by receptor activation. Specifically, we measured the hydration parameter for the probe at positions 5.58 and 6.44 with GHSR assembled into POPC:POPG nanodiscs in the absence or presence of phosphatidylinositol-4,5-bisphosphate (PIP2), a lipid that has been shown to impact on the activity of many different membrane proteins (Hammond and Burke, 2020) including GPCRs (Yen et al., 2018) and, more recently, the ghrelin receptor (Damian et al., 2021). As shown in Figure 5, the H parameter for the two positions that were sensitive to receptor activation was further affected by PIP2. Indeed, adding 2.5% of this lipid to the nanodiscs increased the H parameter for position 5.58 and concomitantly decreased that for position 6.44. PIP2 therefore exalted the effect of the full agonist ghrelin had on the receptor hydration features. This suggests that PIP2 further shifts the conformational equilibrium toward hydration pattern associated with the active-like conformation of GHSR, indicative of an allosteric coupling between the full agonist and PIP2 for stabilizing this state. This effect could explain the impact of PIP2 on GHSR-catalyzed G protein activation (Damian et al., 2021).

Impact of lipids on the local hydration of growth hormone secretagogue receptor (GHSR).

H parameter for L-(7-hydroxycoumarin-4-yl) (7H4MC)-labeled GHSR assembled into nanodiscs containing or not phosphatidylinositol-4,5-bisphosphate (PIP2) (2.5% PIP2-to-total lipids molar ratio), in the presence of 10 µM ghrelin. The data represents the mean value ± SD of three experiments.

Figure 5—source data 1

H parameter as a function of PIP2 in the nanodiscs.

https://cdn.elifesciences.org/articles/63201/elife-63201-fig5-data1-v1.xlsx

Discussion

Local hydration and polar networks have been proposed to play a role in the allosteric activation process of GPCRs, as it is the case for many other proteins (Leitner et al., 2020). However, illuminating this role is complicated by the lack of straightforward experimental approaches that could help delineate the hydration pattern of receptors under a variety of conditions. To decipher the concerted changes in the hydration pattern and conformational repertoire of GHSR, we used here a strategy combining advanced MD to fluorescence spectroscopy, an approach that had been previously developed with a model protein (Amaro et al., 2015). Specifically, 7H4MC-ethylglycine was incorporated at specific positions in GHSR through codon suppression technology. 7H4MC-ethylglycine includes the 7H4MC fluorophore whose emission properties are directly dependent on the water content in its microenvironment (Georgieva et al., 2005). Of importance, this chromophore has been shown not to affect the global hydration levels but, in some particular situations, only to influence the residence time of water molecules (Georgieva et al., 2005). It therefore should not markedly modify the hydration pattern of the protein, with the condition it does not affect its three-dimensional fold (Georgieva et al., 2005).

The fluorescent probe was introduced at different positions along GHSR sequence. Among all the mutants we considered, only those involving I1343.43, Y2325.58, V2686.40, F2726.44, and S3157.45 were essentially neutral with regard to the ability of the receptor to bind ghrelin and activate Gq. This suggests that incorporation of 7H4MC-ethylglycine at these positions does not dramatically perturb the three-dimensional fold of the ghrelin receptor. The absence of major effect on substituting F2726.44 with 7H4MC-ethylglycine on GHSR functioning was puzzling, as mutating this residue to an alanine had been shown to abolish GHSR constitutive activity and reduce ghrelin-induced signaling without affecting ghrelin binding affinity (Valentin-Hansen et al., 2012). However, replacing the phenylalanine with a tyrosine had a far lower impact on both basal- and ghrelin-induced GHSR activity (Valentin-Hansen et al., 2012). As concluded by the authors, a possible mechanism would be that a rigid, aromatic group is required at this position for stabilizing the receptor active state. This could explain why replacing phenylalanine with 7H4MC-ethylglycine did not lead to a major decrease in GHSR activity.

The fluorescence properties of 7H4MC were affected by the activation state of GHSR only when the probe was located at positions 5.58 and 6.44. Besides these two positions, highly hydrated, invariant positions were observed. These could correspond to regions of the receptor directly accessible to the solvent whatever its activation state is. Alternatively, hydration at these positions could include the contribution of water molecules with a structural rather than a functional role. Changes in the hydration pattern of GPCRs have been shown either to involve a direct rearrangement of the water molecules in the receptor ligand-binding pocket upon ligand binding or to result from the conformational changes associated with receptor activation. The two positions where we observed a change in the H parameter upon agonist binding, that is, Y2325.58 and F2726.44, are both located outside the major putative ligand-binding pocket of GHSR (Shiimura et al., 2020). The effects we observed on the H parameter upon ligand binding are thus not likely to be the direct consequence of the rearrangement of water molecules within the GHSR binding pocket but may rather result from differences in the hydration pattern of the different conformational states of GHSR, as fully supported by our MD simulations.

The emission profile of 7H4MC reflects the average, equilibrium contribution of the hydration pattern in the different GHSR states present in the solution. Any change in the emission properties of the probe thus implies (i) a change in the distribution of the different states in the receptor conformational landscape and (ii) a difference in the hydration features of these states. Taken together, the variations we observed therefore demonstrate that the remodeling of the hydration pattern is an integral component of the rearrangement of the GHSR conformational repertoire associated with receptor activation. This remodeling is directly correlated to concerted, specific and well-defined movements in the TM domains of the receptor, as demonstrated by MD. This is fully consistent with previous work demonstrating the role of water molecules in the activation process of other GPCRs such as the GLP-1 receptor (Zhao et al., 2020; Wootten et al., 2016a; Wootten et al., 2016b). Whether our data reflects the fact that water molecules are allosteric players in the activation process, as demonstrated for other proteins including rhodopsin (Chawla et al., 2021), or that the movements of water molecules compensate the changes in the intramolecular voids within the different states involved in receptor activation remains an open question.

In parallel to the fluorescence approach, we used a complementary approach based on MD simulations that was aimed at providing a structural framework to the experimental observables. This method is dedicated to the exploration of the conformational space accessible to the receptor without any bias. Indeed, structures from the Protein Data Bank (PDB) can help in identifying hotspots for protein-water interactions, but in more than half of the available structures the resolution is not good enough to observe any water molecules. Moreover, static structures do not inform about the dynamical behavior of water molecule inside the receptor. Structures often show isolated water molecule in interaction with the protein and give no clue about the global hydration of the receptor’s pockets. MD simulations of inactive and active state of GHSR allowed us to analyze a fully solvated receptor, where the global hydration pattern varied as a function of the conformational state of the receptor. Even though we simulated the wild-type GHSR, sidechain positions and hydration patterns in both states were compatible with our experiment for positions 1343.43, 2325.58, 2686.40, and 2726.44. In contrast, however, the hydration at position 3157.45 did not agree with our experimental results. A possibility would be that the size of the residue itself excludes water molecules from its vicinity, and/or orients it toward the membrane, thus explaining the low hydration in the inactive and the active states of the receptor we measured experimentally.

Interestingly, a difference in the GHSR hydration pattern was observed depending on whether the ligand was a full or a Gq-biased agonist. This indicates a different arrangement of the water network in the conformational states stabilized by these ligands. This is in line with previous observations with NK1R where mutation of the water hydrogen bond network affected Gq- and Gs-mediated signaling in a different way (Valentin-Hansen et al., 2015). In the same way, the central polar network in the GLP1 receptor has been suggested to be critical for G protein-dependent but not for G protein-independent signaling (Wootten et al., 2016a). Taken together, this data indicates that the different states in the GHSR conformational landscape differ in their hydration pattern, as stated above but, in addition, that ligands with different pharmacological profiles, here a full and a partial, biased agonist, have a different impact on the distribution of these states. This conclusion with GHSR is consistent with our previous data using a different conformational reporter, monobromobimane (Mary et al., 2012).

In addition to ligands, other components in the receptor environment allosterically impact on GPCR activation and conformational dynamics. This is the case of the lipid bilayer whose composition has been shown to impact on the structure and function of many different membrane proteins including ion channels (Hille et al., 2015) and receptors (Strohman et al., 2019; Dawaliby et al., 2016; Casiraghi et al., 2016). Among the lipids reported to affect GPCR signaling such as cholesterol (Casiraghi et al., 2016; Zocher et al., 2012) or charged phospholipids (Strohman et al., 2019; Dawaliby et al., 2016), PIP2 has been shown to stabilize the receptor:G protein complex for the adenosine A2a-, β1-adrenergic, and neurotensin receptor 1 (Yen et al., 2018). More recently, we showed that PIP2 could be an allosteric regulator of ghrelin signaling (Damian et al., 2021). Accordingly, we observed here that including PIP2 in the nanodiscs was associated with a further amplification of the changes in 7H4MC fluorescence triggered by ghrelin, reflecting the allosteric coupling between the full agonist and this lipid for stabilizing an active-like hydration pattern of GHSR.

