A single molecule assay to assess GPCR/β-arr1 complex formation
To directly detect and quantitate mGluR/β-arr complexes, we sought to establish a single molecule imaging platform based on visualizing constructs bearing self-labeling tags (e.g., SNAP, Halo) for organic fluorophore conjugation. We turned to previously established N-terminally SNAP-tagged mGluR constructs59 and generated a C-terminally Halo-tagged β-arr1 (see “Methods”). As in our prior work54,55, live cell imaging revealed that SNAP-mGluR3 and SNAP-mGluR8 (labeled with cell impermeable SBG-JF646) recruit cytosolic β-arr1-Halo (labeled with cell permeable CA-JF549 for Halo) to the plasma membrane upon glutamate addition (Fig. 1A, B; and Supplementary Fig. 1A, B). β-arr1-Halo co-internalized with SNAP-mGluR8, but not with SNAP-mGluR3 (Supplementary Fig. 1A, B), further confirming that the unique subtype-specific properties of β-arr recruitment are maintained with our constructs. We also prepared a C-terminally truncated β-arr1 construct (“β-arr1-∆CTD-Halo”) where the autoinhibitory C-terminal domain (aa 383-418) of β-arr1 has been removed. This construct was efficiently recruited to the plasma membrane by both mGluR3 and mGluR8 (Supplementary Fig. 1C, D) and partially impaired internalization of both receptor subtypes, consistent with the removal of motifs which facilitate clathrin binding within the β-arr1-CTD (Supplementary Fig. 1E, F)65,66,67,68,69.
Fig. 1: A single molecule assay for detection of mGluR/β-arr complexes.
Representative images based on at least 3 separate biological replicates showing surface localization and co-internalization of β-arr1 with mGluR3 (A) or mGluR8 (B), respectively. Schematic (C, F), representative images (D, G), and summary quantification (E, H) of SiMPull assay of mGluR/β-arr1 complexes. Yellow circles (D, G) highlight colocalized spots that are indicative of mGluR/β-arr1 complexes. Points represent individual movies from at least 2 biological replicates (E, H). All data shown as mean ± SEM. Scale bar: 5 µm.
We then assessed the ability of SNAP-tagged mGluRs to couple to β-arr1 via single molecule pulldown (“SiMPull”)70,71. Briefly, fresh lysates were prepared from cells transfected with SNAP-tagged mGluRs and Halo-tagged β-arr1 constructs that were labeled with fluorophores and subsequently incubated with saturating concentrations of glutamate. An N-terminal HA-tag on the SNAP-mGluR constructs enabled receptor isolation on passivated glass coverslips via a conjugated anti-HA primary antibody (Fig. 1C–H). We then used total internal reflection fluorescence (TIRF) microscopy to visualize individual receptors and β-arr1 in separate channels and to quantify complex formation via colocalization of spots in both channels.
In the absence of GRK overexpression, minimal complex formation (<10%) with either WT β-arr1 or β-arr1-∆CTD was observed for mGluR8, while mGluR3 showed modest colocalization with β-arr1-∆CTD (~25%) (Fig. 1D–H). Co-transfection with either GRK2, GRK5, or a membrane-tethered GRK2 containing a C-terminal CaaX-box motif (GRK2-CaaX), dramatically increased the extent of β-arr1 colocalization for both mGluR3 and mGluR8 (Fig. 1D–H), likely by enhancing the degree of receptor phosphorylation. Both mGluR3 and mGluR8 exhibited substantial pulldown of β-arr1-∆CTD with ~60% or 75% of receptors, respectively, showing β-arr1 colocalization under optimal conditions with GRK2-CaaX co-transfected. Strikingly, only mGluR8 showed substantial co-localization (~55%) with WT β-arr1 (Fig. 1D–H). The strong dependence on β-arr1 truncation for colocalization with mGluR3, but not mGluR8, was seen across all GRK conditions (Fig. 1E, H). Importantly, in the absence of receptor co-expression, very minimal β-arr1-Halo pulldown was observed (Supplementary Fig. 1G). Overall, these results suggest that mGluR8 has a higher phosphorylation threshold for β-arr1 coupling but can form more stable complexes with β-arr1 than mGluR3, which may explain the ability of mGluR8/β-arr1 complexes to persist following endocytosis.
