1. Biochemistry and Chemical Biology
Download icon

Allosteric ligands control the activation of a class C GPCR heterodimer by acting at the transmembrane interface

  1. Lei Liu
  2. Zhiran Fan
  3. Xavier Rovira  Is a corresponding author
  4. Li Xue
  5. Salomé Roux
  6. Isabelle Brabet
  7. Mingxia Xin
  8. Jean-Philippe Pin  Is a corresponding author
  9. Philippe Rondard  Is a corresponding author
  10. Jianfeng Liu  Is a corresponding author
  1. Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, China
  2. Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, France
  3. MCS, Laboratory of Medicinal Chemistry, Institute for Advanced Chemistry of Catalonia (IQAC-CSIC), Spain
Research Article
  • Cited 0
  • Views 597
  • Annotations
Cite this article as: eLife 2021;10:e70188 doi: 10.7554/eLife.70188

Abstract

G protein-coupled receptors (GPCRs) are among the most promising drug targets. They often form homo- and heterodimers with allosteric cross-talk between receptor entities, which contributes to fine-tuning of transmembrane signaling. Specifically controlling the activity of GPCR dimers with ligands is a good approach to clarify their physiological roles and validate them as drug targets. Here, we examined the mode of action of positive allosteric modulators (PAMs) that bind at the interface of the transmembrane domains of the heterodimeric GABAB receptor. Our site-directed mutagenesis results show that mutations of this interface impact the function of the three PAMs tested. The data support the inference that they act at the active interface between both transmembrane domains, the binding site involving residues of the TM6s of the GABAB1 and the GABAB2 subunit. Importantly, the agonist activity of these PAMs involves a key region in the central core of the GABAB2 transmembrane domain, which also controls the constitutive activity of the GABAB receptor. This region corresponds to the sodium ion binding site in class A GPCRs that controls the basal state of the receptors. Overall, these data reveal the possibility of developing allosteric compounds able to specifically modulate the activity of GPCR homo- and heterodimers by acting at their transmembrane interface.

Editor's evaluation

This manuscript builds upon recent structural insights into the GABAB receptor, an unusual and important member of the G protein-coupled receptor (GPCR) family which functions as an obligate heterodimer. The work investigates positive allosteric modulators (PAMs) of the GABAB receptor that bind to the heterodimeric interface between transmembrane helix 6 of the two protomers. Through functional characterization of a large panel of mutant receptors, a role for this binding site in conveying agonism by the PAMs is tested. The manuscript also provides evidence for a role of residues deep in the transmembrane domain in regulating both constitutive activity and allosteric agonism in GABAB receptors. These principles are likely also relevant for other family C GPCRs, suggesting a strategy for drug development targeting this important GPCR family.

https://doi.org/10.7554/eLife.70188.sa0

Introduction

G protein-coupled receptors (GPCRs) are key players in intercellular communication. They are involved in many physiological functions (Heng et al., 2013), and many mutations (Schoneberg and Liebscher, 2021) and genetic variants (Hauser et al., 2018) of GPCR genes are associated with human diseases. Not surprisingly, GPCRs are major targets for drug development (Hauser et al., 2017). Although GPCRs are able to activate G proteins in a monomeric state, they can also form homo- and heteromers (Ferré et al., 2014), even in native tissues (Albizu et al., 2010; Rivero-Müller et al., 2010), often named homo- and heterodimer for simplicity. Such complexes allow allosteric cross-talk between receptors and contribute to a fine-tuning of transmembrane (TM) signaling. Then, modulating the activity of GPCR dimers could offer a new way of controlling physiological functions.

Several approaches were developed to control GPCR dimer activity by targeting the TM domain (TMD) with ligands that could be of potential interest in vivo (Botta et al., 2020). One approach being tested was the use of bivalent ligands; that is, two ligands attached by a linker able to bind to each protomer within a dimer (Huang et al., 2021). However, each ligand still has the possibility to act on the monomers. Another possibility highly anticipated would be the development of ligands binding at the dimer interface, which would not be able to act on monomers, but instead it would specifically control the dimer activity. Hope for the possible identification of such type of ligands came from the discovery of allosteric modulators binding to sites outside the TM bundle in class A and B GPCRs (Thal et al., 2018; Wang et al., 2021), in regions possibly involved in receptor dimerization and oligomerization (Figure 1A). The first evidence for this type of ligands came from the structure of the class C GABAB receptor, where two different positive allosteric modulators (PAMs) were reported to bind at the TM interface of this heterodimer made of the two homologous subunits GABAB1 (GB1) and GABAB2 (GB2) (Kim et al., 2020; Mao et al., 2020; Shaye et al., 2020; Figure 1B). No other example has been described yet, including in the other class C GPCRs, such as the metabotropic glutamate (mGlu) receptors where the allosteric modulators that target the TMD bind in the ancestral GPCR cavities within the TMD core (Doré et al., 2014; Wu et al., 2014; Figure 1B).

Allosteric binding sites in the different classes of G protein-coupled receptors (GPCRs).

(A, B) Scheme and structure of the transmembrane domain (TMD) representative of the diversity of the binding sites for allosteric modulators (filled blue circles or blue spheres) in selected human class A and B GPCRs (muscarinic M2 receptor PDB 4MQT [1], purinergic P2Y1 receptor PDB 4XNV [2], corticotropin-releasing factor receptor 1 PDB 4K5Y [3], and β2 adrenergic receptor PDB 5X7D [4]) (A), as well as in the class C homodimer mGluR5 (PDB 6N51) bound to a NAM (PDB 4OO9) and the heterodimer GABABR bound to GS39783 (PDB 6UO8) (B). In classes A and B, the allosteric modulators were shown to bind to different sites within and outside of the TM bundle, in contrast to the orthosteric ligand (black triangle).

The GABAB is activated by γ-aminobutyric acid (GABA), the main inhibitory neurotransmitter in the central system linked to various neurological diseases. This receptor is an attractive drug target for brain diseases (Bowery, 2006; Gassmann and Bettler, 2012) with therapeutic drugs such as baclofen (Lioresal) and β-phenyl-γ-aminobutyric acid (phenibut) used to treat spasticity (Chang et al., 2013), alcohol addiction (Agabio et al., 2018), anxiety, and insomnia (Lapin, 2001). Auto-antibodies that target GABAB have previously been identified at the origin of epilepsies and encephalopathies (Dalmau and Graus, 2018), and genetic mutations have previously been associated with Rett syndrome and epileptic encephalopathies (Hamdan et al., 2017; Vuillaume et al., 2018; Yoo et al., 2017). GABAB has a unique allosteric mechanism for signal transduction, in which the binding of an agonist in the extracellular domain of GB1 leads to G protein activation through a rearrangement of the intracellular interface of the TMD of GB2 (Monnier et al., 2011; Shaye et al., 2020; Xue et al., 2019).

PAMs offer a number of advantages over the agonists since they temporally and spatially control enhanced signaling only when the natural ligand is present (Conn et al., 2014; Pin and Prézeau, 2007; Schwartz and Holst, 2007). However, many of these PAMs have an intrinsic agonist activity as they can increase receptor activity in the absence of an orthosteric agonist. These allosteric ligands are called ago-PAMs (Conn et al., 2014), in contrast to pure PAMs that have no intrinsic agonist effect. The ago-PAMs could be of tremendous interest for the treatment of patients with genetic variants of GPCRs as more and more are discovered in the exome sequence studies (Hauser et al., 2018). Finally, the allosteric agonist activity of these molecules is not predictable yet, and the molecular basis of such activity remains unknown.

In this study, we examined if the commonly available GABAB PAMs all bind in the GB1-GB2 TM interface and how these PAMs can control the GABAB heterodimer activity. We show that, indeed, all PAMs bind at the same site, at the active interface of these TMDs, despite their different structures. We also reveal that the agonist activity of these PAMs involves a key region in the central core of the GB2 TMD that also controls the constitutive activity of the receptor. This region is functionally conserved in most GPCRs as it corresponds to the Na+ site found in class A receptors (Zarzycka et al., 2019). These data reveal the possibility of developing allosteric compounds able to specifically modulate the activity of GPCR homo- and heterodimers.

Results

Different functional properties of the PAMs for GABAB receptor

We have evaluated the agonist activity and the allosteric modulation of the most studied PAMs (Urwyler, 2011) on the recombinant wild-type GABAB receptors, rac-BHFF (Malherbe et al., 2008), CGP7930 (Urwyler et al., 2001), and GS39783 (Urwyler et al., 2003; Figure 2A). These compounds had showed pure PAM or ago-PAM effect on GABAB activity in different studies both in vitro and in vivo (Urwyler, 2011). The binding site of both rac-BHFF (Kim et al., 2020; Mao et al., 2020) and GS39783 (Shaye et al., 2020) in the purified full-length GABAB receptor was recently reported from structural analyses, including when the receptor is in complex with the G protein (Shen et al., 2021). They bind into a pocket formed in the active heterodimer by its TM6s at the interface of the TMDs.

Figure 2 with 1 supplement see all
Different functional properties of the positive allosteric modulators (PAMs) for GABAB receptor.

(A) Chemical structures of the three PAMs of GABAB receptor used in the study and commercially available. (B, C) Intracellular Ca2+ responses (B) and inositol-phosphate-1 (IP1) accumulation (C) mediated by the indicated compounds. (D) Schematic representation of the Go protein BRET sensor. (E, F) Kinetics of BRET ratio changes of this sensor (E) upon addition (arrow) of buffer (control condition), 100 μM GABA or 100 μM of the indicated PAMs (blue). Data are from a typical experiment performed three times independently. Changes in BRET ratio (F) were measured 150 s after drug application. Data are shown as means ± SEM of three biologically independent experiments. Data are analyzed using one-way ANOVA test followed by a Dunnett’s multiple comparison test to determine significance (compared with the buffer condition) with ***p<0.0005, ****p<0.0001. (G–I) Intracellular Ca2+ responses mediated by GABA in the absence or presence of the indicated concentrations of rac-BHFF (G), CGP7930 (H), or GS39783 (I). Data are normalized by the response of 1 mM GABA and shown as means ± SEM of 4–15 biologically independent experiments.

In the present study, we show that rac-BHFF and CGP7930 have intrinsic agonist and PAM activity in different functional assays (Figure 2B–F, Figure 2—figure supplement 1). In contrast, the agonist activity of GS39783 is weaker, acting more like a pure PAM. These results are in line with our previously reported data (Lecat-Guillet et al., 2017). The strong agonist effect of rac-BHFF was revealed by its capacity to activate the GABAB receptor even in the absence of GABA. It was measured by intracellular calcium release and inositol-phosphate-1 (IP1) accumulation assays in cells coexpressing the chimeric G protein Gαqi9 (a Gαq protein in which the last nine C-terminal residues have been replaced by those of Gαi2), which allows the coupling of Gi/o-coupled receptors to phospholipase C (Figure 2B–F and Supplementary files 1 and 2; Supplementary file 3). rac-BHFF alone reached more than 60% of the maximal effect of the full agonist GABA (Figure 2B and C, Figure 2—figure supplement 1A, and Supplementary file 1). CGP7930 also has agonist activity in absence of GABA (Figure 2B and C and Supplementary file 1). In contrast to the two other PAMs, the intrinsic agonist activity of GS39783 was only observed in the IP1 accumulation assays (Figure 2C), and not in the calcium release assay (Figure 2B and Supplementary file 1), and using a BRET sensor for the activation of GαoA protein (Figure 2F). The observed discrepancy is most likely related to the nature of these assays, the IP1 accumulation assay being an equilibrium assay while the two other are highly dependent on the kinetics of ligand binding to the receptor (Bdioui et al., 2018). The IP1 assay is thus proposed to be the most sensitive assay for evaluating slow binders and low-efficacy compounds such as the GS39783.

GB1 and GB2 TMDs are sufficient for the agonist activity of the PAMs

Our recent studies have shown that there is strong positive cooperativity between GB1 and GB2 TMDs for receptor activation (Monnier et al., 2011; Xue et al., 2019). Here, we demonstrate that both GB1 and GB2 TMDs are required for the efficient agonist activity of the PAMs. This is consistent with the binding site of rac-BHFF and GS39783 involving both GB1 and GB2 TMD in the active conformation of the receptor (Kim et al., 2020; Mao et al., 2020; Shaye et al., 2020; Shen et al., 2021).

We measured the rac-BHFF-induced intracellular calcium signaling of different GABAB constructs expressed at the cell surface (Figure 3, Figure 3—figure supplement 1A, and Supplementary file 3). The presence of both GB1 and GB2 TMDs was sufficient for its activity (Figure 3A–D). Accordingly, the presence of GB1 or GB2 ECD or both is dispensable for the ago-PAM activity since the relevant constructs in which one or both ECDs were deleted could still be activated by rac-BHFF alone efficiently (Figure 3B–D). However, no agonist activity of rac-BHFF could be measured in GABAB receptors in which the GB1 TMD was replaced by a GB2 TMD or only the seventh helix of the GB1 (GB1-TM7) (Monnier et al., 2011; Figure 3E–F, Figure 3—figure supplement 1B and C), even though these constructs could still be activated by GABA efficiently (Figure 3E–F, Figure 3—figure supplement 1D). Of note, the conformational state of the GB1 TMD is not critical for both orthosteric and allosteric activation since similar results were obtained with the mutant GB1DCRC (Figure 3—figure supplement 1E and F). This mutant is activated by GABA similarly to the wild-type receptor (Figure 3—figure supplement 1E) as previously reported (Monnier et al., 2011). But it was engineered to create a disulfide bond in GB1 TMD, thus expecting to limit the conformational change of this domain upon ligand stimulation of the GABAB receptor. The importance of the intact GB1 TMD in the agonist activity of rac-BHFF was also confirmed by measuring the Go protein activation by BRET (Figure 3—figure supplement 2A). Finally, we obtained a similar conclusion regarding the requirement of both GB1 and GB2 TMDs for the ago-PAM activity of the two other PAMs by comparing all of them in the most sensitive IP1 assays (Figure 3—figure supplement 2B). Of note, for CGP7930 there is an apparent controversy between our data where both GB1 and GB2 TMD are required to observe an agonist effect, while Binet et al., 2004 found GB2 TMD alone was sufficient. It might be due to the endogenous expression of the GB1 subunit that may exist in some cell lines including the HEK293 cells used in both studies (Xu et al., 2014).

