Functional and structural basis of a negative allostery within GABA_B hetero-tetramers

Abstract

G protein coupled receptors (GPCRs) oligomerization may allow signal integration from different GPCR units. The GABA_B receptor, activated by the main inhibitory transmitter, GABA, is an obligatory heterodimer. It is the target of two therapeutic drugs, baclofen and GHB, and can form stable oligomers. The existence, roles, and possible allosteric interaction of GABA_B oligomers remain elusive. Here, we show that GABA_B oligomers exist in neurons. Their function can be specifically affected by human disease-associated mutations, demonstrating their essential role for normal brain function. The cryo-EM structure of a hetero-tetramer in the apo state reveals the heterodimers interacting in an asymmetrical way to prevent one unit from being activated. This represents a nice example of a negative allosteric interaction between GPCRs related to human diseases.

The GABA_B receptor is a mandatory heterodimer. Here authors show the existence of GABA_B dimers of dimers in neurons, and their relevance in human health. By solving the structure of this tetramer, they show the structural basis for a negative cooperativity between the dimers within such a complex.

Introduction

G protein-coupled receptors (GPCRs) play crucial roles in cell signaling and communication, allowing cells to detect and respond to various external signals and, as such, represent the most important drug targets^1–3. Most of them function as monomers, such as those of class A, B, and F, or as constitutive dimers such as the class C receptors^4,5. Accumulating evidence indicates that these GPCRs can form heterodimer/oligomer and regulate each other through direct interaction, either transient or stable^6–11. Such interactions reveal possible allosteric modulations of their function through formation of GPCR hetero-complexes, leading to the allosteric control^12,13 of one receptor by another^14,15, to signaling bias¹⁶ with changes in signal efficacy and potency^17–19. Although a few structures of GPCR homo-oligomers have been reported, such as rhodopsin^20,21, β1-AR²², or CXCR4²³, the molecular and structural basis of the allosteric control of one GPCR by another within a hetero-oligomer, likely important for signal integration, remains poorly understood.

GABA_B receptor is activated by γ-aminobutyric acid (GABA), the primary inhibitory neurotransmitter in the brain²⁴. GABA_B receptor plays crucial roles in regulating brain functions and is implicated in locomotion, learning, memory, and cognition^25–27. Dysregulation of GABA_B receptor activity has been implicated in various diseases, such as epilepsy, anxiety, depression, spasticity, addiction and pain^28,29. Human genetic mutations and auto-antibodies targeting the GABA_B receptor can be responsible for human diseases, including Rett syndrome³⁰, developmental delay, epilepsy, and encephalitis^31–33. Accordingly, GABA_B receptor is the target of Baclofen (Lioresal^®), a specific GABA_B receptor agonist used to treat spasticity in multiple sclerosis and alcohol abuse disorder^28,34, and by γ-hydroxybutyrate (GHB), a drug of abuse, also used for anesthesia³⁵. GABA_B receptor is a member of the class C GPCRs, and an obligatory heterodimer consisting of two subunits, GB1 where agonists bind, and GB2 that couples to G proteins^24,36,37, stabilized by a coiled-coil interaction between their intracellular tails^38,39. Each subunit comprises a large extracellular venus flytrap domain (VFT) that consists of two lobes (LB1 and LB2)⁴⁰ and seven-transmembrane domains (7TMs). In the inactive state, GB1 and GB2 interact through their LB1s and the intracellular tips of TM3 and TM5⁴¹. Agonists (GABA, Baclofen) and antagonists (CGP54626) bind to GB1 VFT, whereas the VFT of GB2 does not bind ligands³⁷. Agonists act by stabilizing a closed state of GB1 VFT, bringing the LB2 of each subunit in contact. The interplay between the two VFTs, essential for receptor activation, triggers a rearrangement from the inactive 7TM interface to an active one formed by the TM6 of both subunits^40,42,43. This active conformation allows the coupling of G_i/o proteins to GB2 in a shallow pocket formed by TM3 and three intracellular loops⁴². Of interest, GABA_B receptor positive allosteric modulator (PAM) such as rac BHFF, binds at the TM6-TM6 interface to stabilize the active conformation of the receptor⁴² and also acts as a positive allosteric agonist (ago-PAM)^42,44,45.

GABA_B heterodimers can assemble into stable hetero-tetramers or higher-order oligomers^{7,17,18,46–48}. Indirect evidence for the existence of such oligomers in the brain exist^17,49. Our previous results showed that the association of GABA_B heterodimers within a hetero-tetramer is mediated by GB1 subunits, an effect leading to a negative control of one heterodimer by the other^17,18,47. However, the structural basis and functional consequences for this allosteric control within GABA_B hetero-oligomer remain unknown.

In this study, we develop nanobodies specific to each subunit to provide direct evidence for the existence of GABA_B hetero-tetramers in primary neurons. We bring further support for a negative control of one GABA_B receptor over the other in hetero-tetramers and show that some disease-related mutations in GB1 VFT mainly affect the GABA_B hetero-tetramer function. These data demonstrating the physiological and pathological relevance of GABA_B hetero-oligomers prompt us to solve its structure. This reveals an asymmetric assembly of the extracellular domains of each receptor units, such that only one can be activated at a time. This study illustrates how GPCRs can control their activities through allosteric interaction within oligomers and provides insights to develop drugs against GPCR oligomers.

Results

Development of nanobodies targeting the GABA_B subunits

To identify nanobodies targeting the GABA_B subunits, HEK293T cells transiently expressing both human and rat GABA_B receptors were injected into an alpaca, and VHH (variable domain of the heavy chain of the camelid heavy-chain antibody) encoding sequences were amplified to generate a phage display library⁵⁰. Nanobodies were further validated through binding experiments in HEK293T cells transfected with various human and rat GABA_B receptor constructs, including two GB1 isoforms (human GB1a, human GB1b), a GB1 mutant in which the ER retention signal RSR (Arg-Ser-Arg) at the C-terminal end of the coiled-coil domain was mutated to ASA (Ala-Ser-Ala), allowing cell surface expression of the construct expressed alone (human GB1^ASA, rat GB1^ASA), GB2 subunit (human GB2, rat GB2) and GB2 truncations that lack the VFT domain (human GB2^ΔVFT, rat GB2^ΔVFT) (Extended Data Fig. 1a). Two nanobodies Nb-4C10 and Nb-5E9 were found to bind to GABA_B receptor with high-affinity (K_d = 4.0 nM and 6.2 nM, respectively) (Extended Data Fig. 1b, Extended Table 1). At a concentration of 1 μM, Nb-4C10 binds to both human and rat GB1a^ASA, but shows no binding to human GB1b, human GB2 or rat GB2 (Extended Data Fig. 1b). This demonstrates its specific binding to the GB1a subunit isoform. As the GB1b isoform lacks the presence of a pair of sushi repeats in the N-terminal ectodomain of GB1a²⁴, it suggests the epitope of Nb-4C10 is within the sushi domains of GB1a. The Nb-5E9 binds to the human and rat GB2 subunit, but shows no binding to human and rat GB1a or GB1b subunit, or human and rat GB2^ΔVFT truncations (Extended Data Fig. 1b), indicating its specific binding to the GB2 subunit, particularly to the VFT of GB2. Both Nb-4C10 and Nb-5E9 display the same affinities for the active (agonist GABA-treated) and inactive (antagonist CGP54626-treated) forms of GABA_B receptor (Extended Data Fig. 1c, d), indicating that Nb-4C10 and Nb-5E9 bind to both conformational states of GABA_B receptor. In all, we have identified one nanobody (Nb-4C10) specific of both human and rat GB1a and another one (Nb-5E9) specific of the VFT of both human and rat GB2 subunits.

To address whether the binding of one nanobody affects the binding of another, we performed competitive binding assays using HEK293 cells expressing human GABA_B receptors. Cells were incubated with 10 μM Alexa Fluor 647-conjugated nanobody, followed by increasing concentrations of the corresponding unlabeled nanobody (Extended Data Fig. 2a). We observed that competition occurred in 2 h after the addition of the competing nanobody only within identical nanobody pairs (Nb-4C10 competed with Nb-4C10^AF647 (Extended Data Fig. 2a), and Nb-5E9 competed with Nb-5E9^AF647 (Extended Data Fig. 2b), indicating that Nb-4C10 and Nb-5E9 bind independently to distinct, non-overlapping epitopes. Therefore, the binding of one nanobody does not interfere with the other, ensuring that our dual labeling strategy allows accurate quantification of dimers and tetramers in cells.