In closing, the combination of incorporation of 7H4MC-ethylglycine into proteins, fluorescence spectroscopy, and advanced MD simulations provided us with a straightforward strategy to delineate conformational events associated with GHSR activation through an unexplored but nevertheless central feature in the functioning of membrane proteins, local hydration. Using this strategy, we found that the hydration pattern in specific regions of TM5 and TM6 is dependent on the activation state of the receptor. This illuminates an unexpected role of water molecules as possible allosteric modulators of GHSR activation, consistent with their general effect on the allosteric regulation of proteins (Leitner et al., 2020). Hence, a model emerges where the activation process of GPCRs and their final signaling output could be the result of the concerted, synergistic, and exquisitely tuned influence of all the components in the receptor environment, including the solvent, on the distribution of the different states composing their conformational landscape. In addition, these observations demonstrate that water movements are tightly correlated to the receptor activation process and could therefore be used as a fingerprint to navigate the conformational landscape of GPCRs.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain backgroundBL21(DE3)
Escherichia coli
Sigma-AldrichCMC0014Chemically competent cells
Recombinant DNA reagentpEvol-aaRSdoi: 10.1021/ja062666k
Recombinant DNA reagentpMSP1E3D1Addgene#20066
Recombinant DNA reagentpET21a-α5-GHSR (transfected construct; Homo sapiens)doi: 10.1074/jbc.M111.288324
Peptide, recombinant proteinGhrelinThis workSynthesis is described in the Materials and methods section
Peptide, recombinant proteinFluorescent ghrelinThis workLabeling is described in the Materials and methods section
Peptide, recombinant proteinThrombinSigmaT7009
Commercial assay or kitGTPase-GloTM assayPromegaV7681
Chemical compound, drug7H4MC-ethylglycineThis workSynthesis is described the Materials and methods section
Chemical compound, drugAmpicillinSigmaA9518
Chemical compound, drugChloramphenicolCalbiochem220551
Chemical compound, drugIPTGSigmaI6758
Chemical compound, drugAmphipol A8-35AnatraceA835 100 MG
Chemical compound, drugβ-DDMAnatraceD310
Chemical compound, drugCholesteryl-hemisuccinateAnatraceCH210
Chemical compound, drugPOPCAvanti Polar Lipids850457C
Chemical compound, drugPOPGAvanti Polar Lipids840457C
Chemical compound, drugPIP2Avanti Polar Lipids850155P
Chemical compound, drugBio-Beads SM-2BIO-RAD1528920
Chemical compound, drugLumi4-Tb NHSCisBio62TBSPEA
Chemical compound, drugDY647P1-maleimideDyomics647P1-03
Chemical compound, drugAmine reactive Tb chelateFisher11563467
Chemical compound, drugNiNTA SuperflowQiagen30430
Chemical compound, drugStreptavidin-agaroseThermofisher20361
Chemical compound, drugSuperdex S200 increase 10×300 GLGE Healthcare (Cytiva)28990944
Chemical compound, drugSource 15Q 4.6×100 PEGE Healthcare (Cytiva)17518101
Chemical compound, drugZebaSpin 40K MWCO columnThermofisher87766
Software, algorithmPrismGraphPadVersion 8.4.3
Software, algorithmVMDdoi: 10.1016/0263-7855(96)00018-5
Software, algorithmBio3Ddoi: 10.1093/bioinformatics/btl461
Software, algorithmPymolSchrodinger LLC
Software, algorithmGromacs 2020.3doi: 10.5281/zenodo.3923645

Materials

MSP1E3D1(-) was expressed and purified in E. coli as described (Ritchie et al., 2009). 7H4MC-ethylglycine was synthesized as described (Amaro et al., 2015) with the exception that the final product was purified using reverse-phase HPLC.

Production of 7H4MC-labeled GHSR

Request a detailed protocol

For labeling with 7H4MC-ethylglycine, the TAG amber codon was introduced at the positions indicated in Figure 1A by site-directed mutagenesis with the pET21a expression vector encoding human GHSR fused to the α5 integrin (Damian et al., 2012). The UAA solution was prepared by dissolving 263 mg of 7H4MC-ethylglycine in 10 mL 200 mM KOH solution and filter-sterilizing. The ghrelin receptor expression vector was co-transformed with the pEvol-aaRS carrying the engineered orthogonal tRNA and aminoacyl-tRNA synthase pair (Wang et al., 2006) in BL21(DE3) E. coli cells. Cultures were grown at 37°C in 2YT medium containing ampicillin and chloramphenicol until the OD600 reached 0.5–0.6. After centrifugation, cell pellets were resuspended in fresh 2YT-ampicillin-chloramphenicol medium containing 10 mL of the UAA solution. The culture was incubated again at 37°C until OD600 reached 1 and protein expression was induced by addition of IPTG and arabinose (1 mM and 0.02%, respectively). Cell growth was continued for 16 hr at 30°C. In all cases, GHSR purification and assembly into nanodiscs was carried out as described for the unlabeled receptor (Damian et al., 2012). Briefly, the α5-GHSR fusion protein was first purified from inclusion bodies as an SDS-unfolded protein using IMAC. After cleavage of the α5 integrin partner with thrombin, the resulting receptor was dialyzed in a 50 mM Tris-HCl, 1% SDS, pH 8 buffer. Amphipol (APol)-mediated folding was then carried out by adding APol A8-35 to the SDS-solubilized receptor at a 1:5 protein/APol weight ratio in the presence of 10 µM of JMV3011. After 30 min incubation at room temperature, GHSR folding was initiated by precipitating dodecyl sulfate as its potassium salt through addition of KCl to a final 200 mM concentration. The potassium dodecyl sulfate precipitate was then removed by two 15 min centrifugations at 16,100×g. The supernatant was extensively dialyzed against a 50 mM potassium phosphate, 150 mM KCl, 10 µM JMV3011, pH 8 buffer. APols were then exchanged to n-dodecyl-β-D-maltopyranoside (β-DDM) in the presence of cholesteryl hemisuccinate (CHS). To this end, the APol/GHSR complex was incubated for 2 hr at 4°C with 0.2% (w/v) β-DDM, 0.02% (w/v) CHS in a 50 mM Tris-HCl pH 8, 150 mM NaCl, 10 µM of the JMV3011 buffer. The sample was then loaded onto a pre-equilibrated HisTrap column and the resin washed with a 50 mM Tris-HCl pH 8, 150 mM NaCl, 0.2% (w/v) β-DDM, 0.02% (w/v) CHS, 10 µM JMV3011 buffer and then with a 50 mM Tris-HCl pH 8, 150 mM NaCl, 0.1% (w/v) β-DDM, 0.02% (w/v) CHS, 10 µM JMV3011 buffer. The protein was finally eluted from the column with the same buffer containing 200 mM imidazole and dialyzed into a 25 mM HEPES, 100 mM NaCl, 2 mM β-DDM, 0.02% (w/v) CHS, 10 µM JMV3011 buffer. For reconstitution into nanodiscs, the His-tagged receptor in 25 mM HEPES, 100 mM NaCl, 2 mM β-DDM was first bound onto a pre-equilibrated Ni-NTA superflow resin at a protein-to-resin ratio at 0.1–0.2 mg of receptor per mL of slurry (batch conditions). The receptor was then mixed with 10 µM of JMV3011, and with MSP1E3D1(-) and a POPC:POPG (3:2 molar ratio) mixture, in the absence or presence of PIP2 (2.5% PIP2-to-total lipid molar ratio), at a 0.1:1:75 receptor:MSP:lipid ratio, with the receptor still immobilized on the Ni-NTA matrix. After 1 hr incubation at 4°C, polystyrene beads (Bio-Beads SM-2) were added to the slurry at an 80% (w/v) ratio and incubated under smooth stirring for 4 hr at 4°C. The resin was then extensively washed with a 50 mM Tris-HCl pH 8, 150 mM NaCl buffer, and the His-tagged receptor eluted with the same buffer containing 200 mM imidazole. After extensive dialysis in a 25 mM HEPES, 150 mM NaCl, 0.5 mM EDTA, pH 7.5 buffer, active receptor fractions were purified using affinity chromatography (Ferré et al., 2019). To this end, the receptor in lipid discs was loaded on a streptavidin-agarose column where the biotinylated JMV2959 antagonist had been bound following manufacturer’s instructions. After washing with 25 mM Tris-HCl, 150 mM NaCl, pH 7.4, the bound proteins were recovered by washing the column with the same buffer containing 1 mM of the low affinity JMV4183 antagonist. This antagonist was then removed through extensive dialysis against a 25 mM Tris-HCl, 150 mM NaCl, 0.5 mM EDTA, pH 7.4 buffer. We previously demonstrated that under such conditions all the ligand is removed from its binding site on GHSR (Ferré et al., 2019). Homogeneous fractions of GHSR-containing discs were finally obtained through a size-exclusion chromatography step on an S200 increase column (10/300 GL) using the 25 mM Tris-HCl, 150 mM NaCl, 0.5 mM EDTA, pH 7.4 buffer as the eluent (Figure 1—figure supplement 1).

