Nanobody-based sensors reveal a high proportion of mGlu heterodimers in the brain

Membrane proteins, including ion channels, receptors and transporters, are often composed of multiple subunits and can form large complexes. Their specific composition in native tissues is difficult to determine and remains largely unknown. In this study, we developed a method for determining the subunit composition of endogenous cell surface protein complexes from isolated native tissues. Our method relies on nanobody-based sensors, which enable proximity detection between subunits in time-resolved Förster resonance energy transfer (FRET) measurements. Additionally, given conformation-specific nanobodies, the activation of these complexes can be recorded in native brain tissue. Applied to the metabotropic glutamate receptors in different brain regions, this approach revealed the clear existence of functional metabotropic glutamate (mGlu)2–mGlu4 heterodimers in addition to mGlu2 and mGlu4 homodimers. Strikingly, the mGlu4 subunits appear to be mainly heterodimers in the brain. Overall, these versatile biosensors can determine the presence and activity of endogenous membrane proteins in native tissues with high fidelity and convenience. Using nanobodies labeled with FRET fluorophores, the authors show the presence and activation of GPCR mGlu2 and mGlu4 dimers in mouse brain samples and reveal that mGlu2–mGlu4 is the major form of mGlu4-containing dimers outside the cerebellum.

Ten years ago, we reported that these mGlu subunits could also assemble into heterodimers with specific combinations, revealing the possible existence of 16 additional mGlu receptors 26 . Since then, specific pharmacological properties of these heterodimers have been reported, providing indirect evidence of their existence in the brain [27][28][29] . However, such data could also be explained by functional cross-talk between colocalized homodimeric receptors. In addition, these studies did not reveal the proportion of such heterodimers in the brain compared to homodimers.
In the present study, using specific nanobodies for both mGlu2 and mGlu4 subunits, we were able to reveal the existence of mGlu2-mGlu4 heterodimers, in addition to mGlu2 and mGlu4 homodimers, in various brain regions. We also confirm their specific pharmacological properties and reveal that mGlu2-mGlu4 is a major type of receptor containing the mGlu4 subunit in the brain outside the cerebellum. These data demonstrate the effectiveness of our approach in deciphering the subunit composition of membrane protein complexes in their native environment and in providing relative quantification of endogenous membrane receptor species in native tissues.

Results
Among the possible 16 mGlu heterodimers observed in recombinant cells 26,30 , the mGlu2-mGlu4 heterodimer is the most investigated [27][28][29] , but its existence and abundance in the brain remain unclear. To clarify this issue in different brain areas, we have developed two kinds of nanobody-based TR-FRET sensors ( Fig. 1): (1) conformational sensors or 'biosensors' that reveal the activation of these receptors upon agonist binding and (2) the 'detectors' that enable the relative quantification of both mGlu homodimers and heterodimers.
An mGlu2 FRET-based conformational biosensor. We first developed a 'biosensor' for the mGlu2 homodimer by taking advantage of a pair of specific and high-affinity nanobodies for the mGlu2 receptor, DN10 and DN1 (ref. 15 ). DN10 specifically recognizes the receptor dimer in its active state, whereas DN1 is not sensitive to the conformational state (Fig. 2a). The DN10 epitope overlaps that of DN13 (ref. 15 ), the nanobody contacting both subunits, which is located at an interface of the two ECDs of the mGlu2 homodimer in its active form exclusively 24 . In contrast to DN13, DN10 can also bind to the active mGlu2-mGlu4 heterodimer, as shown below, likely because the mGlu4 part of the epitope is compatible with DN10 binding. By contrast, the DN1 epitope remains unknown 15 (Supplementary Tables 1 and 2). When DN1 and DN10 were covalently labeled with donor Lumi4-Tb and acceptor d2, respectively, a FRET signal was measured in cells expressing mGlu2 in the presence of the mGlu2 and mGlu3 agonist LY379268 but not with the antagonist LY341495 (Fig. 2b). No signal was measured when DN10-d2 was absent (Extended Data Fig. 1a), and the FRET signal followed a saturation curve with the increase of DN10-d2 and a fixed concentration of DN1-Tb (Extended Data Fig. 1b). The signal was specific to mGlu2, as no signal was measured with other mGlu receptors (Extended Data Fig. 1c). Finally, DN10-Tb and DN1-d2 can also detect endogenous mGlu2 in rat hippocampal neurons by TR-FRET microscopy (Extended Data Fig. 2).
This pair of nanobodies could also be used with dissociated cells from different mouse brain regions (Extended Data Fig. 3a,b). The more cells, the higher the FRET signal with the nanobodies in the presence of agonists (Extended Data Fig. 3c). The slopes, representative of the FRET signal per amount of brain cells, revealed a high signal in the cerebellum and other regions (Fig. 2c), consistent with DN1-d2 staining of brain slices (Extended Data Fig. 4) 31,32 . No signal was detected from Grm2-knockout (Grm2 −/− ; called mGlu2-KO mice in this study) mice ( Fig. 2c and Extended Data Fig. 4). Altogether, these data validate the use of this pair of nanobodies in the detection of endogenous active mGlu2 receptors.
This conformational biosensor is a sensitive tool to report the rearrangement of the mGlu2 ECD upon agonist activation. In transfected cells, the TR-FRET signal generated using increasing concentrations of various full and partial agonists revealed potencies and efficacies consistent with results from a SNAP-tag FRET-based assay 33,34 (Fig. 2d). The potencies also correlated well with those measured by the accumulation of inositol phosphate-1 (IP 1 ) (Fig. 2e,f). Such a good correlation between the potencies of partial and full agonists observed in both assays was not expected, as the amplification resulting from receptor reserve is expected to increase the potencies of full agonists more than those of partial agonists. The good correlation may be due to the fact that the efficacy of partial agonists is closer to a full efficacy in the presence of the G protein bound to the active receptor 33,35 . In sum, these data show that this nanobody-based biosensor constitutes a new generation of untagged mGlu conformational sensors.
This biosensor can also reveal the activation of endogenous mGlu2 receptors in dissociated cells from various brain regions. The agonist-induced change in FRET observed in the cortex, hippocampus and cerebellum was similar to that found with transfected cells (Fig. 2g-i). The antagonist LY341495 was also found to inhibit both the basal signal likely generated by ambient glutamate in the assay and the response evoked by the concentration of agonist required to give 80% of a maximum response (the EC 80 ). Altogether, our data demonstrate that this optical nanobodybased biosensor can be used to reveal the activation of endogenous mGlu2 receptors in native dissociated tissues. It represents an innovative assay for throughput screening of drug efficacy on native mGlu2 receptors in brain tissue.