We further developed a nanobody toolbox based on the Nb-4C10 and Nb-5E9, including various Fc-fusions such as Nb-4C10^Rabbit-Fc (Nb-4C10^rFc), Nb-4C10^Mouse-Fc (Nb-4C10^mFc), and Nb-5E9^Mouse-Fc (Nb-5E9^mFc) (Extended Data Fig. 1e, f) and bivalent formats (Nb-4C10^Bi) (Extended Data Fig. 1g, h), in which two GB1a-targeting nanobodies were linked via a GSG linker. All engineered nanobodies exhibit higher affinity than the monovalent ones (Extended Table 1).

Expression and activation of GABA_B hetero-tetramers

The nanobody-based proximity ligation assay (PLA) method achieves high-sensitivity detection of protein interactions using specific nanobodies engineered with rabbit-Fc and mouse-Fc domains to target proteins, utilizing a pair of DNA probes conjugated to anti-rabbit and anti-mouse antibodies for spatial proximity ligation and amplification⁵¹. Here, we developed a GABA_B receptor Nb-based PLA assay to evaluate the organization of GABA_B heterodimer and hetero-tetramer in the cell membrane. The Nb-4C10^rFc and Nb-4C10^mFc, both targeting GB1a (referred to as GB1 later), were used to detect GABA_B hetero-tetramer, as these results from a direct interaction between the GB1 subunits. As discussed later, GB1 strict homodimers are not expected to represent a large fraction of the GB1 subunits at the cell surface due to their stable association with GB2^52,53, and their retention in the ER in the absence of GB2^38,39. Nb-4C10^rFc and Nb-5E9^mFc, targeting GB1 and GB2, respectively, were used to detect GABA_B heterodimer, either strict heterodimers or heterodimers being part of larger complexes (Fig. 1a). In HEK293 cells transfected with human GB1 and GB2, fluorescence dots were clearly detectable using both pairs of antibodies. The intensity of these signals was significantly higher compared to that of the control antibodies (Nb-4C10^rFc and mouse IgG) (Fig. 1b), demonstrating the formation of both human GABA_B heterodimer and hetero-tetramer in HEK293 cells overexpressing human GABA_B receptor subunits. As a control for the PLA assay, the mGlu2 homodimer was validated by quantifying PLA-dots obtained with the mGlu2 nanobodies, and were found to represent 138% those obtained with the GB1-GB2 nanobodies, suggesting a higher expression level of the mGlu2 homodimers than GABA_B heterodimers (Extended Data Fig. 2c). In contrast, the mGlu2-GB1 number of PLA-dots per µm² represent about 21% of the values measured with GB1-GB2, offering a good estimate of PLA signals resulting from a random proximity between two non-interacting cell surface proteins expressed at this level (Extended Data Fig. 2c). Due to the identical amino acid sequence of the sushi repeats in the GB1 subunit among human, rat and mouse, we extended the use of these nanobodies in primary mouse cerebellar granule neurons to examine the native GABA_B receptor. Similar fluorescence dots were also observed in neurons (Fig. 1c), providing additional evidence for the presence of both native GABA_B heterodimer and hetero-tetramers.

Fig. 1. Detection of GABA_B hetero-tetramer expression and activation.

a Schematic representation of the PLA assay detecting GABA_B heterodimers and hetero-tetramers using Fc-conjugated nanobodies. The antibodies targeting GB1 subunit (rabbit Fc-fused Nb-4C10, referred to as Nb-4C10^rFc) and GB2 (mouse Fc-fused Nb-5E9, referred to as Nb-5E9^mFc) are used to detect GABA_B heterodimer. The rabbit Fc-fused and mouse Fc-fused Nb-4C10 (Nb-4C10^rFc and Nb-4C10^mFc) targeting GB1 are used to detect GABA_B hetero-tetramer. Representative images of PLA experiments performed with Nb-4C10^rFc and Nb-5E9^mFc (anti-GB1/GB2), or Nb-4C10^rFc and Nb-4C10^mFc (anti-GB1/GB1), or Nb-4C10^rFc and non-specific mouse IgG (anti-GB1/IgG) in HEK293 cells expressing human GB1a and human GB2 (b) or in primary mouse cerebellar granule neurons (c). PLA signal is shown in red dots, cell nuclei are stained with 4′,6-diamidino-2-phenylindole (DAPI) in blue. Scale bar: 10 μm. Data are representative of three independent experiments performed on different cultures. The histograms show mean ± s.e.m. (b n = 20 GB1/GB2, n = 19 GB1/GB1, n = 19 GB1/IgG; c n = 28 GB1/GB2, n = 20 GB1/GB1, n = 23 GB1/IgG) of PLA dot number per region of interest cells from three independent biological experiments. Data are analyzed using one-way analysis of variance (ANOVA) with a Dunnett’s post-hoc multiple comparison test to determine significance (compared with GB1/IgG group). ****P < 0.0001. d Schematic representation of the G protein recruitment to human GABA_B heterodimer and hetero-tetramer using CODA-RET system, in the presence of the agonist GABA or an allosteric agonist-positive allosteric modulator (ago-PAM), rac BHFF. e The effect of GABA and rac BHFF on the G protein recruitment to human GABA_B heterodimer (left panel) and hetero-tetramer (right panel). Data are presented as mean ± s.e.m. from five independent biological experiments each performed in triplicates, and the estimated TOP values of GABA and rac BHFF were analyzed using unpaired t-test (two-tailed) to determine significance in the inserted graph. **P = 0.0057 and ****P < 0.0001.

To examine the functional properties of both GABA_B heterodimers and hetero-tetramers, we employed the complemented donor-acceptor bioluminescence resonance energy transfer (CODA-RET)⁵⁴ technique to measure G protein signal exclusively induced by either heterodimers or hetero-tetramers (Fig. 1d and Extended Data Fig. 3). This method relies on the NanoBiT system in conjunction with a bioluminescence resonance energy transfer (BRET) assay. In our experiments, the two fragments derived from Nluc, the 17.6-kDa subunit (LgBiT) and the modified small 9-amino acid subunit (HiBiT)⁵⁵, were fused to the C-terminus of either GB1 or GB2. When GB1_LgBiT interacts with GB2_HiBiT in a heterodimer or with another GB1_HiBit in a hetero-tetramer, the LgBiT and HiBiT fragments come into close proximity, enabling efficient complementation and reconstitution of functional Nluc. Additionally, in order to perfectly control the surface expression of both GB1 and GB2 subunits, an ER retention motif KKXX was added at the C-terminal end of the GB2 subunit⁵⁶. As such, neither GB1 nor GB2_KKXX could reach the cell surface alone. With these subunits, the coiled coil interaction between the C-terminal tails of GB1 and GB2, being able to mask both the natural and introduced ER retention motifs of GB1 and GB2, respectively, only the GB1-GB2 heterodimers could reach the cell surface. In parallel, we expressed a miniGαi protein labeled with the acceptor molecule Venus. Since only the GABA_B hetero-tetramer and heterodimer containing the complemented Nluc can generate a luminescent signal, this approach allows for the specific measurement of G protein recruitment to these receptor populations (Extended Data Fig. 3). Using this approach, we found that GABA, which binds to the VFT of GB1, induced G protein recruitment in both GABA_B heterodimer and hetero-tetramer. Meanwhile, rac BHFF, a positive allosteric agonist (ago-PAM), which binds at the 7TM interface between GB1 and GB2^42,44,45, also induced G protein recruitment to both GABA_B heterodimer and hetero-tetramer (Fig. 1e). However, while the effect of GABA and rac BHFF at 100 μM showed no significant difference in GABA_B heterodimer, the effect of rac BHFF was significantly stronger in GABA_B hetero-tetramer by 3-fold of that of GABA, suggesting a distinct mode of G protein activation controlled by GABA_B hetero-tetramer, with a much lower efficacy of agonists compared to the ago-PAM.