To assess the generalizability of our assay, we investigated β-arr1 coupling to two SNAP-tagged prototypical family A GPCRs, the beta-2 adrenergic receptor (β2AR) and the arginine vasopressin 2 receptor (V2R) (Fig. 2A, B). In the presence of GRK2-CaaX and saturating agonist, each receptor showed pulldown of β-arr1-∆CTD, but V2R was substantially more efficient (~60% vs. 15% colocalization). Similarly to mGluR8 and mGluR3, respectively, V2R was able to produce strong pulldown of WT β-arr1 while β2AR was not (Fig. 2A, B). We confirmed that both receptors recruit β-arr1-Halo in live cells but SNAP-β2AR interacts transiently7,72 while SNAP-V2R co-internalizes7 with with β-arr1-Halo (Supplementary Fig. 2A–D). These results are consistent with class B (V2R, mGluR8), but not class A (β2AR, mGluR3), GPCRs being capable of stable interactions with WT β-arr1 and indicate that our SiMPull assay can provide a general means of quantitatively probing and classifying GPCR/β-arr interactions. We interpret the colocalization percentages of β-arr1 with receptors to reflect the relative stabilities of the complexes, but these in vitro percentages may be underestimates due to dissociation of some complexes upon cell lysis, incomplete fluorescent labeling, and competition from native β-arrs (see “Methods”). Importantly, as these are non-equilibrium experiments, they do not allow us to directly measure or estimate the affinity or energetics of GPCR/β-arr interactions.
Fig. 2: Single molecule analysis of GPCR/β-arr1 coupling reveals a range of efficiencies and stoichiometries, including 2:2 mGluR/β-arr1 complexes.
A Representative images showing variable β-arr1-ΔCTD pulldown efficiencies for V2R, mGluR2, and mGluR5. B Summary graph of β-arr1 colocalization across GPCRs for both WT and β-arr1-ΔCTD. Representative images and bleaching step traces (C, D) and photobleaching step distributions (E, F) for β-arr1-ΔCTD complexes mGluR3 and V2R. Arrows indicate bleaching steps. Summary of % of colocalized molecules showing 2-step bleaching in the receptor (G) and β-arr1 (H) channels across receptors. I, J Schematic and representative image and bleaching step traces for mGluR/Gαi1 SiMPull. K Summary graph of Gαi1 protein colocalization across GPCRs. L Summary of % of colocalized molecules showing 1-step bleaching in the Gαi1 channel. For (A–H) GRK2-CaaX was co-expressed and for all experiments cells were treated with agonist. Circles (A, C, D, J) indicate colocalized spots indicative of GPCR/transducer complexes. Points represent individual movies from at least 2 biological replicates (B, E–H, K, L). All data shown as mean ± SEM.
We next extended our analysis across the mGluR family and tested SNAP-tagged mGluR2, mGluR5, and mGluR7 in the presence of GRK2-CaaX and either β-arr1-Halo or β-arr1-∆CTD-Halo (Fig. 2A, B). SNAP-mGluR2 showed negligible levels (<5%) of co-localization with either β-arr1 or β-arr1-∆CTD (Fig. 2A, B), consistent with the lack of β-arr1 recruitment and receptor internalization observed in live cells (Supplementary Fig. 2E, F)54,55. In contrast, SNAP-mGluR5 showed modest pulldown (~15–20%) of β-arr1-∆CTD, but not WT β-arr1 (Fig. 2A, B). Live-cell imaging mirrored these results, where only β-arr1-∆CTD showed detectable glutamate-induced membrane localization (Supplementary Fig. 2G, H). Finally, SNAP-mGluR7 showed weak (~5-10%) pulldown of both β-arr1-∆CTD, and WT β-arr (Fig. 2B and Supplementary Fig. 2I), in line with live cell imaging showing modest β-arr1 recruitment (Supplementary Fig. 2J, K).
Variable stoichiometries of GPCR/β-arr1 complexes
A major advantage of single molecule imaging is the ability to measure the stoichiometry of protein assemblies via photobleaching step analysis, a technique we have used to confirm the constitutive dimerization of mGluRs60,61,73. We thus asked two key questions about mGluR/β-arr1 complexes: (1) Do mGluRs retain dimerization upon β-arr interaction? And (2) how many β-arr1 molecules can interact with an mGluR at a time? To assess these questions, we analyzed the photobleaching step distribution for both the receptor and β-arr1 channels for all β-arr-colocalized mGluR3, mGluR5, and mGluR8 spots (Fig. 2C–H). We found that ~50% of spots displayed 2 bleaching steps in the receptor channel for each mGluR subtype (Fig. 2G), confirming that mGluRs remain dimeric upon β-arr interaction. This result is in line with our recent study showing that internalized mGluR3 and mGluR8 are strict dimers55. As previously established60,61,73, the subpopulation of mGluRs showing single bleaching steps is an expected phenomenon for a strictly dimeric protein complex due to a variety of factors including incomplete fluorophore labeling and incidental bleaching.
