Multiple intramolecular triggers converge to preferential G protein coupling in the CB2R

Morales-Pastor, Adrian; Miljuš, Tamara; Dieguez-Eceolaza, Miguel; Stępniewski, Tomasz Maciej; Ledesma-Martin, Vicente; Heydenreich, Franziska M.; Flock, Tilman; Plouffe, Bianca; Le Gouill, Christian; Duchaine, Jean; Sykes, David A.; Nicholson, Colin; Koers, Eline J.; Guba, Wolfgang; Rufer, Arne C.; Grether, Uwe; Bouvier, Michel; Veprintsev, Dmitry B.; Selent, Jana

doi:10.1038/s41467-025-60003-0

Download PDF

Article
Open access
Published: 11 June 2025

Multiple intramolecular triggers converge to preferential G protein coupling in the CB₂R

Nature Communications volume 16, Article number: 5265 (2025) Cite this article

6404 Accesses
1 Citations
62 Altmetric
Metrics details

Subjects

Abstract

G protein-coupled receptors (GPCRs) are important therapeutic drug targets for a wide range of diseases. Upon activation, GPCRs can initiate several signaling pathways, each with unique therapeutic implications. Therefore, understanding how drugs selectively engage specific signaling pathways becomes paramount. However, achieving this selectivity remains highly challenging. To unravel the underlying multifaceted mechanisms, we integrate systematic mutagenesis of the CB₂R, comprehensive profiling of G_αi2 and β-arrestin1 engagements and computer simulations to track the effects of mutations on receptor dynamics. Our research reveals multiple triggers within a complex allosteric communication network (ACN) that converge to preferential CB₂R coupling by modulating evolutionarily conserved motifs. Utilizing network path analysis, we find that potent triggers are typically highly connected nodes and are located near regions of high information transmission within the ACN. Our insights highlight the complexity of GPCR signaling and provide a framework for the rational design of drug candidates tailored to evoke specific functional responses, ultimately enhancing the precision and efficacy of therapeutic interventions.

A non-canonical mechanism of GPCR activation

Article Open access 16 November 2024

G protein-coupled receptors (GPCRs): advances in structures, mechanisms and drug discovery

Article Open access 10 April 2024

Functional dynamics of G protein-coupled receptors reveal new routes for drug discovery

Article 02 January 2025

Introduction

G protein-coupled receptors (GPCRs) are the largest family of cell-surface receptors in humans¹. They are involved in virtually every physiological process, which makes them invaluable targets for drug development. The cannabinoid receptor 2 (CB₂R) is a GPCR that plays a crucial role in numerous physiological processes, including immune response and inflammation². This makes the CB₂R a promising therapeutic target for a variety of diseases, such as chronic pain, neuroinflammation, and cancer^3,4.

Currently, synthetic Δ⁹-THC derivatives (e.g. Nabilone⁵) are available for the treatment of chemotherapy-induced nausea and vomits. However, these compounds are non-selective agonists of cannabinoid receptors and may cause psychoactive effects via the CB₁R⁴. Additionally, they can also induce adverse cardiovascular liability in a dose-dependent manner via the CB₁R⁶. In contrast, cannabinoid receptor ligands that do not activate the CB₁R typically lack psychoactive side effects. An example is the FDA-approved phytocannabinoid cannabidiol for the treatment of epilepsy⁷ which acts as a negative allosteric modulator of the CB₁R⁸ and a partial agonist at the CB₂R⁹. In addition, cannabidiol has been widely used without prescription for the relief of anxiety and stress¹⁰ as well as the management of pain¹¹ and inflammation¹².

Presently, there are over 20 selective CB₂R agonists in clinical trials for a variety of inflammatory conditions with good safety profiles. Unfortunately, these compounds have not yet been approved, reflecting their elusive therapeutic effect. A potential reason for this is a broad distribution of CB₂R in different tissues, as well as “on demand” engagement of the endocannabinoid system¹³, making the development of systemically applied drugs challenging. Developing more pathway-selective drugs represents a promising strategy to modulate the endocannabinoid system with greater precision. An encouraging approach to improve the specificity of drug action as well as modulate its efficacy and to reduce potential side effects, is to exploit the concept of signaling bias. Hereby, a drug preferentially activates specific downstream signaling pathways over ones that are related to unwanted side effects^14,15,16.

Recent work provides evidence for the relevance of allosteric communication and implicated networks in GPCR function and signaling bias^17,18. Allosteric communication networks (ACNs) serve as dynamic infrastructure that facilitates communication between remote regions within a GPCR, connecting critical sites like the ligand-binding site to the intracellular effector coupling site. However, the details of how ACNs drive preferential engagement of a specific effector protein remain largely elusive. A main challenge of studying allostery in GPCRs is to monitor ACNs and their alterations related to specific signaling conditions at appropriate spatio-temporal resolution. Molecular dynamics (MD) simulations have emerged as an invaluable tool for complementing static structural, biophysical and biochemical data^19,20,21,22 to access the proper resolution levels. They allow for the exploration of protein dynamics at an atomic level, offering insights into the conformational changes at microsecond timescale and corresponding intramolecular networks under conditions of signaling bias^23,24. Knowledge of the amino acid networks involved in this allosteric communication can not only advance our understanding of GPCR signaling bias but also boost the design of biased ligands with a more precise therapeutic profile.

In this study, we dissect the allosteric communication networks in the CB₂R, combining high-throughput state-of-the-art computational and cell-based assay methods. In a systematic study, we first probe the perturbation of each individual residue within the primary structure of the CB₂R through mutagenesis combined with G protein activation and β-arrestin1 (βarr₁) recruitment. This information is integrated with time-resolved molecular dynamics of the mutation-induced alteration of the CB₂R ACNs using a machine learning pipeline. Most importantly, our study highlights multiple molecular mechanisms and how they interact with the main communication channels between the orthosteric ligand binding site and the intracellular effector binding interfaces via highly conserved residues. Of interest, mutations that preferentially trigger G protein coupling show high connectivity and are typically located closer to ligand-stabilized communication channels of high information transmission. These insights have important implications for the rational development of novel biased ligands targeting not only the orthosteric site but also allosteric sites in the CB₂R with a therapeutic potential to treat inflammation and pain.

Results

Mutant position and sequence conservation impacts receptor cell-surface expression

First, we analyzed how a mutation in each residue position impacts cell-surface expression. To investigate this, we generated 360 mutants changing the corresponding amino acid to alanine or valine if the amino acid was originally an alanine (see Supplementary Data 1, Fig. 1A–C). We find that approximately 60% of all mutants exhibit a cell-surface expression within the wild-type (WT)-like range (i.e., 80% to 120% of WT expression) (Fig. 1B). This finding suggests that most residue positions in CB₂R display remarkable resilience in maintaining their cell-surface expression upon mutation to alanine/valine. Interestingly, approximately 30% of the mutants show a decrease in expression below the WT-like level, while only 10% of the mutants display an increase (Fig. 1B). This observation indicates that expression-impacting mutants in the CB₂R predominantly reduce cell-surface expression rather than increasing it. Such an effect can be attributed to mutation-induced misfolding and subsequent sequestration in the endoplasmic reticulum or the Golgi complex²⁵.

**Fig. 1: Cell surface expression of 360 CB₂R mutants.**

Of note, the distribution of expression groups across the receptor sequence reveals an interesting pattern (Fig. 1A). Mutations that result in high CB₂R expression levels tend to be preferentially located in transmembrane domains (TMs) 2 and 3 (yellow circles). On the other hand, mutants associated with low expression (red circles) are more evenly dispersed throughout the transmembrane core of the protein, with fewer occurrences in TM2 and TM3. This spatial association between expression groups suggests a potential functional significance of these specific transmembrane domains in modulating cell-surface expression levels in CB₂R.

Furthermore, we studied the relationship between sequence conservation and cell-surface expression. The sequence conservation in the CB₂R was computed based on the Jensen-Shannon Divergence (JSD) scoring method²⁶ from a class A human GPCR alignment obtained from GPCRdb²⁷. We observe that residues with lower conservation scores (<0.2) show less impact on cell-surface expression, with most mutants belonging to the WT-like group (gray points, Fig. 1C). In contrast, residues with conservation scores > 0.2 yield larger expression deviations compared to the WT (arrows). This finding supports the widely accepted viewpoint that highly conserved residues are crucial for vital receptor functions, such as protein folding or trafficking to the cell membrane. Of interest are also residues that, despite being highly conserved (conservation score > 0.4), show a WT-like expression level upon mutation (gray points). Those residues are likely involved in relevant receptor functions other than cell-surface expression, e.g., ligand binding, coupling to transducer proteins or signal transduction.

Low mutational resilience of βarr₁ recruitment results in preferential G_αi2 coupling

We also systematically assessed the effect of 360 point mutations within the CB₂R on both G_αi2 coupling and βarr₁ recruitment after stimulation with the agonist HU-210 using the ebBRET-based Effector Membrane Translocation Assay (EMTA) biosensor platform (Supplementary Data 1). Briefly, receptor-mediated G_αi2 activation was monitored through the translocation of a downstream effector Rap1GAP subunit to the plasma membrane, where it selectively interacts with activated G_i/o protein subfamily²⁸. The same plasma membrane translocation principle is used to measure βarr₁ recruitment to the receptor²⁵. Supplementary Data 4 shows BRET response scores of the simulated mutants.