Quantification of the mGlu2 homodimer. We next aimed to quantify mGlu2 homodimers using DN1-Tb and DN1-d2 as a TR-FRET pair, first on the surface of transfected cells (Fig. 3a). As expected, this 'detector' signal was independent of the conformation of the homodimer, whether bound to an agonist or an antagonist (Fig. 3b). The concentrations of the nanobodies were optimized to have a FRET signal proportional to the quantity of mGlu2 over a wide range of receptor amounts (Extended Data Fig. 5a-c). This 'detector' was also specific for mGlu2 among all mGlu homodimers (Fig. 3c). We also believe that the FRET signal was mostly due to the mGlu2

Fig. 2 |
A nanobody-based biosensor to detect the expression and activation of the mGlu2 receptor in both transfected cells and dissociated brain tissues. a, Schematic representation of the TR-FRET-based mGlu2 receptor conformational sensor in the presence of agonist (ago) or antagonist (antago) to stabilize the active and inactive states, respectively. TMD, transmembrane domain. This sensor was made of 7.5 nM donor nanobody DN1-Tb (in pink) (Lumi4-Tb is shown as a circled 'D') and 15 nM acceptor nanobody DN10-d2 (in purple) (d2 is shown as a circled 'A'). b, TR-FRET signal measured in HEK293 cells transiently transfected to express mGlu2 or in mock cells in the presence of the biosensor with agonist LY379268 (1 μM) or antagonist LY341495 (10 μM). Data are represented as mean ± s.e.m. of triplicate measurements in three independent experiments. c, Analysis of the relative expression of the mGlu2 receptor in brain tissues in the presence of LY379268 (1 μM). The TR-FRET signal indicates the slope values of the relative linear quantification experiments. Each dot represents a TR-FRET experiment performed on the indicated brain tissue of one mouse (n = 5 for all samples of wild-type (WT) mice, n = 2 for the cerebellum of the mGlu2-KO mouse group and n = 3 for the cerebellum of mGlu4-KO mouse group). Data are mean ± s.e.m. and were analyzed using one-way ANOVA followed by Dunnett's post hoc test (compared with the cerebellum of the mGlu2-KO group), with ****P < 0.0001 for all, except for the midbrain (**P = 0.0044). d,g-i, Dose-dependent effects of the ligands on the TR-FRET signal of the biosensor measured in HEK293 cells transiently cotransfected to express mGlu2 and the high-affinity glutamate transporter EAAC1 (d) or in dissociated cells from the cortex (g), the hippocampus (h) or the cerebellum (i). DCG-IV, (2S,1′R,2′R,3′R)-2-(2,3-dicarboxycyclopropyl)glycine. e, Dose-dependent effects of the indicated ligands on IP 1 accumulation measured in HEK293 cells cotransfected to express mGlu2, EAAC1 and the chimeric G protein G qi9 . In d,e and g-i, data are mean ± s.e.m. of three independent experiments performed in triplicate and normalized to the response to LY379268. f, Correlation between the half-maximum effective concentration potencies (pEC 50 ) determined with the indicated agonists on the conformational biosensor (d) and IP 1 assay (e).
Data are mean ± s.e.m. of three independent experiments.
homodimer and not to higher-order oligomers. First, the FRET signal between the DN1 nanobodies was proportional to the number of SNAP-mGlu2 subunits on the cell surface over a wide range of receptor amounts (Extended Data Fig. 5c). Second, this linearity was also observed with the 'controlled' mGlu2 homodimer formed by the mGlu2 C1 and mGlu2 C2 subunits as previously described 26,36 (Extended Data Fig. 5d). In these constructs, the C terminus of the mGlu2 subunits was replaced by that of the modified γ-aminobutyric acid (GABA) B1 (C1) or GABA B2 (C2) subunits, respectively, preventing any of these from reaching the cell surface alone. Indeed, only C1-C2 dimers can reach the cell surface 26 . However, when using a similar controlled mGlu2-mGlu4 heterodimer made of mGlu2 C1 and mGlu4 C2 subunits, no FRET signal with the DN1 'detector' was measured (Extended Data Fig. 5d). Under these conditions, the mGlu2-mGlu4 heterodimer is present at the cell surface in the absence of both mGlu2 and mGlu4 homodimers. This is consistent  ). c,f, TR-FRET signal measured in HEK293 cells transfected to express the indicated mGlu receptors or in mock cells with LY341495 (10 μM). In b,c,e,f, data are mean ± s.e.m. of triplicate measurements from one representative experiment of three independent experiments. g,h, Relative expression of the mGlu2 (g) and mGlu4 (h) homodimers in the indicated brain tissues as shown by their respective 'detectors'. The TR-FRET signal indicates the slope values of the relative linear quantification experiments. Each dot represents a TR-FRET measurement from one mouse (n = 5 for all samples of wild-type mice, n = 2 for the cerebellum of the mGlu2-KO mouse group and n = 3 for the cerebellum of the mGlu4-KO mouse group). Data are mean ± s.e.m. and were analyzed using one-way ANOVA followed by Dunnett's post hoc test (the control group is the cerebellum of the mGlu2-KO group (g), with ****P < 0.0001 for all, except for the striatum (*P = 0.0164) and the midbrain (not significant (n.s.), P = 0.9996) or the mGlu4-KO group (insert, h), with n.s., P > 0.05 for the olfactory bulb (>0.9999), the striatum (0.0896), the hippocampus (0.8253) and midbrain (0.0615) and the PFC (**P = 0.0031) or Welch's ANOVA test followed by Dunnett's T3 post hoc test (compared with the cerebellum of the mGlu4-KO group) (h), with n.s., P > 0.05 for the olfactory bulb (>0.9999), the striatum (0.2251), the hippocampus (0.8436) and the midbrain (0.1304) and *P < 0.05 for the PFC (0.0425) and the cerebellum of the wild-type group (0.0435).
with our previous demonstration that the controlled mGlu2-mGlu4 heterodimer 26 as well as the mGlu2 homodimer 26,37,38 do not have the tendency to form oligomers in transfected cells. Third, the highest FRET signal measured for equal concentrations of donor (DN1-Tb) and acceptor (DN1-d2) was also consistent with the presence of strict mGlu2 homodimers (Extended Data Fig. 5b). By comparing the mGlu2 homodimer 'detector' signal in different brain areas ( Fig. 3g and Extended Data Fig. 5e), mGlu2 was found to be more abundant in the cerebellum (Fig. 3g). As a control, no signal was observed in the cerebellum of mGlu2-KO mice. However, this 'detector' appeared less sensitive than the 'biosensor' , but this was expected for two main reasons. First, only half of the mGlu2 homodimers can be labeled with two FRET-compatible DN1 nanobodies (DN1-Tb and DN1-d2). By contrast, each mGlu2 subunit will be labeled with two FRET-compatible nanobodies in the biosensor assay, DN1-Tb and DN10-d2, such that each mGlu2 homodimer carries two pairs of FRET-compatible nanobodies. Second, this biosensor assay detects any active form of the mGlu2 subunits, whether in a homodimer or heterodimer, in contrast to the 'detector' , which reveals mGlu2 homodimers only.