Distinct pharmacogenetic property of GABA_B hetero-tetramer

Several mutations in the GB1 and GB2 encoded genes, which are located in the GB1 VFT and the TM6 of GB2, have been reported to be associated with Rett syndrome, epileptic encephalopathies and milder forms of developmental delay with absence epilepsy^30,31 (Fig. 2a and Extended Data Fig. 6a). The functional characterization of these variants and the investigation of the drug response were performed using GABA_B heterodimer as a model⁴². However, the pharmacological property of these mutations in GABA_B hetero-tetramers remains unknown. Using the CODA-RET system, we tested seven GABA_B receptor variants, including E368D, A397V, and A535T mutations in GB1 VFT (Fig. 2a, b), and A567T, S695I, I705N, and A707T mutations in GB2 TM6 (Extended Data Fig. 6a, b). We found that baclofen-induced G protein coupling by the E368D, A397V, and A535T mutants was more profoundly decreased in the GABA_B hetero-tetramer compared to the heterodimer. While the A397V and A535T mutations had very limited effect on the heterodimer, they profoundly affected the functional response of the hetero-tetramer. Similarly, while E368D did decrease heterodimer function, this mutation almost suppressed activity of the hetero-tetramer (Fig. 2c, d). Similarly, rac BHFF that showed slight or no decrease in its efficacy at heterodimer mutants E368D and A397V (decrease 21% and 18%, respectively), or A535T (no effect), had a large drop in efficacy at the mutant hetero-tetramers (87%, 47%, and 59% decrease, respectively) (Fig. 2c, d). To further quantify ligand efficacy, we reanalyzed the concentration–response curves using the operational model^57–59, which yields the parameter logτ as a measure of functional efficacy assuming a similar receptor expression. The resulting logτ values followed a trend consistent with the differences in maximal response (TOP values) observed between wild-type and mutant receptors (Extended Data Fig. 4a). Similar results were observed for GABA in the GABA_B heterodimer and hetero-tetramer (Extended Data Fig. 4b, c). Most receptors were expressed at a similar level as the WT, except E368D, which showed a two-fold lower expression level compared to WT (Extended Data Fig. 4d). However, E368D of the hetero-tetramer mutant showed significant loss of function (Fig. 2d). Meanwhile, the luminescence signal showed no significant difference between GABA_B WT and mutants (Extended Data Fig. 4e). The effects of these heterodimer mutants on overall G protein signaling were also confirmed using conventional BRET assay detecting the change between GB2-Nluc and Venus-miniG_i1, which was similar as the result in CODA-RET assay (Extended Data Fig. 5a–e). We found the effect of baclofen on the G protein coupling efficacy of mutants A567T, S695I, I705N and A707T in both GABA_B heterodimer and hetero-tetramer were significantly decreased by 21–81% compared to the respective WT (Extended Data Fig. 6c, d). The efficacy of rac BHFF at 100 μM showed no significant change in GABA_B heterodimer mutants in A567T, S695I and I705N, while a slight increase was found in A707T (increase 32% respectively) (Extended Data Fig. 6c). In GABA_B hetero-tetramer, the efficacy of rac BHFF showed slight decrease for A567T, S695I and I705N (decrease 13%, 25% and 40% respectively), while significant increase in A707T by 42% (Extended Data Fig. 6d). Further analysis using the logτ values confirm this difference in receptor efficacy (Extended Data Fig. 6e). And most receptors were expressed at a similar level as the WT (Extended Data Fig. 6f, g). These results showed that the pharmacological and functional properties of GABA_B heterodimers and hetero-tetramers are differently affected by genetic variations. Some mutations affect mainly the hetero-tetramer, revealing its functional importance for normal brain function in humans. Such data highlight the importance of characterizing both forms of GABA_B receptor during drug development, and push us to determine the structure of the hetero-tetramer to get more information on the structural basis for the negative control of one receptor by the other.

Fig. 2. Human disease-associated mutations in GB1 VFT mainly affect the activity of GABA_B hetero-tetramers.

Genetic mutations in GB1 VFT which are reported to participate in epileptic encephalopathies. Location of these GB1 mutations (E368, A397, A535) is shown in the VFT of GB1 (a) within the GABA_B heterodimer and the orthosteric binding site of GB1 VFT (b). Effects of baclofen (upper panel) and rac BHFF (lower panel) on the G protein recruitment to GABA_B heterodimer (c) or hetero-tetramer (d) of WT or indicated mutants (with E368D, A397V, A535T in GB1, respectively). The inserted graphs represent the estimated TOP values of baclofen and rac BHFF on G protein recruitment to GABA_B heterodimer and hetero-tetramer WT or indicated mutants. Data are presented as means ± s.e.m. (c heterodimer baclofen, n = 6 WT, n = 4 E368D, n = 6 A397V, n = 6 A535T, rac BHFF, n = 6 WT, n = 6 E368D, n = 5 A397V, n = 6 A535T; d hetero-tetramer baclofen, n = 6 WT, n = 6 E368D, n = 6 A397V, n = 6 A535T, rac BHFF, n = 6 WT, n = 6 E368D, n = 6 A397V, n = 6 A535T) from at least three biological independent experiments each performed in triplicates, and analyzed using one-way ANOVA with a Dunnett’s post-hoc multiple comparison test to determine significance (compared with WT). ****P < 0.0001, ****P < 0.0001 and ****P < 0.0001; **P = 0.0025, *P = 0.0150 and not significant (ns); not determined (ND), ****P < 0.0001 and ****P < 0.0001; ****P < 0.0001, ****P < 0.0001 and ****P < 0.0001.

Structural determination of GABA_B hetero-tetramer

To determine the structure of the GABA_B hetero-tetramer, we initially co-expressed human GB1 and GB2 under identical conditions previously established for the GABA_B heterodimer^40,42. The size-exclusion chromatography (SEC) revealed distinct peaks corresponding to GABA_B heterodimer and higher-order oligomers. However, the proportion of oligomers was notably low (Extended Data Fig. 7a–c). To address this limitation, we employed a strategy to combine the NanoBiT system⁵⁵ with the LgBiT or HiBiT fused to the C-terminus of GB1 (GB1^LgBiT and GB1^HiBit) and the co-incubation of bivalent nanobody⁶⁰ Nb-4C10^Bi in the membrane extracts (Extended Data Fig. 7d). It significantly facilitated the stabilization of the tetrameric form of GABA_B receptor during purification (Extended Data Fig. 7e, f), thereby enhancing the efficiency of isolating GABA_B hetero-tetramer for the determination of its cryo-electron microscopy (cryo-EM) structure.

We successfully determined the structure of the GABA_B hetero-tetramer in apo state (Fig. 3a). Representative 2D class averages indicated the assembly of two dimeric units (Extended Data Fig. 8a). In total, 55,594 particles were selected for 3D reconstruction, achieving a tetrameric structure with an overall resolution of 3.83 Å (Extended Data Fig. 8 and Extended Table 3). To obtain the positioning of another 7TM-unit and enhance the resolution of the VFT region, we performed the local refinement on these two parts, respectively, based on the overall map. This enabled high-resolution visualization of the VFT domains at 3.70 Å, facilitating accurate placement of most side chains within the VFTs. The density of the sushi domains was not visualized in the cryo-EM structure as reported before^40,43, as well as Nb-4C10, which might be due to their high flexibility. Although the 7TMs were resolved at a lower resolution, the 7TM-unit1 domains at 6.07 Å still permitted unambiguous domain docking, facilitating the construction of the GABA_B hetero-tetramer model (Fig. 3b and Extended Data Fig. 8).

Fig. 3. Cryo-EM structure of the full-length GABA_B hetero-tetramer in apo state.

Cryo-EM map (a) and model (b) of the GABA_B hetero-tetramer, which consists of one heterodimer (unit 1) formed by GB1 (cyan) and GB2 (yellow) subunits, and another heterodimer (unit 2) formed by GB1′ (pink) and GB2′ (purple) subunits. c, d Side view of the unit 1 and unit 2 (left and right) and extracellular view of the 7TMs of the unit1 and unit 2 (middle) compared with GABA_B heterodimer (PDB: 7C7S) when the extracellular domains are superposed.