For this large-scale analysis, we decided to focus on cases with a marked impact on coupling, i.e., the loss of G_αi2 or βarr₁ engagement or both to obtain meaningful results. We defined four classes of receptor mutants (Fig. 2A, see methods): (i) preferential G_αi2 coupling (PrefCoup_Gαi2): no measurable βarr₁ recruitment but preserved G_αi2 coupling (Emax > 50%), (ii) preferential βarr₁ recruitment (PrefCoup_βarr1): no measurable G_αi2 coupling but preserved βarr₁ recruitment (Emax > 50%), (iii) loss of receptor function (NoCoup_{Gαi2_βarr1}): no measurable signal for G_αi2 nor βarr₁ and finally (iv) preserved coupling (Coup_{Gαi2_βarr1}): mutations that preserve G_αi2 coupling and βarr₁ recruitment with an Emax > 50%. Using this classification, we observe that 70% of mutants preserve coupling properties (Emax > 50%) to G_αi2 and βarr₁, whereas 30% result in a loss of G_αi2, βarr₁ or both (Fig. 2B, Supplementary Fig. 1). Interestingly, our data indicate that coupling-impaired mutants predominantly induce the loss of βarr₁ recruitment (19%), leading to preferential G_αi2 coupling, whereas G_αi2 coupling seems to be more robust to mutational alterations (only 2% specifically abrogated G_αi2 activation).

**Fig. 2: Induced coupling profiles by CB₂R mutants.**

To investigate whether a loss of βarr₁ recruitment is correlated with a decrease in G_αi2 coupling, we plotted the G_αi efficacy (Emax_Gi) distribution for mutants with impaired βarr₁ recruitment (i.e., PrefCoup_Gαi2 mutants) (Fig. 2C). The plot reveals that PrefCoup_Gαi2 mutants exhibit G_αi efficacies that are normally distributed around the WT level (Fig. 2C), suggesting that abolishment of βarr₁ recruitment does not necessarily affect the efficacy of G protein signaling. Of note, there is also a significant fraction of receptor mutants (12%) incapable of coupling to both G_αi and βarr₁ (Fig. 2B). These impactful mutations are enriched in highly conserved regions (e.g., microswitches, highlighted with a black circle, Fig. 2A) and represent approximately 50% of highly conserved residues (Fig. 2D).

To obtain a more comprehensive understanding of the implication of residue conservation for receptor functionality, we plotted its distribution for mutants with preserved G_αi2 and βarr₁ coupling (Coup_{Gαi2_βarr1}) compared to the ones with preferential G_αi2 coupling (PrefCoup_Gαi2). To exclude a potential impact through differential receptor expression level, we only plot mutations with a WT-like expression level (i.e., 80% to 120%). We find that PrefCoup_Gαi2 mutants (orange line, Fig. 2E) show a clear shift toward higher conservation scores compared to Coup_{Gαi2_βarr1} mutants (blue line, Fig. 2E). Interestingly, the PrefCoup_Gαi2 mutant distribution shows a bimodal behavior pointing to the existence of two populations. One population is similar to the Coup_{Gαi2_βarr1} mutant distribution with a peak at a conservation score of 0.2 and a second population that is shifted to higher scores with a peak at 0.42. Our data highlight the relationship between the evolutionary conservation of a residue and its impact on the receptor coupling preference.

Preferential G_αi2-coupled mutants are located in proximity to network connections with high information transmission

We hypothesize that the receptor dynamics and the stabilization of specific receptor states through an allosteric communication network (ACN) are crucial for the functional response of the CB₂R. In this framework, PrefCoup_Gαi2 mutants likely exert their effects by perturbing the allosteric communication network in the CB₂R compared to mutants with preserved coupling to G_αi2 and βarr₁ (Coup_{Gαi2_βarr1}).

To test this hypothesis, we performed all-atom molecular dynamics (MD) simulations and sampled the ACN for the WT CB₂R (WT ACN) in complex with the agonist HU-210 over 2 μs accumulated simulation time (5 × 400 ns). The ACN is computed based on contact frequencies including hydrophobic interactions, hydrogen bonds and water-mediated interactions. Then we identified the 100 shortest pathways on this network (LigACN) that mediate the communication from the orthosteric ligand binding site to the intracellular coupling site (Fig. 3A). To estimate information transmission of a contact in the ACN, we computed the frequency of their appearance in 100 shortest pathways (see methods, Fig. 3A, B, and Supplementary Data 3 for transmission information computed as degeneracy for 35 simulated mutants). For our interpretation, we treat the LigACN as the principal contact network, which is stabilized by the ligand to promote a specific receptor state and functional response. In contrast, the entire WT ACN encompasses the complete allosteric connectivity within CB₂R and provides valuable insights into connections that link remote receptor regions to the LigACN. These connections represent a resource that can be harnessed by allosteric modulators, including membrane lipids, small molecules, or other interacting proteins, to fine-tune receptor conformational states and, consequently, modulate its response.

**Fig. 3: The allosteric communication network in the WT CB₂R.**

A first inspection of the LigACN stabilized by agonist HU-210 shows that it comprises several residues crucial for receptor function, known as conserved microswitches (Fig. 3A). These microswitches include residue R131^3x50 from the DRY motif, residue N291^7x45 and S292^7x46 in the sodium binding site, and all five residues in the NPxxY motif N295^7x49 to Y299^7x53 (the GPCRdb numbering scheme is used²⁹). The presence of these highly conserved residues within the LigACN strongly supports its relevance in transmitting information and maintaining the protein’s function. Furthermore, we find that the LigACN elicits a strong communication through the upper and lower part of TM7 as well as the lower parts of TM2, 6 and 3.

We then determined the distance of PrefCoup_Gαi2 mutants to the top connections with high information transmission in the LigACN (LigACN_top, see method section) (Fig. 3C). Interestingly, we find that PrefCoup_Gαi2 mutants are present closer to the LigACN_top (i.e., 0 to 1 connection, Fig. 3D) whereas Coup_{Gαi2_βarr1} mutants are located at larger distances (2 to 4 edges). We verified that this closeness was statistically significant by comparing its mean value to a null distribution built using random sets of residues (p-value 0.006). On top of this proximity, inspection of the protein contact network revealed that PrefCoup_Gαi2 mutants on average have higher connectivity (i.e., a residue has more connections to other residues) than Coup_{Gαi2_βarr1} mutants (Fig. 3B, E). Both proximity to highly information-transmissive connections in the LigACN and the higher connectivity rationalize why these residues have a significant functional impact on the receptor’s coupling profile.

Multiple molecular mechanisms initiate G_αi2 preferential coupling

To obtain a deeper mechanistic understanding of changes in the ACN that drive preferential G_αi2 coupling as a result of the loss of βarr₁ recruitment, we carried out additional MD simulations of a set of 34 mutants from the 360 point mutations including 14 PrefCoup_Gαi2 mutants, 20 Coup_{Gαi2_βarr1} and the wild type, which belongs to Coup_{Gαi2_βarr1} (Fig. 4A, Supplementary Table 1). The selection of mutants was done based on two criteria: (1) mutants show WT-like expression levels, as a reduced/increased receptor expression can significantly alter receptor signaling³⁰ and (2) mutants are located in receptor regions that are structurally resolved. With the aim to understand the initial ACN re-arrangements for preferential G_αi2 coupling as a result of βarr₁ loss, we started from an inactive receptor state (PDB ID 5ZTY³¹, see “Methods” section). For each simulated mutant, we monitored the dynamics of formation and disruption for different interaction types within the entire receptor contact network. The interaction stability for thousands of interatomic connections was quantified by computing the contact frequency for the last 400 ns in 5 replicates (i.e., 2 μs total analysis time per mutant) (Fig. 4B).

**Fig. 4: MD-based classification of PrefCoup_Gαi2 and Coup_{Gαi2_βarr1} mutants.**

First, we investigated if PrefCoup_Gαi2 mutants induce alterations in the ligand-stabilized ACNs. In fact, we find that individual PrefCoup_Gαi2 mutants typically induce larger LigACNs compared to Coup_{Gαi2_βarr1} mutants, i.e., increase the number of nodes in the pathways from the ligand to the intracellular receptor site (Fig. 4C). It seems that PrefCoup_Gαi2 mutants disrupt information transmission pipelines in the WT LigACN. This disruption leads to a branching of the information flow along longer pathways including more residue nodes, likely resulting in a less efficient transmission of information. The functional readout of those mutants suggests that the inclusion of more residues in the communication pipeline and in turn less efficient information transmission is linked to a loss of β-arrestin recruitment in the CB₂R.

To extract further structural information relevant for preferential CB₂R coupling, we reduced the dimensionality of simulated contact networks by applying principal component analysis (PCA) and selected the best performing principal component (PC) using a Support Vector Classifier (SVC). Interestingly, projecting the data on a PC plane defined by PC1 and PC10 reveals a separation of Coup_{Gαi2_βarr1} mutants (blue points) and the PrefCoup_Gαi2 mutants (orange points). In other words, preferential G_αi2 coupling and loss of βarr₁ recruitment in our dataset is described by one PC (PC1) with high explained variability (13%) and a second PC (PC10) with less explained variability (3%). Apparently, there are also contacts with high variability (i.e., PC2 to 9) that are not related to preferential G_αi2 coupling in the CB₂R and that could be linked to other receptor functions. This could include protein folding, trafficking, post-translational modifications, receptor dimerization, as well as interactions with other coupling partners such as kinases (e.g., GRKs), RGS proteins, Filamin A, calmodulin or other natural allosteric modulators like lipid molecules and ions. These interactions enable the CB₂R to mediate its full range of physiological effects.