Receptor labeling for ligand-binding assays

Request a detailed protocol

To avoid any labeling of the scaffolding protein, labeling of the receptor N-terminus with the amine-reactive Tb chelate for the HTRF-monitored ligand-binding assays was carried out in the APol-folded state (Damian et al., 2012), that is, before insertion into the nanodiscs. To this end, the receptor in A8-35 was dialyzed in a 50 mM potassium phosphate, 100 M KCl, pH 7.7 buffer. This pH value was determined from a series of labeling reactions we first carried out at different pH to define the optimal value for labeling essentially the protein N-terminal α-amine and not the lysyl ε-amino groups (Damian et al., 2012), which display a higher pKa value (Grimsley et al., 2009). The amine-reactive chelate was added to the protein solution (dye-to-protein equimolar ratio), and the reaction was carried out overnight at 4°C under constant stirring. The conjugate was separated from any possible unreacted labeling reagent by desalting on a ZebaSpin 40K column. Specific labeling of the N-terminal amine was assessed in a pilot experiment by the absence of fluorescence of the labeled receptor after digestion with TEV of a construct we designed to determine if labeling indeed occurred essentially at the GHSR N-terminus (Figure 1—figure supplement 4). The receptor reconstitution procedure was then continued by exchanging the APol to β-DDM and assembly into nanodiscs, as described above.

Ghrelin1-18-DY647P1 synthesis

Request a detailed protocol

The structure of the fluorescent ghrelin peptide we used in the ligand-binding experiments is shown in Figure 1—figure supplement 5. H1Gly-2Ser-3Asp(n-octanoyl)-4Phe-5Leu-6Ser-7Pro-8Glu-9His-10Gln-11Arg-12Val13Gln-14Gln-15Arg-16Lys-17Glu-18Ser-19Cys-NH2 was synthesized by solid-phase peptide synthesis starting from Agilent Amphisphere 40 RAM resin using Fmoc chemistry, HATU/DIEA system for coupling, and piperidine/DMF for deprotection. All coupling steps (5 eq.) were performed twice for 10 min, except for 15Gln, 12Val, and 11Arg where the first coupling lasted 45 min. Final deprotection was performed with a TFA/TIS/H2O (95/2.5/2.5) mixture for 3 hr. After purification by preparative RP-HPLC, the peptide (0.845 eq.) was dissolved in 1 mL of sodium phosphate solution (pH 7) and 1 mL of acetonitrile and conjugated with 1 mg of DY-647P1-maleimide (Dyomics) for 3 hr. The fluorescent peptide was directly injected on a preparative RP-HPLC column and purified (Figure 1—figure supplement 5). Their identity and purity were evaluated by mass spectrometry analyses (Figure 1—figure supplement 5). Preparative RP-HPLC was run on a Gilson PLC 2250 Purification system instrument (Villiers le Bel, France) using a preparative column (Waters DeltaPak C18 Radial-Pak Cartridge, 100 Å, 40–100 mm, 15 µm particle size) in gradient mode with a flow rate 50.0 mL/min. Buffer A was 0.1% TFA in water, and buffer B was 0.1% TFA in acetonitrile.

LC/MS analyses

Request a detailed protocol

The LC/MS system consisted of an HPLC-ZQ (Waters) equipped with an ESI source. Analyses were carried out using a Phenomenex Kinetex column (C18, 100 Å, 100×2.1 mm2, 2.6 µm). A flow rate of 0.5 mL/min and a gradient of 0–100% B in 5 min were used: eluent A, water/0.1% HCO2H; eluent B, ACN/0.1% HCO2H. Positive ion electrospray (ESI+) mass spectra were acquired from 100 to 1500 m/z with a scan time of 0.2 s. Nitrogen was used for both the nebulizing and drying gas.

MALDI MS and MS/MS analyses

Request a detailed protocol

Samples were analyzed from CHCA or SA matrix deposits, in positive ion mode with a Rapiflex (Bruker Daltonics) instrument. A pulsed Nd:YAG laser at a wavelength of 355 nm was operated at a 66.7 Hz frequency with a laser focus of 29%. Data were acquired with the Flex Control software (version 4.1, Bruker Daltonics). Spectra were integrated with the Flex Analysis software (version 4.0, Bruker Daltonics), the centroid algorithm was used to assign peaks. An acceleration voltage of 25.0 kV (IS1) was applied for a final acceleration of 21.95 kV (IS2) and lense voltage of 9.6 kV. The reflectron mode was used for the ToF analyzer (voltages of 26.3 and 13.8 kV). The delayed extraction time was 30 ns. Acquisitions were performed using a reflector detector voltage of 1.722 kV. MS data were processed with the Flex Analysis software (version 4.0, Bruker Daltonics). External calibration was performed with commercial peptide mixture (Peptide Calibration Standard II, Bruker Daltonics). Fragmentation experiments were performed under laser-induced dissociation conditions with the LIFT cell voltage parameters set at 19.0 kV (LIFT 1) and 3.7 kV (LIFT 2) for a final acceleration of 29.5 kV (reflector voltage) and a pressure in the LIFT cell around 4 × 10–7 mbar. The precursor ion selector was set manually to the first monoisotopic peak of the molecular ion pattern for all analyses. MS/MS data were processed with the Flex Analysis software (version 4.0, Bruker Daltonics). Mass lists were generated according to the following parameters: SNAP as peak detection algorithm, S/N threshold 3.

G protein production

Request a detailed protocol

A Gαqβ1γ2 heterotrimer composed of the wild-type rat Gαq and bovine Gβ1 subunits and of a bovine Gγ2 subunit tagged with a hexahistidine was expressed in sf9 cells and purified as described (Kozasa, 2004). For the functional assay, the protein was further purified by ion-exchange chromatography. To this end, the heterotrimer was isolated using a Source 15Q 4.6×100 PE column. After binding of the protein to the column in a 20 mM HEPES, 30 mM sodium chloride, 1 mM MgCl2, 0.05% DDM, 100 mM TCEP, 20 mM GDP, pH 7.5 buffer and washing with the same buffer, the heterotrimer was eluted with a linear gradient of 30–500 mM NaCl and the fractions containing the G protein trimer were pooled (see SDS-PAGE profile in Figure 1—figure supplement 6).

Functional assays

Request a detailed protocol

Competition ligand-binding assays were performed using fluorescence energy transfer with a purified receptor labeled at its N-terminus with Lumi-4 Tb NHS and the dy647-labeled ghrelin peptide (Damian et al., 2015; Leyris et al., 2011). Increasing concentrations in the competing compound were added to a receptor:ghrelin peptide mixture (100 nM concentration range). After a 30 min incubation at 15°C, fluorescence emission spectra were recorded at the same temperature between 500 and 750 nm (Cary Eclipse spectrofluorimeter, Varian) with excitation at 337 nm. GTP turnover was assessed as described (Hilger et al., 2020). All experiments were carried out at 15°C. The receptor (200 nM) was first incubated with the isolated G protein (500 nM) and, when applicable, the ligand (10 µM) for 30 min in a 25 mM HEPES, 100 mM NaCl, 5 mM MgCl2, pH 7.5 buffer. GTP turnover was then started by adding GTP (1 µM) and the remaining amount was assessed after 15 min incubation at 15°C using the GTP-Glo assay (Promega).

7-H4MC fluorescence measurements

Request a detailed protocol

Fluorescence spectra were recorded with a Cary Eclipse spectrofluorimeter (Varian) equipped with a Peltier-based temperature control device. All experiments were carried out at 15°C. The emission spectra after excitation at 320 nm were recorded between 340 and 600 nm. The normalized emission intensity was fitted by means of nonlinear least-square procedure to the sum of peak function (Amaro et al., 2015). The R-square parameter was used to estimate the goodness of the fit.

Statistical analyses

Request a detailed protocol

Data in different conditions were compared by one-way ANOVA followed by Dunnett’s multiple comparison test and reporting of multiplicity-adjusted p-values and confidence intervals (Michel et al., 2020). As stated in the legends of the corresponding figures, data are presented as mean ± SD of three experiments. All analysis steps, including the sample size, were decided before looking at the data. No data was removed from the analysis. No measure to avoid experimental bias was taken.