Our findings reveal that the GABA_B hetero-tetramer comprises two GABA_B heterodimers through the interaction between GB1 subunits. However, different from the reported GB1 homodimer (PDB: 6W2Y) structure, which is symmetrically organized, our data revealed that two GB1s in the GABA_B hetero-tetramer adopt a more complicated asymmetrical structure. If the two GB2 subunits from the GABA_B heterodimers (PDB: 7C7S) are added to the GB1 homodimer (PDB: 6W2Y) to form a symmetrically organized chimeric hetero-tetramer, significant steric clashes are observed in both the VFT and 7TM domains (Extended Data Fig. 9), also supporting a preference for asymmetric arrangement in the GABA_B hetero-tetramer. When comparing the tetrameric units with the GABA_B heterodimer in its apo state, we observed significant conformational changes in unit 1, where GB2 rotated away from GB1, resulting in a translational shift of approximately 30–37 Å in the 7TM region (Fig. 3c). In contrast, the unit 2 exhibited more modest alterations, with minimal changes in the extracellular region, though conformational adjustments in the stalk region led to a 25° rotation of the 7TM (Fig. 3d). Despite the differing extents of conformational change, the integrity of the TM3-5/TM3-5 interface was preserved in each GABA_B heterodimers.

Asymmetric interaction between GB1 VFTs

The cryo-EM structure of GABA_B hetero-tetramer revealed an asymmetric interface formed by the GB1 VFTs (Fig. 4a). Notably, the LB2 of one GB1 subunit (GB1) closely interacts with the cleft formed between LB1 and LB2 of the other GB1 subunit (GB1’). Detailed analysis revealed that residues V412-E414 in GB1 form a close contact with residues A397-N399 and S458-E460 in GB1’ (Fig. 4b). They are located in the flexible loops (referred to as loop 1-3) of GB1 and GB1’. We then generated two mutants of GB1: GB1^A397N/N399T and GB1^E414N. Both mutants then carry an NXS/T sequence to allow the attachment of a bulky N-glycan moiety to the side chain of the Asn residue, creating a steric wedge to interference the GB1 and GB1’ interaction^17,47 (Fig. 4c). We found that the maximal Ca²⁺ release response induced by GABA increased by 80% for GABA_B receptor formed with GB1^A397N/N399T and GB2 and by 30% for GABA_B receptor formed with GB1^E414N and GB2, compared to the wild type GABA_B receptor (Fig. 4d). The lower effect observed with the E414N mutant compared with GB1^A397N/N399T is likely explained by its 2-fold lower expression level relative to WT (Extended Data Fig. 10a). These results demonstrate that the efficacy of the G protein coupling is negatively controlled within the GABA_B hetero-tetramer by the asymmetric interactions between the two GB1 VFTs.

Fig. 4. Asymmetric interaction of GB1 VFTs within the GABA_B hetero-tetramer.

a Interactions between the VFT domain in GB1 and another VFT in GB1′. b Detailed interaction between the Lobe2 (LB2) of one VFT domain in GB1 and the Lobe1′ (LB1′) and Lobe2′ (LB2′) of another VFT domain in GB1′. c N-glycosylation mutants in three VFT interface loops of the GABA_B hetero-tetramer. The amino acids were mutated into the NXS/T sequence to enable the attachment of a bulky N-glycan moiety to the side chain of the Asn residue, thereby creating a steric wedge. d GABA-induced Ca²⁺ release in GABA_B WT or N-glycolation mutants. Data are presented as mean ± s.e.m. from three independent biological experiments each performed in triplicate, and analyzed using one-way ANOVA with a Dunnett’s post-hoc multiple comparison test to determine significance (compared with the E_max or pEC₅₀ of GABA_B WT). ****P < 0.0001 and **P = 0.0025; ****P < 0.0001 and **P = 0.0036. e Comparision of the three loops involved in the GABA_B hetero-tetramer VFT interface in human and Drosophila GB1. There are six additional residues in loop3 of Drosophila GB1, which may cause steric clashes and affect the interaction with another Drosophila GB1, likely preventing hetero-tetramer formation. f GABA-induced Ca²⁺ release in human GABA_B receptor WT or a mutant with six residues (EEGYDI) from the loop 3 in Drosophila GB1 inserted after K455 in human GB1 (6AA mutant). Data are presented as mean ± s.e.m. from three independent experiments each performed in triplicates, and analyzed using unpaired t-test (two-tailed) to determine significance. ***P = 0.0002; *P = 0.0148.

Tracing the evolutionary history of GABA_B receptors shows that while these receptors are found in various species, five species, including Drosophila, have evolved in specific lineages (Extended Data Fig. 10b, c). Structural comparison for three loops linking the LB1 and LB2 between human and Drosophila GABA_B heterodimer composed by dGB1 and dGB2 reveal that the loop 1 and loop 2 shares high similarity (Fig. 4e). However, loop 3 in Drosophila GB1, which corresponds to E459-E460 in the human ortholog, is longer with six additional residues (EEGYDI) (Extended Data Fig. 10c), likely preventing dGB1-dGB1 interaction to form hetero-tetramer. The effect of this loop was validated by the PLA assay measured between GB1 and GB1, which showed a lower number of dots/µm² (Extended Data Fig. 10d). A similar value was obtained with the Drosophila GB1 constructs in which we added the sushi domains to allow the generation of PLA signal with the nanobodies, indicating that the addition of the drosophila loop in the human GABA_B receptor results in the same PLA signal as that measured with the drosophila receptors (Extended Data Fig. 10d). Taken into account the background signal measured between GB1 and mGlu2, the introduction of the drosophila loop led to a two-fold decrease in the PLA signal, consistent with a lower proportion of GABA_B hetero-tetramers.

Inserting these six residues into the human GB1 also significantly enhanced the GABA-induced Ca²⁺ release response by 50% for maximal efficacy (Fig. 4f), further supporting the pivotal role of the asymmetric VFT interaction in the formation of the GABA_B hetero-tetramer and in the negative control of G protein coupling efficacy. Interestingly, when further comparing the residues of the three loops with the corresponding sequences of other class C family receptors, we found that there are dozens of extra residues near the position of loop 3 for both calcium-sensing receptor and metabotropic glutamate receptors (Extended Data Fig. 11a). Structural comparisons of these class C GPCR dimers also showed that it is very unlikely for these receptors to form tetrameric or multimeric forms through such VFT interaction (Extended Data Fig. 11b, c), consistent with previous studies suggesting that metabotropic glutamate receptors are incapable of forming oligomers in their apo state⁶¹.

We further investigated whether both GB1 are able to bind the ligand and whether both sites are required for function. We first performed structural modeling. Specifically, we individually aligned the closed conformation of the GB1 VFT domain (derived from the GABA_B dimer bound to baclofen) to either the GB1 or GB1′ subunit within our tetrameric structure (Fig. 5a–c). These modeling studies revealed that each GB1 VFT domain could independently adopt a closed conformation without steric clashes, suggesting that each GB1 subunit in the GABA_B tetramer possesses the potential to bind ligand (Fig. 5a, b). However, when we attempted to simultaneously model two closed GB1 VFT domains onto the two GB1/GB1’ subunits within the tetramer, we observed steric hindrance (Fig. 5c). This indicates that while both GB1 subunits can potentially bind ligand, only one subunit can undergo VFT domain closure at a given time due to these steric constraints. Then, we introduced two mutations (S247A and E466A) into the VFT of one GB1 to prevent GABA binding as previously reported¹⁷, to generate a GB1^ΔB mutant (Extended Data Fig. 12a). We co-transfected GB1_LgBiT, GB1^ΔB_HiBiT and GB2_KKXX to express a GABA_B hetero-tetramer mutant with only one VFT of GB1 able to bind GABA (Fig. 5c). Using the CODA-RET system, we found that this mutant was able to recruit G proteins with GABA stimulation, but the G protein coupling efficacy was significantly decreased by half compared with WT hetero-tetramer, while the potency was not affected (Fig. 5d). Assuming, as indicated above, that a single heterodimer can be activated by agonists within a tetramer, these data are consistent with the view that only one binding site in a tetramer is sufficient to activate the receptor. Indeed, in the tetramer bearing a GB1^ΔB mutant subunit, no effect of agonist can be generated when this subunit is the one that can be activated, explaining the 2-fold decrease in efficacy, confirming a coordinated action of GABA_B hetero-tetramer. The GABA-induced G protein recruitment to either GABA_B heterodimer or hetero-tetramer was totally blocked by the antagonist CGP 54626 (Extended Data Fig. 13a–c). However, CGP 54626 decreased rac BHFF-induced G protein recruitment to GABA_B heterodimer and hetero-tetramer only partially (Extended Data Fig. 13d, e), indicating that action at the heterodimer 7TM interface can still lead to G protein recruitment when both GB1 VFTs are maintained in their inactive open state.