Interestingly, the spreading of the mutants that preserve coupling (Coup_{Gαi2_βarr1}, blue points) around the true WT system (red points) indicates that some variability in the contact network is tolerated for preserving coupling with an Emax > 50%. Intriguingly, we find that PrefCoup_Gαi2 mutants (orange points) tend to shift further away from the true WT center (red points) (Fig. 4D). Apparently, these mutations perturb specific elements (i.e., stabilization or destabilization) in the contact networks that fall outside the tolerated region of the Coup_{Gαi2_βarr1} mutants (blue points). In addition, we find that PrefCoup_Gαi2 mutants move in three different directions forming clusters 1 to 3. Hence, each cluster appears to differently disrupt the contact network, leading to impaired βarr₁ recruitment and preferential G_αi2 coupling. Ultimately, this suggests that different mechanisms are implicated in preferential G_αi2 coupling and the loss of βarr₁ recruitment in the CB₂R. This could also explain why some PrefCoup_Gαi2 mutants fall within the blue cluster of preserved coupling as they likely exploit another molecular mechanism that is not projected into the plane of PC1 and PC10.

Furthermore, we observe that the three PrefCoup_Gαi2 clusters (orange points) are located at different distances from the true WT simulations (red points). Specifically, cluster 3 is very close to the true WT simulations compared to clusters 1 and 2. This suggests that cluster 3-specific contact stabilities are more sensitive and smaller network alterations can push the CB₂R from a preserved coupling profile to a PrefCoup_Gαi2 profile. The PCA data separation is confirmed using a machine-learning classification approach with a Support Vector Classifier (SVC), which effectively differentiates mutant populations (blue contour, Fig. 4D). Additionally, K-means clustering analysis further corroborates the presence of three distinct clusters (Supplementary Fig. 2).

Of note, we also simulated the same 34 mutants including 14 PrefCoup_Gαi2 mutants, 20 Coup_{Gαi2_βarr1} and the wild type, which belongs to Coup_{Gαi2_βarr1} in the active CB₂R structure coupled to the G_α subunit (PDB ID 6KPF³²). Interestingly, we were not able to obtain conclusive results discriminating PrefCoup_Gαi2 from the Coup_{Gαi2_βarr1} profile (Supplementary Fig. 3). A possible reason for this is that G protein binding to the CB₂R and its penetration into the receptor core disrupts many intramolecular contacts within the allosteric contact network as TM helices move apart (Supplementary Fig. 4). This in turn reduces the measurable impact of mutants on the contact network within CB₂R.

Contact alterations in conserved molecular switches are involved in G_αi2 preferential coupling

To capture in more detail the structural and dynamic alterations in the receptor contact network that lead to G_αi2 preferential coupling, we fitted a logistic regression for each PrefCoup_Gαi2 cluster to separate them from the Coup_{Gαi2_βarr1} group (see method). Top interaction features (orange color scaled connections) that distinguish preferential G_αi2 from the preserved coupling profile are plotted on the allosteric communication network (WT ACN, gray), including the ligand-stabilized ACN (cyan) (Fig. 5A). The obtained representation indicates that mutational perturbation in preferential G_αi2 coupling for clusters 1 to 3 (orange color scale) involves a complex and multifaceted modulation of the WT ACN (gray plus cyan) and the corresponding LigACN (cyan). To further simplify the interpretation of these complex perturbations of the WT ACN, we focus on highly conserved receptor regions with known relevance for receptor function, such as the CWxP motif, allosteric sodium binding site, NPxxY motif, and the DRY motif. Cluster-specific modulations are depicted in Fig. 5B–D (detailed structural description can be found in Supplementary Note 1). A cluster summary (Fig. 5E) reveals that the sodium binding site and the DRY motif are modulated in all three identified clusters, the CWxP only in clusters 1 and 2, and the NPxxY motif only in cluster 3. Based on the finding that the DRY motif and the sodium binding site are modulated in all three PrefCoup_Gαi2 clusters, it is tempting to speculate that they play a critical role in preserving βarr₁ recruitment. The observed modulation of those motifs occurs indirectly through mutational perturbations from topologically different sites. Unfortunately, we could not assess the direct impact of those motifs on preferential CB₂R coupling as their mutation dramatically impacts receptor expression (see Supplementary Data 1). Another interesting observation is that detected perturbations around the sodium binding site, CWxP and DRY motifs involve contact (de)stabilization effects through different residues (Fig. 5E, Supplementary Fig. 5). This corroborates the finding that multiple molecular mechanisms can converge into the same receptor response, i.e., loss of βarr₁ recruitment while preserving G_αi2 coupling with an Emax > 50%.

**Fig. 5: Structural characterization of PrefCoup_Gαi2 mutant clusters.**

Discussion

Allosteric communication networks are highly complex dynamic entities that drive receptor conformational states and in turn determine their final functional outcome as a response to environmental changes (e.g., mutations or ligand stimulation). Here, we investigated the dynamics of such networks and their modulation related to preferential signaling of the CB₂R. The results obtained from our study shed light on several key aspects that link receptor structure and dynamics to function. We anticipate that our findings will provide a broadly applicable framework for understanding CB₂R function due to the integration of large-scale experimental data (a systematic mutational scan of CB₂R) with computational data (MD simulations of all CB₂R mutants retaining WT-like expression levels). Additionally, the use of graph theory analysis and machine learning-based classification enhances the generalizability of our approach.

First of all, our functional readouts suggest that G_αi2 coupling is more robust to mutational perturbations compared to βarr₁ recruitment in the CB₂R (Fig. 2B). Our observation is in line with a similar study on β₂ adrenergic receptor (β₂AR) where more than 75% of mutational substitutions retain the receptor’s G protein coupling profile³³. However, it is important to mention that this is not a general tendency for GPCRs, as shown recently for the angiotensin II type 1 receptor (AT₁R)¹⁸. For this receptor, β-arrestin recruitment is more resistant to mutations compared to G protein binding. This could possibly be related to the intrinsic propensity of GPCRs to differentially engage in interactions with β-arrestins. While class A GPCRs for arrestin recruitment (such as β₂AR and CB₂R) interact with arrestins only transiently, class B receptors (V₂R and AT₁R) form more stable interactions which are potentially more resistant to disruptions by single-point mutagenesis³⁴.

Plotting the G_αi2 preferential mutants on the static CB₂R structure offers a glimpse into their potential impact based on their location within the receptor (Fig. 2A). Interestingly, mutations that impact β-arrestin recruitment are mainly found in transmembrane helices and only one was found in the receptor C-terminal tail (expanded discussion in Supplementary Note 2). This might be surprising as serine and threonine residues in intracellular regions are known to promote arrestin recruitment upon GRK-mediated phosphorylation. However, this result can be attributed to the ability of CB₂R to recruit arrestin independently of GRK phosphorylation, as previously reported³⁵. We further confirmed this finding using GRK2/3/5/6 knockout cell lines (Supplementary Fig. 6).

Although static representations can be informative, they fall short when studying the molecular mechanism that drives the modulation of a receptor that is in constant motion. Here, we demonstrate that the intrinsic CB₂R flexibility can be captured with atomistic resolution through allosteric communication networks within the receptor in which each interatomic connection is characterized through a distinct contact frequency. Further network processing using graph theory highlights how the agonist HU-210 allosterically communicates with the intracellular coupling site of the CB₂R mainly through TM7 (Fig. 3A). The relevance of TM7-related agonist binding goes along with a previous study for the β₂ adrenergic receptor despite using a different approach. Applying the concept of correlated movements for the detection of allosteric communication pipelines³⁶, the authors find that the inverse agonist-bound inactive states uses mainly TM6, whereas agonists-bound active states favor TM7 as the main communication pipeline.

Investigating the mutant position, their information transmission and connectivity within such networks reveals a complex modulation of CB₂R coupling (Fig. 3B). For a more detailed description, we introduced the concept of the ligand-stabilized ACN (LigACN) which is part of the overall CB₂R ACN. The LigACN is important for stabilizing the receptor in a specific conformational ensemble and therefore directing the functional outcome of the ligand-stimulated receptor. Importantly, we find that G_αi2 preferential mutants are located closer to connections of high transmission in the ligand-stabilized ACN (Fig. 3B, C) and are of higher network connectivity (Fig. 3E). From this, we can conclude that both information transmission and connectivity are critical residue properties for modulating the allosteric network in the CB₂R and in turn its ability to recruit βarr₁. Mechanistically, this can be explained as follows: mutant perturbations of allosteric connections (i.e., loosening or strengthening) in the WT ACN and specifically in the LigACN shift the CB₂R conformational ensemble from states with favorable affinities for multiple transducer proteins (e.g., G_αi, βarr₁, others) to states that lose βarr₁ while preferentially preserving G_αi2 with Emax > 50% (Supplementary Fig. 7).

Our findings are in line with previous studies of GPCRs^24,37 and other protein classes³⁸, showing that disruption of allosteric communication networks through mutations results in a significant impact on protein function. This also has implications for allosteric modulators that can fine-tune function by perturbing allosteric connections between different protein regions³⁹, even from peripheral receptor sites. In this context, the obtained mutational CB₂R map, which resolves each amino acid position and its role for receptor function is highly valuable for identifying novel allosteric sites for drug discovery programs. Nevertheless, it is worth noting that not all mutations occurring within our computed LigACN consistently exhibit the same level of efficiency in altering G_αi2 preferential coupling. This is not surprising as those positions might be involved in other functional consequences (e.g., coupling to other (un)known signaling partners) that are not taken into account in our study. Further investigations are required to better understand these functional relationships.

