Extensive unbiased MD simulations uncover a stepwise inter-subunit rearrangement and milliseconds timescale during mGlu5 activationTwo full-length crystal structures of mGlu5 were available when we performed this study, with one in the apo state (the “inactive structure,” PDB ID: 6N52) and the other in the active conformation bound to a nanobody and a potent agonist (the “active structure,” PDB ID: 6N51)12. To efficiently disclose the conformational landscape of the mGlu5 inter-subunit activation pathway, we first generated a converge minimum energy path (MEP) that bridges the inactive and active structures by inserting numerous “replicas” interconnected through NEB (Supplementary Note 1 and Supplementary Fig. 1). The nanobody in the active structure was removed, and the two L-quisqualates bound to the VFTs were replaced by their natural agonists, L-glutamates. The glutamates were docked into the inactive structure to imitate the agonist-induced mGlu5 activation in vivo. After a simulated annealing procedure, 13 initial conformations spread over the MEP were selected and subsequently embedded into a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer membrane, explicit counterions, and water to perform massively parallel MD simulations in an unbiased manner. Each system underwent 1 µs × 10 independent runs with random initial velocities, leading to a cumulative simulation timescale of 130 µs. A comparable simulation protocol has been successfully employed to investigate kinase and GPCR activation mechanisms33,34.Based on the MD trajectories, we applied time–structure-based independent component analysis (tICA). This enabled the depiction of the conformational landscape in an objective manner, capturing essential dynamic processes and enhancing its suitability for constructing an MSM35,36. A shallow connected free-energy landscape (FEL) was obtained, which exhibited five distinct free-energy basins, each potentially representing a metastable state (Fig. 2A). The convergence of the FEL was proven with varied subsets of the MD samplings, confirming that the sampling has been sufficient to explore the mGlu5 activation (Supplementary Note 2 and Supplementary Fig. 2). To further identify the key intermediate states and determine the activation order and timescale, we constructed an MSM for the mGlu5 system. After identifying the optimal hyperparameters and confirming the convergence of the MSM through the implied timescale test, we applied the robust Perron Cluster Cluster Analysis (PCCA+) algorithm to cluster the conformational ensemble into five MSM metastable states. These states successfully passed the Chapman-Kolmogorov test, indicating that the MSM accurately reproduced the dynamics of the simulation (Supplementary Note 3 and Supplementary Fig. 3). These five-state MSM corresponded well with the observed five-state distribution that was denoted as S1–S5 from left to right in Fig. 2A (we will consistently refer to these states throughout the research). Subsequently, transition path theory (TPT) was utilized to unveil the inter-subunit transition kinetics among the different metastable states. The MSM suggested that the dominant transition pathway followed the sequential order of S1 → S2 → S3 → S4 → S5, with a total timescale of ~1 ms (Fig. 2B).Fig. 2: A stepwise activation landscape of mGlu5 dimer.A Conformational landscape of mGlu5 derived from the application of the first two components resolved by tICA. Each metastable state is labeled within its corresponding energy basin. The free energy scales of the landscape are shown on the right and are expressed in kcal/mol. B The transition time among the five metastable states. C Surface representations of the representative structures extracted from the five metastable states. From left to right: the first to fifth structures represent S1 to S5 states. The second to fifth transparent structures represent S1 to S4 states after superimposition. Large yellow arrows in the second and third structures highlight the significant movement of protomers A and B, respectively. Smaller arrows in the fourth and fifth structures indicate slight movement of the two protomers. D Two CVs that represent the distances between the VFT domain (calculated by LB2-LB2 COM distance) and 7TM domains (calculated by TM6–TM6 COM distance) are marked on mGlu5. E Coarse-grained FEL generated using the CVs in (D). Arrows and direct labels highlight different metastable states within the landscape. The blue pentastars at the top and bottom represent the inactive and active cryo-EM structures, respectively.For a better appreciation of the pivotal mGlu5 conformational states along the transition