Hidden GPCR structural transitions addressed by multiple walker supervised molecular dynamics (mwSuMD)

eLife Assessment

This study describes an improved adaptive sampling approach, multiple-walker Supervised Molecular Dynamics (mwSuMD), and its application to G protein-coupled receptors (GPCRs), which are the most abundant membrane proteins and key targets for drug discovery. The manuscript provides solid evidence that the mwSuMD approach can assist in the sampling of complex binding processes, leading to useful findings for GPCR activity, including resolution of interactions not seen experimentally. The method has the potential to have broad applicability in structural biology and pharmacology.

https://doi.org/10.7554/eLife.96513.4.sa0

Significance of the findings:

Useful: Findings that have focused importance and scope

Landmark
Fundamental
Important
Valuable
Useful

Strength of evidence:

Solid: Methods, data and analyses broadly support the claims with only minor weaknesses

Exceptional
Compelling
Convincing
Solid
Incomplete
Inadequate

During the peer-review process the editor and reviewers write an eLife Assessment that summarises the significance of the findings reported in the article (on a scale ranging from landmark to useful) and the strength of the evidence (on a scale ranging from exceptional to inadequate). Learn more about eLife Assessments

Abstract
Introduction
Results
Discussion
Methods
Appendix 1
Data availability
References
Article and author information
Metrics

Abstract

The structural basis for the pharmacology of human G protein-coupled receptors (GPCRs), the most abundant membrane proteins and the target of about 35% of approved drugs, is still a matter of intense study. What makes GPCRs challenging to study is the inherent flexibility and the metastable nature of interaction with extra- and intracellular partners that drive their effects. Here, we present a molecular dynamics (MD) adaptive sampling algorithm, namely multiple walker supervised molecular dynamics (mwSuMD), to address complex structural transitions involving GPCRs without energy input. We first report the binding and unbinding of the vasopressin peptide from its receptor V₂. Successively, we present the complete transition of the glucagon-like peptide-1 receptor (GLP-1R) from inactive to active, agonist and G_s-bound state, and the guanosine diphosphate (GDP) release from G_s. To our knowledge, this is the first time the whole sequence of events leading from an inactive GPCR to the GDP release is simulated without any energy bias. We demonstrate that mwSuMD can address complex binding processes intrinsically linked to protein dynamics out of reach of classic MD.

Introduction

Supervised molecular dynamics (Cuzzolin et al., 2016; Deganutti et al., 2020b) (SuMD) is an efficient adaptive sampling technique for studying ligand-receptor binding and unbinding pathways; here, we present the multiple walker enhancement (mwSuMD) to study a significantly wider range of structural transitions relevant to the drug design. We validated the method by applying it to G protein-coupled receptors (GPCRs), since their inherent flexibility is essential to their function and because these are the most abundant family of membrane receptors in eukaryotes (de Mendoza et al., 2014) and the target for more than one-third of drugs approved for human use (Sriram and Insel, 2018).

Vertebrate GPCRs are subdivided into five subfamilies (Rhodopsin or class A, Secretin or class B, Glutamate or class C, Adhesion, and Frizzled/Taste2) according to function and sequence (Schiöth and Fredriksson, 2005; G protein-coupled receptors, 2019). X-ray and cryo-electron microscopy (cryo-EM) show that GPCRs possess seven transmembrane (TM) helices connected by three extracellular loops (ECLs) and three intracellular loops (ICLs), with an extended and structured N-terminal extracellular domain (ECD) in all subtypes, but class A. The primary function of GPCRs is transducing extracellular chemical signals into the cytosol by binding and activating four G protein families (G_s/olf, G_i/o, G_12/13, and G_q/11) responsible for decreasing (G_i/o) or increasing (G_s/olf) cyclic adenosine-3’,5’-monophosphate, and generating inositol-1,4,5-triphosphate and diacylglycerol to increase Ca²⁺ intracellular levels (G_q) (Syrovatkina et al., 2016).

Many aspects of GPCR pharmacology remain elusive: e.g., the structural determinants of the selectivity displayed toward specific G proteins or the ability of certain agonists to drive a preferred intracellular signaling pathway over others (i.e. functional selectivity or bias) (Tan et al., 2018). GPCRs are challenging proteins to characterize experimentally due to their inherent flexibility and the transitory nature of the complexes formed with extracellular and intracellular effectors. Importantly, agonists can allosterically modify the receptor selectivity profile by imprinting unique intracellular conformations from the orthosteric binding site. The mechanism behind these phenomena is one of the outstanding questions in the GPCR field (Flock et al., 2017).

Molecular dynamics (MD) is a powerful computational methodology that predicts the movement and interactions of (bio)molecules in systems of variable complexity, at atomic detail. However, classic MD sampling is limited to the microsecond or, in the best conditions, the millisecond time scale (Hollingsworth and Dror, 2018; Durrant and McCammon, 2011). For this reason, different algorithms have been designed to speed up the simulation of rare events such as ligand (un)binding and conformational transitions. Among the most popular and effective ones are metadynamics (Bussi and Laio, 2020) and path collective variables metadynamics (D’Amore et al., 2024), accelerated MD (aMD) (Hamelberg et al., 2004), and Gaussian-accelerated MD (GaMD) (Miao et al., 2015), which introduce an energy potential to overcome the energy barriers preventing the complete exploration of the free energy surface, thus de facto bias the simulation. Energetically unbiased MD protocols, on the other hand, comprise weighted ensemble MD (Zuckerman and Chong, 2017), swarms approach (Fleetwood et al., 2020), AdaptiveGoal (Lovera et al., 2019), and SuMD (Cuzzolin et al., 2016; Sabbadin and Moro, 2014), which have largely been applied to (unbinding) small molecules, peptides, and small proteins (Cuzzolin et al., 2016; Sabbadin and Moro, 2014; Deganutti et al., 2021a; Dong et al., 2020; Deganutti et al., 2021c; Cary et al., 2022). Since SuMD is optimized for (un)bindings, we have designed mwSuMD to address more complex conformational transitions and protein-protein associations. GPCRs preferentially couple to very few G proteins out of 23 possible counterparts (Flock et al., 2017; Wall et al., 2022). It is increasingly accepted that dynamic and transient interactions determine whether the encounter between a GPCR and a G protein results in productive or unproductive coupling (Culhane et al., 2022). MD simulations are considered a useful orthogonal tool for providing working hypotheses and rationalizing existing data on G protein selectivity. However, so far, it has not delivered as expected. Attempts have usually employed energetically biased simulations, have been confined to the Gα subunit, or considered a preformed GPCR:G protein complex (Wall et al., 2022; Mattedi et al., 2020; Miao and McCammon, 2018; Mafi et al., 2023).

First, we validated mwSuMD on the nonapeptide arginine vasopressin (AVP) by simulating binding (dynamic docking) and unbinding paths from the vasopressin 2 receptor (V₂R). Dynamic docking, although more computationally demanding than standard molecular docking, provides insights into the binding mode of ligands in a fully hydrated and flexible environment. Moreover, it informs about binding paths and the complete mechanism of formation leading to an intermolecular complex, delivering in the context of binding kinetics Guo et al., 2017 and structure-kinetics relationship studies (Guo et al., 2015). We then studied the class B1 GPCR glucagon-like peptide-1 receptor (GLP-1R) activation by the small molecule PF06882961. GLP-1R is a validated target in type 2 diabetes and probably the best-characterized class B1 GPCR from a structural perspective. GLP-1R is the only class B1 receptor with structurally characterized non-peptidic orthosteric agonists, which makes it a model system for studying the druggability of the entire B1 subfamily. After GLP-1R agonist binding and activation, the coupling of G_s and the release of guanosine diphosphate (GDP), the rate-limiting step of the G protein activation, was simulated for the first time using an energy-unbiased method.

These results demonstrate the usefulness of mwSuMD for illuminating the molecular events involved in GPCR function.

Results

Short mwSuMD time windows improve the AVP dynamic docking prediction

AVP is an endogenous hormone (Figure 1a) that mediates antidiuretic effects on the kidney by signaling through three class A GPCR subtypes: V_1a and V_1b receptors activate phospholipases via G_q/11, while the V₂R activates adenylyl cyclase by interacting with G_s (Birnbaumer, 2000) and is a therapeutic target for hyponatremia, hypertension, and incontinence (Ball, 2007). AVP is amphipathic and in the bound state interacts with both polar and hydrophobic V₂R residues located on TM helices and ECLs (Figure 1b). Although AVP presents an intramolecular C1-C6 disulfide bond that limits the backbone’s overall conformational flexibility, it has many rotatable bonds, making dynamic docking complicated (Gioia et al., 2017). We compared the performance of mwSuMD to the parent algorithm SuMD in reconstructing the experimental V₂R:AVP complex using different settings, simulating a total of 92 binding events (Table 1). As a reference, the AVP root mean square deviation (RMSD) during a classic (unsupervised) equilibrium MD simulation of the X-ray AVP:V₂R complex was 3.80±0.52 Å (Figure 1—figure supplement 1). SuMD (Cuzzolin et al., 2016; Sabbadin and Moro, 2014) produced a minimum RMSD to the cryo-EM complex of 4.28 Å, with most of the replicas (i.e. distribution’s mode) having an RMSD close to 10 Å (Figure 1—figure supplement 2a). mwSuMD, with the same settings (Figure 1—figure supplement 2b, Table 1) in terms of time window duration (600 ps), metric supervised (the distance between AVP and V₂R), and acceptance method (slope) produced slightly more precise results (i.e. distribution’s mode RMSD = 7.90 Å) but similar accuracy (minimum RMSD = 4.60 Å). Supervising the AVP RMSD to the experimental complex rather than the distance (Figure 1—figure supplement 2c) and using the SMscore (Equation 1) as the acceptance method (Figure 1—figure supplement 2d) worsened the prediction. Supervising distance and RMSD at the same time (Figure 1—figure supplement 2e), employing the DMscore (Equation 2), recovered accuracy (minimum RMSD = 4.60 Å) but not precision (distribution mode RMSD = 12.40 Å). Interestingly, decreasing the time window duration from 600 ps to 100 ps impaired the SuMD ability to predict the experimental complex (Figure 1a), but enhanced mwSuMD accuracy and precision (Figure 1b–d). The combination of RMSD as the supervised metric and SMscore produced the best results in terms of minimum RMSD and distribution mode RMSD, 3.85 Å and 4.40 Å, respectively (Figure 1d, Video 1), in agreement with the AVP deviations in the equilibrium MD simulation of the X-ray AVP:V₂R complex (Figure 1—figure supplement 1). These results confirm the inherent complexity of reproducing the AVP:V₂R complex via dynamic docking and suggest that short time windows can improve mwSuMD performance on this system. However, it is necessary to know the final bound state to employ the RMSD as the supervised metric, while the distance is required to dynamically dock ligands with unknown bound conformation as previously reported (Cuzzolin et al., 2016; Wall et al., 2022). Both distance and RMSD-based simulations delivered insights into the binding path and the residues involved along the recognition route. For example, mwSuMD pinpointed V₂R residues E184^ECL2, P298^ECL3, and E303^ECL3 (Figure 2a) as involved during AVP binding, although not in contact with the ligand in the orthosteric complex. None of them are yet characterized through mutagenesis studies according to the GPCRdb (Isberg et al., 2016).

Figure 1 with 2 supplements see all

Download asset Open asset

Arginine vasopressin (AVP) supervised molecular dynamics (SuMD) and multiple walker SuMD (mwSuMD) binding simulations to vasopressin 2 receptor (V₂R) (100 ps time windows).

(a) Chemical structure and (b) binding more within the V₂R orthosteric binding site; AVP is represented in orange stick, while V₂R is in white ribbon and gray stick. For each set of settings (**c–f**) the root mean square deviation (RMSD) of AVP Cα atoms to the cryo-electron microscopy (cryo-EM) structure 7DW9 is reported during the time course of each SuMD (c) or mwSuMD (**d–f**) replica alongside the RMSD values distribution and the snapshot corresponding to the lowest RMSD values (AVP from the cryo-EM structure 7DW9 is in a cyan stick representation, while AVP from simulations is in a tan stick representation). A complete description of the simulation settings is reported in Table 1 and the Methods section. The dashed red line indicates the AVP RMSD during a classic (unsupervised) equilibrium MD simulation of the X-ray AVP:V₂R complex (Figure 1—figure supplement 1).

Table 1

Summary of all the simulations performed and the settings employed.