Ultimately, the most important finding of our study is related to the observation that there are multiple molecular mechanisms that can induce the same functional response, i.e., preferential G_αi2 coupling. In fact, this goes along with the applied concept of allosteric communication networks that includes regulatory triggers from different receptor regions. Principal component analysis (Fig. 3D) and logistic regression (Fig. 4) pinpoint three separate mechanisms that regulate highly conserved receptor motifs such as CWxP, sodium binding site, NPxxY and DRY. The involvement of those microswitches in impairing β-arrestin recruitment has been reported in different individual studies (i.e., DRY motif^39,40,41,42, NPxxY motif⁴³, the toggle switch⁴⁴, and the sodium binding site⁴⁵). Importantly, we find that the same molecular switch can be perturbed through different combinations of interactions and contact stabilities as highlighted by our PrefCoup_Gαi2 clusters 1 to 3 (Fig. 5B–D). Ultimately, we anticipate the existence of other molecular mechanisms based on the observation that not all PrefCoup_Gαi2 mutants in our dataset (e.g., S61A at the bottom of TM1) fall into one of the three PrefCoup_Gαi2 clusters, but are located in the preserved signaling cluster (Fig. 3C). Moreover, we expect that additional mechanisms will come into play when changing experimental conditions such as the choice of agonist or the type of mutational substitution (e.g., as occurring in single nucleotide polymorphisms). Given the multitude of molecular mechanisms that induce preferential G_αi2 coupling suggests that biased ligands may not all operate through a shared universal mechanism. Instead, it is possible that chemically different ligand classes achieve the same bias response by employing distinct structural mechanisms. A key factor in this process will be specific ligand-receptor contacts within the binding pocket, which are unique to each ligand class and which translate into the stabilization of distinct ACNs.

Interestingly, there is experimental evidence that this proposed concept is common for GPCRs. For instance, it has been shown for two chemically different agonists of the δ opioid receptor that their ability to activate G_αi is differentially impacted by a N^7x45A mutation, a residue that is part of the highly conserved allosteric sodium binding site deep inside the receptor (Supplementary Table 2). Whereas, the peptide agonist DADLE completely loses its ability to activate G_αi, the agonist BW373U86 maintains its G_αi activation properties similar to that of the WT upon a N^7x45A mutation. Overall, these findings corroborate that structurally diverse ligands with unique ligand receptor interaction profiles engage distinct ligand-specific networks to achieve the same functional outcome. This has significant implications for drug responses and personalized medicine. For example, a patient harboring a specific genetic variant may respond differently to a treatment depending on whether the variant affects the ligand-specific ACN or not.

In addition to explaining ‘diverse ligands with the same functional outcome’, our framework also accounts for ‘closely related ligands with different functional responses’ as observed in monoaminergic receptors. Subtle modifications in the ligand’s chemical structure—and consequently ligand-receptor interactions—can convert a natural agonist into a biased agonistst^46,47,48. According to our structural framework, this is due to alterations of the stabilized ACNs triggered by the different ligand receptor contacts.

Ultimately, β-arrestin recruitment for many GPCRs depends on GRK-mediated phosphorylation⁴⁹ including prototypical receptors such as β₂AR⁵⁰ and AT₁R⁵¹. Importantly, since CB₂R recruits βarr₁ independently of GRKs (Supplementary Fig. 6), mutant-induced alterations in the ACNs are unlikely to affect GRK-mediated phosphorylation. Instead, they may impact alternative β-arrestin recruitment mechanisms such as its direct interaction with the receptor core to yield a preferential G protein coupling profile. By delving into these unique mechanisms, we can advance our understanding of biased signaling and target GPCR-mediated signaling pathways more effectively to eventually broaden our repertoire of therapeutic approaches.

In conclusion, our study provides insights into the intricate interplay of multiple triggers on complex allosteric communication networks in the CB₂R and their final functional outcomes. Potent triggers of such changes in ACNs that lead to distinct intracellular signaling responses are characterized by high connectivity and their close vicinity to high information-transmissive connections. Our findings have important implications for elaborating drug design strategies aimed at tailoring receptor response towards more efficacious and safer drugs. However, given the complexity of nature and the involvement of multiple molecular mechanisms, further research is necessary to fully comprehend these intricate processes and their potential applications in drug development.

Methods

Plasmids and mutagenesis

Human CB₂R gene was optimized for expression in mammalian systems and synthesized (Genewiz). The CB₂R gene was fused to an N-terminal signal sequence, SNAP-tag, TwinStrep-tag and a C-terminal 1D4-tag, and cloned into pcDNA4/TO vector. Single-point mutants were generated for all 360 CB₂R residues where non-alanine amino acids were replaced by alanine, and alanines were replaced by valines. For that purpose, PCR with mutation-containing primers was used as described in Heydenreich et al.⁵². Plasmids encoding β-arrestin1-RlucIl⁵³, human GRK2⁵⁴, rGFP-CAAX^25,54, and wild-type G_αi2 protein and an effector protein fused with RlucII (Rap1GAP-RlucII)²⁸ were previously described. Briefly, for the G protein activation assay, receptor and wild-type G_αi2 subunit were co-transfected with Rap1GAP-RlucII effector and rGFP-CAAX, while in β-arrestin₁ recruitment assay, receptor was co-transfected with β-arrestin1-RlucII, wild-type GRK2, and rGFP-CAAX. In both assays an increase in BRET signal was monitored upon ligand-induced Rap1GAP or β-arrestin recruitment to the plasma membrane.

Transfection

Human embryonic kidney (HEK) 293 T cells⁵⁵ were cultured in DMEM, 4.5 g/L glucose, with L-glutamine and phenol red, without sodium pyruvate (#319-015, Wisent Inc., Montreal, Canada), supplemented with 10% Newborn Calf Serum (NCS, iron fortified, #075-350, Wisent Inc., Montreal, Canada) and 1x Penicillin-Streptomycin (PS, #450-201, Wisent Inc., Montreal, Canada) in adherent culture at 37 °C, 5% CO₂. For transfection, medium without phenol red was used (#319-051, Wisent Inc., Montreal, Canada). Cells were dissociated using 0.05% Trypsin with 0.53 mM EDTA (#325-142, Wisent Inc., Montreal, Canada) and transiently transfected with receptor and biosensor DNA using 25 kDa linear polyethylenimine (#23966, Polysciences Inc., Warrington, PA, USA) in Cellstar® PS 96-well cell culture plates (Greiner Bio-One, Germany) at a density of 20,000 cells per well.

BRET experiments

Coelenterazine 400a (#340, Prolume Ltd., Pinetop, AZ, USA) was used as the luciferase substrate. HU-210 was purchased from Toronto Research Chemicals (#TRC-H673500, Toronto, ON, Canada) BRET experiments were performed 2 days post-transfection, when culture medium was replaced by 100 μL Tyrode’s buffer (137 mM NaCl, 1 mM MgCl₂, 1 mM CaCl, 0.9 mM KCl, 3.6 mM NaH, 11.9 mM NaHCO₃, 25 mM HEPES, 5.55 mM _D-glucose, pH = 7.4). A volume of 0.4 μL of ligand was added to cell plates using a BiomekFX Lab Automation Workstation (Beckman) and a previously calibrated pin tool, followed by incubation at 37 °C for 5 min in a Cytomat 6001 (Thermo Scientific). Coelenterazine 400a working solution (50 μM DBC and 1% Pluronic F-127 in Tyrode’s buffer) was added with a Multidrop Combi (Thermo Scientific) to a final concentration of 5 μM and shaken for 10 s. The plates were further incubated for 5 min at room temperature before the BRET signal was read on a Synergy Neo (Biotek) equipped with dual photomultiplier tubes (PMTs) (emission recorded at 410 and 515 nm, gain settings of 150 for each PMT and 1.2 s integration time).

Cell surface ELISA measurements

HEK293T cells were transfected as above using only receptor-coding plasmids in poly-D-lysine-coated 96-well plates and incubated at 37 °C and 5% CO₂ for 2 days. Cells were washed with PBS (200 μL/well) and fixed with 3% paraformaldehyde (50 μL/well) for 10 min, followed by three washing steps (100 μL/well, PBS + 0.5% BSA). Receptor was detected by primary rabbit anti-SNAP antibody (50 μL/well of 1:2000 dilution, #A00684, GenScript, 1 h at RT) and anti-rabbit HRP antibody (50 μL/well of a 1:1000 dilution, #NA934, GE Healthcare, 1 h at RT), with washing steps as above, followed by three wash steps with PBS. SigmaFast solution (100 μL/well) was added, followed by incubation at RT protected from light, until color change was observed. To stop the reaction, 25 μL 3 M HCl were added to each well and 100 μL of the solution were transferred to a new transparent 96-well plate for the measurement. The absorbance at 492 nm was measured using a Tecan GENios Plus microplate reader, reporting on the amount of receptor expressed.