pathway, we extracted the representative structures of the five metastable states and performed a superimposition analysis (Fig. 2C). A stepwise and asymmetrical structural transformation underlying the activation process was observed. In the first 80 µs period from the S1 to S3, each protomer underwent a sequential conformational arrangement. The overall architecture of the protomer B initially moved closer to the protomer A (S2 Cα RMSD of 18.1 Å away from S1), whereas the protomer A exhibited relatively minor conformational fluctuations (S2 Cα RMSD of 7.9 Å away from S1). As such, the protomer A (S3 Cα RMSD of 15.2 Å away from S2) approached protomer B (S3 Cα RMSD of 7.9 Å away from S2) during the transition to the S3 state. Once the two protomers were in relative proximity, conformational fluctuations appeared to be attenuated (S4 Cα RMSD of 7.5 Å in protomer A and 6.0 Å in protomer B away from S3). Furthermore, the contacts and relative orientations between the 7TM domains were likely refined in the S4 state. This step was potentially the rate-limiting stage, and it proceeded at ~0.79 ms. Ultimately, further delicate arrangements of inter-subunit interactions occurred during the conformational transition to the S5. Cumulatively, the complete inter-subunit rearrangement between the mGlu5 subunits occurred within ~1 ms, which is consistent with the recently experimentally validated inter-subunit rearrangement timescale of 1–2 ms for the mGlu1 homodimer. Given that mGlu1 shares roughly 70% sequence identity with mGlu5 and the two receptors represent the evolutionary counterparts among the class I mGlu receptors37, this consistency strongly confirmed the precision and reliability of our computational methodology in capturing the dynamics of mGlu5 activation.As illustrated in Fig. 2C, the major domain-specific motion observed during activation was compaction between the VFT and 7TM domains. To further substantiate the conformational dynamics of mGlu5, a coarse-grained FEL was constructed by mapping conformational ensembles onto the two collective variables (CVs) that outlined the approaching motion of the VFT and 7TM domains. One variable (dLB2-LB2) was defined as the distance between the center of mass (COM) of the lower lobe 2 (LB2) of each VFT. The other variable (dTM6-TM6) was determined using the distance between the COM points of TM6 (Fig. 2D). The five identified metastable states, together with the two experimentally resolved inactive and active structures, were denoted in the coarse-grained FEL. Indeed, these five metastable states were situated within distinct energy minima in the profiles (Fig. 2E). The S1 and S5, which represent the starting and ending points of the activation process, respectively, as revealed by the tICA and MSM analyses, coincided with the energy minima corresponding to the inactive and active cryo-EM structures. The distance between the VFT and 7TM domains decreased significantly (21.9 Å of VFTs and 20.0 Å of 7TMs) throughout the transition from the S1 to S3. This is consistent with the prominent sequential movement shown in Fig. 2C. During the transition from the S3 to S5, the distances were modified slightly (3.3 Å of VFTs and 5.7 Å of 7TMs), aligning with the refinement of inter-contacts and accommodating minor conformational reorientations. Previous studies have revealed that in the monomeric β2-adrenergic receptor, the ligand-binding site exhibits loose coupling with the connector site, which, in turn, is loosely coupled to the G protein-binding site14,38. Similarly, the divergent coarse-grained FEL indicated that the VFT domains (ligand-binding site) and the 7TM domains (G protein-binding site) were loosely connected. An in-depth examination revealed a loose coupling among the three functionally significant regions of mGlu5—VFT domains, CRDs (connector site), and 7TM domains—appear to be loosely coupled to one another (Supplementary Note 4 and Supplementary Fig. 4). These observations indicated that such loose coupling might serve as a general mechanism for allosteric regulation in both monomeric and dimeric GPCRs.Collectively, the tICA-based and coarse-grained FEL efficiently elucidated that mGlu5 underwent a stepwise dimer proximity in ~1 ms timescale during the activation pathway in response to glutamate loading.Activation leads to stepwise compaction of the mGlu5 VFTsAs the agonist-binding site is located between the upper (LB1) and lower (LB2) lobes, the VFT domain is regarded as the starting point of agonist-induced mGlu5 activation. The available crystal structures of VFTs reveal profound conformational changes upon the binding of glutamates or alternative orthosteric