Modeling

Building of an active-like model of GHSR

Request a detailed protocol

The structure of GHSR was first retrieved from the PDB (6KO5) (Shiimura et al., 2020) and used as a starting point for our study. We mutated back to wild-type amino acids the two mutations (T1303.39K and N188Q) that were present in the structure to match the wild-type sequence, and modeled the extracellular loop three ab initio (ECL3 – residue G293 to I300), with MODELLER 9.19 (Webb and Sali, 2016). Two cryo-EM structures of the ghrelin receptor in complex with ghrelin or a synthetic agonist and a Gq mimetic have been posted on the BioRxiv preprint server (https://doi.org/10.1101/2021.06.09.447478). However, the coordinates of these models are not yet available. Hence, in order to capture differences in the receptor hydration pattern upon activation, we generated an active-like model of GHSR by targeted molecular dynamics (TMD) simulations performed in an explicit membrane environment. First, we modeled the target conformation based on the dopamine D2 receptor coupled to Gi (Yin et al., 2020) (D2R:Gi, PDB id: 6VMS; sequence similarity: 33%) by homology modeling using MODELLER 9.19 (Webb and Sali, 2016). The sequence alignment between GHSR and D2 was achieved with ClustalW (Larkin et al., 2007). The best out of 100 models built by MODELLER, regarding DOPE score, was further selected as the target conformation for the subsequent TMD. The TMD simulation was run with NAMD 2.13 (Phillips et al., 2020), where the inactive experimental conformation was pushed toward the newly generated active conformation. Inactive GHSR was embedded in a lipid bilayer containing 156 POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), for a size of 80×80 Å2. The system was then solvated and neutralized with a NaCl concentration around 0.15 M (17,270 water molecules, 46 sodium, and 29 chloride ions) with CHARMM-GUI (Wu et al., 2014; Brooks et al., 2009; Jo et al., 2008). In order to limit the deviation from the initial experimental structure, the force during TMD was only applied to residues of the intracellular part of TM helix 6 (TM6, from S2526.24 to L2776.49), which are known to undergo the largest conformational changes during activation of all known GPCRs. All remaining atoms of GHSR were harmonically restrained in position using a force constant of 1 kcal/mol/Å2, but residues L2395.65 to A251 (ICL3), so that the loop could follow the motion of TM6. Prior to TMD, the system was minimized using 10,000 steps of conjugate gradient as implemented in NAMD 2.13 (Phillips et al., 2020), followed by successive short equilibration procedures in NVT and NPT ensembles, to reach a final temperature of 300 K and a pressure of 1 bar using CHARMM36m force field (Huang et al., 2017). We did not modify the equilibration procedure designed by CHARMM-GUI developers (Wu et al., 2014). The TMD simulation was performed in the NPT ensemble (300 K and 1 bar) over a period of 500 ps using a force constant of 200 kcal/mol/Å2 scaled down by the number of selected atoms in TM6 (477 atoms including hydrogens). Non-bonded interactions were truncated at a distance cut-off of 12 Å applying a switching function in the range 10–12 Å, while long range electrostatics were computed via particle mesh Ewald (PME).

MD simulations of inactive and active-like conformers of wild-type GHSR

Request a detailed protocol

The inactive (experimental) and active (modeled) conformers of GHSR were simulated by MD with Gromacs 2020.3 using the CHARMM36m force field (Huang et al., 2017). To fit to the experimental membrane composition used in this study, each conformer was embedded in a symmetric lipid bilayer of size 80×80 Å2, where each layer was composed of 20 cholesterol, 28 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG), 42 POPC, and 10 phosphatidylinositol-4,5-bisphosphate (PIP2) (including five PIP2 protonated on one phosphate group and five PIP2 protonated on the other phosphate group, named respectively POPI24 and POPI25 in CHARMM36m force field). Systems were solvated and their charges were neutralized with a NaCl concentration around 0.15 M (17,751 water, 170 sodium, and 47 chloride ions). The simulation setup was done with the CHARMM-GUI webserver (Wu et al., 2014; Brooks et al., 2009; Jo et al., 2008). Contrary to the TMD protocol, we did change the default CHARMM-GUI procedure for equilibration. Indeed, we added three additional equilibration steps to the default CHARMM-GUI procedure. We modified the harmonic restraints on atomic positions and the number of simulation steps to allow a smooth relaxation of the systems (Supplementary file 1). We reproduced this protocol five times for each system (active and inactive) modifying the starting velocities so that the convergence of the resulting data could be discussed. The production was run in the NPT ensemble (300 K and 1 bar) for 5 µs (leading to a simulation time of 50 µs in total). It is important to notice that, during production, all restraints and constraints were removed. For all simulations, direct non-bonded interactions were truncated at a distance cut-off of 12 Å applying a switching function in the 10–12 Å range, while long range electrostatics were computed via PME.

PCA of experimental structures

Request a detailed protocol

To delineate the possible motions described by the plethora of available GPCRs’ experimental structures, we retrieved 268 structures of class A GPCRs from the PDB. To homogenize these data, only the part corresponding to a single isolated receptor was conserved for further analysis, for instance removing the intra- and/or extracellular partner(s) if required or other copies of the same receptor in the case of dimeric structures. The sequences of all retrieved structures were then aligned with Clustal Omega (Larkin et al., 2007) with default parameters. To perform PCA of the resulting set of coordinates, the length of the resulting sequences also required to be homogenous. As a compromise between the number of structures considered (increasing the conformational diversity) and the length of the sequence common to all receptors (improving the structural description), we decided to discard a residue at a particular position of the alignment if the latter was missing in at least two structures out of the 268. In addition, a structure was discarded if it was the only one presenting a missing residue at a specific position. Using these criteria, only six structures were deleted from the initial set (PDB id: 5WB2, 4PY0, 5ZKP, 3RZE, 4RWA, and 4DAJ). In summary, 262 structures were considered, together describing a set of 164 conserved amino acids (GHSR numbering: 45, 46, 48–68, 76–102, 120–148, 162–179, 181, 212, 213, 215–219, 221–239, 261, 263–286, 310–324). The list of the considered PDB structures together with useful information, according to GPCRdb (Pándy-Szekeres et al., 2018), can be found in Supplementary file 2. Not surprisingly, the final selection covered most of the TM domains, ensuring a good description of the internal motions coded by the ensemble of experimental structures (Figure 4—figure supplement 5). On the contrary, most residues located in the extra- or intracellular loops were excluded. Because of the variability of residues at each position of the final alignment, the PCA was performed only on the coordinates of the Cα atoms with the R package Bio3D (Grant et al., 2006).

Analysis and figure generation

Request a detailed protocol

All analyses were run with VMD (Humphrey et al., 1996) and the R package Bio3D (Grant et al., 2006). Figures were generated using VMD and Pymol.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files.

References

    1. Hille B
    2. Dickson EJ
    3. Kruse M
    4. Vivas O
    5. Suh B-C
    (2015) Phosphoinositides regulate ion channels
    Biochimica Et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids 1851:844–856.
    https://doi.org/10.1016/j.bbalip.2014.09.010

Decision letter

  1. Lucie Delemotte
    Reviewing Editor; KTH Royal Institute of Technology, Sweden
  2. Richard W Aldrich
    Senior Editor; The University of Texas at Austin, United States
  3. Lucie Delemotte
    Reviewer; KTH Royal Institute of Technology, Sweden

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This manuscript uses un-natural amino acid incorporation into the GHSR1a to examine the exposure of particular residues to changes in polarity (interpreted as solvent exposure) experimentally followed by molecular dynamics simulation. Considering the body of work done on GPCRs, there are surprisingly few studies that carry out a quantitative one-to-one comparison between experimental and simulations. This manuscript presents a convincing attempt at doing so.

Decision letter after peer review:

Thank you for submitting your article "Concerted conformational dynamics and water movements in the ghrelin G protein-coupled receptor" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Lucie Delemotte as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Reviewing Editor and Richard Aldrich as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This manuscript uses un-natural amino acid incorporation into the GHSR1a to examine the exposure of particular residues to changes in polarity (interpreted as solvent exposure) experimentally followed by molecular dynamics simulation.

Considering the body of work done on GPCRs, there are surprisingly few studies that attempt a quantitative one-to-one comparison between experimental and simulations. This manuscript presented an interesting attempt at doing so. However, the manuscript asks the reader to accept a substantial amount of work that has either obscure, or superficial methods, little supporting data and significant normalization, making it difficult to fully judge its merits.