System	# Replicas	TWduration	# Walkers	Metric	Acceptance
V₂R:AVP complex	1 classic MD	500 ns	N/A	N/A	N/A
V₂R:AVP binding	9 (SuMD)	600 ps	N/A	Distance	Slope
	10 (mwSuMD)	600 ps	3	Distance	Slope
	11 (mwSuMD)	600 ps	3	RMSD	Slope
	8 (mwSuMD)	600 ps	3	Distance	SMscore
	9 (mwSuMD)	600 ps	3	Distance and RMSD	DMscore
	12 (mwSuMD)	100 ps	10	Distance and RMSD	DMscore
	11 (mwSuMD)	100 ps	10	Distance	SMscore
	11 (mwSuMD)	100 ps	10	RMSD	SMscore
	10 (SuMD)	100 ps	N/A	Distance	Slope
V₂R:AVP unbinding	5 (SuMD)	100 ps	N/A	Distance	Slope
V₂R:AVP unbinding	5 (mwSuMD)	100 ps	10	Distance	SMscore
β₂ AR:G_s protein binding	3 (mwSuMD)	100 ps	5	Distance and RMSD	DMscore
A1R:G_i binding	1 (mwSuMD)	100 ps	3	RMSD	SMscore
GLP-1R:PF06882961	1 (mwSuMD)	100 ps	5	Distance or RMSD (or a combination)	SMscore or DMscore
GLP-1R:G_s protein binding	3 (mwSuMD)	200 ps	3	Distance or RMSD	SMscore or DMscore
G_s AHD opening	1 (mwSuMD)	100 ps	3	Distance	SMscore
GLP-1R:G_s GDP unbinding	3	50 ps	5	Distance	SMscore

N/A: not applicable; SuMD was not performed.

Figure 2

Download asset Open asset

Multiple walker supervised molecular dynamics (mwSuMD) (un)binding simulations of arginine vasopressin (AVP).

Vasopressin 2 receptor (V₂R) residues involved in mwSuMD simulations of AVP. (a) Binding (dynamic docking); (b) unbinding simulations. (c) Dissociation of AVP from V₂R. The root mean square deviation (RMSD) of AVP to the initial bound state is reported during the time course of five replicas of SuMD and mwSuMD, respectively.

Video 1

Download asset

posterframe for video — Arginine vasopressin (AVP) binding (dynamic docking) to vasopressin 2 receptor (V₂R).

Two-view of the mwSuMD replica better reproducing the AVP:V2R experimental complex. Left side: V2R is represented in white quick surface, while AVP is in transparent quick surface and green stick; right side: V2R is represented in white ribbon and cyan stick, while AVP is a green stick. The bound AVP conformation from PDB 7DW9 is reported as a reference in transparent orange ribbon and stick.

Further to binding, an SuMD approach was previously employed to reconstruct the unbinding path of ligands from several GPCRs (Deganutti et al., 2020b; Atanasio et al., 2020). We assessed mwSuMD’s capability to simulate AVP unbinding from V₂R. Five mwSuMD and five SuMD replicas were collected using 100 ps time windows (Table 1). Overall, mwSuMD outperformed SuMD in terms of time required to complete a dissociation (Figure 2c, Video 2), producing dissociation paths almost 10 times faster than SuMD. Such rapidity in dissociating inherently produces a limited sampling of metastable states along the pathway, which can be compensated by seeding classic (unsupervised) MD simulations from configurations extracted from the unbinding pathway (Deganutti et al., 2021b; Deganutti et al., 2020a). Here, the V₂R residues involved during the dissociation were comparable to the binding (Figure 2a and b), although ECL2 and ECL3 were slightly more involved during the association than the dissociation, in analogy with other class A and B GPCRs (Dong et al., 2020; Deganutti et al., 2021b).

Video 2

Download asset

PF06882961 binding and GLP-1R activation

The GLP-1R has been captured by cryo-EM in both the inactive and the active (G_s-bound) conformations and in complex with either peptide or non-peptide agonists (Zhao et al., 2020; Kawai et al., 2020; Ma et al., 2020; Zhang et al., 2020; Cong et al., 2021; Cong et al., 2022b). In the inactive GLP-1R, residues forming the binding site for the non-peptide agonist PF06882961 are dislocated and scattered due to the structural reorganization of the transmembrane domain (TMD) and ECD (Figure 3—figure supplement 1a) that occurs on activation. Moreover, GLP-1R in complex with GLP-1 or different agonists present distinct structural features, even among structurally related ligands (Figure 3—figure supplement 1b and c). This complicates the scenario and suggests divergent recognition mechanisms among different agonists. We simulated the binding of PF06882961, reaching an RMSD to its bound conformation in 7LCJ of 3.79±0.83 Å (computed on the second half of the merged trajectory, superimposing on GLP-1R Cα atoms of TMD residues 150–390), using multistep supervision on different system metrics (Figure 3) to model the structural hallmark of GLP-1R activation (Videos 3 and 4).

Figure 3 with 2 supplements see all

Download asset Open asset

Multiple walker supervised molecular dynamics (mwSuMD) simulation of PF06882961 binding to glucagon-like peptide-1 receptor (GLP-1R) and receptor activation.

Each panel reports the root mean square deviation (RMSD) to the position of the ligand in the active state (top panel) or a GLP-1R structural element over the time course (all but ECL3 converging to the active state). ECD: extracellular domain; TM: transmembrane helix; ECL: extracellular loop. The mwSuMD simulation was performed with four different settings over 1 µs in total. The red dashed lines show the initial RMSD value for reference.

Video 3

Download asset

Video 4

Download asset

Several metrics were supervised consecutively. First, the distance between PF06882961 and the TMD as well as the RMSD of the ECD to the active state (stage 1); second, the RMSD of ECD and ECL1 to the active state (stage 2); third, the RMSD of PF06882961 and ECL3 to the active state (stage 3); lastly, only the RMSD of TM6 (residues I345-F367, Cα atoms) to the active state (stage 4). The combination of these supervisions produced a conformational transition of GLP-1R toward the active state (Figure 3, Video 4). Noteworthy, mwSuMD, like any other CV-based technique, requires some knowledge of the simulated system. The sequence of these supervisions was arbitrary and does not necessarily reflect the right order of the steps involved in GLP-1R activation. This kind of planned multistep approach is feasible when the end-point receptor inactive and active structures are available, and the inherent flexibility of different domains is known. In class B GPCRs, the ECD is the most dynamic substructure, followed by the ECL1 and ECL3, which display high plasticity during ligand binding (Dong et al., 2020; Cong et al., 2022a). For this reason, we first supervised these elements of GLP-1R, leaving the bottleneck of activation, TM6 outward movement, as the last step. However, the protocol employed can be tweaked to study how each conformational transition takes place and influences the receptor domains. Structural elements not directly supervised, such as TM1 or TM7, were influenced by the movement of supervised helixes or loops and therefore displayed an RMSD reduction to the active state. For example, the supervision of ECL3 (stage 3) and TM6 (stage 4) facilitated the spontaneous rearrangement of the ECD to an active-like conformation after the ECD had previously experienced transient high flexibility during stages 2 and 3 (Figure 3). These results suggest a concerted conformational transition for ECD and ECL1 during the binding of PF06882961 and an allosteric effect between ECL3 and the bottom of TM6. While the intracellular polar interactions were destabilized by the ECL3 transition to an active-like conformation (stages 2 and 3), the outward movement of TM6 (stage 4) did not favor the closure of ECL3 toward PF06882961, which appears to be driven by direct interactions between the ligand and R310^5.40 or R380^7.35. Interestingly, the mwSuMD simulation during stage 4 (TM6 supervision) sampled a counterclockwise helix rotation (Figure 3—figure supplement 2a) consistent with the GLP-1R cryo-EM structures in the active, G_s-coupled state (Zhang et al., 2020; Zhang et al., 2021).

It is worth noting that 6LN2, the only inactive GLP-1R structure available complete with the ECD, was stabilized and solved with an antibody bound to the ECD. This strategy might have forced the ECD into a closed conformation that engages the EC vestibule of GLP-1R and possibly restrained the whole TMD in an altered conformation that deviates from the physiological conditions. This might explain why the RMSD of the TMD elements monitored during the simulation rarely reach values lower than 3 Å or 4 Å.

During the supervision of ECL3 and PF06882961 (stage 3), we observed a loosening of the intracellular polar interactions that stabilize GLP-1R TM6 in the inactive state. As a result, the subsequent supervision of TM6 (residues I345-F367, Cα atoms) rapidly produced the outward movement of TM6 toward the active state, in the last step of the mwSuMD simulation (stage 4). A more detailed analysis revealed that the central polar network, which is pivotal for mediating GLP-1 signaling (Wootten et al., 2016), and the residues at the TM6 kink level adopted active-like conformations during the final stage of the simulation (Figure 4a). In particular, the central polar network (E364^6.53, H363^6.52, and Q394^7.49) experienced side chain rearrangements (Figure 4b) while extensive TM6 kink dislocation occurred at L360^6.49, P358^6.48, L357^6.47, and I357^6.46 (Figure 4c).

Figure 4

Download asset Open asset

Glucagon-like peptide-1 receptor (GLP-1R) key structural motifs during multiple walker supervised molecular dynamics (mwSuMD) GLP-1R activation.

(a) Root mean square deviation (RMSD) to the active state GLP-1R (7LCI) of the residues forming the central polar network (magenta) and TM6 kink (orange) during mwSuMD of receptor activation; at the end of the simulations minimum values were reached (dashed square). (b) The position of the polar network within the core of transmembrane domain (TMD) (left-hand panel) and comparison between inactive, active, and mwSuMD final states for the side chains of the residues forming of the polar network. (c) The position of the TM6 kink (right-hand panel) and comparison between inactive, active, and mwSuMD final states for the side chains of the residues forming the TM6 kink.

G_s protein binding to GLP-1R and GDP release

We then focused on simulating the G_s binding to GLP-1R, after activation, without energy input. For this purpose, we first tested the binding between the prototypical class A receptor β₂ adrenoreceptor (β₂ AR) and the stimulatory G protein (G_s) (Video 5, Figure 5—figure supplement 1a and b) by supervising the distance between G_s helix 5 (α5) and β₂ AR as well as the RMSD of the intracellular end of TM6 to the fully active state of the receptor (see Supplementary Methods). During two out of three replicas, both Gα and Gβ achieved distance values close to 5 Å (minimum RMSD = 3.94 Å and 3.96 Å respectively), in good agreement with the reference (the β₂ AR:G_s complex, PDB 3SN6, Figure 5—figure supplement 1c). A possible pitfall is that G proteins bear potential palmitoylation and myristoylation sites that anchor the inactive trimer to the plasma membrane (Linder et al., 1993; Zhang et al., 2004), de facto restraining possible binding paths to the receptor. To address this point and test different conditions, we prepared the adenosine A1 receptor (A₁R) and its principal effector, the inhibitory G protein (G_i) considering the G_iα residue C3 and Gγ residue C65 as palmitoylated and geranylgeranylated respectively and hence inserted in the membrane. Recently, the G_i binding to A₁R was simulated by combining the biased methods GaMD with SuMD (Li et al., 2022) but without considering membrane-anchoring posttranslational modifications. Both classic (unsupervised) and mwSuMD simulations were performed on this system for comparison (Video 6, Figure 5—figure supplement 1d). In about 50 ns of mwSuMD, the G_iα subunit engaged its intracellular binding site on A₁R and formed a complex in good agreement with the cryo-EM structure (PDB 6D9H, RMSD ≈ 5 Å). For comparison, 1 μs of cMD did not produce a productive engagement as the G_iα remained at RMSD values >40 Å (Figure 5—figure supplement 1d), suggesting the effectiveness of mwSuMD in sampling G protein binding rare events without the input of energy. The membrane anchoring affected the overall G_i binding and the final complex, which was rotated compared to the experimental structure due to the lipidation of G_iα and Gγ (Figure 5—figure supplement 1e).

Video 5

Download asset

Video 6

Download asset

Encouraged by results obtained on G_s and G_i binding to β₂ AR and A₁R (Figure 5—figure supplement 1), we extracted the GLP-1R active conformation described above and simulated the G_s binding to its intracellular side. Starting from the inactive, membrane-anchored G_s, we performed three independent mwSuMD replicas by supervising the distance between G_s helix α5 and GLP-1R residues located at the intracellular binding interface (Figure 5a and e). All three mwSuMD replicas showed the G_s approaching GLP-1R, with two out of three reaching an RMSD of the G_sα subunit close to or less than 10 Å, compared to the experimental complex 7LCI (Figure 5e). Replica 2, in particular, reproduced the cryo-EM GLP-1R:G_s complex with RMSD values to 7LCI of 7.59±1.58 Å, 12.15±2.13 Å, and 13.73±2.24 Å for Gα, Gβ, and Gγ, respectively. Such values do not support convergence with the static experimental structure but are not far from the RMSDs measured in our previous simulations of GLP-1R in complex with G_s and GLP-1 (Deganutti et al., 2022) (Gα=6.18 ± 2.40 Å; Gβ=7.22 ± 3.12 Å; Gγ=9.30 ± 3.65 Å), which indicates overall higher flexibility of Gβ and Gγ compared to Gα, which acts as a sort of fulcrum bound to GLP-1R.

Figure 5 with 2 supplements see all

Download asset Open asset

Glucagon-like peptide-1 receptor (GLP-1R) activation and G_s binding.

(**a–d**) Sequence of simulated events during the multiple walker supervised molecular dynamics (mwSuMD) G_s:GLP-1R simulations. (e) Root mean square deviation (RMSD) of G_sα to the experimental GLP-1R:G_s complex (PDB 7LCJ) during three mwSuMD replicas; the RMSD to the experiential bound conformation (7LCJ) during the second part of Replica 2 (red dashed line) is reported for each G_s subunit. RMSDs were computed on Gα residues 11–43 and 205–394 to the experimental structure 7LCI after superimposition on GLP-1R residues 140–240 Cα atoms. (f) RMSD of the αG-α4 loop (purple), the distance between K342^6.31 and D323^Gα (salmon), and the distance between guanosine diphosphate (GDP) and D295^Gα (green) during G_s binding, alpha-helical domain (AHD) opening and GDP dissociation; (g) and (h) comparison between states extracted from before and after AHD opening. Before AHD opening (a), GLP-1R ICL3 interacted with D323^Gα and D295^Gα interacted with GDP; after AHD opening (b) and αG-α4 loop reorganization (curved red arrow), αG and D295^Gα moved away from GDP (straight red arrow), destabilizing its binding to G_s.