Fig. 5. Allosteric interaction between VFTs of two GB1s within the GABA_B hetero-tetramer.

a, b Superposition of GB1 or GB1’ in the GABA_B hetero-tetramer with the GB1 in closed conformation (PDB: 7EB2). Neither GB1′ (a) nor GB1 (b) exhibits steric hindrance with GB1 in the activated state. c Comparing two activated GB1 subunits along with two GB1 subunits in a GABA_B hetero-tetramer, there was a potential spatial hindrance at the interface of the two GB1s following closure, suggesting that simultaneous activation of both GB1 subunits may be limited. d Schematic representation of the G protein recruitment to the GABA_B hetero-tetramer WT or mutant with only one VFT binding to agonist (S247A and E466A in one GB1) using the CODA-RET assay. e The effect of GABA on the G protein recruitment to GABA_B hetero-tetramer WT or S247A-E466A mutant. The graphs represent the estimated TOP values and pEC₅₀ of GABA on G protein recruitment to GABA_B hetero-tetramer WT or indicated mutants. Data are presented as mean ± s.e.m. from five independent biological replicates each performed in triplicates, and analyzed using unpaired t-test (two-tailed) to determine significance. ****P < 0.0001; ns not significant.

GB1 7TM of one heterodimer contacts GB2 of the other dimer

Different from the VFTs, the 7TMs of the GABA_B hetero-tetramer exhibited an overall symmetric organization in a rhomboidal shape. Both heterodimeric units adopted a TM3/5-TM3/5 interactions between GB1 and GB2, similar to that observed in the apo state of the isolated GABA_B heterodimer⁴⁰. Meanwhile, the TM6s from GB1 and GB1’ formed a symmetric interface between two units, consistent with our previous results⁴⁶ (Fig. 6a). Interestingly, we found another interface between the TM1 and TM7 of GB1 of one unit (referred to as GB1) and the TM4 of GB2 of the other unit (referred to as GB2′) (Fig. 6a). In the GABA_B hetero-tetramer, the bottom of both TM3 helices of the GB1-GB2 subunits maintain tight binding with a distance of 9 Å (Fig. 6b). At the top of TM5, despite a distance of 17 Å, the corresponding position was found to be occupied by cholesterol in the previous heterodimer, acting as a mediator in the middle of the interface indirectly maintaining TM3/5-TM3/5 interactions^62,63, and was also validated by cross-linking⁴⁶. We found that the distances between TM1 and TM7 of GB1 and TM4 of GB2′ are 22 Å and 26 Å, respectively (Fig. 6b). Although about 10 Å longer than the TM3/5-TM3/5 distance, it is much closer than the 37 Å distance between TM6-TM6 of GB1-GB1′ and the 52 Å distance between TM4/5-4/5 of GB2-GB2′ (Fig. 6b), implying that the 7TM regions of GB1 and GB2′ may have a greater propensity for coordinated interactions. We applied the cross-linking strategy as previously to verify this interface^16,46,64,65. We mutated several potential residues into cystine. If these cystines are in close contact, a covalent disulfide bond will form between them under the treatment of oxidative copper-phenanthrolin (CuP), allowing the detection of a high molecular weight band for hetero-tetramer in blots. We observed such a band in cells transfected with GB1-K590^1.38C and GB2-A617^4.52C, or GB1-D831^TM7C and GB2-A617^4.52C (Fig. 6c). We also tested four pairs of additional mutant constructs (GB1-S594^1.42C and GB2-I613^4.48C, GB1-L598^1.46C and GB2-C610^4.45, GB1-A835^7.30C and GB2-C614^4.49, GB1-L839^7.34C and GB2-L609^4.44), and none of these alternative mutants led to crosslinked oligomers (Extended Data Fig. 14a, b), further indicating that a close proximity between TM1 or TM7 of the GB1 from one unit with the TM4 of GB2 from the other unit. These data confirm that the 7TMs of the GABA_B hetero-tetramer are arranged in a symmetrical rhomboid shape, following the connection sequence of GB1-GB2′-GB1′-GB2.

Fig. 6. The arrangement of the 7TMs within the GABA_B hetero-tetramer.

a Extracellular view of the GABA_B hetero-tetramer 7TMs in the apo state conformation. b 7TMs interfaces and distances between the helixes in different subunits. c Cross-linking of cell surface GB2 and GB1′ subunits between the indicated residues mutated into cysteine, without (CuP−) or with CuP (CuP+). The percentage of hetero-tetramer are presented as mean ± s.e.m. from three independent biological replicates, and analyzed using unpaired t-test (two-tailed) to determine significance (compared with CuP−). ns not significant, **P = 0.0056 and *P = 0.0014. d Cartoon illustrates the negative allosteric modulation of G proteins activation within the GABA_B hetero-tetramer, comparing with the GABA_B heterodimer. The GABA_B hetero-tetramer assembles via asymmetric interactions between VFT domains from two GB1 subunits of distinct heterodimers. GABA binding to a single VFT domain activates only one unit within the hetero-tetramer, enabling G protein coupling. The asymmetric interactions provide the negative functional cooperativity between the two heterodimers.

Discussion

Although allosteric interactions within GPCR oligomers have been reported^{15,17,18,47,66}, their existence and function in native tissues and the structural basis for these interactions remain largely unknown. In the present study, we bring clear evidence for the existence of complexes made of two or more GABA_B heterodimers in primary neurons using specific GB1 and GB2 nanobodies, and their likely involvement in human brain diseases. We demonstrate that the effect of an agonist is impaired in GABA_B hetero-tetramer relative to heterodimer, while that of an ago-PAM is not. By determining the cryo-EM structure of the GABA_B hetero-tetramer in its apo state, our study provides much detailed information on the basis of such allosteric interaction within GABA_B hetero-tetramers.

The nanobodies used in the PLA assay allow the detection of GB1 proximity. Accordingly, this approach cannot exclude that part of the signal measured can be due to the presence of GB1 homodimers at the cell surface. However, such homodimers likely represent a very minor fraction of the signal. Indeed, GB1 contains an ER retention sequence that prevents it from reaching the trans-Golgi network, thanks to its recognition by the ER gatekeeper PRAF2⁶⁷. GB1 surface expression is also controlled at a later step by msec7-ARF acting at a second retention site (a di-Leucine motif) within its C-terminal tail⁶⁸. These systems are prevented by the GB1-GB2 association, allowing GB1 to reach the cell surface only when associated with GB2^38,39. Moreover, in the absence of GB2, GB1 is rapidly degraded, leading to a very low amount of GB1 subunits⁶⁹, further indicating that GB1 homodimers, if any, represent a very small fraction of the GB1 subunits at the cell surface. One could propose a possible dissociation of GB1-GB2 heterodimers while at the cell surface, enabling the formation of GB1 homodimers. Again, such a possibility is very unlikely due to the strong association between these two subunits brought by the coiled coil interaction between their C-terminal intracellular domains^52,53. Even the tetramers were shown to be very stable and not capable of exchanging subunits⁴⁸.