agonists20,39,40. In terms of inter-domain configuration, the dual VFTs might reorient from a spatially distant (denoted as ‘relaxed’ or ‘R’) to an adjacent (‘active’ or ‘A’) conformation. Regarding the intra-domain configuration, each VFT exhibits an open conformation (designated as ‘o’) and is prone to assume a closed conformation (‘c’) in the absence or presence of orthosteric agonist binding, respectively. Such transition is imperative for downstream effector coupling and signaling.To investigate the conformational dynamics of the VFT domains, in terms of the inter-domain configuration, dLB2-LB2 was taken as the reaction coordinate. As it decreases, the VFTs transit from the ‘R’ to ‘A’ conformation. In terms of the intra-domain configuration, we calculated the summation of Cα RMSD of two VFTs from the crystal VFT conformation in complex with glutamate (PDB ID: 3LMK). This CV (denoted as RMSDVFT) quantifies the deviation of the VFTs from the ‘c’ conformation. The resulting two-dimensional (2D) FEL revealed a shift in the occupation of the main states following glutamates attachment, with the active state assuming dominance (Fig. 3A). The S3–S5 states were situated approximately in the same regions as the Acc VFT state defined in 6N51 with dLB2-LB2 ranging from 37.2–40.5 Å and RMSDVFT from 2.5–3.4 Å. These indicated that the VFT domains indeed underwent significant conformational changes upon glutamates binding. Moreover, in the S3 state, the carboxyl group of glutamates initiated polar interactions with the LB2 of both protomers, inducing a complete closure of the VFTs, with a RMSDVFTA value of 1.1 Å and RMSDVFTB value of 1.3 Å. In contrast, the VFTs in the S1 state were located in the same free energy minima as the Roo VFT state defined in 6N52 (dLB2-LB2:60.8 to 63.2 Å; RMSDVFT:7.3 to 9.0 Å), also exhibiting a Roo conformation. In the S1 state, LB1 primarily anchored to the glutamates, whereas the LB2 interface was not engaged in glutamates binding (Supplementary Note 5 and Supplementary Fig. 5).Fig. 3: Conformational diversity of the VFT domains upon glutamate binding.A Activation FEL of VFTs onto the COM of LB2s and Cα RMSD from the conformation of the mGlu5 VFTs crystal structure in complex with glutamate (PDB ID:3LMK). The free energy scales of the landscape are shown on the right and are expressed in kcal/mol. B Per-protomer analysis of the S2 metastable state uncovered an alternative configuration of the VFTs, termed S2’. C Elevated view of the helix B-C interface from S1 to S3 states. The crucial hydrophobic residues are presented in stick mode. D Front view of the helix B-C interface from S1 to S3 states. Both helix B and helix C are highlighted. The helix C cross angle, reflecting the releasement of the hydrophobic constraint of the helix B-C interface, is also labeled. The values are 155.4° and 93.1° in the resting and active cryo-EM structures, respectively. E A top view of the helix B-C interface from S1 to S3 states. The helix B is depicted as a cartoon, with key residues shown in stick mode. The crucial polar interactions are represented by yellow dashed lines. F Schematic illustration of the activation pathway of VFTs in terms of intra-protomer (open ‘o’ to close ‘c’) and inter-protomer (relaxed ‘R’ to activate ‘A’) conformational transitions. Upon agonist binding, the co-existence of transient Rco (S2) and Rcc (S2’) intermediate conformations was identified between the Roo and Acc conformations.Intriguingly, S2 represented an intermediate state along the activation pathway between the Roo and Acc conformations of the VFT domains with dLB2-LB2 and CVRMSD values ranging from 50.0–57.4 Å and 3.7–6.5 Å, respectively. Structural analysis uncovered that the VFTs in the representative S2 conformation adopted a ‘co’ conformation, wherein the glutamate in the protomer B assumed an unexplored vertical conformation by forming hydrogen bonds with R61 from LB1 and R284, R310, and D312 from LB2; however, no interactions were established with LB2 in the protomer A (Supplementary Note 4 and Supplementary Fig. 4). As a result, the VFT of protomer B exhibited a partial closure (RMSDVFTB 1.7 Å), whereas that of protomer A remained open (RMSDVFTA 3.2 Å). Considering the relatively minor change in dLB2-LB2 between the intermediate and inactive R states (~8.3 Å) compared with that of the active A state (~14.9 Å), this representative S2 conformation identified by the simulations represents a transient Rco configuration of the VFTs. Such a short-lived Rco state was also observed via single-molecule fluorescence resonance energy transfer (smFRET) experiments on the VFTs of the mGlu2 homodimer and the newly available mGlu2-mGlu3 