Figure 3 with 2 supplements see all
Both GB1 and GB2 transmembrane domains (TMDs) are sufficient for the agonist activity of the positive allosteric modulators (PAMs).

(A–F) Intracellular Ca2+ responses mediated by the indicated subunit compositions (pictograms) upon stimulation with rac-BHFF. The inserted graphs correspond to the responses upon stimulation with GABA. Data are normalized by using the response of 100 μM rac-BHFF or 1 mM GABA, for rac-BHFF and GABA treatment, respectively, on wild-type GABAB receptor and shown as means ± SEM of 3–8 biologically independent experiments. The dotted lines in the main and inserted graphs indicate the dose–responses of the wild-type receptor determined in panel (A).

GB1 and GB2 TM6s interface is the binding site for the different PAMs

A similar binding site in the structure of the purified GABAB was reported for rac-BHFF and GS39783 (Figure 4A; Kim et al., 2020; Mao et al., 2020; Shaye et al., 2020; Shen et al., 2021). This binding site of GS39783 was investigated using receptor mutants expressed at the surface of live cells (Shaye et al., 2020), and the mutagenesis data were consistent with the binding site observed in the structure. In the present study, we have analyzed both the potency and agonist efficacy of the three PAMs on a series of both GB1 and GB2 bearing single mutations in their TM6s. First, the potency of the rac-BHFF for each mutant in the intracellular calcium assay was measured in the presence of GABA at a concentration equivalent to its 20% maximal efficacy concentration (EC20). Two single mutants of GB1 (K792 ICL3A and Y8106.44A) and three single mutants of GB2 (M6946.41A, Y6976.44A, and N6986.45A) strongly impaired the potency of rac-BHFF for the GABAB in the calcium assay (Figure 4B and C). Residues were named according to the class C GPCR TMD nomenclature (Isberg et al., 2015). This loss of activity was not due to a loss of expression of the GABAB mutants at the cell surface as shown by the cell surface quantification of HALO-tagged GB1 labeled with the fluorophore Lumi4-Tb when coexpressed with GB2 (Figure 4D). In addition, the receptor remained functional for all constructs (Figure 4B and C, Figure 4—figure supplement 1A), although GABA had an impaired activity for the GB1 mutant K792 ICL3A (Figure 4B) and the GB2 mutants M6946.41A and Y6976.44A (Figure 4C). Finally, the critical importance of the two TM6s for the agonist activity of rac-BHFF was confirmed by IP1 measurements. rac-BHFF had a strong impaired efficacy on two GB1 mutants (Y8106.44A and 6.43MYN6.45-AAA) and one GB2 mutant (6.43MYN6.45-AAA) (Figure 4E). The data obtained in the intracellular calcium release assay were also consistent (Figure 4—figure supplement 2). This demonstrated that the interface of TM6s is crucial for the agonist activity of rac-BHFF effect. These mutagenesis data are consistent with the binding site of rac-BHFF at the interface between the two GABAB subunits, similar to GS39783 (Shaye et al., 2020), as reported in the GABAB structures (Kim et al., 2020; Mao et al., 2020; Shen et al., 2021).

Figure 4 with 2 supplements see all
GB1 and GB2 TM6s interface is the binding site for the different positive allosteric modulators (PAMs).

(A) Structure of the GABAB receptor (PDB 6UO8) where the binding site of GS39783 (PDB 6UO8) and rac-BHFF (PDB 7C7Q) in the receptor is highlighted, and close-up view of the molecules bound (PAMs shown as sticks, and hydrogen bonds between PAMs and receptor are depicted as dashed yellow lines). The α-carbon of the main residues involving in the binding site for these PAMs is highlighted as a sphere in GB1 (blue) and GB2 (light blue). (B, C) Intracellular Ca2+ responses mediated by the indicated constructs upon stimulation with rac-BHFF in the presence of EC20 GABA, or GABA alone. Data are normalized by wild-type response of 100 μM rac-BHFF + EC20 GABA or 100 μM GABA, for rac-BHFF and GABA treatment, respectively, and shown as means ± SEM of 4–5 biologically independent experiments. (D) Quantification of cell surface-expressed GB1 in HEK293 cells transfected with the indicated HALO-tagged GB1 and SNAP-tagged GB2 constructs after labeling with HALO-Lumi4-Tb. Data are normalized by wild-type receptor and expressed as means ± SEM. (E, F) Inositol-phosphate-1 (IP1) production induced by the indicated PAMs (E) or basal IP1 accumulation (F) in intact HEK293 cells expressing the indicated subunit combinations. Data are normalized by wild-type response and shown as means ± SEM of 4–5 biologically independent experiments. Data are analyzed using one-way ANOVA test followed by a Dunnett’s multiple comparison test to determine significance (compared with the WT) with *p<0.05, **p<0.005, ***p<0.0005, ****p<0.0001.

By using these GB1 and GB2 mutants bearing single mutations in their TM6s, we investigated the previously unknown mode of action of CGP7930. Interestingly, most of the mutations in GB2 strongly impaired the efficacy of CGP7930, whereas such effect was not observed for the mutations in GB1 (Figure 4E). Thus, the GB2 TM6 seems more important than the GB1 TM6 for CGP7930 compared to rac-BHFF. Of note, GS39783 is more sensitive to mutations than the two other PAMs. It might be because the mutated residues are highly important for the binding of GS39783 or alternatively to its weakest agonist activity compared to Rac-BHFF and CGP7930, then resulting in a stronger loss of agonist activity of GS39783 on these mutants. Altogether, the three PAMs have a different agonist activity on the GABAB mutants. Moreover, our mutagenesis data are consistent with the PAMs sharing the same binding pocket at the TM6 TM interface, even though their mode of binding does not seem to involve the same residues in GB1 and GB2.

Finally, we have analyzed the effect of the mutations on the constitutive activity of the GABAB receptor that was reported in transfected cell lines (Grünewald et al., 2002; Lecat-Guillet et al., 2017). Indeed, TM6 is expected to control the conformational landscape of the receptor and to play a key role for G protein activation in class A and B GPCRs (Weis and Kobilka, 2018). In GABAB receptor, contacts between the two TM6s were shown to stabilize the active state of the heterodimer (Kim et al., 2020; Mao et al., 2020; Shaye et al., 2020; Shen et al., 2021; Xue et al., 2019). These contacts are consistent with the allosteric interactions between the GB1 and GB2 TMDs during receptor activation (Monnier et al., 2011). In addition, genetic mutations in the GB2 TM6 (S6956.42I, I7056.52N, and A7076.54T) were reported to induce a high constitutive activity of the receptor (Vuillaume et al., 2018). Interestingly, in the present study most of the GB1 mutants produced a lower constitutive activity of the receptor compared to the wild-type (Figure 4F), for a similar cell surface expression (Figure 4D). In contrast, GB2 mutants had a similar constitutive activity as the wild-type receptor, even though the cell surface expression was lower. Of interest, this constitutive activity was blocked by the competitive antagonist CGP54626, on wild-type and most mutated receptors, but it remained unchanged for the triple mutant (6.43MYN6.45-AAA) in GB2 (Figure 4—figure supplement 1B). This suggested that these three mutations in GB2 have impaired the allosteric coupling between the two TMDs in the receptor.

Altogether, these results confirmed the important role of TM6s in controlling the landscape of conformations of the receptor and its basal state. Moreover, it suggests a key role for TM6 allosteric interactions between the GB1 and GB2 TMDs.

Exploring the GB2 TMD core to clarify the mechanism for agonist activity of the PAMs

How could ago-PAM binding at the active heterodimer interface induce G protein activation by GB2? Binding in the TM6s interface could directly stabilize the active state of the GB2 TMD. Alternatively, and not incompatible with this first mechanism, a second binding site in the GB2 TMD could exist.

A second binding site for the PAM, not yet discovered, could exist in the central core of GB2 TMD as previously proposed (Dupuis et al., 2006; Evenseth et al., 2020). Binding at this second site per se could be responsible for its agonist activity or alternatively could favor agonist activity of the compound bound in TM6s interface. Possible cavities in GB2 TMD have been revealed by the structures of the receptor, such as those occupied by one phospholipid molecule that covers nearly the entire range of ligand binding positions previously reported in class A, B, C, and F GPCRs (Kim et al., 2020; Papasergi-Scott et al., 2020; Park et al., 2020). This lipid is present only in the inactive conformation, and it was proposed to be important for intramolecular signal transduction in GABAB (Papasergi-Scott et al., 2020; Figure 5A). Mutations designed to destabilize phospholipid binding in GB2 TMD resulted in increases in both GABAB basal activity and receptor response to GABA. Therefore, to identify a potential second binding site for the PAM, we have introduced mutations in the upper part of the GB2 TMD, the region involved in this phospholipid binding (Figure 5—figure supplement 1A). Mutations in this region were shown to confer agonistic activity to GS39783 (Dupuis et al., 2006). Single and multiple amino acid substitutions were performed at seven positions in the GB2 (see details in Figure 5A, Figure 5—figure supplement 1A). We have also changed GB2 L5593.36, a residue critical in the ancestral binding pocket of GPCRs, into alanine. It is equivalent to the residue 3.32 in class A GPCRs (Ballesteros–Weinstein numbering scheme [Ballesteros and Weinstein, 1995]) that was shown to be conserved for direct interactions with agonists and antagonists (Venkatakrishnan et al., 2013). It is also well conserved in the class C GPCRs, including GABAB, where it plays an important role for the NAM and PAM binding and function (Doré et al., 2014; Farinha et al., 2015; Feng et al., 2015; Leach et al., 2016; Wu et al., 2014).

Figure 5 with 6 supplements see all
A deep region in GB2 transmembrane domain (TMD) is responsible for agonist activity of the positive allosteric modulator (PAM).

(A) Cartoon highlighting a possible second binding site (dotted oval) for the PAMs in the ancestral ligand binding pocket of the GB2 TMD. In the structure of the inactive state, this pocket is occupied by one molecule of phospholipid (shown as yellow sticks). The residues (α-carbon) of this phospholipid binding pocket (green) and the residues underneath (red) were changed to evaluate their importance in the agonist activity of rac-BHFF. The highly conserved residues L3.36 and Y5.44 were mutated into Ala; G6.53 conserved in GB2 were changed to Thr conserved at this position in GB1, or Phe conserved at this position for other class C GPCRs such as mGlu and CaSR; in Mut 8, the non-conserved residues 7.26NVQ7.28 were mutated in their equivalent in GB2 Drosophila; V7.35 conserved in GB2 (Val or Phe) was changed to Phe. (B) Intracellular Ca2+ responses mediated by the indicated GB2 mutants (M1 to M16) coexpressed with the wild-type GB1 subunit upon stimulation with 30 μM rac-BHFF or 1 mM GABA. Data are normalized by wild-type response and expressed as means ± SEM of three biologically independent experiments. (C) Intracellular Ca2+ responses mediated by the indicated constructs upon stimulation with rac-BHFF in the presence of EC20 GABA of each combination. Data are normalized by wild-type response of 100 μM rac-BHFF + EC20 GABA and shown as means ± SEM. (D) Basal inositol-phosphate-1 (IP1) accumulation mediated by the indicated constructs. Data are normalized by the response of the wild-type and shown as means ± SEM of five biologically independent experiments. Data are analyzed using one-way ANOVA test followed by a Dunnett’s multiple comparison test to determine significance (compared with the WT) with ***p<0.0005, ****p<0.0001. (E) Correlation between the GABAB constitutive activity measured using the Go protein BRET sensor and the rac-BHFF agonist effect for the WT GB1 subunit coexpressed with the indicated GB2 mutants. Data are normalized by the response of the wild-type and shown as means ± SEM. (F) Basal IP1 accumulation mediated by the indicated constructs, including the genetic mutation A567T identified in human GB2 that is equivalent to the mutation A566T in rat. Data are normalized by the response of the WT and shown as means ± SEM of 3–4 biologically independent experiments. Data are analyzed using an unpaired t-test for human, and one-way ANOVA test followed by a Dunnett’s multiple comparison test to determine significance (compared with the WT) for the rat, with *p<0.05, ***p<0.0005, ****p<0.0001. (G) Top view of the sodium binding pocket within the structure of human A2A adenosine receptor (PDB 4EIY) where the three residues important for Na+ interactions are highlighted (Cα in red), the equivalent residues identified in human GB2 TMD (PDB 6UO8), and human mGluR5 TMD (PDB 4OO9); the X.50 numbers shown for A2AAR are equivalent to the numbers in mGluR5 on the basis of X.50 residues defined in Doré et al., 2014; TM2, TM3, and TM7 that contain the residues involving the identified region are highlighted in cyan. (H) Basal IP1 accumulation mediated by the indicated WT and mutated mGlu5 receptors in the presence of the co-transfected glutamate transporter EAAT3. Data are normalized by the response of the WT and shown as means ± SEM of four biologically independent experiments. Data are analyzed using one-way ANOVA test followed by a Dunnett’s multiple comparison test to determine significance (compared with the WT), with **p<0.005. For clarity, the residue numbers for GB2 subunit are based on the sequence of rat GB2. Negative controls (Ctrl) are HEK293 cells co-transfected with the empty vector and Gαqi9 cDNA (B), or Gαo-Rluc and Gβ1γ2-Venus cDNAs in the absence of receptor (E).