Surprisingly, ~25–55% of co-localized complexes also showed 2 bleaching steps in the β-arr1 channel (Fig. 2C, E, H and Supplementary Fig. 3A, B), indicating that both 2:1 and 2:2 mGluR:β–arr1 complexes can form. The proportion of 2-step β-arr1 bleaching steps was lower for mGluR5 than mGluR3 or mGluR8 which may be due to lower overall complex formation for this receptor (Fig. 2H). Consistent with this, the proportion of 2-step β-arr1 bleaching steps was lower for mGluR3 and mGluR8 under conditions with weaker pulldown owing to the lack of GRK overexpression (Supplementary Fig. 3B). As a key control, V2R:β-arr1-∆CTD complexes showed background levels (~10%) of 2-step bleaching in both channels (Fig. 2D, F–H) consistent with formation of a 1:1 complex between monomeric family A GPCRs and a single β-arr1 molecule. To further validate our finding of 2:2 mGluR:β-arr1 complexes we sought to directly image β-arr1 colocalization at individual receptors. To do this, we labeled β-arr1-Halo with a mix of both CA-JF549 and CA-JF646 fluorophores, while leaving the SNAP-tag on mGluR8 unlabeled (Supplementary Fig. 3C). We observed substantial colocalization (~40%) of JF549- and JF646-labeled β-arr1-∆CTD when imaging both channels (Supplementary Fig. 3D), in agreement with our photobleaching step analysis implying the existence of mGluR/β-arr1 complexes containing two β-arr1 molecules bound to one receptor dimer.
As a control, we also used SiMPull to assess the coupling stoichiometry of mGluRs to G proteins using a Halo-tagged Gαi1 (see Methods). Briefly, we followed the same approach as for mGluR/β–arr complexes but replacing GRK and β–arr1-Halo constructs with Gαi1-Halo and Gβ1 and Gγ2 in our co-transfection (Fig. 2I). Both WT Gαi1 and an “Ins4A” complex-stabilizing variant74,75 were able to be pulled down by mGluR2, mGluR3, or mGluR8 with observable bleaching steps for both receptor and G protein (Fig. 2J). While well-above background, we observed much lower receptor colocalization for G proteins (5–10%) than for β–arrs. However, the Ins4A Gαi1 variant increased pulldown efficiency by ~2-fold for all receptors (Fig. 2K). Notably, unlike our results with β–arr1, comparable pulldown efficiency was seen for all three mGluR subtypes, supporting the interpretation that G protein coupling efficiency is relatively constant across receptors while β–arr coupling efficiency is highly variable. Photobleaching step analysis revealed >90% 1-step photobleaching for Gαi-Ins4A (Fig. 2L), indicative of a strict 2:1 mGluR/G protein stoichiometry as seen in all previous cryo-EM structures which predict steric clashes with two G proteins on a single receptor46,47,48,49,50,51,52,53. While the lower overall pulldown efficiency may limit our ability to detect potential 2:2 mGluR: G protein complexes, it is worth nothing that even at a comparably low efficiency (~10% pulldown) 2:2 complexes were observed for mGluR3 and mGluR8 with β–arr1 (Supplementary Fig. 3B). Overall, these results demonstrate the value of SiMPull for assessing GPCR/transducer complexes and highlight major differences between mGluR/G protein and mGluR/β–arr complexes in terms of overall formation propensity, stoichiometry, and receptor subtype variability.
mGluR8 phosphorylation controls the strength of β-arr1 interaction
Having observed particularly strong colocalization of mGluR8 and β-arr1, we decided to focus further analyses on better understanding this complex. Our prior work in cells has pointed to a critical role of phosphorylation sites in the mGluR8 C-terminus55 for β-arr-mediated receptor internalization. Mass spectrometry of purified, agonist-treated mGluR8 showed extensive phosphorylation in the distal serine/threonine (S/T)-rich region (aa888-907), including many double-phosphorylated peptides (Supplementary Figs. 4 and 5). This finding is consistent with the likelihood that multiple phosphorylation sites simultaneously contribute to β-arr recruitment12,14,76,77. While we expected to see peptides with >2 phosphorylated residues on them, these peptides can be challenging to observe for a variety of reasons including high mass-to-charge ratio, additional modifications leading to mass shifts, challenges with isolation via liquid chromatography, and loss of phosphate groups upon ionization.