heterodimer23,24,41. An in-depth analysis of the S2 metastable state identified an alternative configuration of the VFTs (denoted as S2’), in which the VFT of the protomer A transited from an open (S2) to a near closed state (RMSDVFTA = 1.8 Å), closely approaching a Rcc conformation (Fig. 3B). During the peer-review process, two cryo-EM structures of the mGlu5 in the intermediate-Rcc state with quisqualate bound were determined, providing experimental validation to our mode42,43. Moreover, our simulations indicated that the conformational transitions between the Roo and Acc states occurred on a timescale of ~90 µs (Fig. 2B). This timescale is consistent with ~100 µs timescale of the VFT reorientation kinetics observed in all mGlu subtypes, as demonstrated through smFRET studies44.The inter-domain conformational arrangements of VFTs in the mGlu receptors are regulated by a conserved hydrophobic interface between the B and C helices at the apical surface of the VFT, which plays a crucial role in agonist-induced structural changes. A study on mGlu2 demonstrated that introducing mutations disrupting this interface increased glutamate affinity and the proportion of the active conformation, even in the absence of glutatmate45. The cryo-EM structures of mGlu5 also revealed reorganized interaction networks at the interface between the apo and agonist-bound active conformations12,43. Our simulations further supported these findings. In the S1 state, there are tight non-polar interactions between helices B and C (shown in Fig. 3C). These interactions at the interface were gradually attenuated in the following S2, S2’ and S3 states, resulting in an increasingly exposed interface. Specifically, in the S2 and S2’ state, F165 of the protomer A dissociated from the hydrophobic patch (Fig. 3C), whereas, in the S3, F165 of the other protomer flipped to the opposite side of the interface. The results suggested that agonist binding could release the hydrophobic constraint between the helix B-C interface, as demonstrated by the transitions from the S1 (149.4°) to S2 (122.1°), S2’ (124.9°) and then to S3 (98.2°) states, leading to a more compact conformation of the LB2s (Fig. 3D). Aside from non-polar contacts, the disruption of inter-domain and the formation of intra-domain polar interactions at the apical surface of the B helices could further stabilize the active VFT configuration. In the S1 state, inter-domain interactions were maintained between E111 and R114. However, in the following S2 and S3 states, R114 could rotate to form the intra-domain hydrogen bonds with D115 (Fig. 3E).To summarize the intra- and inter-domain conformational dynamics of the VFTs, our simulations revealed that the VFTs in mGlu5 underwent a series of conformational changes, transitioning between a resting (Roo) state, a transient intermediate (Rco or Rcc) state, and an active (Acc) state on a sub-millisecond timescale (Fig. 3F). These dynamics are driven by the reorganized polar and non-polar interaction networks at the interface between the B and C helices.Conformational dynamics of CRDs reveal a flexible apical moduleThe CRD serves as a crucial link between the VFT and 7TM domains in class C GPCRs. It is a semi-rigid structure comprising three antiparallel β-sheets and is highly conserved among all class C GPCRs except for GABAB receptors. The CRD contains nine cysteine residues. One of these forms an inter-domain disulfide bridge with a cysteine residue in the VFT domain, whereas the others participate in the formation of four intra-domain disulfide bridges. The architecture of the CRD can be subdivided into three submodules, informed by the pattern of cysteine residues as identified in the TNF receptor46. The N-terminal apical module (C511 to C534) in the CRD is distinct and lacks resemblance to any established module types. Nevertheless, its structural features are reminiscent of the B module, leading to its denotation as B2’ 39. On the other hand, the middle module (C537–C549) and the C-terminal module (C552–C565) are both categorized within the A1 module (Fig. 4A).Fig. 4: CRD association dynamics along mGlu5 activation.A Architecture of three CRD modules. The B2’ module is colored in orange and the two A1 modules are rendered in purple. Four intradomain disulfide bridges are emphasized using stick models. a1–a5 represent five representative intra-domain CRD conformations during simulations; their relative occupations are illustrated in a pie chart above and the specific numbers are presented at the bottom. The two arrows indicate two distinct mechanisms that influence the orientations of the CRD. B CVs denote the distances between the near-VFT part of