PAM binding at the TM6s interface might be sufficient to stabilize the active state of GB2 TMD through an allosteric agonist effect. In this scenario, key regions in the GB2 TMD, also called microswitches, should play a key role to stabilize the active state, as well as reported in class A GPCRs (Zhou et al., 2019). These microswitches regions remain largely unknown in the class C GPCRs due to the lack of high-resolution structures in the active and inactive states (Gao et al., 2021; Koehl et al., 2019; Lin et al., 2021; Mao et al., 2020; Seven et al., 2021; Shaye et al., 2020; Shen et al., 2021). We explored a region underneath the phospholipid cavity, named ‘deep region’ in this study, that corresponds to the Na+ binding pocket in most class A GPCRs (Zarzycka et al., 2019). This region could be reached by synthetic allosteric modulators, as observed by the mGlu5 NAMs that directly interact in this deep region (Doré et al., 2014). We changed three residues in GB2 (G5262.46, A5663.43, and T7347.43), where A5663.43 interacts directly with the phospholipid (PDB 6WIV) (Figure 5—figure supplement 1B), and that are equivalent to the residues that bind Na+ in class A GCPRs (2.50, 3.39, and 7.49) (Zhou et al., 2019). Since this Na+ binding pocket collapses during class A GPCR activation, GB2 G5262.46 (C2.46 in GB1) and A5663.43 (G3.43 in GB1) were changed for residues larger to fill the cavity (Cys and Phe, respectively). The side chain of T5347.43, conserved in GB1, was changed to Ala to prevent these conformational changes during GB2 TMD activation. Single and multiple mutants were analyzed (Figure 5B).

The region of the GB2 TMD equivalent to the sodium binding site in class A GPCRs is critical for allosteric agonism

We have tested the capacity of rac-BHFF to activate the GB2 mutants described above when coexpressed with the wild-type GB1 subunit using the intracellular calcium (Figure 5B) and IP1 accumulation (Figure 5—figure supplement 2A) assays. rac-BHFF retained its agonist activity for all GABAB mutants, except for two that contains mutations underneath the lipid binding site, GB2 G5262.46C/A5663.43F (M14) and G5262.46C/T7347.43A (M15). The loss of effect of rac-BHFF was not due to the lack of cell surface expression of the mutant receptors (Figure 5—figure supplement 3A and B), nor to their loss of function since they were activated by GABA with a similar efficacy as the wild-type receptor (Figure 5—figure supplement 3C and D). The agonist activity of CGP7930 was also impaired by these two mutants in the intracellular calcium (Figure 5—figure supplement 3E) and IP1 accumulation assays (Figure 5—figure supplement 2A). Most importantly, these mutated residues are not critical for the binding of the PAMs. Indeed, rac-BHFF increased intracellular calcium release on the mutants M14 and M15 in the presence of GABA with a potency similar to the wild-type receptor (Figure 5C), and improved potency of GABA (Figure 5—figure supplement 4A), an effect also observed for CGP7930 and GS39783 (Supplementary file 2). Altogether, these results showed that the deep region of the GB2 TMD is required for the agonist activity of the PAMs, but it is not critical for their allosteric modulation effect. It also indicates that this region is not involved in the direct binding of these allosteric compounds.

The deep region in the GB2 TMD controls constitutive activity and is involved in human genetic diseases

GABAB is known to display a significant constitutive activity (Grünewald et al., 2002). We have measured this constitutive activity for all the GB2 TMD mutants (Figure 5—figure supplement 2B). In contrast to the GB2 mutations in the phospholipid binding site, those in the ‘deep region’ produced a GABAB receptor with a constitutive activity strongly impaired compared to the wild-type receptor for a similar cell surface expression (Figure 5—figure supplement 3A). The strongest reduction of the constitutive activity was for the GB2 mutants M14 and M15, bearing mutations in the deep region, as measured with both IP1 accumulation (Figure 5D, Figure 5—figure supplement 2B–D) and BRET (Figure 5—figure supplement 4B) assays. It revealed an important role of this region in controlling the basal activity of the GABAB receptor.

Interestingly, the maximal agonist activity of rac-BHFF correlated with their constitutive activity (Figure 5E), demonstrating a possible link between allosteric agonism and constitutive activity. To better understand this relationship, we modeled the agonist activity of PAMs in the GABAB by developing a mathematical model (Figure 5—figure supplement 4C), based on our previously reported mechanistic model for the mGlu receptors (Rovira et al., 2008). This previous model was simplified: the extracellular domains were not considered as their effects will be reflected in the basal state of the TMDs. Only one binding site for the allosteric agonist was used and only the GB2 TMD was considered to couple to G proteins. This model coincides then with the two-state model of receptor activation (Leff, 1995). It integrates the constitutive activity (α) of the GABAB as well as the binding affinities of the ago-PAM for the active and inactive states of the TMD. According to this model, when the constitutive activity is very low (α = 100), the studied PAMs do not efficiently activate the receptor (Figure 5—figure supplement 4C and D). This is in agreement with the loss of agonist activity of ago-PAMs on the mutants M14 and M15.

Genetic mutations responsible for human brain diseases such as Rett-like phenotype (Lopes et al., 2016; Vuillaume et al., 2018; Yoo et al., 2017), infantile epileptic spasms (Hamdan et al., 2017), and epileptic encephalopathy (Hamdan et al., 2017; Yoo et al., 2017) were identified in GABBR2 (which encodes GB2), with most of them resulting in missense mutations in TM3 and TM6. Interestingly, one of the human genetic mutations involved in Rett-like phenotype, GB2 A5673.43T, is located in the deep region and also increases receptor constitutive activity. It corresponds to the rat GB2 mutant M12 (A5663.43F) that displays a lower constitutive activity (Figure 5F). This suggested that depending on the residue at this position the constitutive activity of the receptor can be tuned up or down. This further illustrates the role of the deep region in controlling the conformational landscape of the GABAB.

The deep GB2 TMD region controls the constitutive activity of other class C GPCRs

The role of this deep region in the controlling of constitutive activity is reminiscent to the role of the equivalent region in class A GPCRs reported to control their conformational landscape (Manglik et al., 2015; Ye et al., 2016). This is well illustrated by the role of Na+ ions in many class A GPCRs that bind at this topologically equivalent site (Katritch et al., 2014; Ye et al., 2018; Zarzycka et al., 2019; Figure 5G). It is also illustrated by the mGlu5 NAMs that interact directly with the same three positions in TM2, TM3, and TM7 in mGluR5 (I2.46, S3.43, and A7.43) (Doré et al., 2014), and mGlu4 PAMs (Rovira et al., 2015; Figure 5G). We have then investigated how mutations of these three residues in mGluR5, equivalent to the residues in GB2 subunits (G2.46, A3.43, and T7.43) but not conserved, can influence its constitutive activity. While two mutations did not change the constitutive activity of mGluR5, the mutation A8127.43F increased it (Figure 5H, Figure 5—figure supplement 4E and F). It suggests that this region is also controlling the conformational landscape of the mGlu5 receptor. But our data show that it is not possible to predict if mutations of these residues in this deep region will produce or not a change in the constitutive activity. Further studies will be necessary to generalize the role of this deep region to the mGlu receptors.

Discussion

It is a challenging issue to control GPCR dimers’ activity with ligands, without acting at the individual monomers. This requires to develop compounds that are not acting at the traditional orthosteric or allosteric binding pockets in the TMD or on the extracellular domain. Here, we used the GABAB as a model to demonstrate that allosteric modulators can bind in the TMD interface to specifically control GPCR dimer activity. Since these ligands bind to residues that belong to the two protomers, they are unable to act on the individual GABAB subunits, either GB1 or GB2, that were reported to exist separately (Chang et al., 2020). Our study did not identify other binding sites for these PAMs, including those in the GB2 TMD core as it was previously proposed by different groups (Binet et al., 2004; Evenseth et al., 2020; Urwyler, 2011). Our study also reveals that the agonist activity of these PAMs requires a key region in the GB2 TMD core, which is also controlling the constitutive activity of the receptor. This region corresponds to the sodium binding pocket in class A GPCRs.

This extra-TM bundle binding site involving the two TM6s in the GABAB is novel in the GPCR family. A growing number of allosteric binding sites were discovered outside of the TMD at the receptor-lipid bilayer in class A GPCRs (Thal et al., 2018; Chang et al., 2020), but TM6 is usually not directly involved, with the exception of class B GPCRs (Chang et al., 2020). An important feature to note for this GABAB TM6s binding pocket for PAMs is that it forms only in the active state. In addition, it is rather tight (volume of 729 Å3) and then it accommodates only small compounds (volumes of 268 Å3 and 260 Å3 for GS39783 and rac-BHFF, respectively). Our study also reveals that the different PAMs have different binding modes. The CGP7930 agonist effect was only impaired by mutations in GB2 TM6 and not by those in the GB1 subunit, suggesting its binding involves mostly GB2. In contrast, rac-BHFF and GS39783 effects were impaired by mutations in GB1 and GB2 TM6s, indicating that both subunits of the GABAB are involved in the binding of these compounds.

The binding mode of PAMs in the GABAB is novel within the class C GPCRs where all allosteric compounds that bind in the TMD are found in the ancestral TMD binding pocket, as revealed by the structures of mGlu1 (Wu et al., 2014), mGlu2 (Lin et al., 2021; Seven et al., 2021), mGlu5 (Doré et al., 2014), and the calcium sensing receptor CaSR (Gao et al., 2021) receptors in complex with allosteric modulators. Similar conclusions were reached for PAMs and NAMs from structure–activity and modeling studies of mGlu (Bennett et al., 2020; Lundström et al., 2017; Pérez-Benito et al., 2017; Rovira et al., 2015) and other class C GPCRs (Leach and Gregory, 2017; Leach et al., 2016). In particular, class C receptors with a deletion of its ECD have shown that PAMs and NAMs bind in the TM bundle, like class A GPCR agonists, and the PAMs behave as agonists on these constructs (Goudet et al., 2004). However, in GABAB the presence of one phospholipid that occupies the ancestral binding pocket both GB1 and GB2 TMD (Kim et al., 2020; Papasergi-Scott et al., 2020; Park et al., 2020) could preclude the possible binding of the PAMs. These phospholipids were proposed to play a role in stabilizing the inactive conformation of the receptor, especially of the GB2 subunit, acting as a NAM (Papasergi-Scott et al., 2020; Park et al., 2020).

Our study identified a key region where three residues (G2.46, A3.43,and T7.43) are important for the constitutive activity of the GABAB. They are highly conserved during evolution in GB2 subunits (Figure 5—figure supplement 5), suggesting that they play a key role in maintaining the resting conformation of the receptor, thereby limiting its basal activity. In mammalian GB1, the residues are slightly different (C2.46, G3.43, and T7.43), but they are also highly conserved. The importance of these residues in stabilizing the resting state agrees with their direct interaction with the phospholipid (A3.43 in GB2 and T7.43 in GB1) (Kim et al., 2020; Papasergi-Scott et al., 2020; Park et al., 2020; Figure 5—figure supplement 1B). It also corresponds to the equivalent and highly conserved residues D2.50, S3.39, and N7.49 of the class A GPCRs, which are responsible for sodium binding and stabilization of the inactive conformation (Zarzycka et al., 2019). But this sodium binding pocket is no longer accessible for the Na+ ions in the active state due to a slight rearrangement of these residues between the active and inactive states (Zhou et al., 2019). In GABAB, the binding of an ion in this region is most probably excluded since most of the residues are hydrophobic. But a significant rearrangement between the inactive and active state in the GB1 and GB2 TMD might occur in this region, even though only slight changes are observed in the available structures (Figure 5—figure supplement 6).

The present study also reveals key information to propose a model for the molecular mechanism of activation of the GABAB (Figure 6). First, GB1 TMD plays a key role for the agonist activity of PAMs by providing a binding site at the TM6s interface. In the absence of orthosteric agonist, most probably a fraction of the molecules adopts this TM6s interface state, as observed for the mGlu2 receptor that constantly oscillates between the inactive and active conformation (Cao et al., 2021). This TM6s interface provides a binding site for the PAM that further stabilizes this active state. It could explain the agonist activity of the PAM in the absence of another agonist. And in the presence of orthosteric agonist that stabilizes the active TM6 interface (Mao et al., 2020; Shaye et al., 2020; Xue et al., 2019), and also of the coupled G protein (Shen et al., 2021), the binding site of the PAM would be favored, then facilitating its agonist and allosteric activities. The PAM would then act as a glue to stabilize the active interface and the active state of the GABAB receptor. In this simple model, it is difficult to rule out that the PAM induces a new conformation that would not be stabilized by the orthosteric ligands. Finally, the position of both TM6s stabilized by the bound PAM is also consistent with the absence of movement of GB2 TM6 during receptor activation since there is no need to open a cavity for the G protein to bind (Shaye et al., 2020; Shen et al., 2021). This is a major difference with the other classes of GPCRs (A, B, and F) where a large movement of the intracellular part of TM6 is observed during receptor activation (Hilger et al., 2020; Kozielewicz et al., 2020). Second, the deep region identified in GB2 TMD serves as a microswitch for ago-PAM to stabilize the active state of GB2 leading to G protein activation. This region plays a key role for both the constitutive activity and allosteric agonism. But one cannot rule out that the mutations in this region of GB2 stabilize the inactive conformation of GABAB. Finally, since a TM6 active interface is also observed in the mGlu and CaSR dimers (Gao et al., 2021; Koehl et al., 2019; Ling et al., 2021; Liu et al., 2020; Wu et al., 2014; Xue et al., 2015; Zhang et al., 2020), it remains to be established if a similar molecular mechanism of activation could be conserved in the mGlu-like class C GPCRs.