According to the model of G protein activation, the G protein binding has the effect of allosterically stabilizing the orthosteric agonist in complex with the receptor (Deganutti et al., 2022) and destabilizing the GDP bound to Gα, triggering its release and exchange with guanosine triphosphate (Gregorio et al., 2017), upon opening of the G protein alpha-helical domain (AHD). Following this model, PF06882961 and GDP were respectively stabilized and destabilized during the simulated G_s association (Figure 3—figure supplement 2b and c). The analysis of atomic contacts along the binding path of G_s to GLP-1R highlights a few persistent interactions not observed in the equilibrium MD simulations of GLP-1R:G_s cryo-EM complexes Deganutti et al., 2022; e.g., we propose the hidden interaction between D344^6.33 and R385^α5 to be important for G_s coupling (Appendix 1—table 1), which would explain the GLP-1 EC₅₀ reduction upon mutation of position 344 to Ala (Yuan et al., 2023).

Extending Replica 2, we further investigated the G_s activation mechanism by supervising the opening of the G_s AHD, which is considered a necessary step to allow GDP release from the Ras-like domain (Dror et al., 2015). We first easily obtained the opening of AHD (Figure 5b) and successively supervised the GDP unbinding in a further three replicas, seeded after the AHD opening. In one of these three mwSuMD simulations, the nucleotide dissociated from G_s (Figure 5d). Video 7 shows the full G_s binding, AHD opening, and GDP release. mwSuMD suggested several structural changes as implicated in GDP dissociation (Figure 5f–h): (i) the AHD opening; (ii) the conformational change of αG-α4 loop (residues A303^Gα-P332^Gα), in concert with a loosening of interactions between D323^Gα and K342^6.31, and (iii) the rupture of the hydrogen bond between GDP and D295^Gα triggered by the movement of αG away from the GDP binding site (Figure 5f). The involvement of αG through the αG-α4 during the release of GDP from G_s is supported by hydrogen/deuterium exchange experiments (Du et al., 2019), while the role of D275^Gα has been probed with functional assays on the G_i isoform mutant D272^GαiA (Ham et al., 2021). Moreover, a similar αG behavior was recently suggested by MD simulations of the G_s binding pathway to β₂ AR (Batebi et al., 2024). We also note that the αG-α4 loop length and amino acidic composition diverge among G protein isoforms, further suggesting a role in G protein selectivity consistent with the hypothesis that, in different G proteins, distinct domains of the Gα subunit could be responsible for receptor selectivity (Glukhova et al., 2018). However, a more subtle allosteric communication through internal structural elements like the β2-β3 strands, prompted by α5 tilting, could have weakened GDP phosphate binding as previously suggested by other groups (Sun et al., 2018; Kaya et al., 2016; Flock et al., 2015). Interestingly, no significant conformational changes of the β6-α5 loop happened before or during the GDP dissociation, suggesting that its conformational change as captured in the nucleotide-free GLP-1R:G_s complex (Figure 5—figure supplement 2) occurs after the GDP release as a result of the loss of binding stabilization, rather than being an initiator of the GDP dissociation.

Video 7

Download asset

Discussion

Classic MD simulations sample phase space with an efficiency that depends on the energy barrier between neighboring minima. Processes like (un)binding and protein activation require the system to overcome numerous energy barriers, some creating bottlenecks that slow the transition down to the millisecond, or second, time scale. To reduce some of these limits, we have developed and tested on complex structural events characterizing GPCRs, an energetically unbiased adaptive sampling algorithm, namely mwSuMD, which is based on traditional SuMD, while drawing on parallel multiple replica methods (Dickson and Brooks, 2014; Sugita and Okamoto, 1999).

Our simulations propose that remarkable predictivity can be obtained with distance-driven mwSuMD, as demonstrated by the lowest deviation from the experimental AVP:V₂R complex. The dissociation of AVP from V₂R was simulated much more rapidly by mwSuMD than by SuMD, suggesting it is an efficient tool for studying the dissociation of ligands from GPCRs. This is due to the more extensive sampling obtainable by seeding multiple parallel short simulations instead of a single simulation for batch.

mwSuMD performed similarly to SuMD for the dynamic docking of AVP to V2R when time windows of 600 ps were employed. Time windows of 100 ps remarkably improved mwSuMD. Usually, dynamic docking is performed to either predict the geometry of complexes or sample the binding path of an already-known intermolecular complex, or both. The RMSD of AVP to the experimental coordinates as the supervised metric produced the best results. Consequently, the RMSD should be the metric of choice to study the binding path of well-known intermolecular complexes. The distance, on the other hand, is necessary when limited structural information about the binding mode is available. In the absence of structural information regarding the final bound state, it is possible to sample numerous binding events employing mwSuMD and evaluate the final bound states rank by applying end-point free energy binding methods like the molecular mechanics energies combined with the Poisson-Boltzmann or generalized Born and surface area continuum solvation (MM/PBSA and MM/GBSA [Wang et al., 2019]) models.

We increased the complexity of binding simulations by considering GLP-1R and the non-peptide agonist PF06882961. Using mwSuMD, we obtained a binding of PF06882961 in good agreement with the cryo-EM structure, followed by an active-like conformational transition of GLP-1R. The choice of the metrics supervised was driven by the structural data available (Zhang et al., 2020) and extensive preparatory MD simulations. However, binding routes are possible from either the bulk solvent or the membrane (Deganutti et al., 2021b; Stanley et al., 2016; Bokoch et al., 2018). These results show the power of the mwSuMD method, indicating that future applications could include ligand binding from the membrane, or alternative apo receptor conformations to improve the sampling for more difficult receptors.

mwSuMD enabled us to simulate the G_s binding to the active GLP-1R and the subsequent GDP release. Our results suggest a concerted effect on GDP binding produced by AHD opening, αG-α4 loop rearrangement, and αG shift away from the GDP site. The full rotation and elongation of α5 as in cryo-EM structures would occur after the GDP release, supporting the role of hidden, metastable interactions as the driving force of G protein coupling and selectivity, as per recent work on GLP-1R (Yuan et al., 2023).

We stress that a complete understanding of a complex molecular event like G protein coupling requires the collection of numerous G protein binding paths and GDP dissociation events. mwSuMD is designed to yield a mechanistic description of structural events. For this reason, it integrates well with mutagenesis and kinetics experiments. We note that mwSuMD trajectories, since they describe the sequential states along a transition pathway, can represent a precious backbone for further MD sampling aimed at quantifying or predicting the kinetics of the transition. Approaches such as path collective variable metadynamics (Alba Ortíz et al., 2019), Markov state models (Husic and Pande, 2018), or machine learning models (Kokh et al., 2019; Pérez et al., 2018) can be informed by mwSuMD. We will address these novel opportunities created by mwSuMD in future work.

In summary, we showcased the applicability domain of mwSuMD to key, but scarcely understood, aspects of GPCR structural biology, pharmacology, and drug design hitherto unaddressed by unbiased simulations. Given the generality and simplicity of its implementation, we anticipate that mwSuMD can be employed to study a wide range of structural phenomena characterizing both membrane and cytosolic proteins. mwSuMD is an ongoing project updated on https://github.com/pipitoludovico/mwSuMD_multiEngine (copy archived at Pipitò, 2025) and, in the recent implementation, can exploit ACEMD (Harvey et al., 2009), NAMD (Phillips et al., 2005), GROMACS (Berendsen et al., 1995), and OPENMM (Eastman et al., 2017) as graphic processing units (GPU)-based MD engines while offering the option to run also on CPUs.

Methods

mwSuMD protocol

The SuMD is an adaptive sampling method (Deganutti and Moro, 2017) based on a tabu-like algorithm for speeding up the simulation of binding events between small molecules (or peptides [Salmaso et al., 2017; Bower et al., 2018]) and proteins (Cuzzolin et al., 2016; Sabbadin and Moro, 2014) without the introduction of any energetic bias. Briefly, during SuMD, a series of short unbiased MD simulations are performed, and after each simulation, the distances between the centers of mass (or the geometrical centers) of the ligand and the predicted binding site (collected at regular time intervals) are fitted to a linear function. If the resulting slope is negative (showing progress toward the target), the next simulation step starts from the last set of coordinates and velocities, otherwise, the simulation is restarted by randomly assigning the atomic velocities.

mwSuMD is designed to increase the sampling from a specific configuration by seeding user-decided parallel replicas (walkers) rather than one short simulation as per SuMD. Since one replica for each batch of walkers is always considered productive, mwSuMD gives more control than SuMD on the total wall-clock time used for a simulation. On the flip side, to maximize mwSuMD, it is optimal to assign one walker per GPU, requiring multiple GPUs to be effective. However, modern multi-threaded GPUs can still employ mwSuMD with a smaller cost in GPU performance. In the implementation for ACEMD used in this work, mwSuMD needs as input the initial coordinates of the system as a pdb file, the coordinates, and the atomic velocities of the system from the equilibration stage, the topology file of the system, and all the necessary force field parameters. The user can decide to supervise one (X) or two metrics (X’, X’’) of the simulated system over short simulations seeded in batches, called walkers. In the former case, either the slope of the linear function interpolating the metric values or a score can be adopted to decide whether to continue the mwSuMD simulation. When the user decides to supervise two metrics, a specific score is used. In the present work, distances between centroids, RMSDs, or the number of atomic contacts between two selections were supervised (Table 1). The choice of the metrics is system and problem-dependent, as the RMSD is most useful when the final state is known, while the distance is required when the target state is unknown; details on the scores are given below. The decision to restart or continue mwSuMD after any short simulation is postponed until all the walkers of a batch are collected. The best short simulation is selected and extended by seeding the same number of walkers, with the same duration as the step before.

For each walker, the score for the supervision of a single metric (SMscore) is computed as the square root of the product between the metric value in the last frame (X_{last frame}) and the average metric value over the short simulation (X̅):

S M s c o r e = \sqrt{X_{l a s t f r a m e} * \bar{X}}

If the metric is set to decrease (e.g. binding or dimerization) the walker with the lowest SMscore is continued, otherwise (e.g. unbinding or outward opening of domains), the walker with the highest score is continued. Using the SMscore rather than the slope should give more weight to the final state of each short simulation, as it is the starting point for the successive batch of simulations. Considering the average of the metric should favor short simulations consistently evolving in the desired direction along the metric.

If both X’ and X’’ are set to increase during the mwSuMD simulations, the score for the supervision of two metrics (DMscore) on each walker is computed as follows:

D M s c o r e = ((\frac{X_{l a s t f r a m e}^{'}}{{\bar{X}}_{b a t c h w a l k e r s}^{'}} - 1) + (\frac{X_{l a s t f r a m e}^{''}}{{\bar{X}}_{b a t c h w a l k e r s}^{''}} - 1)) * 100

where X’_{last frame} and X’’_{last frame} are the metrics values in the last frame, while X̅^’_{batch walkers} and X̅^’’_{batch walkers} represent the average value of the two metrics over all the walkers in the batch. Subtracting the value 1 to the metric ratio ensures that if one of the two metrics from the last frame (X’_{last frame} or X^’’_{last frame}) is equal to the average (X̅^’’_{batch walkers} or X̅^’’_{batch walkers}), then that metric addend is null, and DMscore depends only on the remaining metric. If any of the two metrics is set to decrease, then the corresponding component in Equation 2 is multiplied by –1 to maintain a positive score. Considering the average value of the two metrics over all the walkers rather than only over the considered walker should be more representative of the system evolution along the defined metric. In other words, the information about the metric is taken from all the walkers to better describe the evolution of the system.

The DMScore is designed to preserve some degree of independence between the two metrics supervised. Indeed, if the variation of one of them slows down and gets close to zero, the other metric is still able to drive the system’s evolution. It should be noted that DMScore works at its best if the two metrics have similar variations over time, as in the case of distance and RMSD (both of which are distance-based). Notably, differently from SuMD, when a walker is extended by seeding a new batch of short simulations and the remaining walkers are stopped, the atomic velocities are not reassigned. This allows the simulations to be as short as a few picoseconds if desired, without introducing artifacts due to the thermostat latency to reach the target temperature (usually up to 10–20 ps when a simulation is restarted, reassigning the velocities of the atoms).

The current implementation of mwSuMD is for Python3 and exploits MDAnalysis (Michaud-Agrawal et al., 2011) and MDTRaj (McGibbon et al., 2015) modules.

Force field, ligands parameters, and general systems preparation

The CHARMM36 (Huang and MacKerell, 2013; Huang et al., 2017) /CGenFF 3.0.1 (Vanommeslaeghe and MacKerell, 2012; Vanommeslaeghe et al., 2012; Yu et al., 2012) force field combination was employed in this work. Initial ligand force field, topology, and parameter files were obtained from the ParamChem webserver (Vanommeslaeghe and MacKerell, 2012). Restrained electrostatic potential (Woods and Chappelle, 2000) partial charges were assigned to all the non-peptidic small molecules but adrenaline and GDP using Gaussian09 (HF/6-31G* level of theory) and AmberTools20.