To understand the role of each putative mGluR8 phospho-site in β–arr1 complex formation, we performed alanine scanning mutagenesis of the 11 potential phospho-sites in the mGluR8 S/T-rich region. We tested each mutant using SiMPull under our highest colocalization condition (GRK2-CaaX, β-arr1-ΔCTD-Halo) and a less efficiently coupled condition (GRK2, β-arr1-Halo), finding similar trends in both conditions with larger effects for the latter (Fig. 3A). We found no pulldown with simultaneous mutation of all 11 phospho-sites (“mGluR8-11xA”), highlighting the need for phosphorylation in the mGluR8 S/T-rich region for β-arr1 coupling. All individual mutants showed impairment with the largest effects seen for T896A, T898A, and T899A, sites that could constitute a recently identified arrestin binding PxPP motif14,78. Consistent with this, a triple mutation of T896A/T898A/T899A (“mGluR8-3xA”) decreased pulldown to the same degree as mGluR8-11xA, suggesting a dominant role for this PxPP motif in β-arr1 coupling (Fig. 3A).
Fig. 3: Defining tail and core interactions that drive mGluR8//β-arr1 coupling.
A Summary graph of SiMPull data for colocalization of all mGluR8 CTD point mutations either with wild type β-arr1-Halo (with GRK2 co-expression) or β-arr1-ΔCTD (with GRK2-CaaX co-expression). 3xA = T896A/T898A/T899A. B Scatter plot showing correlation between effects of mutations on β-arr1 co-localization in SiMPull (y-axis) and internalization in cell-based surface labeling assay (x-axis). Mean values for each mutant are shown with SEM. A linear fit to the data excluding outliers (S888A, T896A, T899A) is shown with the 95% confidence interval shaded. C Scatter plot showing correlation between effects of mutations on β-arr1 co-localization in SiMPull (y-axis) and degradation in cell-based assay (x-axis). A linear fit to the data is shown with the 95% confidence interval shaded. D Summary graphs showing impairment of β-arr1ΔCTD colocalization in SiMPull (left) and mGluR8 internalization (right) with the F777D mutation in ICL3. E Representative image and line scan showing lack of internalization and β-arr recruitment by mGluR8-F777D based on at least 3 separate biological replicates. F Representative image and line scan showing modest, but clear internalization of mGluR8-ΔCTD based on at least 3 separate biological replicates. Arrows highlight intracellular receptor puncta. G Summary graph showing that mGluR8-ΔCTD shows substantially less surface expression compared to WT but still undergoes a small glutamate-induced reduction in surface levels. Points represent individual movies from at least 2 biological replicates (A, D) or individual experiments from biological replicates (D, G). Two-tailed unpaired t-test is used in (D) and one-tailed paired t-test is used in (G). All data shown as mean ± SEM. Scale bar: 5 µm.
We then assessed the effects of each CTD variant using an established live cell imaging-based assay of glutamate-induced mGluR8 internalization54,55 to understand how differential phosphorylation impacts β-arr1-mediated receptor desensitization. We observed clear impairment of receptor internalization for most mutations (Supplementary Fig. 6A, B), including a complete loss of internalization for both mGluR8-11xA and mGluR8-3xA. When comparing our colocalization and internalization datasets for each mGluR8 CTD mutant, we found a mostly linear correlation between the extent of impairment of β-arr1 colocalization and the extent of impairment of ligand-induced internalization (Fig. 3B). However, mutations to three phospho-sites had differing effects in our two assays: S888, T896, T899. Mutant S888A impacted β-arr1 SiMPull colocalization less profoundly than internalization, while mutants T896A and T899A impaired colocalization more than internalization (Fig. 3B).
To better understand the discrepancy between results in our two assays for these three mutants, we directly imaged β-arr1 recruitment and receptor endocytosis in live cells. Consistent with results in our internalization assay, S888A showed an almost complete loss of β-arr1 recruitment and intracellular receptor puncta compared to WT (Supplementary Fig. 6C). One possible explanation is that S888 serves as a gatekeeper residue whose phosphorylation helps drive subsequent phosphorylation of other positions. The minimal effects on colocalization in SiMPull may be due to GRK2-CAAX overexpression which can overcome this potential gatekeeper role of S888. In contrast, T899A showed clear receptor internalization and β-arr1 recruitment to the plasma membrane, but minimal co-localization of β-arr1 with internalized receptors (Supplementary Fig. 6D). A potential explanation is that phosphorylation of T899 is critical for maintaining a conformation of β-arr1 that survives endocytosis but is not required for initial β-arr1 recruitment to the plasma membrane. Similarly, T896A showed impaired recruitment of β-arr1 but still maintained some weak endosomal co-localization (Supplementary Fig. 6E).