Molecular mechanism of GABAB receptor activation and allosteric modulation.

(A) The orthosteric agonists bind within the GB1 VFT and induce a rearrangement of ECD dimer. This conformational change stabilizes the active state of GB2 transmembrane domain (TMD) via both the stalk of GB2 and the interactions between the two TMDs through the TM6 dimer interface. This interface in the active state is further stabilized by the positive allosteric modulators (PAMs), then enhancing the potency and affinity of the orthosteric agonists. In the present study, we show that allosteric agonism requires a region (residues in red) responsible for basal activity of the receptor. (B) This molecular mechanism is further illustrated by pictograms with the WT receptor and with the mutated GB2 TMD deep region. These cartoons further highlight the GB1 TMD that is proposed to serve as a lever for the activation of the receptor both by orthosteric agonists and PAMs.

In conclusion, our study highlights a distinct mode of action of the PAMs in the GABAB in a pocket at the TM6 interface that form only in the active state. We demonstrate the importance of the constitutive activity of the GB2 TMD for allosteric agonism through the TMD dimer interface. Our study reveals possibilities of developing novel allosteric modulators for the GABAB and other class C GPCR dimers through the TM6 interface. Ligands acting at the dimer interface may potentially be interesting tools also for other GPCRs, even if they generally form transient dimers.

Materials and methods

Materials

GABA was purchased from Sigma. rac-BHFF, CGP7930, and GS39783 were obtained from Tocris Bioscience. Lipofectamine 2000 and Fluo-4 AM were from Thermo Fisher Scientific. Coelenterazine h was purchased from Promega. Fetal bovine serum (FBS), culture medium, and other solutions used for cell culture were from Thermo Fisher Scientific.

Plasmids and transfection

Request a detailed protocol

The pRK5 plasmids encoding N-terminal HA-tagged wild-type rat GB1a, GB1ASA, GB1/2, GB1TM7, GB2/1, GB1ΔECD, and GB1DCRC, N-terminal Flag-tagged wild-type rat GB2 and GB2ΔECD, were described previously (Monnier et al., 2011). The mutations of GB2 TMD in the pRK5 plasmid were generated by site-directed mutagenesis using QuikChange mutagenesis protocol (Agilent Technologies). The pRK5 plasmid encoding the N-terminal Flag-tagged wild-type rat mGluR5 was previously described (Goudet et al., 2004), and the mutations were generated by site-directed mutagenesis.

HEK293 cells (ATCC, CRL-1573, lot: 3449904) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS. They were tested negative for mycoplasma contamination. They were transfected by electroporation or by the Lipofectamine 2000 protocol as described elsewhere (Lecat-Guillet et al., 2017). For different mutations within GB2 TMD, 10 million cells were transfected with 8 μg of each plasmid of interest with 1 μg of wild-type GB1a and completed to a total amount of 10 μg with the plasmid encoding the pRK5 empty vector. For the different combinations of chimeric GABAB receptor, 2.5 μg of different GB1, 5 μg of different GB2 completed to a total amount of 10 μg with the plasmid encoding the pRK5 empty vector were used as indicated. To allow efficient coupling of the receptor to the phospholipase C pathway, cells were also transfected with the chimeric G-protein Gαqi9 (1 μg). For cell-surface expression and functional assays of indicated subunits, experiments were performed after incubation for 36 hr (12 hr at 37°C, 5% CO2 and then 24 hr at 30°C, 5% CO2). For the different mutations within mGluR5 TMD, 10 million cells were co-transfected with 6 μg of wild-type or mutant mGlu5 and 3 μg of EAAT3 (also known as EAAC1) and completed to a total amount of 10 μg with the plasmid encoding the pRK5 empty vector. To eliminate the potential effects of the L-glutamine within classical DMEM, culture medium was replaced by DMEM GlutaMAX 24 hr before experiments.

Cell surface quantification

Request a detailed protocol

Cell surface expression of the indicated subunits was detected by ELISA as previously described (Liu et al., 2017). Briefly, HA-tagged different GB1 and Flag-tagged different GB2 were co-transfected into HEK293 cells seeded into 96-well microplates. Cell surface expression and total expression (treated with 0.05% Triton X-100) were detected with the monoclonal rat anti-HA antibody (3F10, Roche) or monoclonal rat anti-Flag (A8592, Sigma) antibody coupled to HRP. Bound antibody was detected by chemiluminescence using SuperSignal substrate (Thermo Fisher Scientific) and a multi-mode microplate reader (FlexStation 3, Molecular Devices). For the mGluR5, Flag-tagged mGlu5 subunit and transporter EAAT3 were co-transfected into HEK293 cells seeded into 96-well microplates.

Cell surface expression of the HALO-tagged GB1 was detected after labeling with the fluorophore Lumi4-Tb, as previously reported (Scholler et al., 2017). Briefly, 24 hr after transfection, HEK-293 cells were labeled with 100 nM HALO-Lumi4-Tb in Tag Lite buffer (PerkinElmer Cisbio) for 1 hr at 37°C. After three washes with Tag Lite buffer, the emission of Lumi4-Tb was measured at 620 nm on a PHERAstar FS microplate reader (BMG LABTECH, Ortenberg, Germany).

Intracellular calcium release measurements

Request a detailed protocol

Intracellular Ca2+ release in HEK293 cells was performed as previously described (Lecat-Guillet et al., 2017). Briefly, transfected cells in the 96-well plate were washed once with HBSS buffer (20 mM HEPES, 1 mM MgSO4, 3.3 mM Na2CO3, 1.3 mM CaCl2, 0.1% BSA, and 2.5 mM probenecid) and pre-incubated with 1 μM Ca2+-sensitive Fluo-4 AM (Thermo Fisher Scientific) prepared in the HBSS buffer for 1 hr at 37°C. Cells were washed once with HBSS buffer and 50 μl of this buffer was added into the wells. The fluorescent signals (excitation at 485 nm and emission at 525 nm) were then measured at intervals of 1.5 s for 60 s after adding of 50 μl of the indicated compounds 20 s after the first reading by the microplate reader (FlexStation 3, Molecular Devices). The Ca2+ response was given as the agonist-stimulated fluorescence increase, normalized according to the indication. Concentration–response curves were fitted using ‘log(agonist) vs. response -- Variable slope (4 parameters)’ by GraphPad Prism software.

Inositol phosphate measurements

Request a detailed protocol

IP1 accumulation in HEK293 cells co-transfected with the indicated subunits was measured by using the ‘IP-One Gq assay kits’ purchased from PerkinElmer Cisbio according to the manufacturer’s recommendations. In brief, the stimulation buffer with indicated compounds was added into the 96-well microplates and microplates were incubated in the incubator (37°C, 5% CO2) for 30 min. After adding the IP1-d2 and anti-IP1 terbium cryptate conjugate reagents, the microplate was kept in a dark place for 1 hr at room temperature before being detected by the Multi-mode plate reader (PHERAstar FSX, BMG LABTECH). For the mGluR5, Flag-tagged mGlu5 subunit and EAAT3 were co-transfected into HEK293 cells seeded into 96-well microplates.

BRET signal measurements

Request a detailed protocol

The BRET sensor used here is composed of GαoA fused to Rluc (GαoA-Rluc), Gβ1 and Gγ2 fused to Venus (Gγ2-Venus), which will lead to a BRET signal decreased upon activation of the G protein. BRET measurements were recorded after indicated compounds stimulation on the Mithras LB 940 plate reader (Berthold Biotechnologies, Bad Wildbad, Germany) as previously described (Comps-Agrar et al., 2011).

Binding pocket volume calculations

Request a detailed protocol

The calculation of the binding pocket volume was performed with Discovery Studio (BIOVIA, Dassault Systèmes, v20.1.0.19295, San Diego; Dassault Systèmes, 2019). Briefly, the ECD of the full-length GABAB crystal structure (PDB 6UO8) co-crystallized with GS39783 was truncated and only the TMDs were kept for further analysis. Then, an automatic identification of the binding site was performed from the receptor cavities using the ‘Define and Edit Binding Site tools’ from the program with the default parameters except the site opening that was set to 8 Å. From all sites identified, the one covering the binding site of PAMs was selected. Then, a structural alignment was performed with the structure co-crystallized with rac-BHFF (PDB 7C7Q). Finally, only those site points at less than 4.5 Å from either GS39783 and rac-BHFF were kept and visually verified. The final volume of the pocket was finally calculated as the product of number of site points and the cube of the grid spacing (0.5 Å).

Statistical analysis

Request a detailed protocol

Data are presented as means ± SEM, and all nonlinear regression analyses of the concentration response curves were performed using GraphPad Prism software. Activation concentration–response curves were fitted to the default equation of ‘Log(agonist) vs. response -- Variable slope (4 parameters)’ within the software, as Y = Bottom + (Top-Bottom)/(1 + 10^((LogEC50-X)*Hill slope)) with no constraints for the 4 parameters (X, log of concentration; Y, response; Top and Bottom, plateaus in same units as Y; LogEC50, same log units as X; and Hill slope, unit less) in the equation. For activation effects, the ‘Top’ values in the equation are considered as the ‘Emax’ for the ligands. Comparison of parameters between different conditions was determined using one-way ANOVA test followed by a Dunnett’s multiple comparisons test. For the comparison only containing two members, the unpaired Student’s t-test was performed. All statistical analyses were performed by GraphPad Prism software, and p<0.05 were considered statistically significant.

Data availability

Figure 2- Source Data 1 contain the numerical data used to generate the figures; Figure 3 - Source Data 1 contain the numerical data used to generate the figures; Figure 4 - Source Data 1 contain the numerical data used to generate the figures; Figure 5 - Source Data 1 contain the numerical data used to generate the figures.

References

    1. Schoneberg T
    2. Liebscher I
    (2021) Mutations in G Protein-Coupled Receptors: Mechanisms
    Pathophysiology and Potential Therapeutic Approaches. Pharmacol Rev 73:89–119.
    https://doi.org/10.1124/pharmrev.120.000011

Decision letter

  1. Andrew C Kruse
    Reviewing Editor; Harvard Medical School, United States
  2. Richard W Aldrich
    Senior Editor; The University of Texas at Austin, United States
  3. Jonathan A Javitch
    Reviewer; Columbia University, United States
  4. Aashish Manglik
    Reviewer; University of California, San Francisco, United States

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

Decision letter after peer review:

Thank you for submitting your article "Allosteric ligands control the activation of a GPCR heterodimer by acting at the transmembrane interface" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by Andrew Kruse as the Reviewing Editor and Richard Aldrich as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Jonathan A Javitch (Reviewer #1); Aashish Manglik (Reviewer #3).

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

Essential revisions:

1) The reviewers all noted that in a number of places the claims are slightly overstated relative to the underlying data, for example in implying the generalizability of the model to G protein coupled receptors outside of family C. The authors should revise the text to temper their claims and clearly indicate for readers which aspects of the proposed model are direct deductions from experimental results and which aspects are more speculative. Please see detailed reviews below for specific areas to revise.

2) In some places certain expected data are missing, for example the pharmacological characterization is more extensive for rac-BHFF than for CGP7930 or GS39783. Is there a reason that all compounds were not profiled similarly? If so, please explain. If not, then including equivalent data for all three compounds would be preferred.

Reviewer #1:

The authors explore the role of an allosteric modulator binding pocket identified in recent structures at the TMD interface of the two GABAB receptor protomers associated with three ago-PAMs and undercover the important role of the TM6s in receptor activation by these PAMs but not by GABA itself. They also identify a key region deep in the GB2 TMD core that is critical not only for the allosteric agonism of these ago-PAMs but also for the constitutive activity of GABAB receptor. Based on these findings, the authors propose a model for the activation of GABAB receptor.

The experiments and their presentation are generally clear and convincing, although there is a tendency in the language to overinterpret the role of mutations as "proving" hypothesis about binding sites rather than supporting the hypothesis, and there is no discussion about a potential role of indirect effects of the mutations. The data are strong and supportive of the inferences drawn, but recognition that they are inferences and not "proof" per se would improve the manuscript further as discussed below in more detail.

By evaluating the agonist activity (ago-PAM) and the positive allosteric modulatory actions of three typical PAMs (rac-BHFF, CGP7930 and GS39783), the authors show both similarities and differences in their functional properties. They show that the GB1 TMD is crucial for the agonist activity of rac-BHFF. Although they also claim that the agonist activity of CGP7930 is also dependent on the GB1 TMD, there are inconsistencies between some of their results and previous literature that need further clarification, as noted below. Through analyzing the potency and agonist efficacy of the three PAMs on a series of both GB1 and GB2 bearing single mutations, the authors provide evidence that the binding sites of rac-BHFF, CGP7930 and GS39783 are at the interface between the two GABAB subunits and that the interface of TM6s is crucial for both their agonist and PAM activity. They use language suggesting that mutations can "prove" that the ligands bind in this position, which should be modified to show that the data support the hypothesis and are consistent with previous crystal structures and at least provide some mention of a caveat that mutagenesis results can of course be confounded by indirect effects and don't prove or even show or demonstrate that this is the binding site but rather support the inference.

The authors also identify a "deep region" within the GB2 TMD core and provide evidence suggesting that it plays a role in stabilizing the active state and is critical for allosteric agonism and constitutive activity. These results are interesting and discussed in the context of the sodium site in Family A receptors. As noted below, there seems to be some selection of the mutants studied here and some additional data would strengthen the argument as discussed in more detail below.

Altogether, this study highlights a distinct mode of action of the PAMs in the GABAB, in a pocket at the TM6 interface that appears to form only in the active state and demonstrates the importance of the GB2-TMD for allosteric agonism through the TMD dimer interface.