Quantification of the mGlu4 homodimer.
To quantify the mGlu4 homodimer, we isolated and characterized the nanobody DN42, highly specific for mGlu4 (Extended Data Fig. 6a,b). For this study, we used the Fc-DN42 dimeric construct (80 kDa), as the DN42 monomer has low affinity after labeling with a fluorophore (~100 nM), which is not compatible with its use in native tissues. Interestingly, Fc-DN42 has subnanomolar affinity for the mGlu4 ECD and a similar affinity for the inactive and active conformations (Supplementary Tables 1 and 2 and Extended Data Fig. 6c). We verified by immunofluorescence that Fc-DN42 was able to specifically detect mGlu4 subunits in brain slices of wild-type mice but not in Grm4-knockout (Grm4 −/− ; called mGlu4-KO mice in this study) mice (Extended Data Fig. 6d).
We used Fc-DN42, from now on referred to as DN42, as a 'detector' to quantify mGlu4 homodimers, similar to what was done with DN1 for the mGlu2 homodimers. With optimized concentrations of DN42 (Extended Data Fig. 7a,b), a strong FRET signal was measured between DN42-Tb and DN42-d2 specifically in cells transfected to express mGlu4 ( Fig. 3d-f), independent of the state of the receptor. As observed with mGlu2, the FRET signal appeared mostly owing to the mGlu4 homodimer and not to higher-order oligomers. First, the signal was proportional to the number of SNAP-mGlu4 subunits on the cell surface over a wide range of receptor amounts (Extended Data Fig. 7c). Second, this linearity was also observed with a controlled mGlu4 homodimer in which the C terminus of mGlu4 subunits was replaced by that of the GABA B receptor (Extended Data Fig. 7d). Lastly, the low FRET signal between mGlu4 subunits measured with the controlled mGlu2-mGlu4 heterodimer (Extended Data Fig. 7d) may be due to random proximity between these heterodimers, although one cannot exclude a very few mGlu2-mGlu4 oligomers. This is also consistent with several reports showing the low probability of mGlu dimers oligomerizing under basal conditions 26,37 .
In native brain tissues, this pair of labeled DN42 nanobodies revealed a detectable signal mainly in the cerebellum of the wild-type and mGlu2-KO mice but not in mGlu4-KO mice ( Fig. 3h and Extended Data Fig. 7e). Significant signal could also be observed in the prefrontal cortex (PFC) (Fig. 3h, inset), but the signals measured in other brain areas were not significant, indicating that our assay was not sensitive enough to detect mGlu4 homodimers in these areas, if any. This agrees with the strong expression of mGlu4 in the cerebellum 31 , as shown also by DN42-stained brain slices, especially in the molecular layer (Extended Data Fig. 6d). Altogether, our results show very low expression of the mGlu4 homodimer in most brain regions except for the cerebellum, which is not really consistent with the immunostaining data 31 (Extended Data Fig. 6d), suggesting that mGlu4 subunits could be associated with other mGlu subunits outside the cerebellum.

An mGlu2-mGlu4 heteromer FRET-based detector.
Recent studies argue in favor of the existence of endogenous mGlu2-mGlu4 receptors in neuronal cell lines 27 as well as in the PFC, striatum and hippocampus [27][28][29] , as suggested by electrophysiological, pharmacological and biochemical data. Although convincing, the results provide indirect evidence of the endogenous mGlu2-mGlu4 heterodimer [27][28][29] . More direct evidence for this heterodimer could come from a proximity assay based on FRET between the mGlu2 and mGlu4 subunits in tissues, owing to the nanobodies described above.
Thus, we used DN1 and DN42 nanobodies to detect mGlu2-mGlu4 heterodimers first on transfected cells. Using the optimized concentrations of DN1 and DN42 (Supplementary Table 1 and Extended Data Fig. 8a,b), a strong FRET signal was measured, proportional to the amount of controlled mGlu2-mGlu4 heterodimers on the cell surface (Extended Data Fig. 8c). We also showed that the FRET signal could not result from mGlu homodimers (Extended Data Fig. 8d). Because mGlu2-mGlu4 is likely coexpressed with mGlu2 and mGlu4 homodimers in native tissues, we used two methods to detect mGlu2-mGlu4 heterodimers in transfected cells. First, we used cells coexpressing mGlu2 C1 and mGlu4 C2 (Extended Data Figs. 5d and 7d), such that only heterodimers could reach the surface (Fig. 4a). Second, cells were cotransfected to express the wild-type mGlu subunits to obtain a mix of mGlu2 and mGlu4 homodimers together with the mGlu2-mGlu4 heterodimer on the cell surface ( Fig. 4b) 26,27 . In both cases, a strong FRET signal was measured whether the receptors were activated with the mGlu2 agonist LY379268 or antagonized with LY341495 ( Fig. 4c,d).
Applied to isolated brain cells, this nanobody-FRET pair generated a strong signal in the olfactory bulb, the PFC, the striatum and the hippocampus (  Fig. 8g). In agreement with these results, co-immunoprecipitation experiments with the olfactory bulb using DN42 revealed the presence of endogenous mGlu2 in the same complexes as mGlu4 (Extended Data Fig. 8f). Notably, the cerebellum did not produce a specific signal between DN1 and DN42 ( Fig. 4e,g), consistent with mGlu2 and mGlu4 subunits being expressed in different cell types 31,39,40 (Extended Data Figs. 4 and 6d). Surprisingly, the slope for the cerebellum was even negative (Fig. 3e,g and Extended Data Fig. 8e,g), most probably due to the relatively high amount of mGlu4 homodimers in this region that was sufficient to titrate the DN42 present at 1.6 nM in the assay, thus resulting in a slight but significant decrease in FRET. In agreement with this hypothesis, the slope was not negative for samples from mGlu4-KO mice, but it was negative in samples from mGlu2-KO mice (Extended Data Fig. 8e,g).

An mGlu2-mGlu4 heterodimer FRET-based biosensor.