Six systems were prepared for MD (Table 1). Hydrogen atoms were added using the pdb2pqr (Dolinsky et al., 2004) and propka (Olsson et al., 2011) software (considering a simulated pH of 7.0); the protonation of titratable side chains was checked by visual inspection. The resulting receptors were separately inserted in a 1-palmitoyl-2-oleyl-sn-glycerol-3-phosphocholine (POPC) bilayer (previously built by using the VMD Membrane Builder plugin 1.1, Membrane Plugin, Version 1.1, at http://www.ks.uiuc.edu/Research/vmd/plugins/membrane/), through an insertion method (Sommer, 2013). Receptor orientation was obtained by superposing the coordinates on the corresponding structure retrieved from the OPM database (Lomize et al., 2006). Lipids overlapping the receptor TM helical bundle were removed and TIP3P water molecules (Jorgensen et al., 1983) were added to the simulation box employing the VMD Solvate plugin 1.5 (Solvate Plugin, Version 1.5. at https://www.ks.uiuc.edu/Research/vmd/plugins/solvate/). Finally, overall charge neutrality was reached by adding Na⁺/Cl^- counterions up to the final concentration of 0.150 M, using the VMD Autoionize plugin 1.3 (Autoionize Plugin, Version 1.3. at https://www.ks.uiuc.edu/Research/vmd/plugins/autoionize/).

System equilibration and general MD settings

The MD engine ACEMD3 (Harvey et al., 2009) was employed for both the equilibration and productive simulations. The equilibration was achieved in isothermal-isobaric conditions (NPT) using the Berendsen barostat (Berendsen et al., 1984) (target pressure 1 atm) and the Langevin thermostat (Loncharich et al., 1992) (target temperature 300 K) with low damping of 1 ps^–1. For the equilibration (integration time step of 2 fs): first, clashes between protein and lipid atoms were reduced through 1500 conjugate-gradient minimization steps, then a positional constraint of 1 kcal mol^–1 Å^–2 on all heavy atoms was gradually released over different time windows: 2 ns for lipid phosphorus atoms, 60 ns for protein atoms other than alpha carbon atoms, 80 ns for alpha carbon atoms; a further 20 ns of equilibration was performed without any positional constraints.

Productive trajectories (Table 1) were computed with an integration time step of 4 fs in the canonical ensemble (NVT). The target temperature was set at 300 K, using a thermostat damping of 0.1 ps^–1; the M-SHAKE algorithm (Forester and Smith, 1998; Kräutler et al., 2001) was employed to constrain the bond lengths involving hydrogen atoms. The cutoff distance for electrostatic interactions was set at 9 Å, with a switching function applied beyond 7.5 Å. Long-range Coulomb interactions were handled using the particle mesh Ewald summation method (Essmann et al., 1995) by setting the mesh spacing to 1.0 Å.

Vasopressin binding simulations

The V₂R in complex with vasopressin (AVP) and the G_s protein (Zhou et al., 2021) was retrieved from the Protein Data Bank (Berman et al., 2000) (PDB 7DW9). The G_s was removed from the system and the missing residues on ECL2 (G185-G189) were modeled from scratch using Modeller 9.19 (Fiser and Sali, 2003), considering the solution with the lowest DOPE score out of 10 conformations produced. AVP was placed away from V₂R in the extracellular bulk and the resulting system was prepared for MD simulations and equilibrated as reported above.

During SuMD simulations, the distance between the centroids of AVP residues C1-Q4 (backbone and side chains), anticipated to bind deep into V₂R, and the V₂R residues lining the peptide binding site Q96, Q174, Q291, and L312 (Cα atoms only) was supervised over time windows of 600 ps or 100 ps (Table 1). mwSuMD simulations considered the same distance, the RMSD of AVP residues C1-Q4 to the experimental bound complex, or the combination of the two during time windows of 600 ps (3 walkers) or 100 ps (10 walkers) (Table 1). Slope, SMscore, or DMscore (see Methods section mwSuMD protocol) was used in the different mwSuMD replicas performed (Table 1). Simulations were stopped after 300 ns (time window duration = 600 ps) or 50 ns (time window duration = 100 ps) of total SuMD or mwSuMD simulation time.

Vasopressin unbinding simulations

The V₂R:AVP complex was prepared for MD simulations and equilibrated as reported above. During both SuMD and mwSuMD simulations (Table 1), the distance between the centroids of AVP residues C1-Q4 (backbone and side chains) and V₂R residues Q96, Q174, Q291, and L312 (Cα atoms only) was supervised over time windows of 100 ps (10 walkers seeded for mwSuMD simulations). Replicas were stopped when the AVP-V₂R distance reached 40 Å.

GLP-1R:PF06882961 binding simulations

The inactive GLP-1R was retrieved from the Protein Data Bank (Berman et al., 2000) (PDB 6LN2) (Wu et al., 2020). Fab and the intracellular negative allosteric modulator were removed, and missing residues in the stalk (129–134) and ICL2 (256–263) were modeled with Modeller 9.19, considering the solutions with the lowest DOPE score out of 10 conformations produced. The PF06882961 initial conformation was extracted from the complex with the fully active GLP-1R (Zhang et al., 2021) (PDB 7LCJ) and placed away from GLP-1R in the extracellular bulk. The resulting system was prepared for MD simulations and equilibrated as reported above. CGenFF dihedral force field parameters of PF06882961 with the highest penalties (dihedrals NG2R51-CG321-CG3C41-CG3C41 [penalty = 143.5] and NG2R51-CG321-CG3C41-OG3C51 [penalty = 152.4]) were optimized (Figure 6a) employing Gaussian09 (geometric optimization and dihedral scan at HF/6-31g(d) level of theory) and the VMD force field toolkit plugin (Mayne et al., 2013).

Figure 6

Download asset Open asset

Preparation to GLP-1R mwSuMD simulations.

(a) Potential energy surface derived from the scan of the PF06882961 rotatable bond through which the dihedrals NG2R51-CG321-CG3C41-CG3C41 (penalty = 143.5) and NG2R51-CG321-CG3C41-OG3C51 (penalty = 152.4) pass. The curve obtained from the original CGenFF parameters (red) was optimized (blue) to resemble the energy profile from quantum-mechanics computations at the HF/6-31g(d) level of theory (black). (b) Distances between glucagon-like peptide-1 receptor (GLP-1R) key residues E138^1.33 and W214^ECL1 (black), H363^6.52 and F390^7.45 (blue) and Q47^ECD and S310^ECL2 (red) during 8 μs of classic molecular dynamics (MD); (c) positions of the three couples of residues reported as two-side views.

Four classic MD replicas, for a total of 8 μs, were performed on the inactive receptor (prepared for MD simulations and equilibrated as reported above) to assess the possible binding path to the receptor TMD and therefore decide the initial position of PF06882961 in the extracellular bulk of the simulation box. A visual inspection of the trajectories suggested three major conformational changes that could allow ligand access to the TMD (Figure 5b and c). Transitory openings of the ECD (distance Q47^ECD-S310^ECL2), TM6-TM7 (distance H363^6.52-F390^7.45), and TM1-ECL1 (distance E138^1.33 and W214^ECL1) were observed. Since the opening of TM1-ECL1 was observed in two replicas out of four, we placed the ligand in a favorable position for crossing that region of GLP-1R.

mwSuMD simulations (Table 1) were performed stepwise to dock the ligand within GLP-1R first and then relax the receptor toward the active state. The PF06882961 binding was obtained by supervising at the same time the distance between the ligand’s heavy atoms centroid and the centroid of GLP-1R TM7 residues L379^7.34-F381^7.36 (Cα atoms only), which are part of the orthosteric site, and the RMSD of the ECD (residues W33^ECD-W120^ECD, Cα atoms only) to the active state (PDB 7LCJ) until the former distance reached 4 Å. In the second phase of mwSuMD, the RMSD of the ECD (residues W33^ECD-W120^ECD, Cα atoms only) and the ECL1 to the active state (PDB 7LCJ, Cα atoms of residues M204^2.74-L224^3.27) were supervised until the latter reached less than 4 Å. During the third phase, the RMSDs of PF06882961, as well as the RMSD of ECL3 (residues A368^6.57-T378^7.33, Cα atoms), were supervised until the former reached values lower than 3 Å. In the last mwSuMD step, only the RMSD of TM6 (residues I345^6.34-F367^6.56, Cα atoms) to the active state (PDB 7LCJ) was supervised until less than 5 Å. RMSDs were computed after superimposition on TM2, ECL1, and TM3 residues 170–240 (Cα atoms), which is the GLP-1R less flexible part (Deganutti et al., 2022).

Membrane-anchored G_s protein:GLP-1R simulations and GDP dissociation

The PDB 6EG8 was processed through Charmm-GUI (Jo et al., 2008) to palmitoylate residue C3^Gαi and geranylgeranylate residue C65^Gγ. The resulting system was inserted into a 120×120 Å POPC membrane and previously built by using the VMD Membrane Builder plugin 1.1, Membrane Plugin, Version 1.1, at http://www.ks.uiuc.edu/Research/vmd/plugins/membrane. Lipids overlapping the palmitoyl and geranylgeranyl groups were removed and TIP3P water molecules (Jorgensen et al., 1983) were added to the simulation box by means of the VMD Solvate plugin 1.5 (Solvate Plugin, Version 1.5. at https://www.ks.uiuc.edu/Research/vmd/plugins/solvate/). Finally, overall charge neutrality was reached by adding Na⁺/Cl^- counterions up to the final concentration of 0.150 M, using the VMD Autoionize plugin 1.3 (Autoionize Plugin, Version 1.3. at https://www.ks.uiuc.edu/Research/vmd/plugins/autoionize/). The first stage of equilibration was performed as reported above (Methods section System equilibration and general MD settings) for 120 ns, followed by a second stage in the NVT ensemble for a further 1 μs without any restraints to allow the membrane-anchored heterotrimeric G_s protein to stabilize within the intracellular side of the simulation box. After this two-stage long equilibration, GLP-1R from the final frame of the activation simulation (in complex with PF06882961) was manually inserted into the equilibrated membrane above the G_s protein using the corresponding structure retrieved from the OPM database as a reference, and the system further equilibrated for 120 ns as reported above (Methods section System equilibration and general MD settings). The GLP-1R-G_s system was then subjected to three simulations (Table 1). Each mwSuMD replica was interrupted by 500 ns of classic MD twice, to relax the system during the transition. In the supervised stages, the distance between residues M386-L394 ^Gαs (all-atoms centroid) of helix 5 (α5) and the GLP-1R intracellular residues R176^2.46, R348^6.37, S352^6.41, and N405^7.60 (Cα atoms only) was monitored, seeding 3 walkers of 200 ps each.

The AHD opening was simulated starting from the GLP-1R:G_s binding mwSuMD replica with the final lowest G_s RMSD, the lowest PF06882961 binding energy and the highest GDP binding energy (Replica 2 in Figure 5e) by supervising the distance between AHD residues G70-R199^Gαs and K300-L394^Gαs (all-atoms centroids) during 3 walkers of 100 ps each. 300 ns of classic MD was performed to relax the system. Finally, the GDP unbinding was supervised as the distance between GDP (all-atoms centroid) and residues E50^Gαs, K52^Gαs, T55^Gαs, K293^Gαs, and V367^Gαs (Cα atoms only) of Gα_s; 5 walkers were used in a 50-ps-long mwSuMD simulations.

MD analysis

Interatomic distances were computed through MDAnalysis Michaud-Agrawal et al., 2011; RMSD were computed using VMD (Humphrey et al., 1996) and MDAnalysis (Michaud-Agrawal et al., 2011). Interatomic contacts and ligand-protein hydrogen bonds were detected using the GetContacts scripts tool (https://getcontacts.github.io), setting a hydrogen bond donor-acceptor distance of 3.3 Å and an angle value of 120° as geometrical cutoffs. Contacts and hydrogen bond persistency are quantified as the percentage of frames (over all the frames obtained by merging the different replicas) in which protein residues formed contacts or hydrogen bonds with the ligand.

The MMPBSA.py (Miller et al., 2012) script, from the AmberTools20 suite (The Amber Molecular Dynamics Package, at http://ambermd.org/), was used to compute MM/GBSA method or the molecular mechanics MM/PBSA approach, after transforming the CHARMM psf topology files to an Amber prmtop format using ParmEd (documentation at https://parmed.github.io/ParmEd/html/index.html).

Supplementary videos were produced employing VMD and avconv (at https://libav.org/avconv.html). Molecular graphics images were produced using the UCSF Chimera (Pettersen et al., 2004) (Version 1.14).