Effect of the receptor expression level on maximal ligand‑induced response

Receptor titrations were performed to determine the effect of receptor expression level on Emax values of both pathways. Various amounts of CB₂R DNA were transfected (0, 0.1, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 150, 200% of the amount used for each biosensor), together with other biosensor components in duplicates. Two days post‑transfection one replicate was used to perform ELISA, while the other replicate was used to measure BRET signal upon stimulation with 1 μM HU-210, as described above. Change in BRET signal (Emax) vs. amount of CB₂R transfected (percentage of the amount used for each biosensor), and cell‑surface expression level (determined by ELISA) vs. amount of CB₂R transfected were measured. Correlations of the Emax and ELISA‑measured expression levels were determined for each pathway, and data fit equation parameters (linear regression for G_αi2 activation and non-linear hyperbolic fit for β-arrestin1 recruitment) were used to correct Emax output of each mutant for its previously measured expression level (Supplementary Data 1).

BRET data analysis

The BRET response was determined by the ratio of the light intensity emitted by the acceptor (rGFP, measured at 515 nm), over the light intensity emitted by the BRET donor (RlucII, measured at 410 nm). Single-replicate concentration-response curves were fitted to a Hill equation with a Hill coefficient of 1. For each concentration-response curve, the maximum response of signal (Emax) and the ligand concentration at half-maximal signal response (as pEC50) was determined using the custom software DataFitter (https://github.com/dbv123w/DataFitter), followed by averaging over biological triplicates. For further analysis, Emax of CB₂R mutants was corrected for the mutant expression levels and normalized to the WT value (Supplementary Fig. 8), while pEC50 of each mutant was subtracted from the WT value. Both raw and corrected values are reported (Supplementary Data 1).

Model generation and molecular dynamics simulation

As a starting point, we have selected the inactive crystal structure of the CB₂R bound to the antagonist AM10257 (PDB code: 5ZTY³¹) as well as the active structure bound to the agonist AM12033 (PDB ID: 6KPF³²). The sequence of the receptor was reverted to the canonical human sequence using data from UNIPROT [P34972]. Residues between positions 222 and 235 of the ICL3 were removed and endings were capped using ACE and CT3 terminal patches following the GPCRm standard protocol²⁰. According to this protocol, experimentally unresolved ICL3 loops that are longer than 10 amino acids are not modeled due to the complication of accurate structural predictions of these highly flexible regions. The approach for this protocol is also based on the observations that GPCRs are able to couple and activate G proteins also in the absence of the ICL3⁵⁶. Nevertheless, to ensure that no artifacts are introduced into TM5 and TM6 dynamics due the lack of ICL3, we carried out a control simulation with the ICL3 modeled. For the ICL3 construction, we used the loop modeling tool and standard settings of the MOE 2022.02 software (Chemical Computing Group). We find that the Root Mean Square Fluctuations (RMSF) of TM5 and TM6 are very similar (Supplementary Fig. 9).

The initial position of the ligand was generated based on the position of the highly similar agonist AM12033 co-solved in the inactive CB₂R CryoEM structure (PDB ID: 6KPC³²) (Supplementary Fig. 10). The resulting complex was embedded in a 54% POPC, 36% POPG and 10% cholesterol membrane, based on the coordinates available in the OPM database⁵⁷. The pre-equilibrated membrane was generated in CHARMM-GUI⁵⁸. Protonation states at a pH of 7.4 were assigned using Propka, and residues facing the membrane were manually curated to avoid charges facing the membrane. The obtained system was solvated with TIP3 waters, and the ionic strength of the solution was maintained at 0.15 M using NaCl ions.

The obtained complex was equilibrated for 100 ns in NPT conditions, with constraints applied to the protein backbone and using the Berendsen barostat⁵⁹. From the equilibrated system, we generated systems for each of the studied mutants (see list in Supplementary Table 1). Every system underwent a short 5 ns NPT equilibration phase to adjust the environment to the introduced mutation. Afterward, we applied a simulation scheme in which each mutant evolved for 1 μs simulation time in an NVT ensemble (Fig. 3B). The last 400 ns were carried out in 5 replicates for the inactive CB₂R reaching a total simulation time of 2.6 μs per mutant (i.e, 600 ns plus 5* 400 ns = 2.6 μs). Running multiple independent replicates provides the advantage of sampling different regions of conformational space and enhances results reproducibility (i.e., no enhanced sampling is required)^20,22,60. In addition, this approach is sufficient to allow the system to comprehensively explore the conformational space around the experimental starting structure which minimizes the dependence on the initial configuration. In addition, we carried out 3 replicates for the active CB₂R reaching a total simulation time of 1.8 μs per mutant (i.e, 600 ns plus 3* 400 ns = 1.8 μs).

All simulations were carried out using the ACEMD engine⁶¹. Parameters were obtained from the Charmm36M forcefield generated in November 2017⁶², with ligand parameters generated in Paramchem^63,64 with the CgenFF forcefield^65,66. The temperature was maintained at 310 K. For NPT runs we have employed a 2 fs timestep, and for NVT runs a 4 fs one. This was possible due to the hydrogen mass repartitioning scheme employed within ACEMD. The reliability and reproducibility checklist for molecular dynamics simulations is found in Supplementary Table 3.

Conservation metrics computation

To calculate the evolutionary conservation score for each residue of the CB₂R, we obtained the structural sequence alignment of class A GPCRs from GPCRdb and input it into Capra et al.‘s online tool with the Jensen-Shannon Divergence (JSD) scoring method²⁶. The default values were used for all parameters, including a window size of 3 and BLOSUM 62 matrices for both the background and target. The JSD measures the similarity between two probability distributions by comparing the distribution of amino acids from a sample in a particular position with the distribution of a set without evolutionary pressure²⁶. The JSD ranges from 0 to 1, where 0 indicates identical distributions, and 1 means that the distributions have no overlap. Hence, a JSD score of 1 indicates that the two probability distributions are entirely dissimilar with no common features. We then used conservation scores to investigate the relationship between the impact of mutations on coupling affinities and the evolutionary pressure on each residue. To achieve this, we calculated the ratio of G_αi2 preferential mutants above each JSD score obtained in the analysis. For example, we found that out of all the residues with a conservation score of 0.36 or higher, 70% are G_αi2 preferential, compared to the 40% found in the whole sequence. A conservation score distribution of the mutants with a WT-like expression level (80% to 100%) can be found in Fig. 2E. A conservation score distribution for all mutants (independently of the expression level) can be found in Supplementary Fig. 11.

Contacts computation

The contact computation in the simulations was done using the GetContacts package [https://getcontacts.github.io/]. All the interaction types supported by the software were used to compute interaction frequencies. Namely, 3 types of hydrogen bonds, salt bridges, pi-cation, pi-stacking, t-stacking, hydrophobic, Van der Waals including hydrophobic contacts, water bridges and extended water bridges. Each of the interaction types has a set of conditions regarding the distance and relative angle of the two groups involved. The complete specifications of these criteria can be found in the GetContacts repository. After computing the interactions in each frame, the overall contact frequency between pairs of residues was computed. This was achieved by counting all the frames where any contact type was established between two residues and dividing it by the total number of frames of the simulation.

ACN computation

We used the AlloViz v0.1 Python package⁶⁷ to construct ACNs for the different simulations and replicas based on the residue contact frequencies. The raw contact frequencies were processed as in Foutch et al.⁶⁸ by (1) removing the contacts between consecutive residues in the amino acid sequence, as their interactions are consistently established and provide limited useful information, and (2) defining a lower frequency threshold of 0.1 to filter out contacts that are rarely established. Finally, we built the networks by connecting the pairs of interacting residues (nodes) with edges, weighted by their contact frequency.

Ligand-stabilized ACN (LigACN)

Taking the resulting networks from the WT simulations, we used the Dijkstra’s algorithm as implemented in NetworkX⁶⁹ (https://networkx.org/) to calculate the 100 shortest pathways connecting the ligand CHEMBL5085420 (source) and four selected residues in the intracellular domain of the receptor (sinks) for each replica (Fig. 3B). This resulted in a total of 2000 pathways (100 pathways × 4 sinks × 5 replicas). The sinks, namely Arg131 (3 × 50), Asp240 (6 × 30), Ser303 (8 × 47) and Ser69 (2 × 39), were selected based on their proximity to the G protein in the CB₂R-G protein complex (Supplementary Fig. 12). The robustness of this approach was evaluated by assessing the impact of selecting alternative sink residues on the shortest pathways (see Supplementary Fig. 13). Subsequently, we computed the information transmission through each edge in the network as their degeneracy, that is, the frequency with which they appear in the shortest pathways.

Computing the distance to the top connections with highest information transmission (LigACN_top)

The top connections with highest information transmission (LigACN_top) are defined setting a degeneracy cutoff ≥ 0.146 (light blue edges in Fig. 3A). Then, we computed the distance of PrefCoup_Gαi2 and Coup_{Gαi2_βarr1} mutants to the LigACN_top (Fig. 3D). The distance is defined as the number of edges in the shortest path between the mutant residue of interest and any residue from the LigACN_top. The obtained plot shows that mutants with coupling bias (PrefCoup_Gαi2) are located overall closer to the LigACN_top at distances 0 and 1 compared to mutants that behave like the WT receptor (Coup_{Gαi2_βarr1}). We tested if the proximity of PrefCoup_Gαi2 mutants to the LigACN_top was higher than could be expected by chance. For this, we calculate the quantile of the distance of PrefCoup_Gαi2 mutants in the distribution of the distance of random residue sets. Finally, to check for a potential bias by the selected degeneracy cutoff, we also computed the mutant distances for a range of different degeneracy thresholds (0–0.2) which further confirmed our observations (Supplementary Fig. 14).