the CRD (B2’-B2’ COM distance) and the near-7TM part (D560-D560 Cα distance) labeled on mGlu5. A free energy profile of CRD inter-domain dynamics is then calculated using the CVs. The free energy scales of the landscape are shown on the right and are expressed in kcal/mol. C Quantification of cell surface expression. The bars indicate the mean ± SEM values from six independent experiments. ns, not statistically significant (one-way ANOVA with Dunnett’s post hoc test). D Crosslinking at the potential surface of the B2’ module by an I523C mutant displays high constitutive activity, as measured using IP accumulation. Alanine mutation of critical residues involved in forming the observed B2’-mediated interface blunt glutamate-induced signaling compared to the WT. Data were from three independent experiments. The bars indicate the mean ± SEM values.Analysis of the intra-domain mobility of CRD and conformational changes revealed distinct dynamics in different modules. The B2’ module in both protomers exhibited higher fluctuations than that of the A1 modules (Supplementary Note 6 and Supplementary Fig. 6). Further cluster analysis based on K-means algorithm was performed to investigate the conformational changes of the B2’ module, leading to the identification of five main populations. As shown in Fig. 4a1–a3, the B2’ module underwent gradual sideways rotation. Figure 4a4 and a5 revealed that the spatial arrangements of the B2’ and the first A1 module were altered, with the first A1 module extending outward. These findings suggested that the CRD can utilize two distinct mechanisms to orient its conformations and interact with the adjacent protomer: One involves a lateral rotation of the B2’ muddle, and the other is characterized by the extension of the first A1 module.Next, to investigate the inter-domain conformational arrangement of the CRDs, we introduced two CVs, namely, dB2’-B2’, which reflects the spacing between the B2’ modules, and dD560-D560, which signifies the distance between the second A1 modules. Utilizing these CVs, the conformational landscape of CRDs was delineated within a 2D space (Fig. 4B). Concentrating on the dominant population zones (under 2 kcal/mol), the inter-domain arrangement of CRDs can be divided into three steps: Firstly, from the S1 to just before entering the S3 state, the closure of each VFT yielded a substantial progressive reduction in the two CVs that were broadly linearly correlating (Supplementary Fig. 7). The dB2’-B2’ value decreased from 61.6 to 29.2 Å and the dD560-D560 value decreased from 81.0 to 51.6 Å. Next, along with the transient transition of the S3 state, the dB2’-B2’ exhibited a significant decrease (~7.8 Å, from 29.2 to 21.4 Å), whereas dD560-D560 showed a modest reduction (~2.1 Å, from 51.6 to 49.5 Å). This change may be attributed to the pronounced flexibility of the B2’ module, which underwent lateral rotation and initiated the “active” CRDs interface mediated by B2’ module (Supplementary Figs. 7 and 8). Finally, from the end of the S3 to S5 states, dD560-D560 gradually decreased (~15.8 Å, from 49.5 to 33.7 Å) while interactions between B2’ modules underwent subtle adjustment (~3.6 Å, from 21.4 to 17.8 Å).Importantly, the role of the B2’ module in mediating the active CRDs interface was confirmed by introducing a cysteine-mediated crosslink at the potential interface of the B2’ module (I523C). The I523C mutant maintained comparable levels of cell surface expression but demonstrated a robust constitutive activity as high as the glutamate-induced activity in the wild type (WT) (Fig. 4C, D). Mutating the I523 residue to alanine or altering its adjacent two residues led to the absence of constitutive activity and a significant reduction in efficacy (Fig. 4D and Supplementary Table 3), yet cell surface expression levels remained similar to those of the WT. Such inter-domain dynamics of the CRDs well explained why smFRET studies on mGlu2, mGlu3, and mGlu5 have observed that the CRDs interconvert multiple states through monitoring distance changes between donor and acceptor fluorophores at the position of D548 in mGlu2 and D560 in mGlu524,43 (Supplementary Fig. 9).Taken together, our findings elucidate the dual role of the B2’ module. On the one hand, it is pivotal in maintaining the spatial alignment between the VFTs and CRDs mainly via an inter-domain disulfide bond. On the other hand, the significant mobility of the B2’ module allows it to act as a structural “pioneer”, adept at receiving, amplifying, and forwarding signals from the VFT domains through the CRDs and onwards to the 7TM domains. This capacity of the B2’ module underscores its indispensable role in orchestrating