Essential revisions:

Related to experiments:

1) Please address the functionality of the expressed, refolded and MSP incorporated GHSR1a. Could the authors please provide original Coomassie stained gels and size exclusion chromatography traces that support their purification and re-incorporation of GHSR1a into the MSP scaffold? The authors report that they have expressed GHSR1a in bacteria then unfolded and refolded this protein, the reviewers are not aware of any other GPCR for which this has been successfully performed, indeed the structural paper that the authors discuss (below), uses a thermostabilized GHSR1a expressed in insect cells in order to obtain crystallizable protein. Additionally, the below paper uses a thermofluor assay to demonstrate the stabilization of their construct. The thermofluor assay, by its nature, indicates that GHSR1a does not spontaneously refold.

Shiimura Y, Horita S, Hamamoto A, Asada H, Hirata K, Tanaka M, Mori K, Uemura T, Kobayashi T, Iwata S, Kojima M. Structure of an antagonist-bound ghrelin receptor reveals possible ghrelin recognition mode. Nat Commun. 2020 Aug 19;11(1):4160. doi: 10.1038/s41467-020-17554-1. PMID: 32814772; PMCID: PMC7438500.

2) As evidence of the functionality of the refolded, purified and MSP incorporated GHSR1a the authors provide a FRET competition binding assay. There is insufficient explanation of this assay for this to be reproduced by another group. In the method the authors indicate that the GHSR1a is labelled at the N-terminus with Lumi-4 Tb NHS. NHS esters typically react with any primary amine, in the authors methods they have the purified GHSR1a in a 25 mM Tris buffer, where the Tris would be expected to preferentially react with the Lumi-4 Tb NHS, the GHSR1a is also incorporated into MSP, which has an N-terminus and both proteins have a number of lysine residues where the Lumi-4 Tb NHS would be expected to react with the epsilon amino group. Perhaps there are some significant details missing from their methods that might enlighten this? In any case, could the authors please provide their in-gel fluorescence (or alternative analysis such as mass spectrometry) that demonstrates specific labelling of GHSR1a (and not MSP) on the N-terminus (and not on lysine side chains)? This assay uses a dy647 labelled ghrelin peptide, could the authors please provide details of how the labelling was performed and either HPLC of mass spec data for the resulting labelled reagent? The cited reference (19) does not appear to contain this reagent, whereas the cited reference (33) does contain a "red-ghrelin" where no details about the chemical composition are readily available. The published affinity of ghrelin for GHSR1a is 400 pM (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=246), the authors need to specifically address why the reported affinity in their assay (Figure 1B) appears to be approximately 250 fold lower at ~100 nM? Could the authors also please provide the original, non-normalised FRET data so that the reader can understand the window for this assay along with a specificity control such as Lumi-4 Tb NHS labelled empty MSP nanosdiscs?

3) In figure 1C the authors provide further evidence for functionality of their purified GHSR1a using a GTP turnover assay. The authors need to provide the full details of how the Galpha/β/γ heterotrimer were expressed and purified for this assay. Please include details of which particular isoform and species each G protein subunit is from, the expression system and how they were purified. A representative Coomassie stained gel that demonstrates stoichiometric equivalence of the subunits in the purified complex should also be supplied. The references provided for this assay to not appear to relate to a GTP depletion style assay as appears to be described here. Could the authors also please describe how 0% was defined for the assay and the relative concentrations of GHSR1a and G proteins heterotrimer that were added to the reactions? The authors show apparent differences in bound GTP in this assay, could they please provide a statistical analysis for these differences?

Related to simulations:

1) Driving large structural changes fast is risky, and the reviewers were not convinced the water populations had equilibrated. Indeed, forces to enhance the sampling were applied along each PC separately. Since these PCs represent a linear decomposition of the overall family-wide conformational change, it didn't appear wise to enhance the sampling along them: linear decomposition of the movement could in principle result in very non-physical motions. Can the authors provide the readers with a supplementary figure showing the comparison between the starting and the 2 end structures, as well as the PDBs of the resulting structures so their quality can be checked?

Relatedly, the mix of active-like and inactive-like structures used in the PCA to derive the biasing forces is expected to have a major effect. The authors need to explicitly list (in the supplement) the structures used, and categorize them by activity state and preferably by GPCR family as well. The first principle vector probably point more or less along the path between inactive and active, but it would be nice to check this.

2) The reviewers also asked for a clearer rationalization of why the authors picked this sampling strategy as opposed to (1) building models of the ghrelin receptor in different states and simulating them or (2) enhancing the sampling using a non-linear approach. Ultimately, what do we learn from the fact that PC1 is the most compatible with the experimental data, given that the overall motion is a combination of all the PCs? It would be wise to replicate the results with a more standard MD simulation protocol to rule out artefacts from this choice of enhanced sampling methods.

3) Additionally, the methods section was unclear about several aspects:

a) it is unclear as to how many replicates were done for each mode: if it's a single replicate for each mode, no conclusion could be drawn about the hydration. To have any confidence in the result, the reviewers would want to see the simulations rerun many times (at least 10x, perhaps more depending on how variable the answer is), preferably starting from different structures within the equilibrated ensemble.

b) Which state does the original model represent? Can the comparison to the recently published structure be more thorough than simply showing a Calpha(?) RMSD (Figure S8)? Are the enhanced sampling MD carried out in presence or absence of ligand and why? Finally the method section as well as the Results section explaining the enhanced sampling protocol should be clarified such that it is not necessary to read the original paper explaining the method to understand.

c) Many key details of the simulations are missing: number of lipids, number of waters, electrostatics method (as written, it sounds like they didn't use Ewald, which would be a huge problem).

d) There is no discussion of statistical convergence. The simulations are very short by today's standards, and the reviewers saw no reason to assume the protein has stopped systematically changing after 300-350 ns (given the uncertainty of starting from a homology model), let alone begun actually sampling. The only evidence is Supp Figure 5, which shows the RMSD is still increasing, while nothing at all is shown for the mutants.

e) As it stands, too many details are missing for these calculations to be repeated. Please collect and document all of the scripts used (building the system, running, and analyzing) and put them either into the supplementary info or better yet into a separate repository (e.g. GitHub, zenodo, etc).

https://doi.org/10.7554/eLife.63201.sa1

Author response

Essential revisions:

Related to experiments:

1) Please address the functionality of the expressed, refolded and MSP incorporated GHSR1a. Could the authors please provide original Coomassie stained gels and size exclusion chromatography traces that support their purification and re-incorporation of GHSR1a into the MSP scaffold?

As requested, the size-exclusion chromatography (SEC) profile of the GHSR-containing nanodiscs, along with the SDS-PAGE profile of the fractions pooled, is presented as Figure 1—figure supplement 1 in the revised version of the manuscript. As evidenced by these SEC and SDS-PAGE profiles, the GHSR-containing discs correspond to a homogeneous fraction composed of the scaffolding protein and receptor only. To be noted, faint bands corresponding to multimeric receptor species are present in the SDS gel, but this does not reflect the occurrence of such species in the nanodiscs. Indeed, we previously demonstrated unambiguously using native gel electrophoresis and fluorescence transfer that the conditions we use for nanodisc assembly (large MSP-to-receptor molar ratio, on-matrix batch assembly with a large resin-to-protein ratio) provide lipid discs with a single receptor (1). Similar conditions were also shown to provide monomeric receptor for a series of other GPCRs, including rhodopsin and the b2-adrenergic receptor (2,3). Oligomerization of a membrane protein on an SDS-PAGE, as in the case of the gel in Figure 1—figure supplement 1, does not reflect its oligomerization state in solution when associated to various surfactants or lipids. Indeed, at the concentration used to run a gel, SDS is known to only partially unfold these membrane proteins and trigger some aggregation, with the exception of β-barrel ones. The consequences most frequently encountered are an anomalously electrophoretic mobility and protein aggregation in the gel (R.B. Gennis “Biomembranes: Molecular Structure and Function”, Springer Science, 1989, page 93"; Rath et al. (2009) (4)).

The authors report that they have expressed GHSR1a in bacteria then unfolded and refolded this protein, the reviewers are not aware of any other GPCR for which this has been successfully performed, indeed the structural paper that the authors discuss (below), uses a thermostabilized GHSR1a expressed in insect cells in order to obtain crystallizable protein. Additionally, the below paper uses a thermofluor assay to demonstrate the stabilization of their construct. The thermofluor assay, by its nature, indicates that GHSR1a does not spontaneously refold.

Shiimura Y, Horita S, Hamamoto A, Asada H, Hirata K, Tanaka M, Mori K, Uemura T, Kobayashi T, Iwata S, Kojima M. Structure of an antagonist-bound ghrelin receptor reveals possible ghrelin recognition mode. Nat Commun. 2020 Aug 19;11(1):4160. doi: 10.1038/s41467-020-17554-1. PMID: 32814772; PMCID: PMC7438500.