Definition of a balanced set of CB₂R alanine mutants for unveiling the determinants G_αi2 preferential signaling

We compiled a balanced dataset of Coup_{Gαi2_βarr1} and G_αi2 preferential mutants according to the following criteria: (i) Coup_{Gαi2_βarr1} with Emax_Gαi2 and Emax_βarr1 > 50%, (ii) preferential G_αi2 coupling with an Emax_Gαi2 > 50% and no measurable arrestin activity (Supplementary Data 1, Supplementary Fig. 15). As cell-surface expression can alter receptor signaling³⁰, we prioritized the selection of mutants with WT-like expression levels. A final requirement for mutant selection was that mutants are in receptor regions that had been structurally resolved previously to allow their investigations in reliable three-dimensional dynamics models. The final set consists of the wild type and 34 mutants of which 20 were Coup_{Gαi2_βarr1} and 14 were G_αi2 preferring mutants (PrefCoup_Gαi2). The location of selected mutants in the CB₂R is indicated in Fig. 4A. Despite the mutant class definition relying exclusively on Emax, it is not ignoring EC50. This is because all the mutants without measurable efficacy also presented a potency of 0. So, independently of which criteria, Emax or EC50, is used, the compositions of the groups would be identical.

Principal component analysis and clustering definition

We conducted a principal component analysis (PCA) on the contact frequencies data in order to reduce its dimensionality and to detect variables that are able to discern between the two mutant groups PrefCoup_Gαi2 and Coup_{Gαi2_βarr1}. The PCA was constructed using contact frequencies taking into account all residues in the CB₂R. The matrix used to calculate the Principal Components (PCs) is 175 ((34 mutants + WT) × 5 replicates) × 4421 (contacts). Then, we explored all 45 pairwise combinations of such components. This inspection revealed the presence of a set of outlier simulations, all of them belonging to the A53V mutant. After the removal of this mutant, we found the best-performing PC plane, defined by PC1 and PC10, by using a Support Vector Classifier (SVC)⁷⁰. The SVC model clearly separates the data into three different clusters for G_αi2 preferential mutants and one cluster for Coup_{Gαi2_βarr1} mutants. Ultimately, we also confirmed this clustering using the K-means algorithm⁷¹.

Cluster characterization

To characterize each cluster, we used logistic regression to differentiate between the wild type-like and the preferential coupling cluster. Because the model could not generalize well with divided training and validation data, we opted to train a regularized model with high bias on the entire dataset. After training, we extracted the coefficients of each predictor variable, sorted them based on absolute value, and selected the top 20 variables. We then associated the coefficient with the residues involved in the interaction that formed the predictor variable. Model and model training was done using the scikit-learn Python module⁷². From all interactions that were highlighted by the highest coefficients in absolute value from the logistic regression we paid special attention to those involving conserved residues and interactions that were densely packed in a specific region of the receptor.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Data supporting the findings of this manuscript are available as a Supplementary Information and Supplementary Data 1 to 4. The MD data generated in this study have been deposited in the GPCRmd database under accession code: https://gpcrmd.org/dynadb/publications/1540/. 3D structures used in this study: 6KPC, 5ZTY, 6KPF. Source data are provided with this paper.

Code availability

The python code for data processing is available at GitHub (https://github.com/GPCRmd/prefcoup_cb2r)⁷³. Additional data supporting the findings are available from the corresponding authors upon request.