complex conformational changes and signal propagation within the receptor, which is consistent with its unique structural singularity within class C GPCRs.Dynamic allostery in the 7TM domainAgonist-induced conformational changes ultimately propagate to the intracellular G protein-coupling site of the 7TM domain. A plethora of Cryo-EM structures and dynamic ensembles underscore the common hallmark of G protein coupling in class A and B GPCRs, which is a pronounced outward movement and rotation of TM6 on the cytosolic side that facilitates the deep insertion of G-proteins. Instead, in class C GPCRs, only a subtle TM6 tilt has been identified so far in Gi-stabilized cryo-EM structures, thereby forming a shallow G-protein-binding cavity, indicative of a distinctive overall activation mechanism18. During the activation of mGlu5, each 7TM undergoes an inter-domain approach and a 20° rotation upon activation while maintaining an intra-domain that is relatively similar to the apo state (RMSD7TM 4.0 Å, which is the sum of the Cα RMSDs of the 7TMs from the apo conformation)12. Consequently, such reorientation engenders a TM6–TM6 interface, appearing to be a hallmark of class C GPCR activation11,47.Initially, we investigated whether the intra-domain configuration of 7TM remained rigid throughout activation. Interestingly, it revealed that as the TM6–TM6 interface began to form (dTM6-TM6 < ~20 Å), the 7TM exhibited increased conformational fluctuations ranging from 4.0 to 6.8 Å (Fig. 5A). Further analysis of the per-residue RMSD demonstrated that this increased flexibility was not localized to a specific region of 7TM but rather affected its overall configuration (Supplementary Fig. 10). These findings imply that mGlu5 dimerization at the TM6 interface seems to enhance the flexibility of the overall 7TM domain, thereby eliciting allosteric effects at the G protein-binding site. Such flexibility-driven allosteric effects (also denoted as “dynamic allostery”) have been recently recognized as crucial mechanisms for protein–protein interactions in class B GPCRs with a likely significant role in class C GPCRs activation48.Fig. 5: Macro-reorientation and micro-switches dynamics at the 7TM interface triggering mGlu5 activation.A Conformation landscape projected onto dTM6-TM6 and Cα RMSD from the inactive mGlu5 cryo-EM structure of the 7TM. After multiple rounds of minimization and equilibration, the well-equilibrated inactive conformation (labeled inactive’) shows slight fluctuations from the inactive cryo-EM structure and is depicted as a pentastar. While the active cryo-EM structure is directly represented on the landscape as a pentastar. The free energy scales of the landscape are shown on the right and are expressed in kcal/mol. B Conformation landscape projected onto dTM6-TM6 and the dihedral calculated by the Cα atoms of E7706.35A, S7956.60A, S7956.60B, and E7706.35B. The labels point out four metastable states within the landscape: S3 (purple), S4’ (green), S4 (orange), and S5 (red). The active cryo-EM structure is also presented as a blue pentastar. This color scheme has been maintained in subsequent illustrations. C Representative structures within the 7TM reorientation process. Each structure is presented in a box-colored corresponding to (B). The top half of the box provides a frontal view of mGlu5, while the bottom half provides a top view of the 7TM domain. TM5 and TM6 are highlighted to display their orientation. Significant movement patterns are indicated by green arrows. D Micro-switch conformations and dynamics of protomer B. The structure of each metastable state is superimposed to demonstrate the difference in the conformation of key micro-switches (inactive cryo-EM structure is colored in gray), including Y7796.44 (orange), M6586.43 (green), and Y6593.44 (magenta). Their conformational distribution before and after the establishment of the TM6–TM6 interface is displayed on the right. Chi1 dihedral is used to present rotamer changes of Y7796.44 and Y6593.44, while the distance between the Cα atom of Y6593.44 and the Cε atom of M6586.43 quantifies the displacement of M6586.43. The superscripts refer to the GPCRdb numbering scheme. E, F Mutation of critical micro-switches significantly reduces glutamate-induced signaling compared to the WT, as measured based on IP accumulation with similar levels of cell surface expression. The bars indicate the mean ± SEM values from six (E) and three (F) independent experiments. ns, not statistically significant (one-way ANOVA with Dunnett’s post hoc test).Next, we attempted to elucidate the intricate process of TM6–TM6 interface establishment from an inter-domain perspective and