Motivated by the complex phenotypes observed for the aforementioned mutants with an apparent disconnect between SiMPull and internalization data, we tested glutamate-induced degradation using our established SNAP-tag based assay55 for all mGluR8 CTD mutants (Supplementary Fig. 6F). We observed a linear correlation between our SiMPull colocalization and receptor degradation datasets for each mGluR8 CTD mutant without major outliers (Fig. 3C). Most notably, T896A, T898A, and T899A all showed larger impairments in degradation compared to their effects on receptor internalization. The better agreement of the SiMPull assay with the degradation assay compared with the internalization assay suggests that coupling efficiency of β-arr1 to mGluR8 directly controls targeting of receptors to degradative pathways while internalization of the receptor is a more complex process. Critically, we cannot rule out the chance that mutating residues in the mGluR8 CTD may affect both the intrinsic properties and dynamics of the mGluR8 CTD or TMD, as well as the resulting phosphorylation of other CTD residues. Taken together, the myriad effects of mutations across the ST-rich region of the mGluR8 CTD, suggest the potential of multiple β-arr1 binding modes and the existence of multiple possible phosphorylation patterns of the mGluR8 CTD.
β-arr1 coupling to the TMD core of mGluR8
We next asked if the mGluR8 TMD core contributes to mGluR8/β-arr1 coupling. We first turned to a well-established mutation of a highly conserved phenylalanine (F777 in mGluR8) in intracellular loop 3 that strongly impairs G protein coupling across mGluR subtypes61,79,80. We hypothesized that if a similar interface is used for β-arr1 core coupling, mutation to this position would also strongly impair mGluR8/β-arr1 coupling and endocytosis54. Colocalization of β-arr1 was strongly reduced in SiMPull and glutamate-driven internalization in live cells was abolished for mGluR8-F777D (Fig. 3D, E and Supplementary Fig. 6G). While such effects could be due, at least in part, to impairment of GRK coupling, this result supports the hypothesis that TMD core interactions contribute to mGluR8/β-arr1 coupling. To further test this hypothesis, we asked if receptor core interactions alone are sufficient to produce detectable mGluR8/β-arr1 coupling in cells by producing an mGluR8 construct lacking a CTD (mGluR8-∆CTD). In line with a role for core coupling, SNAP-mGluR8-ΔCTD showed weak, but clear, glutamate-driven internalization (Fig. 3F, G).
β-arr1 can couple to mGluR8 in both cis and trans configurations
The involvement of both the mGluR8 tail and core in β-arr1 coupling, and the orientation of β-arr1 found in our structural studies81, raises questions about the possible configurations of mGluR8/β-arr1 complexes. Specifically, can coupling occur via interactions of β-arr1 with the tail and TMD core of the same mGluR8 subunit (termed cis), and/or via interactions of β-arr1 with the tail of one subunit and the TMD core of the other (termed trans)? Notably, our active-like, β-arr1-bound mGluR8 structure81 orients the beginning of the CTD of a given mGluR8 protomer near the TMD of the other mGluR8 protomer within the dimer. Furthermore, the mGluR8 S/T-rich region is located at the end of the CTD, which could allow for the tail-bound mGluR8/β-arr1 complex to maintain a high degree of conformational flexibility. Together, the structural features of β-arr1-bound mGluR8 position it such that it is conceivable that β-arr1 coupling to the TMD core of either subunit can occur while remaining bound to the CTD.
To test this, we took advantage of mutations described above that strongly impair either tail interaction (“11xA”) or core interaction (F777D) to prevent mGluR8/β-arr1 co-internalization (Fig. 4A, B). We first co-expressed SNAP-tagged mGluR8-F777D with untagged mGluR8-11xA. These expression conditions should produce a mix of SNAP-tagged F777D homodimers and F777D/11xA heterodimers, as well as 11xA homodimers which are unlabeled. Given that both F777D and 11xA homodimers show no internalization (Fig. 3B, D), we reasoned that any visible intracellular fluorescence or drop in surface fluorescence in the surface labeling assay would indicate heterodimer internalization. Indeed, we observed clear internalization both in cell imaging and surface labeling internalization assays for this mix of receptor subunits (Fig. 4A, B). This finding supports the ability of trans interactions whereby β-arr binds to the CTD of the F777D subunit and the core of the 11xA subunit to induce β-arr-mediated internalization. Similar results were observed upon co-expression of SNAP-mGluR8-11xA with untagged mGluR8-F777D (Fig. 4A, B). Supporting cis interactions of β-arr1 with the CTD and core of the same subunit, co-expression with wild type untagged mGluR8 enabled internalization of SNAP-mGluR8-F777D-11xA (Fig. 4A, B).