Although the above data provide strong evidence for the presence of mGlu2-mGlu4 heterodimers in various brain regions, one cannot exclude the possibility that the signal came from the proximity between mGlu2 and mGlu4 homodimers. To bring further evidence of the existence of the endogenous mGlu2-mGlu4 heterodimer, we developed a nanobody-based FRET conformational sensor for this heterodimer using DN10 and DN42, as DN10 can bind to the mGlu2-mGlu4 heterodimer in the active state only (Extended Data Fig. 9a,b and Supplementary Tables 1 and 2). Of note, L-AP4, a partial agonist of the mGlu2-mGlu4 heterodimer 17,27 , remained partial in promoting DN10 binding (Extended Data Fig. 9a,b), as monitored by the FRET between the nanobody and the SNAP-mGlu4 subunits. Interestingly, DN10 binding to mGlu2-mGlu4 in the presence of the mGlu4 agonist L-AP4 (Extended Data Fig. 9a,c) was strongly potentiated by the mGlu2 agonist LY379268 (Extended Data Fig. 9c,e), whereas L-AP4 did not induce binding of DN10 to the control mGlu2 homodimer (Extended Data Fig. 9d).
By combining DN42, which specifically binds to the mGlu4 subunit, and DN10, which binds to the active form of mGlu2-mGlu4, we could detect mGlu2-mGlu4 activation by FRET (Fig. 5a-f and Extended Data Fig. 9f), a signal that could not result from either homodimers (Extended Data Fig. 9g). Similar data were obtained from cells expressing the controlled mGlu2-mGlu4 (Fig. 5a-c) and  3). For f, n = 3 for the olfactory bulb of wild-type, mGlu2-KO and mGlu4-KO mice. For g, n = 8 for the cerebellum of wild-type mice, n = 3 for mGlu2-KO mice and n = 6 for mGlu4-KO mice. Data are mean ± s.e.m. and were analyzed using one-way ANOVA followed by Dunnett's post hoc test, compared with the cerebellum of the mGlu4-KO group (e), with ****P < 0.0001 for the olfactory bulb and the PFC, **P < 0.01 for the striatum (0.0054), the hippocampus (0.0062) and the cerebellum (0.0045) and n.s., P > 0.05 for the midbrain (0.1533); compared with the olfactory bulb of the mGlu4-KO group (f), with ***P < 0.001 for the wild-type group (0.0002) and n.s., P > 0.05 for the mGlu2-KO group (0.8100); or compared with the cerebellum of the wild-type group (g), with n.s., P > 0.05 for the mGlu2-KO group (0.4120) and ***P < 0.001 for the mGlu4-KO group (0.0005).
from cells expressing both mGlu2 and mGlu4 subunits ( Fig. 5d-f). Under the latter conditions, both homodimers were present on the cell surface along with the heterodimer (Fig. 5d), demonstrating that they do not interfere with the specific signal generated by the active heterodimer. Under these conditions, treatment with the agonist LY379268 generated a large signal that was largely inhibited by the antagonist LY341495, while treatment with L-AP4 generated a smaller signal (Fig. 5b,e). Moreover, as previously reported 17,27,34 , a strong positive cooperativity was observed between the mGlu2 and the mGlu4 agonists on the heterodimer, illustrated here with the large increase in L-AP4 potency caused by a low concentration of LY379268 (Fig. 5c,f and Extended Data Fig. 10d). These data are also consistent with the IP 1 production data obtained from cells expressing controlled mGlu2-mGlu4 or both mGlu2 and mGlu4 (Extended Data Fig. 10a-c). Altogether, these results show that DN10 and DN42 can be used to detect the active form of the mGlu2-mGlu4 heterodimer. This mGlu2-mGlu4 biosensor also detected activation of endogenous mGlu2-Glu4 heterodimer in dissociated brain cells as revealed by the synergy between the mGlu2 and mGlu4 agonists. Activation of endogenous mGlu2-mGlu4 was revealed by the large FRET signal induced by LY379268 in all regions where the mGlu2-mGlu4 heterodimer was detected (that is, the olfactory bulb, PFC, striatum and hippocampus) but not in the cerebellum (Fig. 5g). As expected, the LY379268 effect disappeared in mGlu4-KO brain samples. In addition, in the olfactory bulb (Fig. 5h), L-AP4 potency was increased by a low concentration of LY379268 ( Fig. 5i and Extended Data Fig. 10e), as observed in transfected cells. Finally, a good correlation in agonist potencies measured in transfected cells and in brain samples was observed (Extended Data Fig. 10f and Supplementary Table 3). Altogether, these results bring further direct evidence for the existence of the mGlu2-mGlu4 heterodimer in different brain regions.
mGlu2-mGlu4 heterodimer brain distribution. Intriguingly, we detected higher signals using the mGlu2-mGlu4 'detector' (Fig. 4e) than with the mGlu4 homodimer 'detector' (Fig. 3h) in most brain areas outside the cerebellum. Such a difference in signal intensity was not due to a difference in FRET efficacy between the nanobodies in each dimeric combination. This is best illustrated by the perfect correlation between the FRET obtained with either the mGlu4-specific homodimer DN42 'detector' or the mGlu2-mGlu4-specific heterodimer DN1-DN42 'detector' , relative to the cell surface expression of the SNAP or CLIP subunits (Fig. 6a). These data indicate that there are more mGlu2-mGlu4 heterodimers than mGlu4 homodimers in most regions outside the cerebellum (Fig. 6b).

Discussion
Our study describes an innovative and general method for the quantification and analysis of endogenous multi-subunit membrane proteins using nanobody-based optical sensors. Using this method, we provide direct evidence for the existence of mGlu2-mGlu4 heterodimers in different brain areas. Surprisingly, our results revealed that most mGlu4 subunits are likely associated with another subunit, such as mGlu2, in most brain regions outside the cerebellum. Our method combines the high spatial resolution of TR-FRET technology (<15 nm) with the small size of single-domain nanobodies (~2.5 nm) to detect low amounts of endogenous subunits in native tissues. Obtaining such information in native membranes is essential, as lipid composition and ions likely play an important role in stabilizing protein complexes 41 . No chemical fixation or biochemical treatment of the biological sample is required, in contrast to other analyses, thus preventing conformational changes of the complex. In addition, our results prove that nanobodies have great potential as TR-FRET probes, which help to solve the shortcomings of small molecules in terms of specificity, which limits their use in TR-FRET experiments 4 . Nanobodies have hydrophilic properties, in contrast to small molecules that can be hydrophobic, and help overcome the limitations of classical antibodies in recognizing specific protein conformations. Nanobodies are small antibodies (ten times smaller than immunoglobulin G (IgG) proteins) and easy to engineer and display good and rapid tissue penetrance 42 . They often recognize conformational and cryptic epitopes not accessible to classical antibodies. Our method can be applied to any cell surface protein, including ligand-gated ion channels or transporters 5 . In addition, our method is versatile, as the fluorophores can be covalently attached to small ligands, antibody fragments or common antibodies. Finally, our method does not require a high level of expertise or expensive equipment. It only entails working in microplates with standard biochemical protocols and a standard commercial TR-FRET reader. However, the TR-FRET approach may not be appropriate for the detection of heterodimers using microscopy of brain slices due to the low quantum yield of the donor and the need for special equipment for the time delay between the excitation and the measurement of the emission signal 43 . It can however be used for cultured neurons (Extended Data Fig. 2). However, the use of fluorophores compatible with conventional FRET microscopy may allow the detection of dimers in neuronal subcompartments using microscopy of cultured neurons with better precision.