Major issues:

1. The authors primarily use two different methods (intracellular ca2+ release or IP1 accumulation) to determine the agonist or allosteric modulator activity of three PAMs, and intermittently add BRET data on Go activation. There is one sentence early on about the IP1 assay being more sensitive because of the equilibrium nature of the assay as opposed the kinetic nature of the than ca2+ release. The sensitivity of the BRET assay is not discussed. Does the BRET assay reveal agonism for GS39783 as seen in the IP1 assay but not the ca2+ release assay? In many figures, the authors only present data from one assay, although sometimes the results are confirmed with another assay. Some description or discussion about the difference or limitation of these two methods and why they are used as they are would be helpful. It is not clear whether the entire data set should be provided in the supplement, but some logic as to why the assays are being used as they are seems important. Furthermore, when loss of function for rac-BHFF is shown by mutations, it would seem prudent to include data for the more sensitive IP1 assay and not only ca2+ release.

2. Although the authors claim that they evaluate the agonist activity and the allosteric modulation of all three PAMs, rac-BHFF is the main target of their designed experiments. In some critical results, the data about CGP7930 and GS39783 especially the latter is absent. Since these three PAMs have different functional properties and proposed molecular mechanisms, their data should be presented along with those for rac-BHFF, especially when supporting the conclusion that they have same binding pocket at the TM6s interface.

3. In Figure 3 supplement 2B, the authors show that CGP7930 has no agonist function in the GB1/2-GB2 heterodimer and the GB2 monomer in the IP1 accumulation assay. This conflicts with the data in the paper published in 2004 (Virginie Binet et al., JBC, 2004), which show clearly that CGP7930 can activate GB1/2+GB2 and GB2 monomer without GABA. The authors do not comment upon this discrepancy and how it might be related to the different assay systems – the results seem directly contradictory to this earlier work. Notably the authors select the less sensitive ca2+ release assay for their studies of rac-BHFF in this figure. I think it important that these critical data be validated in the IP1 assay as well to establish whether there is any observable agonism in this assay or the direct agonism of the compound is completely gone. Are these issues of assay sensitivity and size of effect?

Additional issues:

1. Figure 1A, the structure panel, is very confusing and the legend does not match the figure. Which structure is shown here or it is a mixture of 4 GPCRs structure? I think it would be helpful to identify the different binding sites somehow so that they can be associated with the specific PDB structures as this is a helpful composite figure if made clearer and with an appropriate legend.

2. In figure 4B and 4C, the curves are very difficult to see. Furthermore, the effects reflect a mixture of agonist and allosteric activity of rac-BHFF. To determine the PAM activity of rac-BHFF, the GABA dose curves in present certain or multiple concentration of rac-BHFF should be presented as in figure 3—figure supplement 1D.

3. Line 156, in Figure 4E, GB1-Y810A and GB1-MYN-AAA only show the TM6s are crucial for the agonist activity of rac-BHFF. This evidence cannot support that GB1 mutants can reduce the PAM activity of rac-BHFF. PAM activity should be removed from the text unless parallel experiments with GABA are shown.

4. Line 162, The data in Figure 4E can only demonstrate CGP7930 effects were impaired by the GB2 mutations but not by GB1 mutations; we cannot make a conclusion that it can activate GB2 alone from these data. This conclusion is also in conflict with the data presented in Figure 3 and its supplemental data where the GB1-TM is shown to be critical even for CGP7930 (Figure 3 S2B).

5. At line 131 there is a conclusion about the GB1DCRC mutant that is locked in an active state still being able to support allosteric activation. Of note, it is also able to support GABA activation. This should be added to text, along with a brief explanation of what it means to be locked in the inactive state if both orthosteric and allosteric activation is intact.

6. Line 152 – there seem to be substantial effects of some of these mutations on GABAB activation, not just the 2 mutations noted.

7. Line 100 – if the lipid is present only in the inactive conformation, then why in the second half of the sentence is it proposed to be important for signal transduction. This should be clarified.

8. Line 244-246 – is it conceivable that the compounds binding in the TMD of GB1 to exert their effects and that they can no longer bind to the deep region in GB2?

9. Line 249 – do any of the green mutations affect constitutive activity? M11, 12 and 13 have effect but emphasis is only on M14 and M15. Would be nice to see all the mutations in this analysis. Are the effects limited to the deep residue mutations or is this more complex?

10. Line 272 – the language is confusing since the different mutations have opposite effects on constitutive activity and this should be clarified.

11. Line 284 -the effect shown in Mglu5 is opposite that shown in GABAB for the same mutation. And two mutations are completely tolerated in the mGlu5 despite their negative effects on GABAB. Caution in generalizing seems prudent here, and I think these are the weakest data in the paper and either should be deleted or strongly caveated. There are many more differences than similarities here.

12. I don't think the concept of GB1 TM domain acting as a lever is a useful or justifiable metaphor. A lever is "a rigid bar resting on a pivot, used to help move a heavy or firmly fixed load with one end when pressure is applied to the other." How agonist binding is conveyed to GB2 and how a PAM affects this is quite complex is terms of the actual physical forces, and I don't believe any of the data here allow this conclusion. Moreover, the authors write "GB1 TMD serves as a lever to stabilize the active state of GMB2 TMD…" As per the definition quoted above, I am not sure how a lever can stabilize. I would remove the lever metaphor altogether. It is also incompatible with the next statement: "This lever model is compatible with the restriction of the TM6s in the active interface where both TM6s do not move substantially between inactive and active states".

This is a matter of taste, but in Figure 2 red and orange are used in B and C for the different PAMS – then in D-F the same colors are used to represent the PAM concentration. I think this is confusing. I would prefer to see consistency in the color schemes in the same and across figures, but this is obviously just my taste.

As best I can tell, Figure 2B red and Figure 2 supplement 1A show the same experimental design, but in Figure 2B a lower range of concentration is shown and the curve doesn't reach the Emax. In the supplemental data, the curve goes to higher concentrations and reaches the Emax, although only two independent experiments are shown. Would have been nice to have a full curve in the main figure. Regardless the legend in the supplemental figure should clarify that it is the same type of experiment as 2B as not clear now why the same experiment is shown twice.

Figure 3 legend should explain the dotted lines.

Figure 5 In panel E I would use the mutant names that are used in all the other panels. I would include all the mutants from panel B in E so that we can see the relationship for both the green and red mutants. Does the relationship hold for all or just the subset shown?

There are a lot of surface expression data. It would be nice to help the reader by adding panels in the supplements showing normalized emax throughout to make it easier for the reader to see the impact of expression – now one can only see that expression isn't dramatically affected but too hard to do it one at a time by mutant.

Reviewer #2:

GPCRs are major targets for drug development. A current promising avenue for GPCR drug development is to identify strategies to affect receptor signaling beyond simply targeting the endogenous ligand binding site. This would provide larger surface area to explore and better subtype specificity. Allosteric modulators do this by binding elsewhere in the receptor and enhance or attenuate effect of the orthosteric ligand. The general goal of this research was to investigate the mechanism of such effect in GABAB receptors. Finding the mechanism could inform development of future pharmaceutical drugs. GABAB receptors are members of class C GPCRs. This small family of GPCRs and is somewhat unique in that class C receptors form constitutive dimers. Recent structures of GABAB receptors from 4 different groups showed that 2 GABAB receptor PAMS (rac-BHFF and GS39783) bind at the dimeric interface which is unique among GPCRs. In this study, authors first verified this by performing functional assays and mutagenesis. They also showed the previously described constitutive activity of GABAB receptors and the direct agonist effect of these PAMs (called ago-PAM). Using mutation and structure-function approaches they showed the role of TM6 in both of these properties. Next, they authors tried to use similar assays to investigate the mechanism of agonist activity and modulation by these PAMs. Their results demonstrate the contributions of an internal cavity within the 7-transmembrane domain of GB2, dubbed "deep region", in the constitutive activity of GABAB receptor and the agonist effect of the PAMs but not the allosteric modulations.

While these results provide an interesting twist to how allosteric modulators are perceived and that the pharmacological classification of compounds can be intertwined and not orthogonal, it provides few new mechanistic understanding. For example how does PAMs affect activation (does it modify the energy landscape? or induce a new conformation?) and is it possible to synthesize a PAM that is only a modulator and does not have direct agonist effect? Moreover, the methods section lacks practical details.

Specific comments:

Line 110 – Is it possible the effect of GS39783 in the IP1 assay is due to nonspecific effect? Control titration experiments for figure 2B with cells transfected only with the chimeric G protein or the individual GABAB subunits + the chimeric G proteins would be highly relevant to this data.

Figure 3, figure supplement 1- While it is generally suggested that the ER retention signals inhibit trafficking of GABAB1 subunit to the surface, depending on expression level and cell type this could be leaky. Authors should provide GABAB1 only and GABAB2 only surface expression levels.

Figure 4, panel H. Previous work from authors and others (for example Koehl et al., 2019) has shown that for experiments on mGluR5, co-expression of neuronal excitatory amino acid transporter 3 (EAAT3) is necessary to be able to characterize the effect of modulators correctly. Without that, modulators can appear to show intrinsic agonist activity (for example Koehl et al., 2019). In the data in this panel, EAAT3 was not co-expressed.

Authors have previously shown that GABAB receptors form extensive dimers of dimers and have characterized those. What is the effect of those higher order oligomers on the observed PAM effect?

Line 166 – This explanation for the discrepancy is not clear. The different PAMs tested affect certain mutants in a distinct and sometimes opposite trends (potentiation versus reduction of IP1 accumulation). Specifically, Y810A, MYN-AA, M694A show opposing effects of the different PAMs, while the K792A mutant results in GS39783 showing the greatest potentiating effects.

Line 168 – Based on the data provided (figure 4E) and cited literature (Binet et., 2004; Sun et al., 2016), while the authors showed that CGP7930 rely on key residues in TM6 of GB2 to elicit its agonist effects, this does not imply that these residues are in contact with the compound, forming its binding site versus being a distal component of its allosteric network, or preclude the possibility that CGP7930 binds within or at a different external part of the TMD of GB2. Due to the lack of a structure of GABAB in complex with CGP7930 and the fact that past work shows CGP7930 can directly activate GB2 TMD, the assertion that CGP7930 directly binds at the TM6 interface is under-substantiated. In addition, mutations in the "deep region" within the GB2 TMD pocket, specifically mutants M14 and M15, in principle show that CGP7930 have ablated agonist activity – which can either be caused by the disruption of a "microswitch" necessary for CGP7930 agonism or disruption of a secondary binding site.

Line 190- Kind of related to the previous comment, authors propose 2 general mechanisms for the ago-PAM effect of the compounds, essentially stabilization of the active state via binding of the PAM at the TM6 interface or within a second binding site. The type of experiments they present following up does not directly provide mechanistic insights to verify these models. The mutagenesis work generally reveals involvement of certain residues. This result is not very surprising as these compounds are allosteric and as it has been shown for example with deep mutational scanning on bAR, many mutations within the TMDs affect receptor signaling (Jones et al., eLife 2020).

Aside from his argument, these are not the only 2 possible mechanisms for the ago-PAM effect. One very plausible model is that the ago-PAM effect could be due to stabilization of an intermediate state, or even appearance of a new intermediate state upon PAM binding. These reasonable models are not discussed and not included in the mathematical modeling.

Methods section lacks important details for all the assays. For example, Line 397, what concentration of Fluo-4 AM was used? what is the assay buffer? what is the volume? what is the temperature that the assays were performed at?

Line 433 – Fitting equation, parameters, and constraints on the fits should be included. currently it is reported as "Activation- and Inhibition concentration response curves were fitted to the default equations of "log(agonist) vs. response -- Variable slope (four parameters)" and "log(inhibitor) vs. response -- Variable slope (four parameters)" within the software respectively."

Reviewer #3:

Liu et al., investigate the mechanism of action of allosteric ligands that affect signaling at the GABAB receptor, a G protein-coupled receptor important in GABA function. The central question is how do these molecules achieve their unique activity, either as direct agonists or as allosteric potentiators of GABA. This study comes at the heels of several cryo-EM structures of GABAB receptors which have revealed the overall architecture of this heterodimeric receptor and the unique between transmembrane domain binding sites for two allosteric modulators that are the focus of this study (rac-BHFF and CGP7930). The authors perform a number of signaling studies to establish that two of these modulators have agonistic and allosteric activity, while one has only allosteric activity. Consistent with the binding site observed in previously determined structures, the authors determine that the transmembrane regions of the GABAB receptor are sufficient for the activity of agonistic positive allosteric modulators. Careful dissection of binding site residues suggests that the interactions made between the three allosteric molecules tested are slightly unique, though their binding sites are likely similar. The authors then test residues deep in the core of the seven transmembrane region of the GABAB subunits, and find that altering this region affects both constitutive activity and the agonistic properties of the modulators, while leaving their allosteric properties intact. With this, the authors propose that the deep site is crucially important for GABAB activation.

While the conclusions specific to GABAB are supported by the authors data, some aspects need to be clarified or revised.

1) It is unclear how well the authors model applies to GPCRs outside of the family C receptors. The authors suggest that the "deep" region is analogous to the sodium binding site of family A GPCRs. These regions differ substantially between the family A and family C GPCRs, and it is unlikely that they serve a similar structural rearrangement upon receptor activation. Furthermore, recent structures of family C receptors bound to G proteins suggest a very unique and divergent mode of G protein engagement compared to family A GPCRs. The authors' speculation around the similarity of activation mechanisms should take these considerations into account.

2) The authors suggest that their findings regarding GABAB modulators that bind at interface between two transmembrane domains likely has implications outside of the family C receptors. The authors should provide appropriate context that, to date, no family A GPCRs have conclusively been shown to form constitutive dimers similar to GABAB or other family C GPCRs.