We agree that our production procedure, which is based on the expression of the ghrelin receptor as an unfolded protein in E. coli inclusion bodies followed by in vitro folding and purification of the active receptor using ligand-affinity chromatography, is not the most common one in the field. It has nevertheless been successfully used in several occasions by us and others, first for an olfactory receptor (5), and then for different GPCRs such as the receptors for neuropeptides Y1 and Y2 (6,7), ghrelin (1,8,9), chemokine (10,11), leukotriene B4 (12,13) or serotonin (14). In all cases, the receptor obtained was representative of the native one in terms of ligand binding and activation of G proteins. This is the case for the ghrelin receptor we used in the present work that, since our princeps paper in 2012 (1), has been repeatedly shown by us and others to be structurally and pharmacologically relevant (8,9,15-19) (see below). Besides, in our opinion, GHSR produced in sf9 cell for crystallization purposes can be hardly compared to the present situation, as the former study dealt with a significantly different version of GHSR where the receptor was fused to BRIL at its N-terminus, devoid of its N- and C-terminal regions, thermostabilized through specific mutations and purified in a detergent solution.

2) As evidence of the functionality of the refolded, purified and MSP incorporated GHSR1a the authors provide a FRET competition binding assay. There is insufficient explanation of this assay for this to be reproduced by another group. In the method the authors indicate that the GHSR1a is labelled at the N-terminus with Lumi-4 Tb NHS. NHS esters typically react with any primary amine, in the authors methods they have the purified GHSR1a in a 25 mM Tris buffer, where the Tris would be expected to preferentially react with the Lumi-4 Tb NHS, the GHSR1a is also incorporated into MSP, which has an N-terminus and both proteins have a number of lysine residues where the Lumi-4 Tb NHS would be expected to react with the epsilon amino group. Perhaps there are some significant details missing from their methods that might enlighten this?

The description of the labeling procedure was not detailed enough, and we apologize for that. In fact, labeling of the scaffolding protein is not an issue, as the receptor is labeled before its assembly into nanodiscs, i.e., in the amphipol-folded state. This procedure was described in our initial paper reporting production of GHSR in nanodiscs, although for a different version of the fluorophore (amine-reactive derivative of Alexa Fluor 350 instead of Tb chelate) (1). Before its assembly into nanodiscs, the receptor is first folded from an SDS- to an amphipol (APol)-stabilized state using a procedure we developed in close collaboration with J.-L. Popot’s laboratory (20). This procedure, which consists in exchanging SDS for APol through precipitation of the detergent as its potassium salt, allows recovery of the receptor as a stable APol:protein complex (see the SEC profile in the Figure 1—figure supplement 3B,C of the revised manuscript). Labeling with the fluorophore is then carried out at this stage. Briefly, the receptor in A8-35 APol is dialyzed in a 50 mM potassium phosphate, 100 M KCl, pH 7.7 buffer to remove any Tris salt that would indeed bias labeling, as appropriately noted by the reviewer. This pH value was determined from a series of labeling reactions we first carried out at different pH to define the optimal value for labeling essentially the protein N terminal a-amine and not the lysyl e-amino groups (1), which display a significantly higher pKa value (the average pKa values in proteins is 7.7 and 10.5 for the N-terminal a- and lysine e-amino groups, respectively (21)). The conjugate is then incubated with the Apol:protein complex under the conditions described in the Methods section and unreacted labeling reagent is removed by desalting on a ZebaSpin column (ThermoFisher). The receptor reconstitution procedure in continued by exchanging the amphipol to b-DDM and then assembling the labeled receptor into the nanodiscs. The full labeling protocol is now detailed in the revised version of the manuscript.

In any case, could the authors please provide their in-gel fluorescence (or alternative analysis such as mass spectrometry) that demonstrates specific labelling of GHSR1a (and not MSP) on the N-terminus (and not on lysine side chains)?

In-gel fluorescence may not be totally effective in assessing N-terminal specific labeling of the receptor, as we think it cannot discriminate between different positions for the fluorophore. Besides, high-resolution mass spectrometry analysis of integral a-helical membrane proteins in lipid bilayers is still not technically trivial, at least in our hands. Therefore, we applied an alternative strategy we devised previously to ascertain N-terminal labeling of G proteins with a Tb chelate (17). The experiment consists in introducing a TEV cleavage site after the protein N-terminus (see schematic representation in the figure below). This construct was used for the present experiment only. Not to affect labeling, the TEV cleavage site was introduced 11 residues after the N-terminus of the receptor. We then carried out labeling as described above and subsequently cleaved the GHSR N-terminus using TEV protease. Removal of the N-terminal residues is not detrimental to the receptor’s fold, as indicated by the crystal structure of GHSR where 28 residues were removed from the receptor N-terminus (22). The TEV cleavage is almost quantitative, as shown by the SEC profile in Figure 1—figure supplement 3. Besides, this profile shows that most of the labeling occurs at the receptor N-terminus, as the Tb-chelate moiety absorbing at 337 nm is essentially found in the column total volume after TEV digestion. This is confirmed by the UV absorption and fluorescence spectra of the non-cleaved and cleaved proteins. Indeed, essentially all the absorption and emission signatures of the Tb-chelate are lost after cleavage, indicating that most of the labeling indeed occurred at the receptor N-terminus. This figure has been added to the revised version of the manuscript as Figure 1—figure supplement 3. In any case, it is to be noted that even a residual labeling of the receptor at other sites besides the N-terminus would not be deleterious for the binding assay. Indeed, the latter simply consists in assessing the proximity between a fluorescence donor and acceptor on the receptor and the ligand, respectively, as a signature of the binding process, with no more detailed interpretation of the fluorescence transfer signal. Accordingly, equivalent binding plots have been reported using the same procedure and a fluorophore attached to any of six different positions in GHSR (23).

This assay uses a dy647 labelled ghrelin peptide, could the authors please provide details of how the labelling was performed and either HPLC of mass spec data for the resulting labelled reagent? The cited reference (19) does not appear to contain this reagent, whereas the cited reference (33) does contain a "red-ghrelin" where no details about the chemical composition are readily available.

The synthesis of the labeled peptide used in the ligand-binding assays is now described in Materials and Methods section of the revised version of the manuscript. The HPLC profile and mass spectra of the resulting product are given as Figure 1—figure supplement 4.

The published affinity of ghrelin for GHSR1a is 400 pM (https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=246), the authors need to specifically address why the reported affinity in their assay (Figure 1B) appears to be approximately 250 fold lower at ~100 nM?

Different values have been reported in the literature for the affinity of ghrelin for GHSR. It has been proposed that a peculiar behavior of ghrelin in binding assays results, at least in part, from large non-specific binding effects due to its high hydrophobicity that can lead it to bind to low-affinity ‘‘pseudo-specific” sites into the plasma membrane (24). A value as low as 400 pM has been indeed initially reported and is referenced as such in the database mentioned by the reviewer (25). As reported in the original paper, this value was obtained using radiolabeled ghrelin and human tissues, with as much as 60-65% of specific binding. However, higher values have also been reported repeatedly. For instance, a 2.3 nM Kd was obtained using 125I-ghrelin and GHSR-expressing COS cells (26). We ourselves measured a Kd value of 4.4 nM for ghrelin binding to GHSR using 125I-labeled ghrelin and the recombinant receptor expressed in HEK293 cells (24), which is close to the value reported in Holst et al. (2005) (26). In the same paper, we reported a Ki of 4.6 nM for ghrelin, as measured from a FRET-based assay with the receptor expressed at the surface of HEK293 cells and a fluorescent ghrelin peptide. Again, a 3-fold higher value has been recently reported for the Kd of a similarly labeled ghrelin and GHSR-expressing HEK cells (12.47 nM) (27). In any case, as perfectly noted by the reviewer, the Ki we report for the isolated receptor (77 nM, as inferred for the wild type receptor from the plot in Figure 1B) is about 15-fold lower than that measured for the same receptor in a native environment, i.e. in HEK293 cells (24). However, it must be kept in mind that, as is the case for most GPCRs, isolated GHSR in the absence of G proteins is in a low affinity state for agonists. Accordingly, we previously demonstrated that high affinity ghrelin binding, very similar to that observed in HEK cells, could be recovered upon reconstituting GHSR in nanodiscs with its cognate Gq protein (see Figure 1D in (19)). This data established full functionality for the ghrelin receptor assembled into nanodiscs using exactly the same procedure as the one we used in the present study.