In our study, we found that the PLA signal (Dots/ROI) measured between GB1 and GB2 is about twice as big as that measured between GB1 and GB1, both in transfected cells and in cultured neurons. According to the high proximity between heterodimers within a tetramer, we can assume that a single dot will be generated by a tetramer when using the heterodimer detection tools. Accordingly, the number of dots detected with the GB1-GB2 detector, very likely represents the total number of strict heterodimers plus the tetramers. In contrast, the dots generated by the GB1-GB1 PLA tools will only reveal tetramers. Such observations suggest that about half of the GABA_B complexes are tetramers, the other half being strict heterodimers. However, more work is needed to validate the linearity of the PLA signal under our conditions, as a function of the protein expression, to make sure such data can be used as a quantification method. Moreover, the GB1-GB1 PLA signal may also account for some heterodimers in close enough proximity, but not assembled into tetramers, as this is allowed by the size of the a nanobody + secondary antibody. Although we cannot provide a good estimate of the proportion of tetramers, our and previous data suggest that they do represent a significant fraction of the receptors.

The GABA_B hetero-tetramer structure reveals a direct contact between the GB1 subunits of each heterodimer at the level of their VFTs. Such interaction is found to be asymmetric, specifically, LB2 of the GB1 subunit from one heterodimer interacts with LB1 and LB2 of the GB1 subunit from the other heterodimer. Such a finding is consistent with GB1 mutants impairing the formation of hetero-tetramers reported previously^17,47 and in the present study. This is in contrast to the classical view of allosteric complexes that are mainly viewed as symmetric cooperative machines^13,20–23.

In contrast to the asymmetric organization of the VFT in hetero-tetramer, the four 7TM units are organized in a symmetrical rhomboid-like structure, with an interface involving TM5 of GB1 and GB2 subunits in both heterodimer units, but additional contact between GB1 from one unit and GB2 from the other unit is observed. This assembly is consistent with the cross-linking experiments between GB1 and GB2. Such additional interaction between the heterodimers at the 7TM level can explain the high stability of GABA_B hetero-tetramer⁴⁸. Indeed, neither mutations at the GB1 VFTs interface nor the insertion of glycosylation moiety can fully disrupt the two heterodimers, even though the negative allostery between the GB1 VFTs is largely decreased⁴⁷. Such an organization appears not to prevent each GB2 7TM from activating G proteins, as the effect of the ago-PAM appears similar on heterodimers and hetero-tetramers. Such 7TM organization in hetero-tetramer is made possible through a major change in the relative position of the GB1-GB2 7TMs relative to the VFTs in unit 1, in agreement with the asymmetric organization of the four VFTs. Such an organization provides an explanation for the negative allostery between the two GABA_B heterodimers. Indeed, although both GB1 VFTs are open and ready to bind an agonist, the closure of one of them is likely to prevent the closure of the other one, as steric clashes are observed in a closed-closed model. As such, only one heterodimer within the hetero-tetramer will likely be able to generate a signal, consistent with the present and previous findings^17,47.

Although the detailed functional roles of GABA_B hetero-tetramers are unknown, we found mutations reported in humans, and associated to brain diseases that have differential consequences on the function of GABA_B hetero-tetramers versus heterodimers. Of particular interest, some mutations found in GB1 VFT mainly affect hetero-tetramer function, with a very limited effect on heterodimers. Considering that when heterodimer function is examined, one cannot rule out a certain proportion of tetramers, this reinforces the conclusion that these mutations mainly affect hetero-tetramer function. Such observation then highlights the importance of a normal activity of GABA_B hetero-tetramer for brain function. They also highlight the need to better characterize the effect of GABA_B receptor auto-antibodies^32,33 responsible for human diseases on both heterodimers and hetero-tetramers. These observations underscore the critical need for a thorough understanding of GABA_B hetero-tetramers in drug development. The distinct pharmacological characteristics of these receptor complexes may have important implications for therapeutic strategies targeting GABA_B receptor, particularly in the context of receptor mutations associated with disease.

In summary, our functional and structural characterization of the GABA_B hetero-tetramer reveals a form of negative cooperativity within the oligomer, driven by the asymmetric organization of the VFT and symmetric organization in 7TMs. These findings bring much evidence for the existence of such complexes in vivo. These findings not only deepen our understanding of GABA_B receptor signal integration but also highlight the critical importance of targeting receptor oligomers in drug development.

Methods

This research complies with all relevant ethical regulations. All animal experiments were approved by the Animal Experimentation Ethics Committee of the College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China.

Constructs

For functional experiments, the pCDNA3.1 plasmids encoding the HA-tagged wild-type GB1 and Flag-tagged GB2 from human were used as previously^40,42. A large fragment (LgBiT) and a small fragment (HiBiT) of the NanoBiT luciferase were cloned after the C-terminal of GB1 (WT or mutant). The Venus was fused before miniGi protein as previous reported⁷⁰. The site-directed mutations of GABA_B receptor were generated using the QuikChange mutagenesis protocol (Agilent Technologies, Santa Clara, CA, USA). All the constructs were confirmed by sequencing.

For protein production, human GB1a (residues 15–860) and GB2 (residues 42–780) cDNAs were cloned into the pEGBacMam vector as previously described. In brief, the original GB1 and GB2 signal peptides were substituted by haemagglutinin (HA) signal sequence. To facilitate expression and purification, an 8 × histidine tag and 3C protease cleavage site were inserted at the C-terminus of GB1a subunit, while a Flag epitope tag (DYKDDDD) and a 2 × GSG linker were added to the N-terminus of the GB2 subunit. To assist the formation of the GABA_B hetero-tetramer, LgBiT or HiBiT were added to the C-terminus of GB1a as previously⁵⁵, to increase the stability of the hetero-tetramer.

Alpaca immunization, VHH library construction and phage display

The membrane extraction of HEK293 cells overexpressing human GABA_B receptor were used as the antigen for the generation of specific nanobodies as our previously published protocol⁷¹. Briefly, a healthy young Alpaca was immunized subcutaneously with a solution of cell membrane extract. Peripheral blood was collected during and after immunization. The peripheral blood mononuclear cells (PBMCs) were isolated and combined. Total RNA was extracted from the PBMCs and reverse-transcribed into cDNA. VHH genes were amplified from cDNA by a two-step nested PCR, digested with restriction enzymes, and ligated into the pHEN1 phagemid vector. The pelB signal sequence is located at the N-terminus of the VHH gene and leads to the periplasmic expression of nanobodies. The ligated phagemids were transformed into E. coli TG1 competent cells. All transformants were collected and stored as an immune VHH library.

The immune VHH library was used to select and screen nanobodies targeting the human GABA_B receptor. A representative fraction of the VHH library was cultured and rescued using the KM13 helper phage to display nanobodies at the tips of the M13 phage. Phages were extracted from the E. coli TG1 strain and used for phage display biopanning in HEK293 cells overexpressing the human GABA_B receptor. During the bio-panning process, negative phage-displayed VHHs were removed by HEK293 cells depletion with mock transfection. The phage was enriched and eluted with a trypsin solution. After phage infection with E. coli TG1, a representative fraction of the bacteria was plated, and hundreds of colonies were randomly inoculated and cultured for positive clonal VHH screening. These gIIIp-fused VHHs were expressed by isopropyl-β-D-thio galactopyranoside (IPTG) induction, and the periplasmic extracts were screened for specific binding to receptors using flow cytometry. The results of these selected clones were compared with those of negative and non-specific binding controls. Clones with a specific fluorescence intensity distribution of the anti-His-tag antibody detected by flow cytometer were recognized as positive and subjected to Sanger sequencing. The remaining phages were rescued and amplified for the next round of phage display.

Production and purification of nanobodies

DNA sequencing results from the positive clones were translated into amino acids for analysis. All amino acid sequences were aligned, repeated or redundant sequences were removed, and only clones with unique amino acid sequences were retained for further analysis. The pHEN1 phagemids encoding unique VHH genes were extracted from the cultured E. coli TG1 and transformed into E. coli BL21(DE3) competent cells, as the amber codon stops translation before the gIIIp gene⁷². High-purity soluble nanobodies were extracted from the periplasm of cultured E.coli inducing by IPTG and purified using a His-tagged protein purification protocol, followed by size-exclusion chromatography.