3) A central new observation in this study is that the deep site is important for constitutive activity and agonistic allosteric modulator activity. The authors propose that this site is a hub for the activation mechanism. Another potential explanation, however, is that the mutations simply stabilize the inactive conformation of the GABAB compared to the wild-type receptor, which has a combined affect on both constitutive activity and the relatively weak agonistic activity of allosteric modulators like rac-BHFF. Indeed, Figure5-supplement 2B shows a rightward shift for GABA dose response for these mutants. The authors should more explicitly discuss potential caveats to their interpretation of the signaling data.

I recommend that the authors revise their speculation that the observations made in this work have general implications for the broader GPCR family outside of family C GPCRs.

https://doi.org/10.7554/eLife.70188.sa1

Author response

Essential revisions:

1) The reviewers all noted that in a number of places the claims are slightly overstated relative to the underlying data, for example in implying the generalizability of the model to G protein coupled receptors outside of family C. The authors should revise the text to temper their claims and clearly indicate for readers which aspects of the proposed model are direct deductions from experimental results and which aspects are more speculative. Please see detailed reviews below for specific areas to revise.

We thanks the reviewers for their constructive comments and justified criticism. In the revised version, we have strongly tuned down the generalizability of our discoveries outside the class C GPCRs, including in the title where we have replaced “GPCR heterodimer” by “class C GPCR heterodimer”. In the revised text, we have clarified the interpretation of the data and the aspects that are more speculative, by using "suggest", "support our hypothesis", or "is consistent with" when it was necessary.

2) In some places certain expected data are missing, for example the pharmacological characterization is more extensive for rac-BHFF than for CGP7930 or GS39783. Is there a reason that all compounds were not profiled similarly? If so, please explain. If not, then including equivalent data for all three compounds would be preferred.

In the revised version, we have now included the equivalent data for the three PAMs. For the wildtype GABAB receptor, the agonist effect of Rac-BHFF, CGP7930 and GS39783 are shown in the three different functional assays, intracellular calcium release (Figure 2B), Go protein rearrangement BRET assay (see new Figure 2D-F) and the most sensitive assay IP1 (Figure 2C and new Figure 2 figure supplement 1). rac-BHFF showed the strongest agonist effect in the three assays, meanwhile GS39783 showed an agonist effect only in the IP1 assay. Now, we have also characterized the agonist effect of the three PAMs for the different GABAB constructs (see new Figure 3—figure supplement 2), and for all the GABAB TMD mutants (Figure 4E and new Figure 5—figure supplement 2) using the IP1 accumulation assay.

Reviewer #1:

[…]

Major issues:

1. The authors primarily use two different methods (intracellular ca2+ release or IP1 accumulation) to determine the agonist or allosteric modulator activity of three PAMs, and intermittently add BRET data on Go activation. There is one sentence early on about the IP1 assay being more sensitive because of the equilibrium nature of the assay as opposed the kinetic nature of the than ca2+ release. The sensitivity of the BRET assay is not discussed. Does the BRET assay reveal agonism for GS39783 as seen in the IP1 assay but not the ca2+ release assay? In many figures, the authors only present data from one assay, although sometimes the results are confirmed with another assay. Some description or discussion about the difference or limitation of these two methods and why they are used as they are would be helpful. It is not clear whether the entire data set should be provided in the supplement, but some logic as to why the assays are being used as they are seems important. Furthermore, when loss of function for rac-BHFF is shown by mutations, it would seem prudent to include data for the more sensitive IP1 assay and not only ca2+ release.

We agree with the reviewer. It is important to compare the effect of the three PAMs in the three functional assays. In the revised version, our results with the wild-type GABA-B receptor show that the agonist activity of GS39783 is detected only in the IP-1 assay (Figure 2C and new Figure 2 —figure supplement 1), and not in the BRET assay (see new Figure 2D-F) and in the calcium release (Figure 2B), indicating a higher sensitivity of the IP1 accumulation assay. For the GABA-B constructs, the agonist activity of the three PAMs was compared only in the IP-1 assay (see new Figure 3 —figure supplement 2, panel B), while in the BRET assay only the agonist effect of rac-BHFF could be detected (see new Figure 3 —figure supplement 2, panel A). Altogether it revealed that the IP-1 assay was more sensitive that the BRET and calcium release assays. The lower sensitivity of this calcium and BRET assays is most probably due to the slow binding properties of the PAMs investigated. This feature is much less limiting in the IP-1 assay that relies on the accumulation of IP-1 induced by the PAM for 30 min before measurement.

But for measuring the allosteric modulation of the PAMs, the calcium release assays is enough sensitive. Accordingly, an allosteric modulator activity is observed including for GS39783 (Figure 2I).

Finally, we would like to point out that all the mutants of the GABA-B receptor have been tested in IP-1 assay (see Figure 4E, 5D and new Figure 5 —figure supplement 2, panel A-D). We can thus conclude that the loss of function of the mutants observed, was not due to the poor sensitivity of the assay used.

Accordingly, a better explanation for the sensitivity of the assays has been described in the Results section of the manuscript.

2. Although the authors claim that they evaluate the agonist activity and the allosteric modulation of all three PAMs, rac-BHFF is the main target of their designed experiments. In some critical results, the data about CGP7930 and GS39783 especially the latter is absent. Since these three PAMs have different functional properties and proposed molecular mechanisms, their data should be presented along with those for rac-BHFF, especially when supporting the conclusion that they have same binding pocket at the TM6s interface.

As stated in Point #1 above, the agonist activity of the three PAMs has been now shown for the wild-type receptor in the three functional assays (see Figure 2B-F, and the new Figure 2 —figure supplement 1). We also added the data on the agonist activity of the three PAMs in IP-1 assay for: (i) the different GABA-B constructs (new Figure 3 —figure supplement 2); (ii) the mutants in the TMD interface (Figure 4E); (iii) the mutants in the deep region of GB2 TMD (new Figure 5 —figure supplement 2, panel A, C and D). In addition, the allosteric modulation effect of the three PAMs has been shown for the wild-type receptor by measuring intracellular calcium release (Figure 2G-I).

3. In Figure 3 supplement 2B, the authors show that CGP7930 has no agonist function in the GB1/2-GB2 heterodimer and the GB2 monomer in the IP1 accumulation assay. This conflicts with the data in the paper published in 2004 (Virginie Binet et al., JBC, 2004), which show clearly that CGP7930 can activate GB1/2+GB2 and GB2 monomer without GABA. The authors do not comment upon this discrepancy and how it might be related to the different assay systems – the results seem directly contradictory to this earlier work. Notably the authors select the less sensitive ca2+ release assay for their studies of rac-BHFF in this figure. I think it important that these critical data be validated in the IP1 assay as well to establish whether there is any observable agonism in this assay or the direct agonism of the compound is completely gone. Are these issues of assay sensitivity and size of effect?

The reviewer highlights an importance point that has been clarified in the revised version. Indeed there is an apparent controversy between the data reported in Binet et al., (2004) and our current data for CGP7930. We think we did not reproduce the data reported in Binet et al., possibly due to the fact that the cells used in this previous study may have enough endogenous GB1 expression. Indeed, it was shown that the endogenous expression of GB1 in cell lines may vary, including in HEK293 cells (Xu et al., (2014) Angewandte Chemie). This has been now discussed in the Results section and this reference has been added in the revised manuscript.

As discussed above, regarding the sensitivity of the assay used, we now compared the three PAMs in the IP-1 assay (the most sensitive one) for all the GABA-B constructs and mutants to avoid to miss any agonism effect. In the revised version, we thus added the agonist activity of the three PAMs in the IP-1 assay for all the GABA-B constructs (new Figure 3 —figure supplement 2, panel B), including the GB1 and GB2 TMD mutants (Figure 4E and new Figure 5—figure supplement 3, panel A).

Regarding the constructs GB1/2+GB2 and GB2 alone, in contrast to the other constructs, none of the three PAMs were able to induce IP-1 accumulation. It further suggests that the simplest view to explain the difference with Binet et al., is the level of endogenous expression of GB1 in HEK293 cells that was probably higher in this previous study.

Additional issues:

1. Figure 1A, the structure panel, is very confusing and the legend does not match the figure. Which structure is shown here or it is a mixture of 4 GPCRs structure? I think it would be helpful to identify the different binding sites somehow so that they can be associated with the specific PDB structures as this is a helpful composite figure if made clearer and with an appropriate legend.

We agree. We added numbers in the Figure 1A to clarify for each allosteric compound its corresponding PDB structure, and we modified the legend accordingly.

2. In figure 4B and 4C, the curves are very difficult to see. Furthermore, the effects reflect a mixture of agonist and allosteric activity of rac-BHFF. To determine the PAM activity of rac-BHFF, the GABA dose curves in present certain or multiple concentration of rac-BHFF should be presented as in figure 3—figure supplement 1D.

To simplify the Figure 4B and 4C we have removed the GB1 and GB2 triple mutants in the revised figure. Indeed, since Rac-BHFF and GS39783 have no agonist activity on these triple mutants in the IP-1 assay (Figure 4E), it is not useful to measure their potency in calcium release assay.

As suggested by the reviewer, for all the mutants at the TMD interface we added the allosteric modulator activity of rac-BHFF on the GABA dose curves in presence of two different concentrations of rac-BHFF in the calcium release assay (see new Figure 4—figure supplement 2). Interestingly, these results revealed a loss of agonist activity of rac-BHFF for all GABA-B mutants, except GB1-M807A. It is in contrast to the lower reduction of this agonist activity measured in the IP-1 assay for all these TMD interface mutants (Figure 4E). This is consistent again with a higher sensitivity of the IP-1 assay.

These data showed that the residue M807 in GB1 plays only a minor role in rac-BHFF activity, as observed in Figure 4B (Rac-BHFF has a similar agonist potency for the WT and mutant GB1-M807A) and Figure 4E (Rac-BHFF has a similar agonist efficacy for the WT and mutant GB1-M807A). This residue M807 seems also no critical for the agonist activity of the two others PAMs (Figure 4E).

3. Line 156, in Figure 4E, GB1-Y810A and GB1-MYN-AAA only show the TM6s are crucial for the agonist activity of rac-BHFF. This evidence cannot support that GB1 mutants can reduce the PAM activity of rac-BHFF. PAM activity should be removed from the text unless parallel experiments with GABA are shown.

We agree for this misunderstanding. We have removed the PAM activity in this sentence.

4. Line 162, The data in Figure 4E can only demonstrate CGP7930 effects were impaired by the GB2 mutations but not by GB1 mutations; we cannot make a conclusion that it can activate GB2 alone from these data. This conclusion is also in conflict with the data presented in Figure 3 and its supplemental data where the GB1-TM is shown to be critical even for CGP7930 (Figure 3 S2B).

We agree with the reviewer’s comments. We have removed the last part of the sentence (Line 163164; “in agreement with CGP7970 being able to active GB2 alone as previously reported”).

5. At line 131 there is a conclusion about the GB1DCRC mutant that is locked in an active state still being able to support allosteric activation. Of note, it is also able to support GABA activation. This should be added to text, along with a brief explanation of what it means to be locked in the inactive state if both orthosteric and allosteric activation is intact.

We agree with the reviewer’s comments that it requires a clarification. We modified the sentence as follow:

“…the conformational state of the GB1 TMD is not critical for both orthosteric and allosteric activation since similar results were obtained with the mutant GB1DCRC (Figure 3—figure supplement 1E-F). This mutant is activated by GABA similarly to the wild-type receptor (Figure 3figure supplement 1E) as previously reported (Monnier et al., 2011). But it was engineered to create a disulphide bond in GB1 TMD thus expecting to limit the conformational change of this domain upon ligand stimulation of the GABAB receptor.”

6. Line 152 – there seem to be substantial effects of some of these mutations on GABAB activation, not just the 2 mutations noted.

We agree. We have included the mutant GB1-K792A in this sentence. We did not include the GB2 triple mutant since we have removed it from Figure 4C (see Point #2 above).

7. Line 100 – if the lipid is present only in the inactive conformation, then why in the second half of the sentence is it proposed to be important for signal transduction. This should be clarified.

We think the reviewer refers to Line 201 (and not 100). Skiniotis’s group was the only one to investigate the possible importance of this lipid for the functioning of the receptor. Mutations designed to destabilize phospholipid binding in GB2 TMD resulted in the increase in both GABA-B basal activity and response to GABA (Papasergi-Scott et al., (2020) Nature). One sentence has been added after Line 201 to clarify this point in the revised manuscript.

8. Line 244-246 – is it conceivable that the compounds binding in the TMD of GB1 to exert their effects and that they can no longer bind to the deep region in GB2?

Currently, it is not possible to exclude that these PAMs bind in the TMD of GB1. But we don’t think so. As scientists, we tend to always favor the simplest explanation (binding of these PAMs at the TMD heterodimer interface, as indicated by our results), but we must be open to alternative.

9. Line 249 – do any of the green mutations affect constitutive activity? M11, 12 and 13 have effect but emphasis is only on M14 and M15. Would be nice to see all the mutations in this analysis. Are the effects limited to the deep residue mutations or is this more complex?

We agree with the reviewer. To give a better picture of what’s happen in the GB2 TMD, we now provide the results for the constitutive activity of all mutants of GB2 (from M1 to M16; see new Figure 5—figure supplement 2B). While all the mutants in the “deep region” (“red mutations”) have a strongly impaired or abolished constitutive activity, many mutants in the lipid binding pocket (“green mutations”) have a similar constitutive activity that the WT receptor. Only the green mutant M9 has no constitutive activity while it is normally expressed at the cell surface (see ELISA for GB1 and GB2-M9 in the Figure 5—figure supplement 3A). It suggested that the three mutations in GB2-M9 have strongly impaired the ability of the GABAB receptor to adopt an active state in the basal conditions, or alternatively, they have stabilized the inactive state of the receptor. Further studies will be necessary to clarify this effect. In contrast, the low constitutive activity of the mutants M8 and M10 might be due, at least partially, to the lower expression of the mutated receptors (see ELISA for GB1 and GB2-M9 in the Figure 5—figure supplement 3A).