As requested by the reviewer, the normalized binding plot has been replaced in Figure 1B by the non-normalized FRET data. As the MSP protein is not labeled with the fluorophore, a similar control experiment with empty nanodiscs could not be carried out, however. Instead, we carried out a control experiment with an unrelated receptor, the leukotriene B4 receptor, we inserted into the same nanodiscs (N-terminal labeling of the leukotriene B4 receptor with a fluorophore was initially described in (28)). This receptor is a typical class A GPCR whose general architecture is close to that of GHSR. As shown in Figure 1—figure supplement 2 of the revised manuscript, essentially no FRET was observed between BLT1-containing nanodiscs and fluorescent ghrelin, indicating very limited non-specific binding or, alternatively, that any possible residual binding of the fluorescent peptide does not contribute to the FRET signal we use to monitor receptor:ligand interaction. This plot is provided as Figure 1—figure supplement 2 in the revised version of the manuscript.

3) In figure 1C the authors provide further evidence for functionality of their purified GHSR1a using a GTP turnover assay. The authors need to provide the full details of how the Galpha/β/γ heterotrimer were expressed and purified for this assay. Please include details of which particular isoform and species each G protein subunit is from, the expression system and how they were purified. A representative Coomassie stained gel that demonstrates stoichiometric equivalence of the subunits in the purified complex should also be supplied. The references provided for this assay to not appear to relate to a GTP depletion style assay as appears to be described here.

In these functional assays, we used the heterotrimer composed of rat Gaq and bovine Gb1g2. This trimer has been repeatedly used in the literature for both functional and structural studies (e.g. (29)). It was expressed in sf9 cells using the viruses initially provided to us by T. Kozasa and T. Kawano. This system allows production of a trimer composed of Gaq and Gb1 with a Gg2 subunit tagged with an hexahistidine, so that the complex can be readily purified using Ni-NTA chromatography. Purification was carried out as described in Kozasa (2004) (30). This reference has been added to the revised version of the manuscript with a description of the additional ion exchange chromatography step we used after IMAC to further purify the recombinant trimer for the functional assays. We have also included a representative Coomassie blue-stained gel as Figure 1—figure supplement 5.

Could the authors also please describe how 0% was defined for the assay and the relative concentrations of GHSR1a and G proteins heterotrimer that were added to the reactions?

Luminescence signals were normalized relative to the signal obtained under the same conditions for the Gaqb1g2 alone, in the absence of any receptor. A 200 nM and 500 nM concentration in receptor and G protein were used in the assays. All these data have been added in the Materials and methods section of the revised version of the manuscript.

The authors show apparent differences in bound GTP in this assay, could they please provide a statistical analysis for these differences?

A statistical analysis has been carried out and is now reported in Figure 1C of the revised version of the manuscript. This analysis shows that the difference in GTP binding in the absence and presence of agonist/inverse agonist is indeed significative. For the sake of clarity, we show in the revised version only the data for the apo, agonist and inverse agonist-loaded states, not to overload the figure. The initial figure with all the ligands (antagonist and Gq-biased agonist in addition to the full- and inverse-agonist) is now given in Figure 1—figure supplement 2.

Related to simulations:

1) Driving large structural changes fast is risky, and the reviewers were not convinced the water populations had equilibrated. Indeed, forces to enhance the sampling were applied along each PC separately. Since these PCs represent a linear decomposition of the overall family-wide conformational change, it didn't appear wise to enhance the sampling along them: linear decomposition of the movement could in principle result in very non-physical motions. Can the authors should provide the readers with a supplementary figure showing the comparison between the starting and the 2 end structures, as well as the PDBs of the resulting structures so their quality can be checked?

Relatedly, the mix of active-like and inactive-like structures used in the PCA to derive the biasing forces is expected to have a major effect. The authors need to explicitly list (in the supplement) the structures used, and categorize them by activity state and preferably by GPCR family as well.

We fully agree with the reviewer that functional motions are usually a combination of multiple PC. The objective here was to identify blindly all separate motions that could explain the experimental observations. It is very important to note that pulling was done with membrane and water. The pulling was slow and soft enough not to generate any nonphysical models thanks to a realistic environment (including membrane) described by a physical atomic force field (i.e. CHARMM36m). Our previous experiences with such kind of exploration along biasing vectors (e.g. normal modes) revealed that no nonphysical conformations are obtained and the structural quality of the conformational states sampled was preserved. Obtaining nonphysical motions would be related to the introduction of very strong forces, stronger than forces which maintain the topology of the receptor. In our case, we observe a wide range of amplitudes for each mode explored. This reveals that the resulting displacement does not necessarily follow precisely the biasing vector, therefore being able to diverge from the pulling direction was due to the restrictions imposed by the realistic environment. The differences in the amplitudes obtained are due to the distinct energy barriers observed for each displacement. Unfortunately, considering this very soft pulling and the statistical nature of results from MD, we were unable to reproduce our observations among different replicas using exactly the same pulling direction. We then chose to use a more classical approach by running free MD simulations in different activation states, as suggested by the reviewer.

Despite we did not use anymore the PCs inferred from experimental structures to explore GHSR motions, we used them to analyze our new MD simulations. We provide the structure list in this regard. The list of all experimental structures used to compute the PCs, together with information some readers might find useful is now available as Supplementary file 2.

The first principle vector probably point more or less along the path between inactive and active, but it would be nice to check this.

The reviewer guessed correctly the nature of the first eigenvector. It indeed represented a transition between inactive and active states and more specifically a transition between inactive adenosine receptor 2 (AA2AR) and opsin (OPSD) coupled to a G-protein (see Author response image 1).

Author response image 1
Projections along the first-two eigenvectors inferred from a Principal Component Analysis (PCA) of all experimental structures.

Names correspond to uniprot names of GPCR and have been placed in respect to the projection. Red names correspond to activated receptors, purple to intermediate conformations and blue names to inactivated receptors (according to the classification of the GPCRdb).

2) The reviewers also asked for a clearer rationalization of why the authors picked this sampling strategy as opposed to 1) building models of the ghrelin receptor in different states and simulating them or 2) enhancing the sampling using a non-linear approach. Ultimately, what do we learn from the fact that PC1 is the most compatible with the experimental data, given that the overall motion is a combination of all the PCs? It would be wise to replicate the results with a more standard MD simulation protocol to rule out artefacts from this choice of enhanced sampling methods.

3) Additionally, the methods section was unclear about several aspects:

a) it is unclear as to how many replicates were done for each mode: if it's a single replicate for each mode, no conclusion could be drawn about the hydration. To have any confidence in the result, the reviewers would want to see the simulations rerun many times (at least 10x, perhaps more depending on how variable the answer is), preferably starting from different structures within the equilibrated ensemble.

We initially ran 1 replica per mode, per direction and per mutant. After running two extra replicas, we could not reproduce our observations among the different replicas using the same pulling direction. This informed about how the protein could freely adapt to our soft pulling protocol and join our reply to the reviewer’s previous comment. Considering this behavior, we suspected that we would need many more than 10 replicas per pulling to observe significant trends for GHSR hydration in respect to its conformation. We thus decided to follow the reviewer’s suggestions by using a more classical approach with non-biased Molecular Dynamics simulations. We then ran 5 replicas of MD simulations (different starting velocities) starting either from an active model and the newly available inactive structure of GHSR. Our new results are described in the new version of the manuscript.

b) Which state does the original model represent? Can the comparison to the recently published structure be more thorough than simply showing a Calpha(?) RMSD (Figure S8)? Are the enhanced sampling MD carried out in presence or absence of ligand and why? Finally the method section as well as the Results section explaining the enhanced sampling protocol should be clarified such that it is not necessary to read the original paper explaining the method to understand.

The initial state of the receptor in the first submission was inactive without any ligand (PDB id: 4GRV). However, we did not use this structure anymore in the current version of the manuscript, as it was a homology model from neurotensin receptor (NTS1R). We decided to start all over from the newly available GHSR inactive structure (this structure was not released when we designed the initial study).

Considering the orthosteric site and the presence of the ligand, only the structure of GHSR bound to a synthetic antagonist is available so far. We previously published GHSR bound to its endogenous agonist ghrelin (19), but it relies only on molecular modeling with a coarse-grained description. The choice not to have any ligand involved is thus based first on the lack of structural information with agonist, and second on the fact that GHSR has a large constitutive activity (above 50%) and can be activated easily without ligand. We hypothesize as well that presence of an agonist/inverse agonist would mainly decrease the energetic barrier associated with conformational changes, which we circumvented by starting from both inactive and active models of GHSR.

c) Many key details of the simulations are missing: number of lipids, number of waters, electrostatics method (as written, it sounds like they didn't use Ewald, which would be a huge problem).