Fig. 4: β-arr coupling in cis and trans configurations to mGluR8.
A Surface labeling assay internalization data showing evidence for trans (SNAP-F777D/11xA; SNAP-11xA/F777D) and cis (SNAP-F777D,11xA/WT) β-arr1 coupling driving mGluR8 internalization. B Representative images showing internalization and β-arr colocalization for a range of mGluR8 homodimer conditions based on at least 3 separate biological replicates. Arrows represent sites of receptor/β-arr1 intracellular co-localization (B). C–E Representative SiMPull images (C, D) and summary quantification (E) showing evidence for cis and trans mGluR8/β-arr1 coupling with higher efficiency for cis compared to trans. Points represent individual movies from at least 2 biological replicates (E) or individual experiments from biological replicates (A). 1-way ANOVA is used in (A) and two-tailed unpaired t-test is used in (E). All data shown as mean ± SEM. Scale bar: 5 µm.
To directly detect and compare the propensities of cis and trans mGluR8/β-arr1 interactions, we turned to SiMPull. First, we generated an HA-tagged mGluR8-11xA construct that did not contain a SNAP-tag and should be unable to pulldown any β-arr1-Halo based on our prior results (Fig. 3A). To detect trans interactions, we tested the ability of this construct to pulldown with WT β-arr1 upon co-expression with SNAP-mGluR8-F777D and observed substantial ~20% colocalization (Fig. 4C), presumably via tail interactions with the F777D subunit. Importantly, the colocalization percentage for this trans-coupling mGluR8 heterodimer was larger than pulldown for homodimeric mGluR8-F777D (Fig. 3D). To detect cis interactions, we produced an HA-mGluR8-F777D-11xA construct and tested its ability to pulldown with β-arr1 upon co-expression with wild type SNAP-mGluR8, and, again, observed substantial colocalization (Fig. 4D), suggesting a complex where both tail and core interactions occur via the WT subunit. Importantly, in these experiments we are only able to detect heterodimers (F777D/11xA or F777D-11xA/WT) since only one subunit is labeled with HA for pulldown and the other subunit contains a SNAP-tag for fluorophore conjugation. Quantification of trans and cis SiMPull conditions showed clear colocalization in both cases with enhanced efficiency in cis (Fig. 4E), suggesting a weak preference for this mode of tail and core interactions.
β-arr1 can couple to mGluR8-containing heterodimers in both cis and trans configurations
The ability of β-arr1 to couple to mGluRs in either cis or trans configurations may be particularly relevant for heterodimeric mGluRs which introduce asymmetry into the cytoplasmic face of the receptor and may be regulated in complex ways by subunit-specific pharmacology or post-translational modifications. We focused on mGluR2/8 heterodimers as these combine a strong β-arr coupled mGluR (mGluR8) with a β-arr resistant mGluR (mGluR2) and we have recently shown that these heterodimers undergo β-arr dependent endocytosis and lysosomal degradation55. We first confirmed that co-expression with untagged mGluR8 enabled internalization of SNAP-mGluR2 (Fig. 5A, B). Introducing the 11xA mutation into mGluR8 abolished mGluR2 internalization but introducing the F777D mutation into mGluR8 only partially impaired mGluR2 internalization (Fig. 5A, B), suggesting that while mGluR8 tail coupling is strictly required, core coupling to the mGluR8 subunit is not. Internalization was fully abolished when the equivalent mutation, F756D, was also introduced into mGluR2 (Fig. 5A, B), suggesting that both cis and trans mGluR2/8 β-arr coupling can occur via the mGluR8 tail and either the mGluR2 or mGluR8 core. Using SiMPull, we found that co-expression of untagged mGluR8 and HA-SNAP-mGluR2 enables colocalization with β-arr1-Halo (Fig. 5C, D). Importantly, the vast majority of immobilized SNAP-mGluR2 and colocalized β-arr1-Halo showed single-step photobleaching (Fig. 5E, F), consistent with a 1:1:1 mGluR2: mGluR8: β-arr1 complex. Finally, we observed a decrease in β-arr1-Halo colocalization when we introduced the F756D mutation into mGluR2, further supporting both cis and trans β-arr1 coupling to mGluR2/8 heterodimers (Fig. 5D). Importantly, β-arr1 pulled down by these mutated mGluR2/8 heterodimers still showed single-step photobleaching, indicating the introduction of F756D in mGluR2 did not alter the stoichiometry of β-arr1 in the complex (Fig. 5F).