References

Rosenbaum, D. M., Rasmussen, S. G. F. & Kobilka, B. K. The structure and function of G-protein-coupled receptors. Nature 459, 356–363 (2009).
Article ADS CAS PubMed PubMed Central Google Scholar
Maresz, K. et al. Direct suppression of CNS autoimmune inflammation via the cannabinoid receptor CB1 on neurons and CB2 on autoreactive T cells. Nat. Med. 13, 492–497 (2007).
Article CAS PubMed Google Scholar
Ashton, J. C. & Glass, M. The cannabinoid CB2 receptor as a target for inflammation-dependent neurodegeneration. Curr. Neuropharmacol. 5, 73–80 (2007).
Article CAS PubMed PubMed Central Google Scholar
Maccarrone, M. et al. Goods and bads of the endocannabinoid system as a therapeutic target: lessons learned after 30 years. Pharmacol. Rev. 75, 885–958 (2023).
Article CAS PubMed PubMed Central Google Scholar
Tsang, C. C. & Giudice, M. G. Nabilone for the management of pain. Pharmacotherapy 36, 273–286 (2016).
Article CAS PubMed Google Scholar
Pacher, P., Steffens, S., Haskó, G., Schindler, T. H. & Kunos, G. Cardiovascular effects of marijuana and synthetic cannabinoids: the good, the bad, and the ugly. Nat. Rev. Cardiol. 15, 151–166 (2018).
Article CAS PubMed Google Scholar
Abu-Sawwa, R., Scutt, B. & Park, Y. Emerging use of Epidiolex (cannabidiol) in epilepsy. J. Pediatr. Pharmacol. Ther. 25, 485–499 (2020).
PubMed PubMed Central Google Scholar
Pandey, P. et al. Determination of the negative allosteric binding site of cannabidiol at the CB1 receptor: A combined computational and site-directed Mutagenesis study. ACS Chem. Neurosci. 16, 311–328 (2025).
Article CAS PubMed Google Scholar
Tham, M. et al. Allosteric and orthosteric pharmacology of cannabidiol and cannabidiol-dimethylheptyl at the type 1 and type 2 cannabinoid receptors. Br. J. Pharmacol. 176, 1455–1469 (2019).
Article CAS PubMed Google Scholar
Berger, M., Amminger, G. P. & McGregor, I. S. Medicinal cannabis for the treatment of anxiety disorders. Aust. J. Gen. Pract. 51, 586–592 (2022).
Article PubMed Google Scholar
Mohammed, S. Y. M. et al. Effectiveness of cannabidiol to manage chronic pain: a systematic review. Pain. Manag. Nurs. 25, e76–e86 (2024).
Article PubMed Google Scholar
Lowin, T. et al. Cannabidiol (CBD): a killer for inflammatory rheumatoid arthritis synovial fibroblasts. Cell Death Dis. 11, 714 (2020).
Article CAS PubMed PubMed Central Google Scholar
Di Marzo, V. Targeting the endocannabinoid system: to enhance or reduce?. Nat. Rev. Drug Discov. 7, 438–455 (2008).
Article PubMed Google Scholar
Smith, J. S., Lefkowitz, R. J. & Rajagopal, S. Biased signalling: from simple switches to allosteric microprocessors. Nat. Rev. Drug Discov. 17, 243–260 (2018).
Article CAS PubMed PubMed Central Google Scholar
Luttrell, L. M. Minireview: More than just a hammer: ligand ‘bias’ and pharmaceutical discovery. Mol. Endocrinol. 28, 281–294 (2014).
Article PubMed PubMed Central Google Scholar
Sivertsen, B., Holliday, N., Madsen, A. N. & Holst, B. Functionally biased signalling properties of 7TM receptors - opportunities for drug development for the ghrelin receptor. Br. J. Pharmacol. 170, 1349–1362 (2013).
Article CAS PubMed PubMed Central Google Scholar
Morales-Pastor, A. et al. In silico study of allosteric communication networks in GPCR signaling bias. Int. J. Mol. Sci. 23, 7809 (2022).
Article CAS PubMed PubMed Central Google Scholar
Cao, Y. et al. Unraveling allostery within the angiotensin II type 1 receptor for Gα and β-arrestin coupling. Sci. Signal. 16, eadf2173 (2023).
Article CAS PubMed PubMed Central Google Scholar
Torrens-Fontanals, M., Stepniewski, T. M., Gloriam, D. E. & Selent, J. Structural dynamics bridge the gap between the genetic and functional levels of GPCRs. Curr. Opin. Struct. Biol. 69, 150–159 (2021).
Article CAS PubMed Google Scholar
Rodríguez-Espigares, I. et al. GPCRmd uncovers the dynamics of the 3D-GPCRome. Nat. Methods 17, 777–787 (2020).
Article PubMed Google Scholar
Lopez-Balastegui, M. et al. Relevance of G protein-coupled receptor (GPCR) dynamics for receptor activation, signalling bias and allosteric modulation. Br. J. Pharmacol. https://doi.org/10.1111/bph.16495 (2024).
Aranda-García, D. et al. Large scale investigation of GPCR molecular dynamics data uncovers allosteric sites and lateral gateways. Nat. Commun. 16, 2020 (2025).
Article PubMed PubMed Central Google Scholar
Suomivuori, C.-M. et al. Molecular mechanism of biased signaling in a prototypical G protein-coupled receptor. Science 367, 881–887 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Nivedha, A. K. et al. Identifying functional hotspot residues for biased ligand design in G-protein-coupled receptors. Mol. Pharmacol. 93, 288–296 (2018).
Article CAS PubMed PubMed Central Google Scholar
Namkung, Y. et al. Monitoring G protein-coupled receptor and β-arrestin trafficking in live cells using enhanced bystander BRET. Nat. Commun. 7, 12178 (2016).
Article ADS PubMed PubMed Central Google Scholar
Capra, J. A. & Singh, M. Predicting functionally important residues from sequence conservation. Bioinformatics 23, 1875–1882 (2007).
Article CAS PubMed Google Scholar
Pándy-Szekeres, G. et al. GPCRdb in 2018: adding GPCR structure models and ligands. Nucleic Acids Res 46, D440–D446 (2018).
Article PubMed Google Scholar
Avet, C. et al. Effector membrane translocation biosensors reveal G protein and βarrestin coupling profiles of 100 therapeutically relevant GPCRs. Elife 11, e74101 (2022).
Article CAS PubMed PubMed Central Google Scholar
Isberg, V. et al. Generic GPCR residue numbers - aligning topology maps while minding the gaps. Trends Pharmacol. Sci. 36, 22–31 (2015).
Article CAS PubMed Google Scholar
Li, A., Liu, S., Huang, R., Ahn, S. & Lefkowitz, R. J. Loss of biased signaling at a G protein-coupled receptor in overexpressed systems. PLoS One 18, e0283477 (2023).
Article CAS PubMed PubMed Central Google Scholar
Li, X. et al. Crystal Structure of the Human Cannabinoid Receptor CB2. Cell 176, 459–467.e13 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hua, T. et al. Activation and signaling mechanism revealed by cannabinoid receptor-G complex structures. Cell 180, 655–665.e18 (2020).
Article CAS PubMed PubMed Central Google Scholar
Heydenreich, F. M. et al. Molecular determinants of ligand efficacy and potency in GPCR signaling. Science 382, eadh1859 (2023).
Article CAS PubMed PubMed Central Google Scholar
Oakley, R. H., Laporte, S. A., Holt, J. A., Caron, M. G. & Barak, L. S. Differential affinities of visual arrestin, beta arrestin1, and beta arrestin2 for G protein-coupled receptors delineate two major classes of receptors. J. Biol. Chem. 275, 17201–17210 (2000).
Article CAS PubMed Google Scholar
Patel, M. et al. Delineating the interactions between the cannabinoid CB2 receptor and its regulatory effectors; β-arrestins and GPCR kinases. Br. J. Pharmacol. 179, 2223–2239 (2022).
Article CAS PubMed Google Scholar
Bhattacharya, S. & Vaidehi, N. Differences in allosteric communication pipelines in the inactive and active states of a GPCR. Biophys. J. 107, 422–434 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Ma, N., Nivedha, A. K. & Vaidehi, N. Allosteric communication regulates ligand-specific GPCR activity. FEBS J. 288, 2502–2512 (2021).
Article CAS PubMed PubMed Central Google Scholar
Laine, E., Auclair, C. & Tchertanov, L. Allosteric communication across the native and mutated KIT receptor tyrosine kinase. PLoS Comput. Biol. 8, e1002661 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Rivalta, I. et al. Allosteric communication disrupted by a small molecule binding to the imidazole glycerol phosphate synthase protein-protein interface. Biochemistry 55, 6484–6494 (2016).
Article CAS PubMed Google Scholar
Hegron, A. et al. Identification of key regions mediating human melatonin type 1 receptor functional selectivity revealed by natural variants. ACS Pharm. Transl. Sci. 4, 1614–1627 (2021).
Article CAS Google Scholar
Al-Zoubi, R., Morales, P. & Reggio, P. H. Structural insights into CB1 receptor biased signaling. Int. J. Mol. Sci. 20, 1837 (2019).
Article CAS PubMed PubMed Central Google Scholar
Morales, P., Goya, P. & Jagerovic, N. Emerging strategies targeting CB cannabinoid receptor: biased agonism and allosterism. Biochem. Pharmacol. 157, 8–17 (2018).
Article CAS PubMed Google Scholar
Wootten, D., Christopoulos, A., Marti-Solano, M., Babu, M. M. & Sexton, P. M. Mechanisms of signalling and biased agonism in G protein-coupled receptors. Nat. Rev. Mol. Cell Biol. 19, 638–653 (2018).
Article CAS PubMed Google Scholar
Cong, X. et al. Molecular insights into the biased signaling mechanism of the μ-opioid receptor. Mol. Cell 81, 4165–4175.e6 (2021).
Article CAS PubMed PubMed Central Google Scholar
Fenalti, G. et al. Molecular control of δ-opioid receptor signalling. Nature 506, 191–196 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Stepniewski, T. M. et al. Mechanistic insights into dopaminergic and serotonergic neurotransmission - concerted interactions with helices 5 and 6 drive the functional outcome. Chem. Sci. 12, 10990–11003 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kossatz, E. et al. G protein-specific mechanisms in the serotonin 5-HT2A receptor regulate psychosis-related effects and memory deficits. Nat. Commun. 15, 4307 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Martí-Solano, M. et al. Detection of new biased agonists for the serotonin 5-HT2A receptor: modeling and experimental validation. Mol. Pharmacol. 87, 740–746 (2015).
Article PubMed Google Scholar
Gutkind, J. S. & Kostenis, E. Arrestins as rheostats of GPCR signalling. Nat. Rev. Mol. Cell Biol. 19, 615–616 (2018).
Article CAS PubMed Google Scholar
Nobles, K. N. et al. Distinct phosphorylation sites on the β(2)-adrenergic receptor establish a barcode that encodes differential functions of β-arrestin. Sci. Signal. 4, ra51 (2011).
Gareri, C. et al. Phosphorylation patterns in the AT1R C-terminal tail specify distinct downstream signaling pathways. Sci. Signal. 17, eadk5736 (2024).
Article CAS PubMed PubMed Central Google Scholar
Heydenreich, F. M. et al. High-throughput mutagenesis using a two-fragment PCR approach. Sci. Rep. 7, 6787 (2017).
Article ADS PubMed PubMed Central Google Scholar
Quoyer, J. et al. Pepducin targeting the C-X-C chemokine receptor type 4 acts as a biased agonist favoring activation of the inhibitory G protein. Proc. Natl Acad. Sci. Usa. 110, E5088–97 (2013).
Article CAS PubMed PubMed Central Google Scholar
Beautrait, A. et al. Mapping the putative G protein-coupled receptor (GPCR) docking site on GPCR kinase 2: insights from intact cell phosphorylation and recruitment assays. J. Biol. Chem. 289, 25262–25275 (2014).
Article CAS PubMed PubMed Central Google Scholar
Beautrait, A. et al. A new inhibitor of the β-arrestin/AP2 endocytic complex reveals interplay between GPCR internalization and signalling. Nat. Commun. 8, 15054 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Rubenstein, R. C., Wong, S. K. & Ross, E. M. The hydrophobic tryptic core of the beta-adrenergic receptor retains Gs regulatory activity in response to agonists and thiols. J. Biol. Chem. 262, 16655–16662 (1987).
Article CAS PubMed Google Scholar
Lomize, M. A., Pogozheva, I. D., Joo, H., Mosberg, H. I. & Lomize, A. L. OPM database and PPM web server: resources for positioning of proteins in membranes. Nucleic Acids Res. 40, D370–6 (2012).
Article CAS PubMed Google Scholar
Jo, S., Kim, T., Iyer, V. G. & Im, W. CHARMM-GUI: a web-based graphical user interface for CHARMM. J. Comput. Chem. 29, 1859–1865 (2008).
Article CAS PubMed Google Scholar
Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F., DiNola, A. & Haak, J. R. Molecular dynamics with coupling to an external bath. J. Chem. Phys. 81, 3684–3690 (1984).
Article ADS CAS Google Scholar
Knapp, B., Ospina, L. & Deane, C. M. Avoiding false positive conclusions in molecular simulation: the importance of replicas. J. Chem. Theory Comput. 14, 6127–6138 (2018).
Article CAS PubMed Google Scholar
Harvey, M. J., Giupponi, G. & Fabritiis, G. D. ACEMD: accelerating biomolecular dynamics in the microsecond time scale. J. Chem. Theory Comput. 5, 1632–1639 (2009).
Article CAS PubMed Google Scholar
Huang, J. et al. CHARMM36m: an improved force field for folded and intrinsically disordered proteins. Nat. Methods 14, 71–73 (2017).
Article CAS PubMed Google Scholar
Vanommeslaeghe, K. & MacKerell, A. D. Jr. Automation of the CHARMM General Force Field (CGenFF) I: bond perception and atom typing. J. Chem. Inf. Model. 52, 3144–3154 (2012).
Article CAS PubMed PubMed Central Google Scholar
Vanommeslaeghe, K., Raman, E. P. & MacKerell, A. D. Jr. Automation of the CHARMM General Force Field (CGenFF) II: assignment of bonded parameters and partial atomic charges. J. Chem. Inf. Model. 52, 3155–3168 (2012).
Article CAS PubMed PubMed Central Google Scholar
Vanommeslaeghe, K. et al. CHARMM general force field: a force field for drug-like molecules compatible with the CHARMM all-atom additive biological force fields. J. Comput. Chem. 31, 671–690 (2010).
Article CAS PubMed PubMed Central Google Scholar
Yu, W., He, X., Vanommeslaeghe, K. & MacKerell, A. D. Jr. Extension of the CHARMM General Force Field to sulfonyl-containing compounds and its utility in biomolecular simulations. J. Comput. Chem. 33, 2451–2468 (2012).
Article CAS PubMed PubMed Central Google Scholar
Nerín-Fonz, F. et al. AlloViz: A tool for the calculation and visualisation of protein allosteric communication networks. Comput. Struct. Biotechnol. J. 23, 1938–1944 (2024).
Article PubMed PubMed Central Google Scholar
Foutch, D., Pham, B. & Shen, T. Protein conformational switch discerned via network centrality properties. Comput. Struct. Biotechnol. J. 19, 3599–3608 (2021).
Article CAS PubMed PubMed Central Google Scholar
Hagberg, A. A., Schult, D. A. & Swart, P. J. Exploring network structure, dynamics, and function using NetworkX. In Proceedings of the Python in Science Conference 11–15 (SciPy, 2008).
Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20, 273–297 (1995).
Article MATH Google Scholar
Lloyd, S. Least squares quantization in PCM. IEEE Trans. Inf. Theory 28, 129–137 (1982).
Article MathSciNet MATH Google Scholar
Pedregosa, F. et al. Scikit-learn: machine learning in python. J. Mach. Learn. Res. abs/1201.0490, (2011).
Morales-Pastor, A., Dieguez-Eceolaza, M. & Selent, J. Multiple Intramolecular Triggers Converge to Preferential G Protein Coupling in the CB2R. https://doi.org/10.5281/ZENODO.15270434. (Zenodo, 2025).
D’Agostino, R. & Pearson, E. S. Tests for Departure from Normality. Empirical Results for the Distributions of b 2 and √b 1. Biometrika 60, 613 (1973).
MathSciNet MATH Google Scholar
Hodges, J. L. The significance probability of the Smirnov two-sample test. Ark. Mat. 3, 469–486 (1958).
Article MathSciNet MATH Google Scholar