We added more information about systems and simulation setups to the method section of the manuscript. All simulations were run with Ewald summation. We modified the text accordingly: “direct non-bonded interactions were truncated at a distance cut-off of 12Å applying a switching function in the 10:12Å range, while long range electrostatics were computed via Particle Mesh Ewald (PME)”.

d) There is no discussion of statistical convergence. The simulations are very short by today's standards, and the reviewers saw no reason to assume the protein has stopped systematically changing after 300-350 ns (given the uncertainty of starting from a homology model), let alone begun actually sampling. The only evidence is Supp Figure 5, which shows the RMSD is still increasing, while nothing at all is shown for the mutants.

as stated above, we decided to run several long MD simulations for the revised version manuscript, which (we hope) now respects today’s standards (5 replicas of 5 µs for both inactive and active states, for a total of 50 µs). These simulations converged in terms of hydration patterns, as they are very similar in all simulations and moreover specific to each starting state. We also checked the structural divergence from the starting structure through Root Mean Square Deviation (RMSD) calculations (see Author response image 2). We observed expected values with higher RMSD when starting from the activated model (generated by Targeted Molecular Dynamics – TMD) in orange in comparison to simulations starting from the experimental inactive structure in blue.

Author response image 2
Root Mean Square Deviation (RMSD) of all MD simulations with initial structures as references.

The 5 MD simulations from the inactive state are represented in blue, and the 5 MD simulations from the active state are represented in orange. Running averages on top of curves are shown for clarity.

e) As it stands, too many details are missing for these calculations to be repeated. Please collect and document all of the scripts used (building the system, running, and analyzing) and put them either into the supplementary info or better yet into a separate repository (e.g. GitHub, zenodo, etc).

We used a classical MD approach for this revised version of the manuscript. We generated our systems using CHARMM-GUI, where all scripts are available on their website. We did not modify these scripts except for the equilibration phase (see Materials and methods section in the revised manuscript).

https://doi.org/10.7554/eLife.63201.sa2

Article and author information

Author details

  1. Maxime Louet

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  2. Marina Casiraghi

    Laboratoire de Biologie Physico-Chimique des Protéines Membranaires, UMR 7099, CNRS, Université de Paris, Institut de Biologie Physico-Chimique (FRC 550), Paris, France
    Present address
    1. Department of Molecular and Cellular Physiology, Stanford University School of Medicine, Stanford, United States
    2. IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0627-2288
  3. Marjorie Damian

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Data curation, Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Mauricio GS Costa

    1. Laboratoire de Biologie et Pharmacologie Appliquées, UMR 8113 CNRS, Ecole Normale Supérieure Paris-Saclay, Gif-sur-Yvette, France
    2. Programa de Computação Científica, Fundação Oswaldo Cruz, Rio de Janeiro, Brazil
    Contribution
    Data curation, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5443-286X
  5. Pedro Renault

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Present address
    Laboratory of Molecular Neuropharmacology and Bioinformatics, Institut de Neurociències, Universitat Autònoma de Barcelona, Barcelona, Spain
    Contribution
    Data curation, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Antoniel AS Gomes

    1. IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    2. Laboratório de Física Biológica, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  7. Paulo R Batista

    Programa de Computação Científica, Fundação Oswaldo Cruz, Rio de Janeiro, Brazil
    Contribution
    Data curation, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3399-174X
  8. Céline M'Kadmi

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Data curation, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  9. Sophie Mary

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  10. Sonia Cantel

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Data curation, Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  11. Severine Denoyelle

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Data curation, Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  12. Khoubaib Ben Haj Salah

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2313-2773
  13. David Perahia

    Laboratoire de Biologie et Pharmacologie Appliquées, UMR 8113 CNRS, Ecole Normale Supérieure Paris-Saclay, Gif-sur-Yvette, France
    Contribution
    Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  14. Paulo M Bisch

    Laboratório de Física Biológica, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil
    Contribution
    Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  15. Jean-Alain Fehrentz

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Formal analysis, Supervision, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6064-3118
  16. Laurent J Catoire

    Laboratoire de Biologie Physico-Chimique des Protéines Membranaires, UMR 7099, CNRS, Université de Paris, Institut de Biologie Physico-Chimique (FRC 550), Paris, France
    Contribution
    Data curation, Formal analysis, Supervision, Writing - original draft
    Competing interests
    No competing interests declared
  17. Nicolas Floquet

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Conceptualization, Formal analysis, Supervision, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3883-2852
  18. Jean-Louis Banères

    IBMM, Univ Montpellier, CNRS, ENSCM, Montpellier, France
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    jean-louis.baneres@umontpellier.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7078-1285

Funding

Agence Nationale de la Recherche (ANR-17-CE11-0011)

  • Jean-Louis Banères

Agence Nationale de la Recherche (ANR-17-CE18-0022)

  • Jean-Alain Fehrentz

Labex (EpiGenMed)

  • Khoubaib Ben haj salah

Campus France (Ph-C882/17)

  • Nicolas Floquet

Agence Nationale de la Recherche (ANR-17-CE11-0022)

  • Jean-Louis Banères

European Union’s Horizon 2020 (Marie Sklodowska-Curie grant agreement No 799376)

  • Marina Casiraghi

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

PG Schultz and L Supekova (TSRI, La Jolla, CA) are greatly acknowledged for providing genetic material. This work was supported by CNRS, Université de Montpellier, Agence Nationale de la Recherche (ANR-17-CE11-0011, ANR-17-CE11-22, ANR-17-CE18-0022), EpiGenMed Labex (post-doctoral fellowship to KBH) and DYNAMO Labex (post-doctoral fellowship to MC). This program received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement n° 799376. Mass spectrometry analyses were performed on the instruments located in the IBMM platform of instrumentation, Laboratoire de Mesures Physiques (LMP) of Université de Montpellier. We thank GENCI (Grand Équipement National de Calcul Intensif), CINES (Centre Informatique National de l'Enseignement Supérieur), and IDRIS (Institut du développement et des ressources en informatique scientifique) for computational resources. We also thank CAMPUS FRANCE for promoting the French-Brazilian collaboration via the CAPES-COFECUB project number Ph-C882/17.

Senior Editor

  1. Richard W Aldrich, The University of Texas at Austin, United States

Reviewing Editor

  1. Lucie Delemotte, KTH Royal Institute of Technology, Sweden

Reviewer

  1. Lucie Delemotte, KTH Royal Institute of Technology, Sweden

Publication history

  1. Received: September 17, 2020
  2. Accepted: July 23, 2021
  3. Version of Record published: September 3, 2021 (version 1)

Copyright

© 2021, Louet et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 370
    Page views
  • 55
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    Xavier Portillo et al.
    Research Article Updated

    An RNA polymerase ribozyme that has been the subject of extensive directed evolution efforts has attained the ability to synthesize complex functional RNAs, including a full-length copy of its own evolutionary ancestor. During the course of evolution, the catalytic core of the ribozyme has undergone a major structural rearrangement, resulting in a novel tertiary structural element that lies in close proximity to the active site. Through a combination of site-directed mutagenesis, structural probing, and deep sequencing analysis, the trajectory of evolution was seen to involve the progressive stabilization of the new structure, which provides the basis for improved catalytic activity of the ribozyme. Multiple paths to the new structure were explored by the evolving population, converging upon a common solution. Tertiary structural remodeling of RNA is known to occur in nature, as evidenced by the phylogenetic analysis of extant organisms, but this type of structural innovation had not previously been observed in an experimental setting. Despite prior speculation that the catalytic core of the ribozyme had become trapped in a narrow local fitness optimum, the evolving population has broken through to a new fitness locale, raising the possibility that further improvement of polymerase activity may be achievable.

    1. Biochemistry and Chemical Biology
    Gajanan S Patil et al.
    Research Article Updated

    Fatty acyl-AMP ligases (FAALs) channelize fatty acids towards biosynthesis of virulent lipids in mycobacteria and other pharmaceutically or ecologically important polyketides and lipopeptides in other microbes. They do so by bypassing the ubiquitous coenzyme A-dependent activation and rely on the acyl carrier protein-tethered 4′-phosphopantetheine (holo-ACP). The molecular basis of how FAALs strictly reject chemically identical and abundant acceptors like coenzyme A (CoA) and accept holo-ACP unlike other members of the ANL superfamily remains elusive. We show that FAALs have plugged the promiscuous canonical CoA-binding pockets and utilize highly selective alternative binding sites. These alternative pockets can distinguish adenosine 3′,5′-bisphosphate-containing CoA from holo-ACP and thus FAALs can distinguish between CoA and holo-ACP. These exclusive features helped identify the omnipresence of FAAL-like proteins and their emergence in plants, fungi, and animals with unconventional domain organizations. The universal distribution of FAALs suggests that they are parallelly evolved with FACLs for ensuring a CoA-independent activation and redirection of fatty acids towards lipidic metabolites.