10. Line 272 – the language is confusing since the different mutations have opposite effects on constitutive activity and this should be clarified.

We agree with the reviewer. We have clarified the text of this paragraph in the revised version as follow:

“Interestingly, one of the human genetic mutations involved in Rett-like phenotype, GB2 A5673.43T, is located in the deep region and also increases receptor constitutive activity. It corresponds to the rat GB2 mutant M12 (A5663.43F) that displays a lower constitutive activity (Figure 5F). This suggested that depending of the residue at this position, the constitutive activity of the receptor can be tuned up or down. This further illustrates the role of the deep region in controlling the conformational landscape of the GABAB.”

11. Line 284 -the effect shown in Mglu5 is opposite that shown in GABAB for the same mutation. And two mutations are completely tolerated in the mGlu5 despite their negative effects on GABAB. Caution in generalizing seems prudent here, and I think these are the weakest data in the paper and either should be deleted or strongly caveated. There are many more differences than similarities here.

We agree with the reviewer. We have clarified the text of this paragraph in the revised version and we avoid to generalize the results obtained with GABA-B to the mGluRs. The text on the mGluR5 results is now as follow:

“We have then investigated how mutations of these three residues in mGluR5, equivalent to the residues in GB2 subunits (G2.46, A3.43 and T7.43) but not conserved, can influence its constitutive activity. While two mutations did not change the constitutive activity of mGluR5, the mutation A8127.43F increased it (Figure 5H, Figure 5—figure supplement 4E-F). It suggests that this region is also controlling the conformational landscape of the mGlu5 receptor. But our data show that is not possible to predict if mutations of these residues in this deep region will produce or not a change in the constitutive activity. Further studies will be necessary to generalize the role of this deep region in the mGlu receptors.”

12. I don't think the concept of GB1 TM domain acting as a lever is a useful or justifiable metaphor. A lever is "a rigid bar resting on a pivot, used to help move a heavy or firmly fixed load with one end when pressure is applied to the other." How agonist binding is conveyed to GB2 and how a PAM affects this is quite complex is terms of the actual physical forces, and I don't believe any of the data here allow this conclusion. Moreover, the authors write "GB1 TMD serves as a lever to stabilize the active state of GMB2 TMD…" As per the definition quoted above, I am not sure how a lever can stabilize. I would remove the lever metaphor altogether. It is also incompatible with the next statement: "This lever model is compatible with the restriction of the TM6s in the active interface where both TM6s do not move substantially between inactive and active states".

We agree with the arguments of the reviewer. Accordingly, we have removed the lever metaphor, and we have rephrased the corresponding paragraph in the Discussion.

This is a matter of taste, but in Figure 2 red and orange are used in B and C for the different PAMS – then in D-F the same colors are used to represent the PAM concentration. I think this is confusing. I would prefer to see consistency in the color schemes in the same and across figures, but this is obviously just my taste.

We agree. We have changed the colors used in Figure 2B and C, and in the new Figure 2E and 2F. Now, red and orange are used only to represent the PAM concentrations (see Figure 2G-I).

As best I can tell, Figure 2B red and Figure 2 supplement 1A show the same experimental design, but in Figure 2B a lower range of concentration is shown and the curve doesn't reach the Emax. In the supplemental data, the curve goes to higher concentrations and reaches the Emax, although only two independent experiments are shown. Would have been nice to have a full curve in the main figure. Regardless the legend in the supplemental figure should clarify that it is the same type of experiment as 2B as not clear now why the same experiment is shown twice.

We agree. We have transferred all the data of this former Figure 2-suppplement 1 in the main figure (see new Figure 2) of the revised manuscript.

Figure 3 legend should explain the dotted lines.

The dotted lines in the main and inserted graphes indicate the dose-responses of the wild-type receptor determined in panel A. It is now clearly indicated in the revised version.

Figure 5 In panel E I would use the mutant names that are used in all the other panels. I would include all the mutants from panel B in E so that we can see the relationship for both the green and red mutants. Does the relationship hold for all or just the subset shown?

We agree. In panel E, we used the mutant names that are used in all the other panels of Figure 5. We have also performed additional IP1 assays to show that the agonist activity of the three PAMs is correlated with the constitutive activity of the GABAB receptor for the mutants of the “deep region”, but also in some extent for those of the phospholipid binding pocket (see new Figure 5- Supplement 2C-D).

There are a lot of surface expression data. It would be nice to help the reader by adding panels in the supplements showing normalized emax throughout to make it easier for the reader to see the impact of expression – now one can only see that expression isn't dramatically affected but too hard to do it one at a time by mutant.

To better see the effect of the GB2 “green” and “red” mutations, we have plotted the expression of the mutants versus the activation by GABA or by Rac-BHFF (agonist effect of the PAM in absence of GABA) in calcium release assays (see new Figure 5 – supplement 3, panel B and C).

Reviewer #2:

GPCRs are major targets for drug development. A current promising avenue for GPCR drug development is to identify strategies to affect receptor signaling beyond simply targeting the endogenous ligand binding site. This would provide larger surface area to explore and better subtype specificity. Allosteric modulators do this by binding elsewhere in the receptor and enhance or attenuate effect of the orthosteric ligand. The general goal of this research was to investigate the mechanism of such effect in GABAB receptors. Finding the mechanism could inform development of future pharmaceutical drugs. GABAB receptors are members of class C GPCRs. This small family of GPCRs and is somewhat unique in that class C receptors form constitutive dimers. Recent structures of GABAB receptors from 4 different groups showed that 2 GABAB receptor PAMS (rac-BHFF and GS39783) bind at the dimeric interface which is unique among GPCRs. In this study, authors first verified this by performing functional assays and mutagenesis. They also showed the previously described constitutive activity of GABAB receptors and the direct agonist effect of these PAMs (called ago-PAM). Using mutation and structure-function approaches they showed the role of TM6 in both of these properties. Next, they authors tried to use similar assays to investigate the mechanism of agonist activity and modulation by these PAMs. Their results demonstrate the contributions of an internal cavity within the 7-transmembrane domain of GB2, dubbed "deep region", in the constitutive activity of GABAB receptor and the agonist effect of the PAMs but not the allosteric modulations.

While these results provide an interesting twist to how allosteric modulators are perceived and that the pharmacological classification of compounds can be intertwined and not orthogonal, it provides few new mechanistic understanding. For example how does PAMs affect activation (does it modify the energy landscape? or induce a new conformation?) and is it possible to synthesize a PAM that is only a modulator and does not have direct agonist effect? Moreover, the methods section lacks practical details.

We agree with the reviewer. Our study clarifies how the current allosteric modulators of the GABAB receptor work. As other GABA-B ligands, the PAM are expected to modify the energy landscape of the receptor, since membrane receptors are highly dynamic protein, and both orthosteric and allosteric ligands are expected to modify, even slightly, its structural dynamics. Even though the most simple view is that GABAB PAMs favor active conformations stabilized by the orthosteric agonists, it is difficult to exclude that these PAMs induce a new conformation that would not be stabilized by the orthosteric ligands. This has been discussed in the revised part of the Discussion. Other studies would be necessary to address this issue.

The three GABAB PAMs investigated in this study have all an agonist activity, even though that of the GS39783 can be revealed only by the most sensitive functional assay (IP1 accumulation). Neural allosteric ligands (NAL) have been described for other GPCRs (Christopolous et al., (2014) Pharm Rev; Hellyer et al., (2018) Mol Pharm) indicating that allosteric modulators that do not have an agonist effect should be possible to obtain. To our knowledge, such compounds were not yet reported for the GABA-B receptor.

The method section (intracellular calcium assay and curve fitting) was better described in the revised version.

Specific comments:

Line 110 – Is it possible the effect of GS39783 in the IP1 assay is due to nonspecific effect? Control titration experiments for figure 2B with cells transfected only with the chimeric G protein or the individual GABAB subunits + the chimeric G proteins would be highly relevant to this data.

We agree with the reviewer. We have now included IP1 data with chimeric G protein or the individual GABAB subunits + the chimeric G proteins for the GS39783 and the two other PAMs (see new Figure 2 – supplement 1).

Figure 3, figure supplement 1- While it is generally suggested that the ER retention signals inhibit trafficking of GABAB1 subunit to the surface, depending on expression level and cell type this could be leaky. Authors should provide GABAB1 only and GABAB2 only surface expression levels.

The results with the wild-type GABAB1 only and GABAB2 only have been shown in the revised Figure 3—figure supplement 1, panel A.

Figure 4, panel H. Previous work from authors and others (for example Koehl et al., 2019) has shown that for experiments on mGluR5, co-expression of neuronal excitatory amino acid transporter 3 (EAAT3) is necessary to be able to characterize the effect of modulators correctly. Without that, modulators can appear to show intrinsic agonist activity (for example Koehl et al., 2019). In the data in this panel, EAAT3 was not co-expressed.

It is a good remark from the reviewer. In Figure 5, panel H, the experiments have been performed in presence of the co-transfected glutamate transporter EAAT3 (also known as EAAC1), as well as in the cell surface quantification performed with the mGlu5 receptors and glutamate induced IP1 accumulation (Figure 5 —figure supplement 4E-F). It is now clearly written in this figure legend, in the revised text and in the Materials and methods section.

Authors have previously shown that GABAB receptors form extensive dimers of dimers and have characterized those. What is the effect of those higher order oligomers on the observed PAM effect?

It is a good question from the reviewer. Indeed, in our experimental conditions, we probably have a mix of heterodimers, dimer of dimers and higher order oligomers. Our lab recently showed that a strong negative effect between the GABAB1 binding sites exists within the GABAB oligomer, that limits G protein activation (Comps-Agrar et al., (2012) EMBO J; Stewart et al., (2018) Neuropharmacol). Accordingly, the effect of the PAM on signaling might be lower on the oligomers compared to the heterodimer. Experimental evidence would require to test the PAM effects on GABAB1 mutants (Comps-Agrar et al., (2012) EMBO J; Stewart et al., (2018) Neuropharmacol) that have less tendency to form dimer of dimers. But these experiments are difficult to perform since these mutants also impair the cell surface expression of the GABAB receptor, and we think they are outside the scope of the present study.

Line 166 – This explanation for the discrepancy is not clear. The different PAMs tested affect certain mutants in a distinct and sometimes opposite trends (potentiation versus reduction of IP1 accumulation). Specifically, Y810A, MYN-AA, M694A show opposing effects of the different PAMs, while the K792A mutant results in GS39783 showing the greatest potentiating effects.

We clarify the text in the revised version: “Of note, GS39783 is more sensitive to mutations than the other PAMs. It might be because the mutated residues are highly important for the binding of GS39783, or alternatively to its weakest agonist activity compared to Rac-BHFF and CGP7930 then resulting in a stronger loss of agonist activity of GS39783 on these mutants.”

Line 168 – Based on the data provided (figure 4E) and cited literature (Binet et., 2004; Sun et al., 2016), while the authors showed that CGP7930 rely on key residues in TM6 of GB2 to elicit its agonist effects, this does not imply that these residues are in contact with the compound, forming its binding site versus being a distal component of its allosteric network, or preclude the possibility that CGP7930 binds within or at a different external part of the TMD of GB2. Due to the lack of a structure of GABAB in complex with CGP7930 and the fact that past work shows CGP7930 can directly activate GB2 TMD, the assertion that CGP7930 directly binds at the TM6 interface is under-substantiated. In addition, mutations in the "deep region" within the GB2 TMD pocket, specifically mutants M14 and M15, in principle show that CGP7930 have ablated agonist activity – which can either be caused by the disruption of a "microswitch" necessary for CGP7930 agonism or disruption of a secondary binding site.

We agree with the reviewer. The possibility of another binding site for CGP7930 in GB2 TMD and the apparent controversy between the previous data reported by Binet et al., (2004) and our current data on CGP7930 need to be clarified. We think we did not reproduce the data reported in Binet et al., possibly due to the fact that the cells used in this previous study may have enough endogenous GB1 expression. Indeed, it was shown that the endogenous expression of GB1 in cell lines may varies, including in HEK293 cells (Xu et al., (2014) Angewandte Chemie). This reference has been added in the revised manuscript. Our current hypothesis is thus that the previous studies from Binet et al., (2004) and Sun et al., (2016) have proposed a wrong interpretation of the results, most probably due to endogenous expression of the GB1 subunit.

The reviewer suggests that the loss of agonist activity of CGP7930 on the GB2 mutants M14 and M15 could be due to the disruption of a binding site for this ago-PAM. Although we cannot totally exclude the possibility of an allosteric binding site of CGP7930 in the GB2 TMD, a similar loss of agonist activity were measured for the three PAMs on the same mutants (see new Figure 5supplement 3), while it is clear that GS39783 and rac-BHFF bind at the TM6 interface in the heterodimer as reported in the cryo-EM structures (Shaye et al., Nature, 2020; Mao et al., Cell Res 2020; Shen et al., Nature, 2021). Based on our mutagenesis analysis, we can thus reasonably propose that the main functional binding site for CGP7930 is most probably in the TM6s interface in the heterodimer.

Line 190- Kind of related to the previous comment, authors propose 2 general mechanisms for the ago-PAM effect of the compounds, essentially stabilization of the active state via binding of the PAM at the TM6 interface or within a second binding site. The type of experiments they present following up does not directly provide mechanistic insights to verify these models. The mutagenesis work generally reveals involvement of certain residues. This result is not very surprising as these compounds are allosteric and as it has been shown for example with deep mutational scanning on bAR, many mutations within the TMDs affect receptor signaling (Jones et al., eLife 2020).