capture the intermediate conformations during the process. A dihedral φTM6-TM6 was calculated by the Cα atoms of E7706.35A, S7956.60A, S7956.60B, and E7706.35B to monitor the conformational orientations of the TM6–TM6 interface. The reliability of φTM6-TM6 was confirmed by observing a rotation of ~20° between the apo (28.2°) and active (8.7°) cryo-EM structures. In conjunction with dTM6-TM6, the conformational landscape of the 7TM orientation was obtained (Supplementary Fig. 11). Focusing on the region where the TM6–TM6 interface begins to form, the S3 to S5 states were located within three predominant free-energy basins (Fig. 5B). In the S3, with the φTM6-TM6 of 25.3°, the B2’ modules of CRDs initiated mutual interactions, while the A1 modules remained separate (Supplementary Fig. 7). Transitioning to the S4, the first A1 module of protomer A assumed the configuration depicted in Fig. 4a4, yielding a φTM6-TM6 of −8.7°. Given that the representative conformation of the S4 metastable state was situated in the lower portion of the basin, further analysis revealed an alternative conformation, S4’, in the upper portion. In this state, the first A1 module of protomer B also adopted the Fig. 4a4 configuration, leading to an intermediate φTM6-TM6 of 9.8° (Fig. 5C). These sequential conformational changes of CRDs in both S4’ and S4 might be amplified through their crucial interactions with ECL2s12, ultimately triggering a 20° rotation in the 7TM domain during activation. Furthermore, the free-energy basin where S5 is located suggests that the TM6–TM6 orientation seen in 6N51 likely denotes a pre-activation state, supported by the trend towards a −8.7° orientation in S5 and even extending to −20° (Fig. 5B). The negative orientations have been observed in recent experimental findings with G protein-stabilized fully active conformations of class C GPCRs ranging from −2.3° in the mGlu2-Gi complex to −20.6° in the mGlu4-Gi complex (Supplementary Table 4)41,49,50. In line with the enhanced flexibility along activation, S3 exhibited an RMSD7TM value of 5.0 Å, with the intracellular half of TM5 exhibiting an asymmetrical upward tilt (Supplementary Fig. 12). This tilt changes the configuration of ICL3, potentially orchestrating the conformational arrangement of the G protein-binding site. Indeed, S4’ showed even greater conformational fluctuations (RMSD7TM 5.9 Å), with the intracellular segment of TM3 in protomer B being symmetrically displaced outward, moving K6653.50B away from E7706.35B (4.1 Å) and R6683.53B away from S6142.35B (8.7 Å). The disrupted ‘ionic lock’ motifs in the protomer B could engender a potential opening cavity to facilitate one G protein binding51. Nevertheless, in the absence of G protein binding, these altered ionic locks might revert, reestablishing the TM6–TM6 interface’s equilibrium orientation.Another significant difference before and after the establishment of the TM6–TM6 interface is the rotamer reorientation of Y7796.44. In the apo crystal structures, Y7796.44 adopted a gauche+ conformation in both protomers12. However, upon interface formation in both S5 and the active cryo-EM structure, Y7796.44 reoriented to a trans conformation, tucking between F7766.41 and I7836.48 (Fig. 5D and Supplementary Fig. 13). During the TM6–TM6 reorientation process, the asymmetric conformations of Y7796.44A and Y7796.44B were captured as the intermediates S3 and S4’. Remarkably, in S4’, protomer B assumed a trans conformation, which correlated with the disrupted salt bridges. Such asymmetric behavior, particularly involving residue Y6.44, has also been captured in a recently released structure of PAM-bound mGlu5. Moreover, this phenomenon is not unique to mGlu5 but has also been observed across G protein-bound class C GPCR structures such as mGlu4-Gi and mGlu2-Gi. The abrogation of Y6.44 in mGlu2 resulted in a decrease in the glutamate-stimulated G protein response49,50. This asymmetric behavior was also evident in heterodimers involving G proteins, including mGlu2-mGlu3-Gi and mGlu2-mGlu4-Gi (Supplementary Fig. 14)41. Our simulations suggested that the rotamer changes of Y7796.44 might modify its neighboring residues, particularly M6586.43 and Y6593.44 (Fig. 5D and Supplementary Fig. 13). Y6593.44 is a well-known crucial activation micro-switch in the class A GPCRs and the mGlu receptors, and its conformational changes are essential for the activation process51,52,53. By modifying the side chain of the neighboring residue M6586.43, the conformation of Y6593.44 may reorient and regulate the activation process. Consistent with this signal transmission pathway, Y7796.44, M6586.43, and Y6593.44 mutations significantly