Our approach has two major advantages in investigating endogenous mGlu heterodimers [27][28][29]44 . First, it analyzes the heterodimer entity directly and not its downstream signaling that could result from cross-talk between signaling pathways 45 . Second, our biosensors are good reporters of the conformational change of the receptor during activation, as are other sensors of the mGlu receptors 27,33,34 . However, whether this approach can be used to detect mGlu receptor activation in real time remains to be tested. For this, the use of fluorophores compatible with conventional FRET measurement will be necessary. It will also be essential to take into consideration the 'ON' rate of binding of the nanobody that recognizes the active form of the receptor, as this may be much slower than the 'ON' rate of mGlu receptor activation that occurs in the submillisecond time scale 35,46 . It is clear that this second point will generate limitations for such analysis. Finally, the pharmacological signature of our new sensors using orthosteric ligands could be defined in transfected cells and could then be observed in native brain samples.
Our study reveals an intriguing distribution of mGlu4 homodimers, mainly found in the cerebellum, where they do not form detectable heterodimers with mGlu2 as expected, as these two subunits are expressed in different types of neurons in the cerebellum. These data appear to be an excellent control for our assay. The absence of significant detection of the mGlu4 homodimer in most brain regions does not exclude the fact that some homodimers may be present. Indeed, mGlu4 homodimers were proposed at hippocampal-medial PFC (mPFC) and amygdala-mPFC synapses 28 and at corticostriatal synapses 29 , suggesting that our approach is not  Fig. 6 | mGlu2-mGlu4 heterodimers are predominant over mGlu4 homodimers outside the cerebellum. a, Relative quantification of mGlu4-mGlu4 homodimers and mGlu2-mGlu4 heterodimers based on the TR-FRET signal between the two subunits in the dimer using either the SNAP-tag or CLIP-tag fluorescent substrates or the 'detector' nanobodies. Data are mean ± s.e.m. of triplicate measurements from one of three experiments. b, Relative amounts of the mGlu4 homodimer compared with the mGlu2-mGlu4 heterodimer, as indicated by the respective circle size, in the adult mouse brain according to the quantification with their respective 'detectors'. The relative amounts of the mGlu2 homodimer determined with its 'detector' are indicated in a separate brain because the sensitivity of this detector is different. sensitive enough to detect these homodimers. However, this conclusion was based on the use of mGlu4 positive allosteric modulators inactive at mGlu2-mGlu4, such as N-phenyl-7-(hydroxyimino) cyclopropa[b]chromen-1a-carboxamide (PHCCC). As the effect of these compounds on mGlu4-mGlu3, mGlu4-mGlu7 or mGlu4-mGlu8 heterodimers 26,30 is not known, further pharmacological studies of these heterodimers will be necessary to clarify this issue. Regardless, our data clearly show that, in many brain regions, there are more mGlu2-mGlu4 dimers than mGlu4-mGlu4 dimers, as a larger signal could be detected with the mGlu2-mGlu4 detector, despite a very similar FRET efficacy per dimer. Interestingly, an astonishing distribution of the mGlu2-mGlu4 heterodimer was observed, with high expression in the olfactory bulb and the PFC, in agreement with the demonstration of mGlu2-mGlu4 at thalamo-mPFC synapses 28 , where they would coexist with mGlu2 homodimers without excluding low amounts of mGlu4 homodimers. This is consistent with the link between the mGlu2 subunit and psychiatric diseases involving the PFC 47 .
Future studies on the existence of other mGlu heterodimers are crucial for assessing the physiological role of mGlu receptors in the brain, potential new drug targets. Indeed, mGlu4 heterodimers could explain the effect of an mGlu4 allosteric modulator acting in the basal ganglia 48 , which had no effect on the mGlu4 homodimer 29 . In addition, mGlu7 heterodimers could also contribute to the enigmatic function of the mGlu7 subunit due to its very low glutamate potency 49 and the effect of mGlu7 negative allosteric modulators with context-dependent activity 50 . Further studies are necessary to clarify these issues, as well as the functional role and therapeutic potential of these mGlu heterodimers, a step that will require the development of specific ligands for these receptor species.
In conclusion, we have reported a general and versatile approach compatible with the quantification and functional analysis of membrane proteins from endogenous native tissues without disrupting the membrane environment, but the availability of specific ligands is a major limitation. However, the number of antibodies targeting these proteins, including nanobodies, is rapidly expanding, even those selective for a conformational state 13,16,18 .

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41589-022-01050-2.

Methods
Animal ethics. This project followed the Animal Welfare Body guidelines and was approved by the internal ethics committee of the Institut de Génomique Fonctionnelle. Wild-type mice were purchased from Janvier Labs, and mGlu2-KO mice 51 were kindly provided by G. Maeso (Virginia Commonwealth University School of Medicine), while mGlu4-KO mice 52 were available at the Institut de Génomique Fonctionnelle. Animals were housed under a 12-h light-dark cycle at 23 ± 2 °C with a relative humidity of 53% ± 10%. Mice had access to water and food ad libitum.
HEK293 cells (ATCC, CRL-1573, lot 3449904) were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% (vol/vol) FBS (Sigma-Aldrich) at 37 °C with 5% CO 2 . HEK293F cells were cultured in suspension in SMM 293-TI animal-free and serum-free medium at 37 °C with 5% CO 2 and shaking at 200 r.p.m. to produce the nanobodies DN42 and Fc-GFP. We routinely checked for mycoplasma using the MycoAlert Mycoplasma Detection kit (Lonza) in accordance with the manufacturer's protocol.
The pRK5 plasmids encoding wild-type rat mGlu1-mGlu8 with the HA-tag and SNAP-tag next to the human mGlu5 signal peptide have been described previously 26 . For the quality-control system, we used the mGlu2/4 C1 or C2 plasmids tagged to express HA or Flag, with or without SNAP-tags 36 . The sequence coding for C1 and C2 contains the coiled-coil region of the C terminus of GABA B1a or GABA B2 and the endoplasmic reticulum-retention signal KKTN, which allows control of the trafficking of the constructs to the cell surface. The other plasmids, namely those encoding the glutamate transporter EAAC1 and the chimeric G protein Gα qi9 , were previously described 26 .