We agree with the reviewer. Site-directed mutagenesis could induce long distance effects such as mutations in the TM6s interface, or in the GB2 TMD “deep region”, could impair a remote binding site in the GB2 TMD core. But it is interesting to observe that the three PAMs have similar effects on all the mutants in GB2 TMD, that could be explained mostly by a common binding site for the three PAMs in GB2 TMD. But due to the different chemical structures of the three PAMs, it is difficult to propose such a common binding site in GB2 TMD. A binding pocket that can accommodate these three PAMs is probably more possible at the TM6s interface.

Aside from his argument, these are not the only 2 possible mechanisms for the ago-PAM effect. One very plausible model is that the ago-PAM effect could be due to stabilization of an intermediate state, or even appearance of a new intermediate state upon PAM binding. These reasonable models are not discussed and not included in the mathematical modeling.

We agree with the reviewer, the ago-PAM effect could be due to the stabilization of an intermediate state of the receptor. We have now included this possibility in the revised version of the Discussion. But we think that including this intermediate state of the receptor in the mathematical model is beyond the scope of this study.

Methods section lacks important details for all the assays. For example, Line 397, what concentration of Fluo-4 AM was used? what is the assay buffer? what is the volume? what is the temperature that the assays were performed at?

In the revised version, the Materials and methods section for the intracellular calcium measurement was clarified as suggested by the reviewer.

“Intracellular ca2+ release in HEK293 cells was performed as previously described (Lecat-Guillet et al., 2017). Briefly, transfected cells in the 96 well plate were washed once with HBSS buffer (20 mM Hepes, 1 mM MgSO4, 3.3 mM Na2CO3, 1.3 mM CaCl2, 0.1% BSA and 2.5 mM probenecid) and pre-incubated with 1 μM ca2+-sensitive Fluo4 AM (Thermo Fisher Scientific) prepared in the HBSS buffer for 1 h at 37°C. Cells were washed once with HBSS buffer and 50 μL buffer was added into the wells before measuring in the multimode microplate reader (FlexStation 3, Molecular Devices). The fluorescence signals (excitation at 485 nm and emission at 525 nm) were then measured at intervals of 1.5 s for 60 s with the adding of 50 μL indicated compounds 20 s after the first reading by the reader automatically. The ca2+ response was given as the agonist-stimulated fluorescence increase, normalized according to the indication. Concentration response curves were fitted using “log(agonist) vs. response -- Variable slope (four parameters)” by GraphPad Prism software.”

Line 433 – Fitting equation, parameters, and constraints on the fits should be included. currently it is reported as "Activation- and Inhibition concentration response curves were fitted to the default equations of "log(agonist) vs. response -- Variable slope (four parameters)" and "log(inhibitor) vs. response -- Variable slope (four parameters)" within the software respectively."

To answer reviewer’s comment, the Materials and methods section was clarified:

“Activation concentration response curves were fitted to the default equation of “Log(agonist) vs. response – Variable slope (four parameters)” within the software, as Y=Bottom + (Top-Bottom)/(1+10^((LogEC50-X)*Hill slope)) with no constraints for the four parameters (X, log of concentration; Y, response; Top and Bottom, plateaus in same units as Y; LogEC50, same log units as X; and Hill slope, unit less) in the equation.”

Reviewer #3:

Liu et al., investigate the mechanism of action of allosteric ligands that affect signaling at the GABAB receptor, a G protein-coupled receptor important in GABA function. The central question is how do these molecules achieve their unique activity, either as direct agonists or as allosteric potentiators of GABA. This study comes at the heels of several cryo-EM structures of GABAB receptors which have revealed the overall architecture of this heterodimeric receptor and the unique between transmembrane domain binding sites for two allosteric modulators that are the focus of this study (rac-BHFF and CGP7930). The authors perform a number of signaling studies to establish that two of these modulators have agonistic and allosteric activity, while one has only allosteric activity. Consistent with the binding site observed in previously determined structures, the authors determine that the transmembrane regions of the GABAB receptor are sufficient for the activity of agonistic positive allosteric modulators. Careful dissection of binding site residues suggests that the interactions made between the three allosteric molecules tested are slightly unique, though their binding sites are likely similar. The authors then test residues deep in the core of the seven transmembrane region of the GABAB subunits, and find that altering this region affects both constitutive activity and the agonistic properties of the modulators, while leaving their allosteric properties intact. With this, the authors propose that the deep site is crucially important for GABAB activation.

While the conclusions specific to GABAB are supported by the authors data, some aspects need to be clarified or revised.

1) It is unclear how well the authors model applies to GPCRs outside of the family C receptors. The authors suggest that the "deep" region is analogous to the sodium binding site of family A GPCRs. These regions differ substantially between the family A and family C GPCRs, and it is unlikely that they serve a similar structural rearrangement upon receptor activation. Furthermore, recent structures of family C receptors bound to G proteins suggest a very unique and divergent mode of G protein engagement compared to family A GPCRs. The authors' speculation around the similarity of activation mechanisms should take these considerations into account.

We agree with the reviewer. In the revised manuscript, we have strongly tuned down the generalizability of our discoveries outside the class C GPCRs. In the title, we have replaced “GPCR heterodimer” by “class C GPCR heterodimer”. In the abstract, we also removed the idea that the “deep region” reported here in GB2 is functionally conserved in class A GPCR.

2) The authors suggest that their findings regarding GABAB modulators that bind at interface between two transmembrane domains likely has implications outside of the family C receptors. The authors should provide appropriate context that, to date, no family A GPCRs have conclusively been shown to form constitutive dimers similar to GABAB or other family C GPCRs.

We agree with the reviewer. In the revised manuscript, we have strongly tuned down the generalizability of our discoveries outside the class C GPCRs. At the end of the Discussion, we just mention that ligands acting at a dimer interface may potentially be interesting tools for other GPCRs, even if they generally form for transient dimers.

3) A central new observation in this study is that the deep site is important for constitutive activity and agonistic allosteric modulator activity. The authors propose that this site is a hub for the activation mechanism. Another potential explanation, however, is that the mutations simply stabilize the inactive conformation of the GABAB compared to the wild-type receptor, which has a combined affect on both constitutive activity and the relatively weak agonistic activity of allosteric modulators like rac-BHFF. Indeed, Figure5-supplement 2B shows a rightward shift for GABA dose response for these mutants. The authors should more explicitly discuss potential caveats to their interpretation of the signaling data.

It is a very good remark from the reviewer. We could not exclude that the mutations in the “deep region” of GB2 TMD stabilize the inactive conformation of GABA-B (this possibility is now considered in the revised Discussion). As proposed by the reviewer, it could explain the rightward shift of GABA dose response for these mutants (see new Figure 5-supplement 3). But this hypothesis is difficult to test experimentally. In the revised text, we better discussed the potential caveats to the interpretation of the signaling data.

I recommend that the authors revise their speculation that the observations made in this work have general implications for the broader GPCR family outside of family C GPCRs.

We agree with the reviewer. This was done to make the manuscript more attractive to the editor. In the revised version, we have strongly tuned down the generalizability of our discoveries outside the class C GPCRs, including in the title where we have replaced “GPCR heterodimer” by “class C GPCR heterodimer”. We also removed the idea that the “deep region” observed here in GB2 is functionally conserved in class A GPCR. In the Discussion on class C GPCRs, and we just mention that ligands acting at a dimer interface may potentially be interesting tools for other GPCRs, even if they generally form for transient dimers.

https://doi.org/10.7554/eLife.70188.sa2

Article and author information

Author details

  1. Lei Liu

    1. Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    2. Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Conceptualization, Data curation, Formal analysis, Writing – original draft
    Contributed equally with
    Zhiran Fan
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9824-9570
  2. Zhiran Fan

    Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    Contribution
    Data curation, Formal analysis
    Contributed equally with
    Lei Liu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9543-1211
  3. Xavier Rovira

    MCS, Laboratory of Medicinal Chemistry, Institute for Advanced Chemistry of Catalonia (IQAC-CSIC), Barcelona, Spain
    Contribution
    Conceptualization, Data curation, Formal analysis, Writing – original draft
    For correspondence
    xavier.rovira@cid.csic.es
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9764-9927
  4. Li Xue

    1. Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    2. Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  5. Salomé Roux

    Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6106-4863
  6. Isabelle Brabet

    Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  7. Mingxia Xin

    Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  8. Jean-Philippe Pin

    Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Writing – original draft, Writing – review and editing
    For correspondence
    jean-philippe.pin@igf.cnrs.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1423-345X
  9. Philippe Rondard

    Institut de Génomique Fonctionnelle, Université de Montpellier, CNRS, INSERM, Montpellier, France
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    philippe.rondard@igf.cnrs.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1134-2738
  10. Jianfeng Liu

    Cellular Signaling Laboratory, International Research Center for Sensory Biology and Technology of MOST, Key Laboratory of Molecular Biophysics of MOE, and College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Supervision, Writing – original draft
    For correspondence
    jfliu@mail.hust.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0284-8377

Funding

Ministry of Science and Technology of the People's Republic of China (2018YFA0507003)

  • Jianfeng Liu

National Natural Science Foundation of China (81720108031)

  • Jianfeng Liu

National Natural Science Foundation of China (81872945)

  • Jianfeng Liu

National Natural Science Foundation of China (31721002)

  • Jianfeng Liu

National Natural Science Foundation of China (31420103909)

  • Jianfeng Liu

Ministry of Education of the People's Republic of China (B08029)

  • Jianfeng Liu

Centre National de la Recherche Scientifique (PICS n°07030)

  • Philippe Rondard

Centre National de la Recherche Scientifique (PRC n°1403)

  • Philippe Rondard

Institut National de la Santé et de la Recherche Médicale (IRP Brain Signal)

  • Philippe Rondard

Agence Nationale de la Recherche (ANR-09-PIRI-0011)

  • Philippe Rondard

Fondation pour la recherche médicale FRM (FRM team: DEQ20170326522)

  • Jean-Philippe Pin

Ministerio de Economía, Industria y Competitividad, Gobierno de España (SAF2015-74132-JIN)

  • Xavier Rovira

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

Acknowledgements

We thank Jordi Haubrich for his comments on the manuscript. JL was supported by the Ministry of Science and Technology (grant number 2018YFA0507003), the National Natural Science Foundation of China (NSFC) (grant numbers 81720108031, 81872945, 31721002, and 31420103909), the Program for Introducing Talents of Discipline to the Universities of the Ministry of Education (grant number B08029), and the Mérieux Research Grants Program of the Institut Mérieux. PR and J-PP were supported by the Centre National de la Recherche Scientifique (CNRS, PICS n°07030, PRC n°1403), the Institut National de la Santé et de la Recherche Médicale (INSERM; International Research Program « Brain Signal »), and by grants from the Agence Nationale de la Recherche (ANR-09-PIRI-0011), the FRM (FRM team: DEQ20170326522). XR by the Spanish Ministry of Economy, Industry and Competitiveness (SAF2015-74132-JIN).

Senior Editor

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

Reviewing Editor

  1. Andrew C Kruse, Harvard Medical School, United States

Reviewers

  1. Jonathan A Javitch, Columbia University, United States
  2. Aashish Manglik, University of California, San Francisco, United States

Publication history

  1. Received: May 9, 2021
  2. Accepted: December 2, 2021
  3. Accepted Manuscript published: December 6, 2021 (version 1)
  4. Accepted Manuscript updated: December 10, 2021 (version 2)
  5. Version of Record published: December 23, 2021 (version 3)

Copyright

© 2021, Liu et al.

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

Metrics

  • 597
    Page views
  • 165
    Downloads
  • 0
    Citations

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

Download links

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

Downloads (link to download the article as PDF)

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

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

Further reading

    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Lindsay B Case et al.
    Research Article

    Integrin adhesion complexes (IACs) are integrin-based plasma membrane-associated compartments where cells sense environmental cues. The physical mechanisms and molecular interactions that mediate initial IAC formation are unclear. We found that both p130Cas ('Cas') and Focal adhesion kinase ('FAK') undergo liquid-liquid phase separation in vitro under physiologic conditions. Cas- and FAK- driven phase separation is sufficient to reconstitute kindlin-dependent integrin clustering in vitro with recombinant mammalian proteins. In vitro condensates and IACs in mouse embryonic fibroblasts (MEFs) exhibit similar sensitivities to environmental perturbations including changes in temperature and pH. Furthermore, mutations that inhibit or enhance phase separation in vitro reduce or increase the number of IACs in MEFs, respectively. Finally, we find that the Cas and FAK pathways act synergistically to promote phase separation, integrin clustering, IAC formation and partitioning of key components in vitro and in cells. We propose that Cas- and FAK- driven phase separation provides an intracellular trigger for integrin clustering and nascent IAC formation.

    1. Biochemistry and Chemical Biology
    2. Computational and Systems Biology
    Dhruva Katrekar et al.
    Tools and Resources

    Adenosine deaminases acting on RNA (ADARs) can be repurposed to enable programmable RNA editing, however their enzymatic activity on adenosines flanked by a 5' guanosine is very low, thus limiting their utility as a transcriptome engineering toolset. To address this issue, we first performed a novel deep mutational scan of the ADAR2 deaminase domain, directly measuring the impact of every amino acid substitution across 261 residues, on RNA editing. This enabled us to create a domain wide mutagenesis map while also revealing a novel hyperactive variant with improved enzymatic activity at 5'-GAN-3' motifs. However, exogenous delivery of ADAR enzymes, especially hyperactive variants, leads to significant transcriptome wide off-targeting. To solve this problem, we engineered a split ADAR2 deaminase which resulted in 1000-fold more specific RNA editing as compared to full-length deaminase overexpression. We anticipate that this systematic engineering of the ADAR2 deaminase domain will enable broader utility of the ADAR toolset for RNA biotechnology and therapeutic applications.