Fig. 5: β-arr coupling in cis and trans configurations to mGluR8/2 heterodimers.
A Surface labeling assay internalization data showing evidence for mGluR2/8 heterodimer internalization via mGluR8 CTD coupling and core coupling to either subunit. B Representative images showing internalization and β-arr colocalization for a range of mGluR2/8 heterodimer conditions based on at least 3 separate biological replicates. Arrows represent sites of receptor/β-arr1 intracellular co-localization. C Representative SiMPull image (left) and photobleaching step traces (right) showing the formation of 1:1:1 mGluR2: mGluR8: β-arr1 complexes. D Quantification of SiMPull showing evidence for mGluR2/8/ β -arr1 complexes that are sensitive to core mutation (F756D) in mGluR2. E Quantification of the prevalence of spots containing one or two bleaching steps for mGluR2/8/β-arr1 complexes in SiMPull experiments, confirming 1:1:1 stoichiometry shown in (C). F Summary graph showing strict 1-step photobleaching for β-arr1 bound to mGluR2/8 heterodimers. Points represent individual movies from at least 2 biological replicates (D–F) or individual experiments from biological replicates (A). 1-way ANOVA is used in (A, D). All data shown as mean ± SEM. Scale bar: 5 µm.
SiMPull reveals differences in β-arr1 and β-arr2 coupling propensities to GPCRs
While β-arr1 has served as the model β-arr subtype for many structural and biophysical studies, β-arr2 also strongly contributes to GPCR regulation. As β-arr2 has known structural and functional differences compared to β-arr124,33,34,35,36, we first confirmed that β-arr2-Halo co-internalizes with both mGluR8 and V2R upon agonist stimulation (Fig. 6A and Supplementary Fig. 7A). Using SiMPull, we observed clear pulldown of WT β-arr2 by mGluR8 and V2R, but not mGluR2, although the efficiency was reduced in comparison to that seen with WT β-arr1 (Fig. 6B–D). mGluR8-mediated pulldown of WT β-arr2 was dramatically reduced with the 11xA or F777D mutations (Supplementary Fig. 7B), suggesting contributions from both tail and core interactions. To determine whether we could improve the pulldown of β-arr2, we produced two “pre-activated” β-arr2-Halo constructs36: one containing 3 alanine mutations (I386A, V387A, F388A) in the proximal CTD of β-arr2 (β-arr2-3A) and one containing these mutations and a subsequent truncation of the β-arr2 CTD (β-arr2-3A-∆CTD). Both β-arr2 variants had little impact on colocalization percentage for mGluR8 but increased pulldown for V2R (Fig. 6C). The differential effect of pre-activating β-arr2 modification across receptor subtypes highlights the complexity and receptor subtype-specificity of GPCR/β-arr coupling. Photobleaching step analysis showed both 2:1 and 2:2 mGluR8/β-arr2 complexes (Fig. 6C, E) indicating that the same stoichiometries are observed for complexes of mGluR8 with β-arr1 or β-arr2, although a larger proportion of 2:1 complexes was seen with β-arr2. As was observed with β-arr1, V2R/β-arr2 complexes showed a 1:1 stoichiometry (Fig. 6E). Finally, we further confirmed the potential for 2:2 mGluR8/β-arr2 assembly by observing substantial colocalization between β-arr2-Halo molecules when labeled with two different colors upon pulldown by HA-mGluR8 (Supplementary Fig. 7C).
Fig. 6: Single molecule analysis of mGluR/β-arr2 complexes.