G_s protein:β₂ AR binding simulations

mwSuMD simulations started from a not fully active, agonist-bound conformation of the β₂ AR and the inactive G_s to resemble pre-coupling conditions. The full-length model of the adrenergic β₂ receptor (β₂ AR) in a not fully active state was downloaded from GPCRdb (https://gpcrdb.org/). The full agonist adrenaline (ALE) was inserted in the orthosteric site by superposition with the PDB ID 4LDO (fully active β₂ AR) (Ring et al., 2013). The structure of the inactive, GDP-bound G_s protein (Liu et al., 2019) was retrieved from the Protein Data Bank (Berman et al., 2000) (PDB ID 6EG8) and placed in the intracellular bulk. The resulting system (G_s >50 Å away from β₂ AR) was prepared for MD simulations and equilibrated as reported above. The PDB ID 3SN6 (fully active β₂ AR in complex with G_s [Rasmussen et al., 2011]) was used as the reference for RMSD computations. Three mwSuMD replicas (Table 1) were performed supervising at the same time the distance between the helix 5 (α5) Gα_s residues R385-L395 (Cα atoms centroid) and the β₂ AR (residues V31-P330 Cα atoms centroid) as well as the RMSD (superimposing on β₂ AR residues 70–170 Cα atoms) of β₂ AR TM6 residues C265-I278 (Cα atoms only) to the fully active state, during 100 ps time windows (5 walkers). To monitor the progression of the simulations, we computed the RMSD of the Cα atoms of the Gα (residues 11–43 and residues 205–394) and Gβ subunits (residues 3–340) to the experimental complex (Rasmussen et al., 2011; Video 3, Figure 5—figure supplement 1). The flexibility of G_sβ is backed by both MD and cryo-EM data suggesting G protein rocking motions around G_sα:receptor interactions (Dong et al., 2020; Liang et al., 2020).

Membrane-anchored G_i protein:A₁R simulations

Since the full-length structure of the inactive human G_i protein has not yet been resolved by X-ray or cryo-EM, it was modeled by superimposing the AlphaFold2 (Jumper et al., 2021) AI models of the Gα_i (P63096-F1), Gβ (Q9HAV0-F1), and Gγ (P50151-F1) subunits to the PDB file 6EG8 (a G_s heterotrimer). The resulting homotrimer (without GDP) was processed through Charmm-GUI (Jo et al., 2008) to palmitoylate residue C3^Gαi and geranylgeranylate residue C65^Gγ (Linder et al., 1993; Mystek et al., 2019). The side chains of these two lipidated residues were manually inserted into a 120×120 Å POPC membrane and previously built by using the VMD Membrane Builder plugin 1.1, Membrane Plugin, Version 1.1. at http://www.ks.uiuc.edu/Research/vmd/plugins/membrane/. Lipids overlapping the palmitoyl and geranylgeranyl groups were removed and TIP3P water molecules (Jorgensen et al., 1983) were added to the simulation box by means of the VMD Solvate plugin 1.5 (Solvate Plugin, Version 1.5. at https://www.ks.uiuc.edu/Research/vmd/plugins/solvate/). Finally, overall charge neutrality was reached by adding Na⁺/Cl^- counterions up to the final concentration of 0.150 M, using the VMD Autoionize plugin 1.3 (Autoionize Plugin, Version 1.3. at https://www.ks.uiuc.edu/Research/vmd/plugins/autoionize/). The first stage of equilibration was performed as reported above (Methods section System equilibration and general MD settings) for 120 ns, followed by a second stage in the NVT ensemble for a further 1 μs without any restraints to allow the membrane-anchored heterotrimeric G_i protein to stabilize within the intracellular side of the simulation box. After this two-stage long equilibration, the active state A₁R in complex with adenosine (PDB 6D9H) was manually inserted into the equilibrated membrane above the G_i protein using the corresponding structure retrieved from the OPM database as a reference, and the system further equilibrated for 120 ns as reported in the Methods section System equilibration and general MD settings. The A₁R-G_i system was then subjected to a 1-μs-long classic MD simulation and an mwSuMD simulation (Table 1). During the mwSuMD simulation, the RMSD (superimposing on A₁R residues 40–140 Cα atoms) of helix 5 (α5) Gα_i residues 329–354 to the PDB 6D9H was supervised, seeding 3 walkers of 100 ps each until the productive simulation time reached 50 ns (total simulation time 150 ns).

Numbering system

Throughout the manuscript, the Ballesteros-Weinstein residues numbering system for class A (Ballesteros and Weinstein, 1995) and the Wootten residues numbering system for class B GPCRs (Wootten et al., 2013) are adopted.

Appendix 1

Appendix 1—table 1

GLP-1R:G_s contacts during multiple walker supervised molecular dynamics (mwSuMD) G_s binding simulations.

GLP-1R residue	G_s residue	Occupancy (% MD frames)
ASP344 (6.33)	ARG385	87.7
ARG348 (6.37)	LEU394	83
GLU408 (7.63)	ARG356	64.4
LEU254 (3.57)	TYR391	58.1
ARG419 (7.83)	ASP312	55.3
LYS415 (7.70)	ASP312	47.6
TYR250 (3.53)	TYR391	46.3
LEU254 (3.57)	LEU393	43.7
ARG419 (7.83)	HIS311	43.2
LEU254 (3.57)	LEU388	42.4
VAL259 (ICL2)	ARG38	39.4
SER352 (6.41)	LEU393	37.4
LYS351 (6.40)	LEU394	36
PHE257 (3.60)	GLN384	35.2
GLU262 (4.38)	ARG38	35
ARG348 (6.37)	ARG385	30.8
ARG348 (6.37)	LEU393	30.6
LEU251 (3.54)	LEU393	29.4
ILE345 (6.34)	ARG385	23.8
LEU349 (6.38)	LEU394	23.2
HIS171 (ICL1)	ASP312	22.3
LYS342 (6.31)	ASP323	21.6
PHE260 (ICL2)	ARG38	21.5
LEU254 (3.57)	GLN384	20.6
ILE345 (6.34)	LEU394	19.9
GLU262 (4.38)	TYR391	19,7
SER261 (ICL2)	GLN35	19.7
ARG348 (6.37)	ARG389	18.8
LEU254 (3.57)	HIS387	18.6
HIS171 (ICL1)	ASN313	17.6
VAL259 (ICL2)	HIS387	16.8
SER352 (6.41)	LEU394	16.7
PHE257 (3.60)	HIS387	16.3
LYS351 (6.40)	LEU393	16.3
TYR250 (3.53)	LEU393	15.7
LEU255 (3.58)	GLN384	15.6
GLU412 (7.67)	ARG356	15
LYS415 (7.70)	ASP354	14
HIS173 (ICL1)	PHE335	13.9
VAL259 (ICL2)	GLN35	13.8
PHE257 (3.60)	LEU388	13.6
HIS171 (ICL1)	ASP333	13.5
SER258 (ICL2)	HIS387	13.1
SER261 (ICL2)	ARG38	12.9
ASN407 (7.62)	ARG356	12.5
VAL259 (ICL2)	TYR391	12.4
ARG170 (ICL1)	ASP312	12.3
LEU255 (3.58)	LEU388	11.6
LEU411 (7.66)	ARG356	11.5
ARG176 (2.46)	GLU392	11.5
GLN263 (4.39)	TRP332	11.5
ILE345 (6.34)	LEU388	11.3
HIS171 (ICL1)	ARG314	11.2
VAL259 (ICL2)	ALA39	10.7
TRP264 (4.40)	PHE335	10.5
ARG267 (4.43)	ASP333	10.5
HIS171 (ICL1)	TRP332	10.5
PHE257 (3.60)	ARG38	10.2
LEU349 (6.38)	LEU393	10
GLN263 (4.39)	ASP333	10
GLN263 (4.39)	ASN313	9.9
ARG419 (7.83)	ASP354	9.7
THR343 (6.32)	ARG385	9.1
SER258 (ICL2)	ARG38	9
LYS342 (6.31)	GLU322	8.7
SER261 (ICL2)	ASP333	8.6
LEU255 (3.58)	LEU394	8.6
PHE260 (ICL2)	TYR391	8.5
ASN406 (7.61)	GLU392	8.4
SER258 (ICL2)	HIS41	8.4
HIS171 (ICL1)	ARG337	8.4
SER258 (ICL2)	GLN384	8.1
ARG348 (6.37)	LEU388	7.9
ARG267 (4.43)	ASN313	7.9
HIS173 (ICL1)	ARG337	7.8
PHE260 (ICL2)	GLN35	7.4
ARG176 (2.46)	TYR391	7.3
ARG267 (4.43)	PHE335	7.2
ALA256 (3.59)	GLN384	7
VAL259 (ICL2)	HIS41	7
CYS174 (2.44)	ARG356	6.5
GLU408 (7.63)	GLU392	6.1
PHE257 (3.60)	TYR391	6.1
SER261 (ICL2)	LYS34	6
HIS173 (ICL1)	ARG314	6
LEU251 (3.54)	LEU388	6
HIS173 (ICL1)	ASP312	5.9
HIS173 (ICL1)	TRP332	5.9
VAL259 (ICL2)	LEU55	5.9
ASN407 (7.62)	ARG389	5.8
HIS171 (ICL1)	GLY310	5.8
TYR178 (2.48)	ASP312	5.7
SER258 (ICL2)	VAL217	5.7
SER261 (ICL2)	GLN31	5.7
LYS342 (6.31)	THR325	5.6
LYS342 (6.31)	ALA324	5.5
LEU254 (3.57)	LEU394	5.4
CYS341 (6.30)	ASP323	5.4
TRP264 (4.40)	ARG52	5.3
GLU408 (7.63)	ARG389	5.2
SER258 (ICL2)	TYR391	5.1

Data availability

MD trajectories available at https://zenodo.org/records/7944479. mwSuMD software is available at https://github.com/pipitoludovico/mwSuMD_multiEngine (copy archived at Pipitò, 2025).

The following data sets were generated

1. Deganutti G
(2023) Zenodo
Updated MD trajectories for "Hidden GPCR structural transitions addressed by multiple walker supervised molecular dynamics (mwSuMD)".

https://doi.org/10.5281/zenodo.7944479

References

(2019) The adaptive path collective variable: a versatile biasing approach to compute the average transition path and free energy of molecular transitions
Methods in Molecular Biology 01:255–290.

https://doi.org/10.1007/978-1-4939-9608-7_11
- Google Scholar
(2020) Addressing free fatty acid receptor 1 (FFAR1) activation using supervised molecular dynamics
Journal of Computer-Aided Molecular Design 34:1181–1193.

https://doi.org/10.1007/s10822-020-00338-6
- PubMed
- Google Scholar
1. Ball SG
(2007) Vasopressin and disorders of water balance: the physiology and pathophysiology of vasopressin
Annals of Clinical Biochemistry 44:417–431.

https://doi.org/10.1258/000456307781646030
- PubMed
- Google Scholar
Book
1. Ballesteros JA
2. Weinstein H
(1995) Integrated methods for the construction of three-dimensional models and computational probing of structure-function relations in G protein-coupled receptors
In: Ballesteros JA, Weinstein H, editors. In Receptor Molecular Biology; Methods in Neurosciences. Elsevier. pp. 366–428.

https://doi.org/10.1016/S1043-9471(05)80049-7
- Google Scholar
1. Batebi H
2. Pérez-Hernández G
3. Rahman SN
4. Lan B
5. Kamprad A
6. Shi M
7. Speck D
8. Tiemann JKS
9. Guixà-González R
10. Reinhardt F
11. Stadler PF
12. Papasergi-Scott MM
13. Skiniotis G
14. Scheerer P
15. Kobilka BK
16. Mathiesen JM
17. Liu X
18. Hildebrand PW
(2024) Mechanistic insights into G-protein coupling with an agonist-bound G-protein-coupled receptor
Nature Structural & Molecular Biology 31:1692–1701.

https://doi.org/10.1038/s41594-024-01334-2
- PubMed
- Google Scholar
(1984) Molecular dynamics with coupling to an external bath
The Journal of Chemical Physics 81:3684–3690.

https://doi.org/10.1063/1.448118
- Google Scholar
(1995) GROMACS: a message-passing parallel molecular dynamics implementation
Computer Physics Communications 91:43–56.

https://doi.org/10.1016/0010-4655(95)00042-E
- Google Scholar
1. Berman HM
2. Westbrook J
3. Feng Z
4. Gilliland G
5. Bhat TN
6. Weissig H
7. Shindyalov IN
8. Bourne PE
(2000) The protein data bank
Nucleic Acids Research 28:235–242.

https://doi.org/10.1093/nar/28.1.235
- PubMed
- Google Scholar
1. Birnbaumer M
(2000) Vasopressin receptors
Trends in Endocrinology and Metabolism 11:406–410.

https://doi.org/10.1016/s1043-2760(00)00304-0
- PubMed
- Google Scholar
1. Bokoch MP
2. Jo H
3. Valcourt JR
4. Srinivasan Y
5. Pan AC
6. Capponi S
7. Grabe M
8. Dror RO
9. Shaw DE
10. DeGrado WF
11. Coughlin SR
(2018) Entry from the lipid bilayer: a possible pathway for inhibition of a peptide G protein-coupled receptor by a lipophilic small molecule
Biochemistry 57:5748–5758.

https://doi.org/10.1021/acs.biochem.8b00577
- PubMed
- Google Scholar
1. Bower RL
2. Yule L
3. Rees TA
4. Deganutti G
5. Hendrikse ER
6. Harris PWR
7. Kowalczyk R
8. Ridgway Z
9. Wong AG
10. Swierkula K
11. Raleigh DP
12. Pioszak AA
13. Brimble MA
14. Reynolds CA
15. Walker CS
16. Hay DL
(2018) Molecular signature for receptor engagement in the metabolic peptide hormone amylin
ACS Pharmacology & Translational Science 1:32–49.