Flow cytometry-based binding assay

HEK293T cells were transiently transfected with HA-tagged GB1 and Flag-tagged GB2. Twenty four hours after transfection, cells were detached with trypsin-EDTA and resuspended in ice-cold assay buffer (PBS, 2%BSA) at 4 million live cells/ ml. Cells were incubated for 1 h at RT with Nb at various concentrations and 1 h anti-His-tag Alexa Fluor® 647-conjugated Antibody (R&D SYSTEMS, catalog no. IC0501R). Cells were washed twice with PBS and resuspended in FACS buffer (PBS, BSA 0.5%). The binding was recorded on viable single cells and on the allophycocyanine (APC) channel using a CytoFLex Flow Cytometer (Beckman Coulter, Brea, CA, USA). Data were analyzed with FlowJoTM v10 software and cells were gated for (i) singlets and (ii) living cells. The analyzed data were plotted using GraphPad Prism Version 9.2 software.

Conjugation of nanobodies to fluorescent dye

Nanobodies (1 mg in PBS) were incubated with 5 molar equivalents of Alexa Fluor 647-NHS (Thermo Fisher, USA). Unreacted dye was removed by PD-10 desalting column (Cytiva, USA). Conjugated nanobodies were concentrated and the degree of labeling was calculated per the manufacturer’s protocol.

Competition test of nanobodies by flow cytometry

HEK293 cells overexpressing human GABA_B receptor were incubated with 10 μM Alexa Fluor 647-conjugated nanobody, followed by the addition of increasing concentrations of unlabeled nanobody for 2 h to allow competitive binding. The percentage of positive events were recognized as the statistical value to characterize their competitive binding to the human GABA_B receptor.

HEK293 cell culture and transfection

HEK293 cells (ATCC, CRL-1573, lot: 3449904) were cultured in DMEM supplemented with 10% FBS at 37 °C in a humidified incubator containing 5% CO₂. For transfections, cells were suspended and transfected using Lipofectamine 2000 with the appropriate expression constructs following the instructions.

Primary cerebellar granule neuron culture

For the primary culture of neurons, all experiments were specifically designed to minimize the number of animals used and were approved by the Animal Experimentation Ethics Committee of the College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China. Kunming mice were obtained from the Center for Disease Control and Prevention of Hubei Province. The mice were raised in a specific pathogen-free (SPF) environment with an ambient temperature of 18–22 °C, a humidity of 50–60%, and a 12 h light-dark cycle.

Primary cerebellar granule neuronal cultures were prepared from 1-week-old newborn Kunming mice as previously described⁷³. The dissected tissue was gently triturated after Versene (15040066; Gibco, Shanghai, China) treatment for 5 min at 37 °C, and the homogenate was centrifuged at 170 × g for 5 min. The pellet was re-suspended in culture medium DMEM-F12 (Ref 11320-033; Gibco) supplemented with 2 mM glutamine, 30 mM KCl, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% fetal bovine serum (FBS) and seeded into dishes previously coated with poly-L-ornithine (Sigma-Aldrich).

Proximity ligation assay (PLA)

For experiments performed in cell cultures, HEK293 cells and primary mouse cerebellar granule neurons were blocked with 10% horse serum and permeabilized with 0.2% Triton X-100 (v/v). Experiments were then performed using PLA kit (Sigma, Cat. No. DUO92101) according to the manufacturer’s instructions (Duolink & PLA Technology, Olink-Bioscience)⁷⁴. Briefly, cells were incubated with Nb-4C10^rFc and Nb-5E9^mFc, or Nb-4C10^rFc and Nb-4C10^mFc, or Nb-4C10^rFc and mouse IgG at 37 °C for 1 h. PLA probes anti-rabbit plus and anti-mouse minus, which are secondary antibodies conjugated with oligonucleotides, were added and incubated for 1 h at 37 °C. Amplification template oligonucleotides were hybridized to pairs of PLA and circularized by ligation. The formed DNA circle was then amplified using rolling circle amplification, and detection of the amplicons was carried out using the PLA kit, resulting in red fluorescence signals. Cells were mounted and analyzed under an oil immersion objective (×100) using a confocal microscope (Olympus FV3000). All acquisitions were made the same day under the same microscope setting and laser power to minimize variability among samples. Distinct bright spots contained within an area of the cells designated by the experimenter were counted using an ImageJ macro. Briefly, regions of interest (ROIs) corresponding to individual cells were defined, and image segmentation was performed by applying a threshold to generate binary images. The Analyze Particles function in ImageJ was then utilized to count particles larger than 0.198 μm² within the binary images, which were identified as the number of PLA dots. The resulting data were imported to Prism software (GraphPad) for further statistical analysis and visualization.

Bioluminescence resonance energy transfer (BRET) assay

HEK293 cells (ATCC, CRL438 1573, lot: 3449904) were transfected with GABA_B heterodimer (GB1_LgBiT + GB2_HiBiT) or mutants of chimeric NanoBiT GABA_B receptor (GB1_LgBiT + GB1_HiBiT + GB2_KKXX) WT or indicated mutants, together with Venus-miniG_i1. Lipofectamine 2000 was used for the cDNA transfection into HEK293 cells. After 36 h transfection, cells were washed and starved in PBS at 37 °C for 1.5 h before the BRET experiments. BRET measurements were performed as previously described using the PHERAstar FS (BMG Labtech, USA)⁷⁵ with the programme PHERAstar control (Version 4.00 R4). The signals emitted by the donor (460–500 nm band-pass filter, Em 480 nm) and the acceptor (510–550 nm band-pass filter, Em 530 nm) were recorded after the addition of 10 μM furimazine. All measurements were performed at 37 °C. The BRET signal was determined by calculating the ratio between the emission of acceptor and donor (BRET = Em 530 nm/Em 480 nm). The change in BRET ratio (net BRET) was obtained by subtracting the BRET ratio after agonist treatment from the basal BRET (net BRET = BRET_basal− BRET_agonist).

Determining logτ using the operational model of agonism

Concentration-response data were analyzed using the operational model of agonism (Black and Leff, 1983) to quantify agonist efficacy, expressed as logτ. For each experimental batch containing wild-type and mutant receptor data, a global nonlinear least-squares fit was performed. In this approach, the maximum system response (E_max) and the transducer slope factor (n) were constrained as shared parameters across all groups, while logτ and logKA were estimated independently for each receptor variant. The fitting was implemented with the Levenberg–Marquardt algorithm via the minpack.lm (v.1.2-4) package in R (V.4.3.0). For mutants exhibiting low responses (e.g., E368D), the upper bound of logKA was constrained to avoid ambiguous parameter solutions, thereby ensuring that the estimated logτ reflected true differences in efficacy rather than artifactual shifts in affinity. From the model fit, logτ and its standard error were extracted directly for subsequent statistical comparison and interpretation of agonist efficacy across receptor constructs.

Cell surface quantification by ELISA

Cell surface and total expression of the indicated constructs were quantified by an enzyme-linked immunosorbent assay (ELISA). Cell culture medium was removed from 96-well plates 48 h after transfection, then cells were washed and fixed with 4% paraformaldehyde (Sigma-Aldrich), then blocked with 10% FBS (for the total expression group, 0.1% Triton X-100 (Thermo Fisher Scientific) was required before this step to increase the permeability of the cell membrane). Flag-tagged constructs were detected with the monoclonal mouse anti-Flag antibody coupled with horseradish peroxidase (1:20,000; F1804, Sigma-Aldrich). HA-tagged constructs were detected with the rat anti-HA antibody coupled with horseradish peroxidase (1:20,000; 3F10, Roche, Indianapolis, USA). Bound antibodies were detected by chemoluminescence using a Super Signal substrate (Thermo Fisher Scientific) and a 2103 EnVision Multilabel Plate Reader (Perkin Elmer, Waltham, MA, USA).

Protein expression and purification of GABA_B hetero-tetramer

The purified plasmid DNA HA-GB1a-LgBit-His, HA-GB1a-HiBit-His, and HA-flag-GB2 were mixed with PEI in a 3:1 ratio of PEI to DNA (w/w) followed by addition to HEK293F cells when the density reached around 3.0 million/ml. Ten micromolar sodium butyrate was added 16–18 h post-transfection. Cells were grown for 3 days at 30 °C, then harvested. Cells were collected and washed with phosphate-buffered saline buffer. Finally, the cell pellets were rapidly frozen in liquid nitrogen and stored at 80 °C for later use.