To construct the cDNA for the DN42 and Fc-GFP nanobodies 53 , the monomeric DN42 and anti-GFP sequences were subcloned into pcDNA3.1 (+). The secreted signal peptide of human interleukin 2 (IL-2) was introduced before the N terminus of the nanobody, allowing the secretion of the protein into the culture medium. The sequence encoding the C terminus of the nanobody was fused with the sequence encoding the Fc region of human immunoglobulin IgG 1 . Cloning resulted in the 6× His-tagged sequence fused at the C terminus for the purification and separation of nanobodies by affinity chromatography with metal chelation.
Library construction and DN42 selection. The DN42 nanobody targeting the mGlu4 subunit was selected from the V H H library after all procedures including llama immunization, library construction and selection of nanobodies targeting mGlu4 were performed as described previously 17 .
Bacteria were then infected with the KM13 helper phage, and phage-containing pellets were purified by two selection rounds with 2 × 10 7 HEK293T cells transfected to express rat mGlu4 receptor. Each round was preceded by a depletion step for cells that were not transfected, and positive selection was performed in the presence of an excess of antibodies reacting on control HEK293 cells. Escherichia coli TG1 bacteria were infected with eluted phages and used for sequencing and production of the nanobody.
Production and purification of nanobodies. For the nanobodies DN1 and DN10, production and purification were performed as previously described 15 . Briefly, an E. coli BL21DE3 colony transformed with the pHEN phagemid carrying cDNA for the nanobody of interest was grown in LB medium. The bacteria were cultured in large scale at 37 °C, and nanobody expression was induced with 1 mM IPTG at 28 °C. Bacteria were then collected after centrifugation and treated with different buffers to extract the periplasmic proteins. Nanobody in the periplasmic space was collected from the supernatant after centrifugation at 4 °C.
For large-scale production of the DN42 and Fc-GFP nanobodies, HEK293F cells were cultured at a density of 0.6 × 10 6 ml −1 with 180 ml fresh medium in a 2-L culture bottle at 37 °C with 5% CO 2 and shaken at 200 r.p.m. Of the cells, 1-1.5 × 10 6 ml −1 were transfected with a mixture of Fc-DN42 or Fc-GFP plasmids (225 μg in 12 ml OMEM) and PEI (675 μg in 12 ml OMEM). Cells were cultured for 4-7 d at 37 °C with 5% CO 2 and shaking at 200 r.p.m. The supernatant was collected after a 10-min centrifugation at 2,000g and 4 °C.
The His-tagged nanobodies from both bacteria and HEK293F cells were then purified from the supernatant using Ni-NTA purification (Qiagen) in accordance with the manufacturer's instructions. Finally, the nanobodies were purified by size-exclusion chromatography on a Superdex 200 10/300 column for DN42 and Fc-GFP and a Superdex 75 10/300 column for DN1 and DN10 (GE Healthcare) in PBS (pH 7.4). Nanobody labeling. Nanobodies were dialyzed overnight at 4 °C and incubated (250 μg nanobody per 2 mg ml −1 ) at 20 °C with d2-NHS (PerkinElmer Cisbio) in 0.1 M carbonate buffer (pH = 9) and Lumi4-Tb-NHS in 50 mM phosphate buffer (pH = 8) at a molar ratio of 6 or 12 for 45 or 30 min, respectively. The nanobodies were then purified using a gel filtration column (in 100 mM phosphate buffer (pH = 7). The final molar ratio, namely, the number of fluorophores per nanobody, was calculated as the fluorophore concentration/conjugated nanobody concentration, and conditions were set for a ratio between 2 and 3 (for DN1 and DN10), 2 and 4 (for DN42 labeled with d2) or 5 and 8 (for DN42 labeled with Lumi4-Tb). The concentration of fluorophores in the labeled fraction was calculated as optical density (OD)ε −1 for each fluorophore (OD at 340 nm and ε = 26,000 M −1 cm −1 for Lumi4-Tb, OD at 650 nm and ε = 225,000 M −1 cm −1 for d2), while that of the nanobodies was determined by the OD at 280 nm (OD 280 ). The conjugated concentration was calculated as OD 280 − (OD fluo Rz max TR-FRET binding measurements. The FRET signal was determined by measuring the sensitized d2 acceptor emission (emission, 665 nm) and Lumi4-Tb donor emission (emission, 620 nm) using a 50-μs delay and 450-μs integration upon excitation at 337 nm. All data were obtained using a PHERAstar FS reader (BMG LABTECH). The TR-FRET ratio was calculated as emission at 665 nm ÷ emission at 620 nm × 10 4 , as previously described 15 .
HEK293 cells were cotransfected to express rat SNAP-tagged mGlu and EAAC1 (unless otherwise indicated) using Lipofectamine in a 100-mm cell culture dish in accordance with the manufacturer's instructions. Twenty-four hours after transfection, cells were plated in polyornithine-coated, white 96-well plates (Greiner Bio-One) at 10 5 cells per well and cultured overnight at 37 °C with 5% CO 2 for adherent cell experiments. Cells were labeled with 100 nM SNAP-Lumi4-Tb in DMEM GlutaMAX (Thermo Fisher Scientific) for 2 h at 37 °C and then washed three times with Krebs buffer (10 mM HEPES, pH 7.4, 146 mM NaCl, 4.2 mM KCl, 1 mM CaCl 2 , 0.5 mM MgCl 2 , 5.6 mM glucose and 0.1% BSA). For suspension experiments, cells were frozen at −80 °C with 10% DMSO and FBS and then washed three times with Krebs buffer before use. Five microliters of cells were plated in a white, small-volume 384-well plate (Greiner Bio-One) at 2 × 10 4 cells per well.
To determine the selectivity of DN42 for the mGlu1-mGlu8 receptors, 100 nM nanobody and 200 nM anti-His antibody labeled with d2, and the agonists (1 μM quisqualic acid for mGlu1; 5 or 100 μM L-AP4 for mGlu4, mGlu6, mGlu7 and mGlu8; 100 nM LY379268 for mGlu2 and mGlu3) or the antagonist (10 μM LY341495) were applied to labeled cells in adherent cell experiments, with a total volume of 60 μl per well. After overnight incubation at 25 °C, the TR-FRET signal between Lumi4-Tb and d2 was determined.
To determine the affinity of the c-Myc-tagged nanobodies (DN1 and DN10) and the His-tagged nanobody DN42, 200 nM anti-c-Myc and 100 nM anti-His antibodies, both labeled with d2, were used. The different reagents were applied to labeled cells in adherent or suspension experiments, with a total volume of 20 μl per well incubated overnight at 25 °C. The TR-FRET signal between Lumi4-Tb and d2 was determined.