https://doi.org/10.1021/acsptsci.8b00002
- PubMed
- Google Scholar
1. Bussi G
2. Laio A
(2020) Using metadynamics to explore complex free-energy landscapes
Nature Reviews Physics 2:200–212.

https://doi.org/10.1038/s42254-020-0153-0
- Google Scholar
1. Cary BP
2. Deganutti G
3. Zhao P
4. Truong TT
5. Piper SJ
6. Liu X
7. Belousoff MJ
8. Danev R
9. Sexton PM
10. Wootten D
11. Gellman SH
(2022) Structural and functional diversity among agonist-bound states of the GLP-1 receptor
Nature Chemical Biology 18:256–263.

https://doi.org/10.1038/s41589-021-00945-w
- PubMed
- Google Scholar
1. Cong Z
2. Chen LN
3. Ma H
4. Zhou Q
5. Zou X
6. Ye C
7. Dai A
8. Liu Q
9. Huang W
10. Sun X
11. Wang X
12. Xu P
13. Zhao L
14. Xia T
15. Zhong W
16. Yang D
17. Eric Xu H
18. Zhang Y
19. Wang MW
(2021) Molecular insights into ago-allosteric modulation of the human glucagon-like peptide-1 receptor
Nature Communications 12:3763.

https://doi.org/10.1038/s41467-021-24058-z
- PubMed
- Google Scholar
1. Cong Z
2. Liang YL
3. Zhou Q
4. Darbalaei S
5. Zhao F
6. Feng W
7. Zhao L
8. Xu HE
9. Yang D
10. Wang MW
(2022a) Structural perspective of class B1 GPCR signaling
Trends in Pharmacological Sciences 43:321–334.

https://doi.org/10.1016/j.tips.2022.01.002
- PubMed
- Google Scholar
1. Cong Z
2. Zhou Q
3. Li Y
4. Chen LN
5. Zhang ZC
6. Liang A
7. Liu Q
8. Wu X
9. Dai A
10. Xia T
11. Wu W
12. Zhang Y
13. Yang D
14. Wang MW
(2022b) Structural basis of peptidomimetic agonism revealed by small- molecule GLP-1R agonists Boc5 and WB4-24
PNAS 119:e2200155119.

https://doi.org/10.1073/pnas.2200155119
- PubMed
- Google Scholar
(2022) Kinetic model of GPCR-G protein interactions reveals allokairic modulation of signaling
Nature Communications 13:1202.

https://doi.org/10.1038/s41467-022-28789-5
- PubMed
- Google Scholar
1. Cuzzolin A
2. Sturlese M
3. Deganutti G
4. Salmaso V
5. Sabbadin D
6. Ciancetta A
7. Moro S
(2016) Deciphering the complexity of ligand-protein recognition pathways using supervised molecular dynamics (SuMD) simulations
Journal of Chemical Information and Modeling 56:687–705.

https://doi.org/10.1021/acs.jcim.5b00702
- PubMed
- Google Scholar
(2024) Minute-timescale free-energy calculations reveal a pseudo-active state in the adenosine A_2A receptor activation mechanism
Chem 10:3678–3698.

https://doi.org/10.1016/j.chempr.2024.08.004
- PubMed
- Google Scholar
(2014) The evolution of the GPCR signaling system in eukaryotes: modularity, conservation, and the transition to metazoan multicellularity
Genome Biology and Evolution 6:606–619.

https://doi.org/10.1093/gbe/evu038
- PubMed
- Google Scholar
1. Deganutti G
2. Moro S
(2017) Estimation of kinetic and thermodynamic ligand-binding parameters using computational strategies
Future Medicinal Chemistry 9:507–523.

https://doi.org/10.4155/fmc-2016-0224
- Google Scholar
Preprint
(2020a) A multisite model of allosterism for the adenosine A1 receptor
bioRxiv.

https://doi.org/10.1101/2020.10.14.338822
- Google Scholar
(2020b) A Supervised molecular dynamics approach to unbiased ligand-protein unbinding
Journal of Chemical Information and Modeling 60:1804–1817.

https://doi.org/10.1021/acs.jcim.9b01094
- PubMed
- Google Scholar
(2021a) Exploring ligand binding to calcitonin gene-related peptide receptors
Frontiers in Molecular Biosciences 8:720561.

https://doi.org/10.3389/fmolb.2021.720561
- PubMed
- Google Scholar
1. Deganutti G
2. Barkan K
3. Preti B
4. Leuenberger M
5. Wall M
6. Frenguelli BG
7. Lochner M
8. Ladds G
9. Reynolds CA
(2021b) Deciphering the agonist binding mechanism to the adenosine A1 receptor
ACS Pharmacology & Translational Science 4:314–326.

https://doi.org/10.1021/acsptsci.0c00195
- PubMed
- Google Scholar
(2021c) Supervised molecular dynamics for exploring the druggability of the SARS-CoV-2 spike protein
Journal of Computer-Aided Molecular Design 35:195–207.

https://doi.org/10.1007/s10822-020-00356-4
- PubMed
- Google Scholar
1. Deganutti G
2. Liang YL
3. Zhang X
4. Khoshouei M
5. Clydesdale L
6. Belousoff MJ
7. Venugopal H
8. Truong TT
9. Glukhova A
10. Keller AN
11. Gregory KJ
12. Leach K
13. Christopoulos A
14. Danev R
15. Reynolds CA
16. Zhao P
17. Sexton PM
18. Wootten D
(2022) Dynamics of GLP-1R peptide agonist engagement are correlated with kinetics of G protein activation
Nature Communications 13:92.

https://doi.org/10.1038/s41467-021-27760-0
- PubMed
- Google Scholar
1. Dickson A
2. Brooks CL
(2014) WExplore: hierarchical exploration of high-dimensional spaces using the weighted ensemble algorithm
The Journal of Physical Chemistry. B 118:3532–3542.

https://doi.org/10.1021/jp411479c
- PubMed
- Google Scholar
(2004) PDB2PQR: an automated pipeline for the setup of Poisson-Boltzmann electrostatics calculations
Nucleic Acids Research 32:W665–W667.

https://doi.org/10.1093/nar/gkh381
- PubMed
- Google Scholar
1. Dong M
2. Deganutti G
3. Piper SJ
4. Liang YL
5. Khoshouei M
6. Belousoff MJ
7. Harikumar KG
8. Reynolds CA
9. Glukhova A
10. Furness SGB
11. Christopoulos A
12. Danev R
13. Wootten D
14. Sexton PM
15. Miller LJ
(2020) Structure and dynamics of the active Gs-coupled human secretin receptor
Nature Communications 11:4137.

https://doi.org/10.1038/s41467-020-17791-4
- PubMed
- Google Scholar
1. Dror RO
2. Mildorf TJ
3. Hilger D
4. Manglik A
5. Borhani DW
6. Arlow DH
7. Philippsen A
8. Villanueva N
9. Yang Z
10. Lerch MT
11. Hubbell WL
12. Kobilka BK
13. Sunahara RK
14. Shaw DE
(2015) Structural basis for nucleotide exchange in heterotrimeric G proteins
Science 348:1361–1365.

https://doi.org/10.1126/science.aaa5264
- PubMed
- Google Scholar
1. Du Y
2. Duc NM
3. Rasmussen SGF
4. Hilger D
5. Kubiak X
6. Wang L
7. Bohon J
8. Kim HR
9. Wegrecki M
10. Asuru A
11. Jeong KM
12. Lee J
13. Chance MR
14. Lodowski DT
15. Kobilka BK
16. Chung KY
(2019) Assembly of a GPCR-G protein complex
Cell 177:1232–1242.

https://doi.org/10.1016/j.cell.2019.04.022
- PubMed
- Google Scholar
1. Durrant JD
2. McCammon JA
(2011) Molecular dynamics simulations and drug discovery
BMC Biology 9:71.

https://doi.org/10.1186/1741-7007-9-71
- PubMed
- Google Scholar
1. Eastman P
2. Swails J
3. Chodera JD
4. McGibbon RT
5. Zhao Y
6. Beauchamp KA
7. Wang LP
8. Simmonett AC
9. Harrigan MP
10. Stern CD
11. Wiewiora RP
12. Brooks BR
13. Pande VS
(2017) OpenMM 7: rapid development of high performance algorithms for molecular dynamics
PLOS Computational Biology 13:e1005659.

https://doi.org/10.1371/journal.pcbi.1005659
- PubMed
- Google Scholar
1. Essmann U
2. Perera L
3. Berkowitz ML
4. Darden T
5. Lee H
6. Pedersen LG
(1995) A smooth particle mesh Ewald method
The Journal of Chemical Physics 103:8577–8593.

https://doi.org/10.1063/1.470117
- Google Scholar
1. Fiser A
2. Sali A
(2003) Modeller: generation and refinement of homology-based protein structure models
Methods in Enzymology 374:461–491.

https://doi.org/10.1016/S0076-6879(03)74020-8
- Google Scholar
(2020) Energy landscapes reveal agonist control of G protein-coupled receptor activation via microswitches
Biochemistry 59:880–891.

https://doi.org/10.1021/acs.biochem.9b00842
- PubMed
- Google Scholar
1. Flock T
2. Ravarani CNJ
3. Sun D
4. Venkatakrishnan AJ
5. Kayikci M
6. Tate CG
7. Veprintsev DB
8. Babu MM
(2015) Universal allosteric mechanism for Gα activation by GPCRs
Nature 524:173–179.

https://doi.org/10.1038/nature14663
- PubMed
- Google Scholar
1. Flock T
2. Hauser AS
3. Lund N
4. Gloriam DE
5. Balaji S
6. Babu MM
(2017) Selectivity determinants of GPCR-G-protein binding
Nature 545:317–322.

https://doi.org/10.1038/nature22070
- PubMed
- Google Scholar
1. Forester TR
2. Smith W
(1998) SHAKE, rattle, and roll: efficient constraint algorithms for linked rigid bodies
Journal of Computational Chemistry 19:102–111.

https://doi.org/10.1002/(SICI)1096-987X(19980115)19:1<102::AID-JCC9>3.0.CO;2-T
- Google Scholar
(2017) Dynamic docking: a paradigm shift in computational drug discovery
Molecules 22:2029.

https://doi.org/10.3390/molecules22112029
- PubMed
- Google Scholar
(2018) Rules of engagement: GPCRs and G proteins
ACS Pharmacology & Translational Science 1:73–83.

https://doi.org/10.1021/acsptsci.8b00026
- PubMed
- Google Scholar
Website
1. G protein-coupled receptors
(2019) BPS Guide to PHARMACOLOGY
Accessed November 11, 2023.

https://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=694
1. Gregorio GG
2. Masureel M
3. Hilger D
4. Terry DS
5. Juette M
6. Zhao H
7. Zhou Z
8. Perez-Aguilar JM
9. Hauge M
10. Mathiasen S
11. Javitch JA
12. Weinstein H
13. Kobilka BK
14. Blanchard SC
(2017) Single-molecule analysis of ligand efficacy in β₂AR-G-protein activation
Nature 547:68–73.

https://doi.org/10.1038/nature22354
- PubMed
- Google Scholar
(2015) The role of target binding kinetics in drug discovery
ChemMedChem 10:1793–1796.

https://doi.org/10.1002/cmdc.201500310
- PubMed
- Google Scholar
(2017) Kinetic aspects of the interaction between ligand and G protein-coupled receptor: the case of the adenosine receptors
Chemical Reviews 117:38–66.

https://doi.org/10.1021/acs.chemrev.6b00025
- PubMed
- Google Scholar
1. Ham D
2. Ahn D
3. Ashim J
4. Cho Y
5. Kim HR
6. Yu W
7. Chung KY
(2021) Conformational switch that induces GDP release from Gi
Journal of Structural Biology 213:107694.

https://doi.org/10.1016/j.jsb.2020.107694
- PubMed
- Google Scholar
(2004) Accelerated molecular dynamics: a promising and efficient simulation method for biomolecules
The Journal of Chemical Physics 120:11919–11929.

https://doi.org/10.1063/1.1755656
- PubMed
- Google Scholar
(2009) ACEMD: accelerating biomolecular dynamics in the microsecond time scale
Journal of Chemical Theory and Computation 5:1632–1639.

https://doi.org/10.1021/ct9000685
- PubMed
- Google Scholar
1. Hollingsworth SA
2. Dror RO
(2018) Molecular dynamics simulation for all
Neuron 99:1129–1143.

https://doi.org/10.1016/j.neuron.2018.08.011
- PubMed
- Google Scholar
1. Huang J
2. MacKerell AD
(2013) CHARMM36 all-atom additive protein force field: validation based on comparison to NMR data
Journal of Computational Chemistry 34:2135–2145.

https://doi.org/10.1002/jcc.23354
- PubMed
- Google Scholar
(2017) CHARMM36m: an improved force field for folded and intrinsically disordered proteins
Nature Methods 14:71–73.

https://doi.org/10.1038/nmeth.4067
- PubMed
- Google Scholar
(1996) VMD: visual molecular dynamics
Journal of Molecular Graphics 14:33–38.