Download references

Acknowledgements

The authors thank Carsten Hoffmann for sharing the GRK knockout cell line for validating the dependence of arrestin recruitment on GRK-mediated receptor phosphorylation. A.M.P. acknowledges support from the Instituto de Salud Carlos III FEDER (FI19/00037). J.S. acknowledges funding from the Horizon Europe Project Obelisk under the grant agreement 101080465, the ERA-NET NEURON & Ministry of Economy, Industry, Competitiveness (AC18/00030) and the MICIU/AEI/10.13039/501100011033 and FEDER, UE (PID2022-137161OB-I00). Furthermore, this work was supported by the Swiss National Science Foundation grants 135754 and 159748 and SNF Sinergia CRSII3_141898 to D.B.V. T.M. was awarded the Swiss National Science Foundation Doc. Mobility supplement and the EMBO Short-term Fellowship (420‑2016). DAS, EK and DBV gratefully acknowledge funding by the Medical Research Council (MR/Y003667/1).

Author information

These authors contributed equally: Adrian Morales-Pastor, Tamara Miljuš.

Authors and Affiliations

Research Programme on Biomedical Informatics, Hospital del Mar Research Institute and Department of Medicine and Life Science, Universitat Pompeu Fabra, Barcelona, Spain
Adrian Morales-Pastor, Miguel Dieguez-Eceolaza, Tomasz Maciej Stępniewski, Vicente Ledesma-Martin & Jana Selent
Laboratory of Biomolecular Research, Paul Scherrer Institute, 5232, Villigen, PSI, Switzerland
Tamara Miljuš, Tilman Flock & Dmitry B. Veprintsev
Department of Biology, ETH Zürich, 8093, Zürich, Switzerland
Tamara Miljuš & Dmitry B. Veprintsev
Centre of Membrane Proteins and Receptors (COMPARE), University of Birmingham and University of Nottingham, Midlands, UK
Tamara Miljuš, David A. Sykes, Colin Nicholson, Eline J. Koers & Dmitry B. Veprintsev
Department of Metabolism and Systems Science, College of Medicine and Health, University of Birmingham, Birmingham, UK
Tamara Miljuš
Department of Biochemistry and Molecular Medicine, Institute for Research in Immunology and Cancer (IRIC), Université de Montréal, Montréal, Québec, Canada
Tamara Miljuš, Bianca Plouffe, Christian Le Gouill, Jean Duchaine & Michel Bouvier
InterAx Biotech AG, PARK InnovAARE, Villigen, Switzerland
Tomasz Maciej Stępniewski
Department of Pharmaceutical Chemistry, University of Marburg, Marburg, Germany
Franziska M. Heydenreich
Wellcome-Wolfson Institute for Experimental Medicine, Queen’s University Belfast, Belfast, UK
Bianca Plouffe
Division of Physiology, Pharmacology & Neuroscience, School of Life Sciences, University of Nottingham, Nottingham, UK
David A. Sykes, Colin Nicholson, Eline J. Koers & Dmitry B. Veprintsev
Z7 Biotech Ltd, Nottingham, UK
David A. Sykes
Roche Innovation Center Basel, F. Hoffmann‑La Roche Ltd., Basel, 4070, Switzerland
Wolfgang Guba, Arne C. Rufer & Uwe Grether

Authors

Adrian Morales-Pastor
View author publications
Search author on:PubMed Google Scholar
Tamara Miljuš
View author publications
Search author on:PubMed Google Scholar
Miguel Dieguez-Eceolaza
View author publications
Search author on:PubMed Google Scholar
Tomasz Maciej Stępniewski
View author publications
Search author on:PubMed Google Scholar
Vicente Ledesma-Martin
View author publications
Search author on:PubMed Google Scholar
Franziska M. Heydenreich
View author publications
Search author on:PubMed Google Scholar
Tilman Flock
View author publications
Search author on:PubMed Google Scholar
Bianca Plouffe
View author publications
Search author on:PubMed Google Scholar
Christian Le Gouill
View author publications
Search author on:PubMed Google Scholar
Jean Duchaine
View author publications
Search author on:PubMed Google Scholar
David A. Sykes
View author publications
Search author on:PubMed Google Scholar
Colin Nicholson
View author publications
Search author on:PubMed Google Scholar
Eline J. Koers
View author publications
Search author on:PubMed Google Scholar
Wolfgang Guba
View author publications
Search author on:PubMed Google Scholar
Arne C. Rufer
View author publications
Search author on:PubMed Google Scholar
Uwe Grether
View author publications
Search author on:PubMed Google Scholar
Michel Bouvier
View author publications
Search author on:PubMed Google Scholar
Dmitry B. Veprintsev
View author publications
Search author on:PubMed Google Scholar
Jana Selent
View author publications
Search author on:PubMed Google Scholar

Contributions

A.M.P., V.L., and T.M.S. carried out the molecular dynamics simulations and data analysis under the supervision of J.S. T.M. carried out the alanine scan for the CB₂R under the supervision of D.B.V. and M.B. and the signaling assays under the supervision of M.B. and D.B.V. F.M.H., B.P., and J.D. contributed to data collection. C.L.G. designed some of the biosensors used in this study. T.F. contributed to the initial data analysis. W.G., A.C.R., and U.G. participated in ligand selection for the alanine scanning signaling screen and provided unique CB₂-selective compounds. D.B.V. wrote the DataFitter software for high-throughput data analysis. D.A.S., C.N., and E.J.K. validated the dependence of arrestin recruitment on GRK-mediated receptor phosphorylation. A.M.P. and J.S. wrote the manuscript. T.M. and D.B.V. contributed to the manuscript writing. M.D.E. carried out major manuscript revision including additional data analysis, representation and discussion under the supervision of J.S. All authors revised the manuscript. D.B.V. and J.S. supervised and coordinated the whole project.

Corresponding authors

Correspondence to Dmitry B. Veprintsev or Jana Selent.

Ethics declarations

Competing interests

M.B. is the president of the Scientific Advisory Board of Domain Therapeutics, a biotech company to which biosensors used in this study were licensed for commercial use. D.B.V. is founder of Z7 Biotech Ltd, an early-stage drug discovery contract research organization. All other authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Nadine Jagerovic who co-reviewed with Paula Morales; Arnau Cordomí, Francesco Raimondi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Reporting Summary

Transparent Peer Review file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Morales-Pastor, A., Miljuš, T., Dieguez-Eceolaza, M. et al. Multiple intramolecular triggers converge to preferential G protein coupling in the CB₂R. Nat Commun 16, 5265 (2025). https://doi.org/10.1038/s41467-025-60003-0

Download citation

Received: 29 February 2024
Accepted: 09 May 2025
Published: 11 June 2025
DOI: https://doi.org/10.1038/s41467-025-60003-0

Subjects

Abstract

Similar content being viewed by others

Introduction

Results

Mutant position and sequence conservation impacts receptor cell-surface expression

Low mutational resilience of βarr1 recruitment results in preferential Gαi2 coupling

Preferential Gαi2-coupled mutants are located in proximity to network connections with high information transmission

Multiple molecular mechanisms initiate Gαi2 preferential coupling

Contact alterations in conserved molecular switches are involved in Gαi2 preferential coupling

Discussion

Methods

Plasmids and mutagenesis

Transfection

BRET experiments

Cell surface ELISA measurements

Effect of the receptor expression level on maximal ligand‑induced response

BRET data analysis

Model generation and molecular dynamics simulation

Conservation metrics computation

Contacts computation

ACN computation

ACN computation

Ligand-stabilized ACN (LigACN)

Computing the distance to the top connections with highest information transmission (LigACNtop)

Definition of a balanced set of CB2R alanine mutants for unveiling the determinants Gαi2 preferential signaling

Principal component analysis and clustering definition

Cluster characterization

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links

Low mutational resilience of βarr₁ recruitment results in preferential G_αi2 coupling

Preferential G_αi2-coupled mutants are located in proximity to network connections with high information transmission

Multiple molecular mechanisms initiate G_αi2 preferential coupling

Contact alterations in conserved molecular switches are involved in G_αi2 preferential coupling

Computing the distance to the top connections with highest information transmission (LigACN_top)

Definition of a balanced set of CB₂R alanine mutants for unveiling the determinants G_αi2 preferential signaling