decreased both the potency and efficacy of glutamate-induced activation compared with that in the WT (Fig. 5E, F and Supplementary Table 3). This confirms the importance of these micro-switches in mediating the activation process.In the aforementioned analysis of the 7TM domains, it was observed that the state of the interaction networks (the intracellular ionic lock motif and asymmetric TM6–TM6 interface) exists in a dynamic equilibrium. This indicated that the conformational shift triggered by glutamate alone might not adequately stabilize its fully active state. This aligned with observations from smFRET and cryo-EM experiments, where the fully active structures of the mGlu receptors have only been identified in the presence of an intracellular effector protein, typically in conjunction with a positive allosteric modulator (PAM)40,41,49,50,54,55,56. To delve deeper into how the PAM and G protein (primarily Gq in mGlu5) impact the dynamics of the receptor and enhance our understanding of mGlu5 full activation, leveraging the recently released cryo-EM structures and Alphafold2-multimer technologies, we conducted unbiased all-atom MD simulations on the two full-length mGlu5 systems: one bound with glutamate and PAM CDPPB (1 µs × 10), and the other with glutamate, PAM, and Gq (1 µs × 10). The results demonstrated that the presence of PAM not only stabilized mGlu5 by tightening the active inter-domain 7TM interface but also maintained the appropriate open and extended intra-domain ICL2 conformation (Supplementary Note 7 and Supplementary Fig. 15). Furthermore, to explore the detailed interaction variations at the mGlu5-7TM-Gq interface in addition to the ICL2, we generated the FEL on the interface contacts via the first two tICs (Supplementary Note 8 and Supplementary Fig. 16). Three metastable states were identified: a pre-coupled G1 state, an intermediate G2 state and a fully activated G3. In the fully activated G3 state, the active-state asymmetric TM6–TM6 interface was formed, with an upward shift in the TM6 of the Gq-coupled protomer facilitating the rearrangement of ICL3. This conformational rearrangement further orchestrated the cytoplasmic tip of TM3, ICL2, and H8 (potentially the C terminus) to envelop the α5 helix of Gq. This Gq binding mode observed in the mGlu5 is entirely distinct from those found in class A GPCRs, providing complementary insights into the full activation mechanism of the dimeric GPCRs.Long-range allosteric coupling between residues in mGlu5Allosteric coupling plays a pivotal role in GPCR activation, relying on spatially long-range communication networks both within and between domains18. To investigate how agonist binding at the VFT domain transmits signals over a distance of 120 Å to the intracellular G protein-coupling site in the 7TM domain, motion correlation analysis was calculated based on extensive MD simulations. The Pearson correlation map primarily showed linearly correlated motions within protomers, except for LB1 because the LB1s moved apart and the LB1–LB1 interface was disrupted upon glutamates binding (as previously demonstrated) (Fig. 6A). Specifically, after glutamates binding, the LB2 motion displayed a high correlation with the movement of the 7TM domain in the same protomer, suggesting that the glutamate-bound VFT could potentially activate the 7TM domain of the same protomer (denoted as cis-activation).Fig. 6: Motional correlations among mGlu5 residues during activation.A Pairwise Pearson correlation coefficients matrices using the Cα atoms of mGlu5. Regions of positive correlation (depicted in red) represent correlated movements, whereas negative correlation regions (blue) indicate anti-correlated movements. B Pairwise mutual information matrices of mGlu5. Only significant correlations, defined by values above 0.2, are displayed.However, given that inter-domain motions often exhibit nonlinear correlations that exceed the descriptive capacity of the Pearson correlation map, we opted for mutual information analysis57. Even at separations of 120 Å, significant correlations (>0.6) were observed in the fluctuation of mutual information as a function of pairwise Cα atom distance (Supplementary Fig. 17). In addition to the strong local coupling observed within the VFTs, CRDs, and 7TMs of mGlu5 in the Pearson correlation matrix, closer examination of the mutual information matrix revealed substantial long-range allosteric coupling between the 7TM domain and the VFT, CRD, and 7TM domains of the counterpart protomer (Fig. 6B). This suggests that the glutamate-induced VFT arrangement may also facilitate the activation of the adjacent protomer (denoted as trans-activation)21,58,59.