SNAP subunit quantification. Transfected HEK293 cells were plated in polyornithine-coated, white 96-well plates (Greiner Bio-One) at 10 5 cells per well and cultured overnight at 37 °C with 5% CO 2 ; the cells were then labeled with 100 nM SNAP-Lumi4-Tb in DMEM GlutaMAX for 2 h at 37 °C, followed by three washes with Krebs buffer. The signal was determined by measuring the emission intensity of Lumi4-Tb at 620 nm with a 50-μs delay and 450-μs integration upon excitation at 337 nm. All data were obtained using a PHERAstar FS reader (BMG LABTECH).
Tissue cell sample preparation. To obtain dissociated cells from brain tissue, 6-8-week-old C57BL/6 mice (including wild-type or Grm4 −/− male and female mice) were euthanized; the whole brain was dissected in cold PBS (pH = 7.4) to obtain the regions of interest, according to the Allen Mouse Brain Atlas (https:// mouse.brain-map.org/) 54 . Tissues were quickly cut into small pieces using a scalpel, collected in a 1.5-ml cryogenic tube (Thermo Fisher Scientific Nalgene system 100) in 1 ml cold cryopreservation medium (DMEM GlutaMAX supplemented with 10% FBS and 10% DMSO), frozen at −80 °C in a freezing box and stored at −80 °C if not used immediately. When tissues were needed, the samples were rapidly thawed in a water bath at 37 °C. The cryopreservation medium was replaced with precooled medium (DMEM GlutaMAX supplemented with 10% FBS), and the tissues were washed with cold PBS. Frozen tissues or fresh tissues were digested with 400 μl Versene solution (Thermo Fisher) for 10 min at 20 °C, and the cells were dissociated by pipetting. Next, 200 μl DMEM supplemented with 10% FBS was added. After 5-8 min of incubation, the supernatant was transferred into a new 2-ml tube. The remaining precipitate was resuspended in 300 μl DMEM GlutaMAX supplemented with 10% FBS. After 3 min of incubation, the supernatants were transferred and combined with the first supernatant. This step was repeated two more times to obtain the largest number of dissociated cells. Finally, the total dissociated cells were centrifuged at 3,000g for 5 min, and the cell pellet was resuspended in 400 μl cold PBS.
TR-FRET measurement with the nanobody-based sensors. For TR-FRET measurements in attached cells, transfected HEK293 cells were plated in polyornithine-coated, white 96-well plates (Greiner Bio-One) at 10 5 cells per well and cultured overnight at 37 °C with 5% CO 2 . Cells were starved in DMEM GlutaMAX for 2 h at 37 °C and then washed once with Krebs buffer to reduce ambient glutamate levels. Donor and acceptor nanobodies and ligands were prepared in Krebs buffer and added to the plate to reach a total volume of 60 μl per well. For TR-FRET measurements in cell suspensions, 5 μl HEK293 cells alone or transfected with plasmids encoding mGlu receptors were added to a low-volume 384-well microplate (Greiner Bio-One) at 2 × 10 4 cells per well. All reagents were added to reach a total volume of 20 μl per well. For TR-FRET measurements in dissociated brain cells, 10 μl cells in PBS, corresponding to different amounts of total protein, were added to a half-area 96-well microplate (Greiner Bio-One); all reagents were added to reach a total volume of 40 μl. In all cell preparations, the plates were incubated at 25 °C for 4 h or overnight, and then the TR-FRET signal was measured as previously described 15 .
Bicinchoninic acid assay for protein quantification. The dissociated cells (30 μl) in PBS were diluted twice with PBS containing 2% Triton X-100 and incubated at 20 °C for 1 h. The total protein quantity was measured in triplicate using a bicinchoninic acid kit (BCA1, Sigma-Aldrich) in accordance with the manufacturer's instructions using the Infinite F500 microplate reader (Tecan).
Relative quantification of the mGlu4-mGlu4 homodimer and the mGlu2-mGlu4 heterodimer by TR-FRET. Twenty-four hours after transfection, cells (100,000 cells per well in a 96-well microplate) were labeled in Tris-Krebs buffer for 1 h at 37°C with 5% CO 2 and the 'detector' nanobodies (1.6 nM DN42-Tb and 1.6 nM DN42-d2 for the SNAP-tagged mGlu4 homodimer; 1.6 nM DN42-Tb and 25 nM DN1-d2 for the coexpressed wild-type CLIP-tagged mGlu2 and SNAP-tagged mGlu4 to obtain the mGlu2-mGlu4 heterodimer) or the SNAP or CLIP substrates (100 nM SNAP-Lumi4-Tb and 100 nM SNAP-d2 for the SNAP-tagged mGlu4 homodimer; 100 nM CLIP-Lumi4-Tb and 100 nM SNAP-d2 for the coexpressed wild-type CLIP-tagged mGlu2 and SNAP-tagged mGlu4). After labeling, the cells were washed three times with Tris-Krebs buffer, and the signal was recorded with 100 μl Tris-Krebs buffer per well. The TR-FRET signal was measured using a PHERAstar FS reader, by the emission of d2 at 665 nm with a 50-μs delay and an integration time of 450 μs after excitation at 337 nm.
Measurements of inositol phosphate concentration. HEK293 cells were transiently cotransfected to express mGlu receptors, EAAC1 and the chimeric G qi9 protein using Lipofectamine 2000. Sixteen hours after transfection, the cells were incubated with the indicated ligands and 10 mM LiCl for 30 min. The accumulated IP 1 concentration was quantified using a PHERAstar FS reader and the IP-One HTRF assay kit (PerkinElmer Cisbio) in accordance with the manufacturer's instructions.
Neuronal culture and TR-FRET microscopy imaging. The primary hippocampal neurons were cultured following the procedures described previously 27 , and, after 17 d of culture, neurons were imaged. Neurons were labeled with 120 nM DN1-d2 and 80 nM DN10-Tb in the presence or absence of ligands (150 nM LY379268 or 1 μM LY341495) for 2 h at 37 °C. After labeling with nanobodies, the neurons were washed four times before imaging to remove the unbound nanobodies. Labeling and washing were performed in imaging buffer (127 mM NaCl, 2.8 mM KCl, 1.1 mM MgCl 2 , 1.15 mM CaCl 2 , 10 mM d-glucose, 10 mM HEPES, pH 7.3), supplemented with 1% BSA. Images were acquired with a home-built TR-FRET microscope 43 following a previously described protocol 27 .
Brain collection and fixation. Mice were euthanized with 140 mg per kg sodium pentobarbital (Euthasol Vet, Dômes Pharma Vétérinaire TVM) followed by cardiac perfusion with PBS. Brains were extracted and incubated overnight at 4 °C in a 4% paraformaldehyde solution (Euromedex), cryoprotected for 4 d at 4 °C with a 30% sucrose solution, included in an optimal cutting temperature compound (Tissue-Tek O.C.T., Sakura Finetek) and quickly frozen in ethanol cooled on dry ice. Brains were stored at −80 °C until use. Frozen brains were mounted on a cryostat (Leica Biosystems), and 16-μm sagittal sections were obtained. Sections were mounted on Superfrost Plus glass slides (Microm France) and kept at −20 °C until use.