https://doi.org/10.1016/0263-7855(96)00018-5
- PubMed
- Google Scholar
1. Husic BE
2. Pande VS
(2018) Markov state models: from an art to a science
Journal of the American Chemical Society 140:2386–2396.

https://doi.org/10.1021/jacs.7b12191
- PubMed
- Google Scholar
1. Isberg V
2. Mordalski S
3. Munk C
4. Rataj K
5. Harpsøe K
6. Hauser AS
7. Vroling B
8. Bojarski AJ
9. Vriend G
10. Gloriam DE
(2016) GPCRdb: an information system for G protein-coupled receptors
Nucleic Acids Research 44:D356–D364.

https://doi.org/10.1093/nar/gkv1178
- PubMed
- Google Scholar
1. Jo S
2. Kim T
3. Iyer VG
4. Im W
(2008) CHARMM-GUI: A web-based graphical user interface for CHARMM
Journal of Computational Chemistry 29:1859–1865.

https://doi.org/10.1002/jcc.20945
- PubMed
- Google Scholar
(1983) Comparison of simple potential functions for simulating liquid water
The Journal of Chemical Physics 79:926–935.

https://doi.org/10.1063/1.445869
- Google Scholar
1. Jumper J
2. Evans R
3. Pritzel A
4. Green T
5. Figurnov M
6. Ronneberger O
7. Tunyasuvunakool K
8. Bates R
9. Žídek A
10. Potapenko A
11. Bridgland A
12. Meyer C
13. Kohl SAA
14. Ballard AJ
15. Cowie A
16. Romera-Paredes B
17. Nikolov S
18. Jain R
19. Adler J
20. Back T
21. Petersen S
22. Reiman D
23. Clancy E
24. Zielinski M
25. Steinegger M
26. Pacholska M
27. Berghammer T
28. Bodenstein S
29. Silver D
30. Vinyals O
31. Senior AW
32. Kavukcuoglu K
33. Kohli P
34. Hassabis D
(2021) Highly accurate protein structure prediction with AlphaFold
Nature 596:583–589.

https://doi.org/10.1038/s41586-021-03819-2
- PubMed
- Google Scholar
1. Kawai T
2. Sun B
3. Yoshino H
4. Feng D
5. Suzuki Y
6. Fukazawa M
7. Nagao S
8. Wainscott DB
9. Showalter AD
10. Droz BA
11. Kobilka TS
12. Coghlan MP
13. Willard FS
14. Kawabe Y
15. Kobilka BK
16. Sloop KW
(2020) Structural basis for GLP-1 receptor activation by LY3502970, an orally active nonpeptide agonist
PNAS 117:29959–29967.

https://doi.org/10.1073/pnas.2014879117
- PubMed
- Google Scholar
1. Kaya AI
2. Lokits AD
3. Gilbert JA
4. Iverson TM
5. Meiler J
6. Hamm HE
(2016) A conserved hydrophobic core in Gαi1 regulates G protein activation and release from activated receptor
The Journal of Biological Chemistry 291:19674–19686.

https://doi.org/10.1074/jbc.M116.745513
- PubMed
- Google Scholar
1. Kokh DB
2. Kaufmann T
3. Kister B
4. Wade RC
(2019) Machine learning analysis of τRAMD trajectories to decipher molecular determinants of drug-target residence times
Frontiers in Molecular Biosciences 6:36.

https://doi.org/10.3389/fmolb.2019.00036
- PubMed
- Google Scholar
(2001) A fast SHAKE algorithm to solve distance constraint equations for small molecules in molecular dynamics simulations
Journal of Computational Chemistry 22:501–508.

https://doi.org/10.1002/1096-987X(20010415)22:5<501::AID-JCC1021>3.0.CO;2-V
- Google Scholar
1. Li Y
2. Sun J
3. Li D
4. Lin J
(2022) The full activation mechanism of the adenosine A₁ receptor revealed by GaMD and Su-GaMD simulations
PNAS 119:e2203702119.

https://doi.org/10.1073/pnas.2203702119
- PubMed
- Google Scholar
1. Liang YL
2. Belousoff MJ
3. Fletcher MM
4. Zhang X
5. Khoshouei M
6. Deganutti G
7. Koole C
8. Furness SGB
9. Miller LJ
10. Hay DL
11. Christopoulos A
12. Reynolds CA
13. Danev R
14. Wootten D
15. Sexton PM
(2020) Structure and dynamics of adrenomedullin receptors AM₁ and AM₂ reveal key mechanisms in the control of receptor phenotype by receptor activity-modifying proteins
ACS Pharmacology & Translational Science 3:263–284.

https://doi.org/10.1021/acsptsci.9b00080
- PubMed
- Google Scholar
1. Linder ME
2. Middleton P
3. Hepler JR
4. Taussig R
5. Gilman AG
6. Mumby SM
(1993) Lipid modifications of G proteins: alpha subunits are palmitoylated
PNAS 90:3675–3679.

https://doi.org/10.1073/pnas.90.8.3675
- PubMed
- Google Scholar
1. Liu X
2. Xu X
3. Hilger D
4. Aschauer P
5. Tiemann JKS
6. Du Y
7. Liu H
8. Hirata K
9. Sun X
10. Guixà-González R
11. Mathiesen JM
12. Hildebrand PW
13. Kobilka BK
(2019) Structural insights into the process of GPCR-G Protein Complex Formation
Cell 177:1243–1251.

https://doi.org/10.1016/j.cell.2019.04.021
- PubMed
- Google Scholar
(2006) OPM: orientations of proteins in membranes database
Bioinformatics 22:623–625.

https://doi.org/10.1093/bioinformatics/btk023
- PubMed
- Google Scholar
(1992) Langevin dynamics of peptides: the frictional dependence of isomerization rates of N-acetylalanyl-N’-methylamide
Biopolymers 32:523–535.

https://doi.org/10.1002/bip.360320508
- PubMed
- Google Scholar
(2019) Reconstruction of apo A2A receptor activation pathways reveal ligand-competent intermediates and state-dependent cholesterol hotspots
Scientific Reports 9:14199.

https://doi.org/10.1038/s41598-019-50752-6
- PubMed
- Google Scholar
1. Ma H
2. Huang W
3. Wang X
4. Zhao L
5. Jiang Y
6. Liu F
7. Guo W
8. Sun X
9. Zhong W
10. Yuan D
11. Xu HE
(2020) Structural insights into the activation of GLP-1R by a small molecule agonist
Cell Research 30:1140–1142.

https://doi.org/10.1038/s41422-020-0384-8
- PubMed
- Google Scholar
1. Mafi A
2. Kim SK
3. Goddard WA
(2023) The dynamics of agonist-β₂-adrenergic receptor activation induced by binding of GDP-bound Gs protein
Nature Chemistry 15:1127–1137.

https://doi.org/10.1038/s41557-023-01238-6
- PubMed
- Google Scholar
(2020) A combined activation mechanism for the glucagon receptor
PNAS 117:15414–15422.

https://doi.org/10.1073/pnas.1921851117
- PubMed
- Google Scholar
(2013) Rapid parameterization of small molecules using the force field toolkit
Journal of Computational Chemistry 34:2757–2770.

https://doi.org/10.1002/jcc.23422
- PubMed
- Google Scholar
1. McGibbon RT
2. Beauchamp KA
3. Harrigan MP
4. Klein C
5. Swails JM
6. Hernández CX
7. Schwantes CR
8. Wang LP
9. Lane TJ
10. Pande VS
(2015) MDTraj: a modern open library for the analysis of molecular dynamics trajectories
Biophysical Journal 109:1528–1532.

https://doi.org/10.1016/j.bpj.2015.08.015
- PubMed
- Google Scholar
(2015) Gaussian accelerated molecular dynamics: unconstrained enhanced sampling and free energy calculation
Journal of Chemical Theory and Computation 11:3584–3595.

https://doi.org/10.1021/acs.jctc.5b00436
- PubMed
- Google Scholar
1. Miao Y
2. McCammon JA
(2018) Mechanism of the G-protein mimetic nanobody binding to a muscarinic G-protein-coupled receptor
PNAS 115:3036–3041.

https://doi.org/10.1073/pnas.1800756115
- PubMed
- Google Scholar
(2011) MDAnalysis: a toolkit for the analysis of molecular dynamics simulations
Journal of Computational Chemistry 32:2319–2327.

https://doi.org/10.1002/jcc.21787
- PubMed
- Google Scholar
(2012) MMPBSA.py: an efficient program for end-state free energy calculations
Journal of Chemical Theory and Computation 8:3314–3321.

https://doi.org/10.1021/ct300418h
- PubMed
- Google Scholar
(2019) Gγ and Gα identity dictate a G-protein heterotrimer plasma membrane targeting
Cells 8:1246.

https://doi.org/10.3390/cells8101246
- PubMed
- Google Scholar
(2011) PROPKA3: consistent treatment of internal and surface residues in empirical pKa predictions
Journal of Chemical Theory and Computation 7:525–537.

https://doi.org/10.1021/ct100578z
- PubMed
- Google Scholar
(2018) Simulations meet machine learning in structural biology
Current Opinion in Structural Biology 49:139–144.

https://doi.org/10.1016/j.sbi.2018.02.004
- PubMed
- Google Scholar
1. Pettersen EF
2. Goddard TD
3. Huang CC
4. Couch GS
5. Greenblatt DM
6. Meng EC
7. Ferrin TE
(2004) UCSF Chimera--A visualization system for exploratory research and analysis
Journal of Computational Chemistry 25:1605–1612.

https://doi.org/10.1002/jcc.20084
- PubMed
- Google Scholar
1. Phillips JC
2. Braun R
3. Wang W
4. Gumbart J
5. Tajkhorshid E
6. Villa E
7. Chipot C
8. Skeel RD
9. Kalé L
10. Schulten K
(2005) Scalable molecular dynamics with NAMD
Journal of Computational Chemistry 26:1781–1802.

https://doi.org/10.1002/jcc.20289
- PubMed
- Google Scholar
Software
1. Pipitò L
(2025) mwSuMD - Molecular Dynamics Supervision Tool, version swh:1:rev:b5cd2dacea7b003df96ee88a34919d87b937b744
Software Heritage.

https://archive.softwareheritage.org/swh:1:dir:3a780e7322852c01cf6ca71a61a2b161b5ccda6c;origin=https://github.com/pipitoludovico/mwSuMD_multiEngine;visit=swh:1:snp:17cec8ea7b03cfd8e5c4cfddc542c49b3796d350;anchor=swh:1:rev:b5cd2dacea7b003df96ee88a34919d87b937b744
1. Rasmussen SGF
2. DeVree BT
3. Zou Y
4. Kruse AC
5. Chung KY
6. Kobilka TS
7. Thian FS
8. Chae PS
9. Pardon E
10. Calinski D
11. Mathiesen JM
12. Shah STA
13. Lyons JA
14. Caffrey M
15. Gellman SH
16. Steyaert J
17. Skiniotis G
18. Weis WI
19. Sunahara RK
20. Kobilka BK
(2011) Crystal structure of the β2 adrenergic receptor-Gs protein complex
Nature 477:549–555.

https://doi.org/10.1038/nature10361
- PubMed
- Google Scholar
1. Ring AM
2. Manglik A
3. Kruse AC
4. Enos MD
5. Weis WI
6. Garcia KC
7. Kobilka BK
(2013) Adrenaline-activated structure of β2-adrenoceptor stabilized by an engineered nanobody
Nature 502:575–579.

https://doi.org/10.1038/nature12572
- PubMed
- Google Scholar
1. Sabbadin D
2. Moro S
(2014) Supervised molecular dynamics (SuMD) as a helpful tool to depict GPCR-ligand recognition pathway in a nanosecond time scale
Journal of Chemical Information and Modeling 54:372–376.

https://doi.org/10.1021/ci400766b
- PubMed
- Google Scholar
(2017) Exploring protein-peptide recognition pathways using a supervised molecular dynamics approach
Structure 25:655–662.

https://doi.org/10.1016/j.str.2017.02.009
- Google Scholar
1. Schiöth HB
2. Fredriksson R
(2005) The GRAFS classification system of G-protein coupled receptors in comparative perspective
General and Comparative Endocrinology 142:94–101.

https://doi.org/10.1016/j.ygcen.2004.12.018
- PubMed
- Google Scholar
1. Sommer B
(2013) Membrane packing problems: a short review on computational membrane modeling methods and tools
Computational and Structural Biotechnology Journal 5:e201302014.

https://doi.org/10.5936/csbj.201302014
- PubMed
- Google Scholar
1. Sriram K
2. Insel PA
(2018) G protein-coupled receptors as targets for approved drugs: how many targets and how many drugs?
Molecular Pharmacology 93:251–258.

https://doi.org/10.1124/mol.117.111062
- PubMed
- Google Scholar
(2016) The pathway of ligand entry from the membrane bilayer to a lipid G protein-coupled receptor
Scientific Reports 6:22639.

https://doi.org/10.1038/srep22639
- PubMed
- Google Scholar
1. Sugita Y
2. Okamoto Y
(1999) Replica-exchange molecular dynamics method for protein folding
Chemical Physics Letters 314:141–151.