The cells were centrifuged, washed with PBS, and resuspended in HEPES buffer (pH 7.5) containing NaCl, glycerol, MgCl₂, and CaCl₂, then homogenized. Membranes were solubilized with 0.5% (w/v) LMNG and 0.1% (w/v) CHS for 3 h at 4 °C. After centrifugation, the supernatant was purified by Ni-NTA and anti-FLAG affinity chromatography. The protein was eluted with HEPES buffer (pH 7.5) containing NaCl, LMNG, CHS, EGTA, and FLAG peptide. The GABA_B receptor was concentrated and purified by size-exclusion chromatography on a Superose 6 Increase column. The sample was pre-screened for sample quality by negative-stain transmission electron microscopy and then immediately prepared on cryo-EM grids for structural determination. For the GABA_B hetero-tetramer, protein was collected and concentrated to 10 mg/mL for electron microscopy studies.

Cryo-EM grid preparation and data collection

For the preparation of Cryo-EM grids, 3 μl of the purified sample was applied onto a glow-discharged 200 mesh holey carbon grid (Quantifoil R1.2/1.3). The grids were blotted for 3.5 s with a blot force of 10 at 4 °C and 100% humidity, followed by plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). Cryo-EM data collection was performed on a Titan Krios microscope (Thermo Fisher Scientific) at 300 kV accelerating voltage equipped with a Falcon 4 direct electron detector and a SelectrisX energy filter performed at 10 eV slit width. Data was automatically collected by EPU software in electron event representation (EER) mode with a raw pixel size of 0.93 Å on the image plane. Image stacks were obtained at a dose rate of ~5.56 electrons per Ų per second with a defocus ranging from −0.8 to −2.4 μm. The total exposure time was 7.36 s, resulting in a total dose of ~50 electrons per Ų. A total of 58,261 movies were collected for GABA_B hetero-tetramer.

Imaging processing and 3D reconstruction

Image stacks were processed by CryoSPARC software (v 4.4.0)⁶⁸. The 58,261 movies were subjected to patch motion correction and patch CTF estimation. Particles were picked by Blob picker and selected by multiple rounds of 2D classification. By using 2D classification and the stochastic gradient descent algorithm in CryoSPARC, we obtained five initial models for subsequent processing. The particles in clearly defined 2D classes were then subjected to several rounds of heterogeneous refinement to isolate unique conformational states and eliminate particles in poorly defined classes. Final global refinement of the hetero-tetramer maps was performed with the CryoSPARC non-uniform refinement and local refinement with masks including only the VFTs or TMs. Composite maps were generated with Chimera (v.1.15)⁷⁶. Map sharpening was performed with CryoSPARC and deepEMhancer⁷⁷. Validation of cryo-EM maps and models was performed with Phenix (v.1.20.1-4487). UCSF Chimera (v.1.15) and Chimera X (v.1.5 and v1.6.1)⁷⁸ were used for map/model visualizations and figure preparation.

Model building and refinement

The initial model for the GABA_B hetero-tetramer was generated based on the structure of antagonist-bound GABA_B (PDB 7C7S)⁴². All of the models were docked into the electron microscopy density maps in UCSF Chimera⁷⁸. The docked model was subjected to flexible fitting using Rosetta⁷⁹ and was further rebuilt in COOT⁸⁰, followed by real-space refinement in Phenix⁸¹. The final refinement statistics were validated using the module “comprehensive validation (cryo-EM)” in Phenix⁸¹. The goodness-of-fit of the model to the map was determined using a global model-versus-map Fourier shell correlation. Structure figures were created using UCSF Chimera and the UCSF Chimera X package.

Intracellular calcium release measurements

After 36 h transfection with plasmids encoding the indicated GABA_B receptor subunits and a chimeric protein G_qi9, HEK293 cells were loaded with 1 mM Ca²⁺-sensitive fluorescent dye Fluo-4 AM (Molecular Probes, Eugene, OR, USA) for 1 h at 37 °C. After a wash, cells were incubated with buffer and GABA solution at various concentrations was added after 40 s of recording. Fluorescence signals (excitation 485 nm, emission 525 nm) were measured by using the fluorescence microplate reader FLIPR Tetra (Molecular Devices, Sunnyvale, CA, USA) at sampling intervals of 1 s for 120 s. The response of Ca²⁺ release was generated by the peak value of fluorescence signals after drug injection subtracting the mean value of fluorescence signals before drug injection. All the values of drug treatment were subtracted the value in buffer treatment as the net Ca²⁺ release. Data were analyzed with the program Soft Max Pro (Molecular Devices, Sunnyvale, CA, USA).

Cross-linking and fluorescent-labeled blot experiments

HEK293 cells were transfected with human HA-Halo-GB1a and Flag-GB2 or indicated mutations by lipofectamine 2000 and plated in 12-well plates for 36 h. Then cells were labeled with 100 nM Halo-660 non-cell permeant reagent in culture medium at 37 °C for 2 h. Afterwards, cross-link buffer (1.5 mM Cu (II)-(o-phenanthroline), 1 mM CaCl₂, 5 mM Mg²⁺, 16.7 mM Tris HCl, pH 8.0 and 100 mM NaCl) was added at 37 °C for 20 min. After incubation with 10 mM N-ethylmaleimide at 4 °C for 15 min to stop the crosslinking reaction, cells were lysed with lysis buffer (containing 50 mM Tris pH 7.4, 150 mM NaCl, 1% NP-40 and 0.5% sodium deoxycholate) at 4 °C for 1.5 h. After centrifugation at 12,000 × g for 30 min at 4 °C, supernatants were mixed with no DTT loading buffer at 37 °C for 10 min before loading the samples. Equal amounts of proteins were resolved by 4–12% SDS-PAGE. Proteins were transferred to nitrocellulose membranes (Millipore). Membrane was imaged on an Odyssey CLx imager (LI-COR Bioscience, Lincoln, NE, USA) at 700 nm and 800 nm.

Statistics and reproducibility

Statistical analysis was performed on at least three individual data sets and analyzed in GraphPad Prism software version 8. Data are present as mean ± s.e.m from at least three independent experiments, with the number of independent experiments indicated in the figure legend. For dose-response curves, data are analyzed using nonlinear curve fitting for the log (agonist) vs. Response (three parameters) curves. Data are analyzed using the one-way ANOVA with Dunnett’s multiple comparison test or unpaired student t test to determine significance. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, not significant (ns) > 0.05.

Reporting summary

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

Author contributions

J.L. conceived and supervised the whole project; C.S. expressed and purified the GABA_B hetero-tetramer; H.D. generated the constructs and performed the functional experiments; S.Z. prepared nanobodies with the help of Y.-Z.L.; C.X. designed the functional experiment protocols; C.X., C.S., and H.D. designed the constructs for functional assays; B.Z., C.S., and S.J. prepared the cryo-EM grids, data collection, data processing and structure determination under the supervision of Y.Z. and J.H.; D.-D.S. evaluated the sample by negative-stain EM; Y.-R.L. cultured cells with the help of R.Z.; J-Y.L helped with Elisa experiment; Y-X.L. and X.C. helped with the BRET experiments; C.S., C.X., and H.D. prepared the figures and participated in manuscript writing advised by J.L.; P.R. and J-P.P. participated in the interpretation of the data and the preparation of the manuscript; J.L., J.P.P. and C.X. wrote the manuscript with inputs from all the authors.

Peer review

Peer review information

Nature Communications thanks Andreas Bock, Ross Cheloha, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data generated in this study are included in the main text or the Supplementary Information. Cryo-EM map has been deposited in the Electron Microscopy Data Bank under accession codes: EMD-67312 for GABA_B hetero-tetramer. The atomic coordinates have been deposited in the Protein Data Bank under accession codes: 9XVZ for GABA_B hetero-tetramer. PDB: 6W2Y, 7C7S, 7EB2, 7SIN, 7DGD, 7EPA, 7WI8, 8WGD, 6N52, 7EPC. Source Data can be found in the Source Data File. Source data are provided with this paper.

Competing interests

J.L., S.Z., and Zhiran Fan have filed a patent application (application no: 2025100458291, patent authorized). The patent protects any possible use of the GABA_B nanobodies described in this manuscript. The remaining authors declare no other competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70640-8.