Tissue immunofluorescence. Fifteen to twenty sections per mouse (n = 2-3 per genotype) were used. Sections were rinsed with PBS and incubated for 1 h at 20 °C with a blocking solution (3% BSA and 0.1% Triton X-100 in PBS). Sections were then incubated with the appropriate nanobody: 200 nM DN1-d2 (overnight at 4 °C) or 300 nM DN42-d2 (1.5 h at 20 °C). Sections were first washed with PBS and then with distilled water and mounted in Fluoroshield mounting medium with DAPI (Sigma-Aldrich). All images were taken with a slide scanner Axio scan Z1 microscope (Carl Zeiss Microscopy) by performing full-section mosaics at 20× magnification.
Co-immunoprecipitation and immunoblotting. Mice were decapitated, brains were rapidly removed, and the olfactory bulb was separated. Samples from four to five mice were combined and homogenized in a tissue grinder with 10 μl lysis buffer (25 mM HEPES, 150 mM NaCl, 1.0% LMNG, protease inhibitor cocktail, pH 8.0) per mg of tissue and incubated for 1 h at 4 °C. The lysate was then centrifuged at 12,000g for 45 min at 4 °C, and the supernatant was collected. The lysate was precleared with protein A beads (Millipore 16-125, Sigma-Aldrich) for 2 h at 4 °C. DN42 or Fc-GFP nanobodies were bound to protein A beads by rotating at 4 °C for 3 h. The precleared lysate was then added to the antibody-bound protein A beads and incubated overnight at 4 °C. The beads were washed four times with wash buffer (25 mM HEPES, 150 mM NaCl, 0.01% LMNG, protease inhibitor cocktail, pH 8.0), and proteins were eluted with unheated SDS sample buffer. Samples were subjected to SDS-PAGE using 10% polyacrylamide gels and then transferred to nitrocellulose membranes (GE Healthcare). After the transfer, membranes were blocked in TBST (25 mM Tris, 150 mM NaCl and 0.1% Tween- 20) containing 5% BSA at 20 °C for 1 h. Anti-mGlu receptor 2 (Abcam, ab15672) and anti-mGlu receptor 4 antibodies (Abcam, ab184302) were diluted 1:1,500 in blocking buffer and incubated with the membranes at 4 °C overnight. Membranes were washed four times with TBST and then incubated in HRP-conjugated goat anti-mouse IgG secondary antibody (Cell Signaling Technology, 7076S) or incubated in HRP-conjugated goat anti-rabbit IgG secondary antibody (Cell Signaling Technology, 7074S) diluted 1:10,000 in TBST containing 5% nonfat milk for 1 h at 20 °C. Membranes were then washed four times with TBST, and an enhanced chemiluminescent assay (Thermo Scientific) was performed to detect immunoreactive proteins. The membrane was scanned using an imager.
Statistical analysis. All data are presented as mean ± s.e.m. and were initially analyzed using GraphPad Prism (version 9.1.2 for Windows, GraphPad software) using Shapiro-Wilk's normality test. Normally distributed datasets (P > 0.05) were analyzed using parametric tests, two-tailed Student's t-tests or one-way ANOVA followed by Dunnett's or Tukey's post hoc analysis or two-way ANOVA followed by Tukey's post hoc test, depending on the experiments analyzed. For data analyzed using one-way ANOVA with a significant Brown-Forsythe test (P < 0.05, meaning that there was unequal variance between the different groups), datasets were analyzed using Welch's ANOVA test followed by Dunnett's T3 post hoc test (recommended for n < 50 per group). For all statistical analyses, a probability of 0.05 was defined as a significant difference. The exact P values are indicated in figures or in figure legends. Data are mean ± SEM of three independent experiments performed in triplicates and normalized to the maximum. (c and d) TR-FRET intensity and cell surface expression were measured on various expression levels of the indicated cell samples. Experiments were performed with HEK-293 cells transiently transfected with SNAP mGlu2 which fused a SNAP tag at the N terminus of receptors (c) or transiently co-transfected with SNAP mGlu2 C1 and mGlu2 C2 or mGlu4 C2 (d). The surface expression of receptors was measured as the specific Tb emission at 620 nm after labeling by substrate SNAP-Lumi4-Tb and exciting at 337 nm. TR-FRET intensity was measured in the presence of 10 μM LY341495. (e) Relative quantification of mGlu2 receptor in the indicated brain tissues from a same mouse in the presence of 10 μM LY341495. For a and c-e, Data are mean ± SEM of triplicate determinations from one representative out of three experiments. Fig. 6 | Nanobody DN42 specifically targets mGlu4 receptors. (a) Cartoon illustrating the principle of the TR-FRET binding assay. The receptor fused to a SNAP-tag (dark gray circled labeled 'S') is labeled with donor fluorescent dye Lumi4-Tb (blue circled 'D') while the nanobody DN42 (dark blue) bearing a 6xHis tag epitope at its C-terminus is labeled with 100 nM of anti-His antibody (bright blue) coupled to d2 fluorophores (red circled 'A'). Binding of the nanobody to the receptor is then measured by TR-FRET. Data are mean ± SEM of three independent experiments. Two-way ANOVA with Tukey's multiple comparisons test, with ***P < 0.001 and *P ≤ 0.05. (d) TR-FRET potencies (pEC 50 ) for the indicated ligands on the conformational sensor with the constructs mGlu2 C1 and mGlu4 C2 , or when both wild-type mGlu2 and mGlu4 subunits are co-transfected. Data are presented as the mean ± SEM of four independent experiments. Two-way ANOVA followed by a Tukey's post-hoc test, with ***P < 0.001, **P ≤ 0.01, and ns P > 0.05. (e) TR-FRET potencies (pEC 50 ) of the ligands on the conformational sensor measured in dissociated olfactory bulb cells. Data are presented as the mean ± SEM of four independent experiments. One-way ANOVA followed by a Tukey's post-hoc test, with ****P ≤ 0.0001 and ***P ≤ 0.001. (f) Correlation between the potencies (pEC 50 ) determined with the indicated agonist conditions by the heterodimer conformational sensor and IP 1 assay. Data are mean ± SEM of at least three independent experiments (TR-FRET of olfactory bulb (n = 4), mGlu2 + mGlu4 (n = 3) and mGlu2 C1 + mGlu4 C2 (n = 3) and IP 1 accumulation of mGlu2 C1 + mGlu4 C2 (n = 3)). LY37 means LY379268. All exact P values are indicated in the panels c-e.