https://doi.org/10.1016/S0009-2614(99)01123-9
- Google Scholar
1. Sun X
2. Singh S
3. Blumer KJ
4. Bowman GR
(2018) Simulation of spontaneous G protein activation reveals a new intermediate driving GDP unbinding
eLife 7:e38465.

https://doi.org/10.7554/eLife.38465
- PubMed
- Google Scholar
(2016) Regulation, signaling, and physiological functions of G-proteins
Journal of Molecular Biology 428:3850–3868.

https://doi.org/10.1016/j.jmb.2016.08.002
- PubMed
- Google Scholar
1. Tan L
2. Yan W
3. McCorvy JD
4. Cheng J
(2018) Biased ligands of G protein-coupled receptors (GPCRs): structure–functional selectivity relationships (SFSRs) and therapeutic potential
Journal of Medicinal Chemistry 61:9841–9878.

https://doi.org/10.1021/acs.jmedchem.8b00435
- Google Scholar
1. Vanommeslaeghe K
2. MacKerell AD
(2012) Automation of the CHARMM general force field (CGenFF) I: bond perception and atom typing
Journal of Chemical Information and Modeling 52:3144–3154.

https://doi.org/10.1021/ci300363c
- PubMed
- Google Scholar
(2012) Automation of the CHARMM general force field (CGenFF) II: assignment of bonded parameters and partial atomic charges
Journal of Chemical Information and Modeling 52:3155–3168.

https://doi.org/10.1021/ci3003649
- PubMed
- Google Scholar
1. Wall MJ
2. Hill E
3. Huckstepp R
4. Barkan K
5. Deganutti G
6. Leuenberger M
7. Preti B
8. Winfield I
9. Carvalho S
10. Suchankova A
11. Wei H
12. Safitri D
13. Huang X
14. Imlach W
15. La Mache C
16. Dean E
17. Hume C
18. Hayward S
19. Oliver J
20. Zhao F-Y
21. Spanswick D
22. Reynolds CA
23. Lochner M
24. Ladds G
25. Frenguelli BG
(2022) Selective activation of Gαob by an adenosine A₁ receptor agonist elicits analgesia without cardiorespiratory depression
Nature Communications 13:4150.

https://doi.org/10.1038/s41467-022-31652-2
- PubMed
- Google Scholar
1. Wang E
2. Sun H
3. Wang J
4. Wang Z
5. Liu H
6. Zhang JZH
7. Hou T
(2019) End-point binding free energy calculation with MM/PBSA and MM/GBSA: strategies and applications in drug design
Chemical Reviews 119:9478–9508.

https://doi.org/10.1021/acs.chemrev.9b00055
- PubMed
- Google Scholar
1. Woods RJ
2. Chappelle R
(2000) Restrained electrostatic potential atomic partial charges for condensed-phase simulations of carbohydrates
Theochem 527:149–156.

https://doi.org/10.1016/S0166-1280(00)00487-5
- PubMed
- Google Scholar
(2013) Polar transmembrane interactions drive formation of ligand-specific and signal pathway-biased family B G protein-coupled receptor conformations
PNAS 110:5211–5216.

https://doi.org/10.1073/pnas.1221585110
- PubMed
- Google Scholar
1. Wootten D
2. Reynolds CA
3. Koole C
4. Smith KJ
5. Mobarec JC
6. Simms J
7. Quon T
8. Coudrat T
9. Furness SGB
10. Miller LJ
11. Christopoulos A
12. Sexton PM
(2016) A hydrogen-bonded polar network in the core of the glucagon-like peptide-1 receptor is a fulcrum for biased agonism: lessons from class B crystal structures
Molecular Pharmacology 89:335–347.

https://doi.org/10.1124/mol.115.101246
- PubMed
- Google Scholar
1. Wu F
2. Yang L
3. Hang K
4. Laursen M
5. Wu L
6. Han GW
7. Ren Q
8. Roed NK
9. Lin G
10. Hanson MA
11. Jiang H
12. Wang MW
13. Reedtz-Runge S
14. Song G
15. Stevens RC
(2020) Full-length human GLP-1 receptor structure without orthosteric ligands
Nature Communications 11:1272.

https://doi.org/10.1038/s41467-020-14934-5
- PubMed
- Google Scholar
(2012) Extension of the CHARMM general force field to sulfonyl-containing compounds and its utility in biomolecular simulations
Journal of Computational Chemistry 33:2451–2468.

https://doi.org/10.1002/jcc.23067
- PubMed
- Google Scholar
1. Yuan S
2. Xia L
3. Wang C
4. Wu F
5. Zhang B
6. Pan C
7. Fan Z
8. Lei X
9. Stevens RC
10. Sali A
11. Sun L
12. Shui W
(2023) Conformational dynamics of the activated GLP-1 receptor-G_s complex revealed by cross-linking mass spectrometry and integrative structure modeling
ACS Central Science 9:992–1007.

https://doi.org/10.1021/acscentsci.3c00063
- PubMed
- Google Scholar
1. Zhang Z
2. Melia TJ
3. He F
4. Yuan C
5. McGough A
6. Schmid MF
7. Wensel TG
(2004) How a G protein binds a membrane
Journal of Biological Chemistry 279:33937–33945.

https://doi.org/10.1074/jbc.M403404200
- Google Scholar
1. Zhang X
2. Belousoff MJ
3. Zhao P
4. Kooistra AJ
5. Truong TT
6. Ang SY
7. Underwood CR
8. Egebjerg T
9. Šenel P
10. Stewart GD
11. Liang YL
12. Glukhova A
13. Venugopal H
14. Christopoulos A
15. Furness SGB
16. Miller LJ
17. Reedtz-Runge S
18. Langmead CJ
19. Gloriam DE
20. Danev R
21. Sexton PM
22. Wootten D
(2020) Differential GLP-1R binding and activation by peptide and non-peptide agonists
Molecular Cell 80:485–500.

https://doi.org/10.1016/j.molcel.2020.09.020
- PubMed
- Google Scholar
1. Zhang X
2. Johnson RM
3. Drulyte I
4. Yu L
5. Kotecha A
6. Danev R
7. Wootten D
8. Sexton PM
9. Belousoff MJ
(2021) Evolving cryo-EM structural approaches for GPCR drug discovery
Structure 29:963–974.

https://doi.org/10.1016/j.str.2021.04.008
- PubMed
- Google Scholar
1. Zhao P
2. Liang YL
3. Belousoff MJ
4. Deganutti G
5. Fletcher MM
6. Willard FS
7. Bell MG
8. Christe ME
9. Sloop KW
10. Inoue A
11. Truong TT
12. Clydesdale L
13. Furness SGB
14. Christopoulos A
15. Wang MW
16. Miller LJ
17. Reynolds CA
18. Danev R
19. Sexton PM
20. Wootten D
(2020) Activation of the GLP-1 receptor by a non-peptidic agonist
Nature 577:432–436.

https://doi.org/10.1038/s41586-019-1902-z
- PubMed
- Google Scholar
1. Zhou F
2. Ye C
3. Ma X
4. Yin W
5. Croll TI
6. Zhou Q
7. He X
8. Zhang X
9. Yang D
10. Wang P
11. Xu HE
12. Wang MW
13. Jiang Y
(2021) Molecular basis of ligand recognition and activation of human V2 vasopressin receptor
Cell Research 31:929–931.

https://doi.org/10.1038/s41422-021-00480-2
- PubMed
- Google Scholar
1. Zuckerman DM
2. Chong LT
(2017) Weighted ensemble simulation: review of methodology, applications, and software
Annual Review of Biophysics 46:43–57.

https://doi.org/10.1146/annurev-biophys-070816-033834
- PubMed
- Google Scholar

Article and author information

Author details

Giuseppe Deganutti

Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom

Contribution
Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

For correspondence
ad5288@coventry.ac.uk

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-8780-2986
Ludovico Pipito

Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom

Contribution
Software, Investigation, Methodology, Writing – review and editing

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-1824-7541
Roxana Maria Rujan

Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom

Contribution
Investigation

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-4335-3338
Tal Weizmann

Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom

Contribution
Investigation

Competing interests
No competing interests declared
Peter Griffin

Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom

Contribution
Investigation

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0003-0966-800X
Antonella Ciancetta

Dipartimento di Scienze Chimiche, Farmaceutiche ed Agrarie, University of Ferrara, Ferrara, Italy

Contribution
Methodology, Writing – review and editing

Competing interests
No competing interests declared
Stefano Moro

Molecular Modeling Section (MMS), Dipartimento di Scienze del Farmaco, University of Padua via Marzolo, Padova, Italy

Contribution
Methodology, Writing – review and editing

Competing interests
No competing interests declared
Christopher Arthur Reynolds
1. Centre for Health and Life Sciences, Coventry University, Coventry, United Kingdom
2. School of Life Sciences, University of Essex, Wivenhoe Park, Colchester, United Kingdom
Contribution
Conceptualization, Resources, Methodology, Writing – review and editing

For correspondence
ad5291@coventry.ac.uk

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-9267-5141

Funding

Biotechnology and Biological Sciences Research Council (BB/W016974/1)

Christopher Arthur Reynolds

Diabetes UK (BDA 20/0006307)

Giuseppe Deganutti

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

Acknowledgements

GD is a member of the GPCRs-focused European COST action ERNEST. CAR is grateful for a Royal Society Industry Fellowship. GD and CAR are grateful for support from the BBSRC (BB/W016974/1) and Diabetes UK (BDA 20/0006307).

Version history

Preprint posted: January 25, 2024
Sent for peer review: February 21, 2024
Reviewed Preprint version 1: May 14, 2024
Reviewed Preprint version 2: December 17, 2024
Reviewed Preprint version 3: April 14, 2025
Version of Record published: April 30, 2025

Cite all versions

You can cite all versions using the DOI https://doi.org/10.7554/eLife.96513. This DOI represents all versions, and will always resolve to the latest one.

Copyright

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

735

views
41

downloads
2

citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

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)

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

Mendeley

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

Giuseppe Deganutti
Ludovico Pipito
Roxana Maria Rujan
Tal Weizmann
Peter Griffin
Antonella Ciancetta
Stefano Moro
Christopher Arthur Reynolds

(2025)

Hidden GPCR structural transitions addressed by multiple walker supervised molecular dynamics (mwSuMD)

eLife 13:RP96513.

https://doi.org/10.7554/eLife.96513.4

Categories and tags

Research organism

Human

Share this article

Cite this article

Arginine vasopressin (AVP) supervised molecular dynamics (SuMD) and multiple walker SuMD (mwSuMD) binding simulations to vasopressin 2 receptor (V2R) (100 ps time windows).

Summary of all the simulations performed and the settings employed.

Multiple walker supervised molecular dynamics (mwSuMD) (un)binding simulations of arginine vasopressin (AVP).

Arginine vasopressin (AVP) binding (dynamic docking) to vasopressin 2 receptor (V2R).

Arginine vasopressin (AVP) unbinding from vasopressin 2 receptor (V2R).

Multiple walker supervised molecular dynamics (mwSuMD) simulation of PF06882961 binding to glucagon-like peptide-1 receptor (GLP-1R) and receptor activation.

Morph of the PF06882961 binding and glucagon-like peptide-1 receptor (GLP-1R) activation.

PF06882961 binding and glucagon-like peptide-1 receptor (GLP-1R) activation.

Glucagon-like peptide-1 receptor (GLP-1R) key structural motifs during multiple walker supervised molecular dynamics (mwSuMD) GLP-1R activation.

Gs binding to β2 adrenoreceptor (β2 AR).

Gi binding to A1 receptor (A1R).

Glucagon-like peptide-1 receptor (GLP-1R) activation and Gs binding.

Gs binding to glucagon-like peptide-1 receptor (GLP-1R) and guanosine diphosphate (GDP) release.

Preparation to GLP-1R mwSuMD simulations.

GLP-1R:Gs contacts during multiple walker supervised molecular dynamics (mwSuMD) Gs binding simulations.

Author details

Giuseppe Deganutti

Contribution

For correspondence

Competing interests

Ludovico Pipito

Contribution

Competing interests

Roxana Maria Rujan

Contribution

Competing interests

Tal Weizmann

Contribution

Competing interests

Peter Griffin

Contribution

Competing interests

Antonella Ciancetta

Contribution

Competing interests

Stefano Moro

Contribution

Competing interests

Christopher Arthur Reynolds

Contribution

For correspondence

Competing interests

Downloads (link to download the article as PDF)

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

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

Categories and tags

Research organism

Further reading

Arginine vasopressin (AVP) supervised molecular dynamics (SuMD) and multiple walker SuMD (mwSuMD) binding simulations to vasopressin 2 receptor (V₂R) (100 ps time windows).

Arginine vasopressin (AVP) binding (dynamic docking) to vasopressin 2 receptor (V₂R).

Arginine vasopressin (AVP) unbinding from vasopressin 2 receptor (V₂R).

G_s binding to β₂ adrenoreceptor (β₂ AR).

G_i binding to A1 receptor (A₁R).

Glucagon-like peptide-1 receptor (GLP-1R) activation and G_s binding.

G_s binding to glucagon-like peptide-1 receptor (GLP-1R) and guanosine diphosphate (GDP) release.

GLP-1R:G_s contacts during multiple walker supervised molecular dynamics (mwSuMD) G_s binding simulations.