A Representative images and line scans showing co-internalization of either mGluR8 with β-arr2 based on at least 3 separate biological replicates. Schematic (B) and representative image (C) showing colocalization and representative bleaching steps of mGluR8 and β-arr2. D Colocalization percentages of β-arr2 and β-arr2 variants across a panel of GPCRs. E Summary graph showing substantial formation of complexes with 2 β-arr2 bleaching steps for mGluR8, but not V2R. Points represent individual movies (D, E). GRK2-CaaX was co-expressed for all SiMPull experiments shown in this figure. All data shown as mean ± SEM. Points represent individual movies from at least 2 biological replicates (D, E).
mGluR8 can simultaneously bind β-arr1 and β-arr2 to form megacomplexes
As mGluR8 interacts with both β-arr1 and β-arr2 and can form a 2:2 stoichiometry, we next asked if a single mGluR8 dimer can simultaneously bind both β-arr1 and β-arr2. We first analyzed single cell RNA sequencing (scRNAseq) data from the mouse anterior cingulate cortex82 to evaluate whether β-arr1 (gene name: Arrb1) and β-arr2 (gene name: Arrb2) are co-expressed in the same neurons. We found subtle differences in β-arr1 versus β-arr2 expression across neuronal subtypes (Fig. 7A) with a consistent trend towards more β-arr1 expression in most classes. Overall, 37.8% of neurons showed co-expression of β-arr1 and β-arr2 (Fig. 7B). Notably, a wide range of RNA ratios between mGluR8, β-arr1, and β-arr2 is seen across neurons with a large proportion (~24% of the total and ~40% of mGluR8 expressing neurons) expressing all three (Fig. 7C; Supplementary Fig. 8A), raising the possibility that under biologically relevant conditions β-arr1 and β-arr2 can either compete for the same GPCR or bind in a combinatorial fashion. Accordingly, live cell imaging revealed endosomal co-localization of CLIP-mGluR8, β-arr1-SNAP, and β-arr2-Halo (Fig. 7D), pointing towards this combinatorial possibility. To address this directly, we performed SiMPull from cells expressing HA-mGluR8 lacking a SNAP tag along with GRK2-CaaX (Fig. 7E) and found a large proportion of spots with colocalized β-arr1-SNAP and β-arr2-Halo (Fig. 7F, G), confirming the presence of an mGluR8/β-arr1/β-arr2 megacomplex. Importantly, β-arr1-SNAP and β-arr2-Halo channels showed a single bleaching step in each channel, consistent with a 2:1:1 complex (Fig. 7F). In contrast, negligible background levels of β-arr1-SNAP and β-arr2-Halo colocalization were observed with pulldown of V2R (Fig. 7G and Supplementary Fig. 8B). This result represents an example of two different arrestin subtypes simultaneously interacting with a single GPCR, further expanding the molecular diversity of mGluR/β-arr coupling.
Fig. 7: Transcriptomic and single molecule analysis of formation of mGluR8/β-arr1/β-arr2 megacomplexes.
A Violin plot showing Arrb1 (β-arr1) and Arrb2 (β-arr2) expression levels across neuronal classes. CPM=counts per million. B Pie chart showing proportion of neurons expressing 0, 1, or both β-arr genes determined from single cell RNA sequencing data using a cutoff of 5 counts per million. C 3-dimensional scatter plot showing Arrb1, Arrb2, and Grm8 (mGluR8) expression levels across all Grm8-expressing cells. D Representative image and line scan showing colocalization of intracellular mGluR8, β-arr1, and β-arr2 (Scale bar is 10 µm; scale bar in the zoomed image is 2 µm) based on at least 3 separate biological replicates. Schematic (E) and representative image (F) showing co-localization of β-arr1 and β-arr2 upon pulldown via mGluR8. G Summary graph showing β-arr1/β-arr2 colocalization upon pulldown by mGluR8, but not V2R. Two-tailed, unpaired t-test used in (G). Points represent individual movies (G). GRK2-CaaX was co-expressed for all SiMPull experiments shown in this figure. All data shown as mean ± SEMs. Points represent individual movies from at least two biological replicates (G).
mGluR8 can simultaneously bind β-arr1 and G proteins to form megacomplexes
Given that mGluR8 can couple to β-arrs and G proteins via similar active receptor conformations and the flexible, dynamic β-arr coupling modes observed in this study, we asked whether mGluR8 can form a megacomplex containing both β-arr and G proteins. GPCR/β-arr/G protein megacomplexes have been observed for monomeric GPCRs and are thought to be important for initiating intracellular signaling83,84,85,86. To address this, we asked if mGluR8 activation can drive G protein re-localization, as internalized mGluR8 can drive localization of β-arrs to endosomes via formation of stable complexes (Fig. 1B)55. We co-transfected cells with SNAP-mGluR8, Gαi1-Halo (WT or Ins4A), and β-arr1-CLIP (along with GRK2, Gβ1, and Gɣ2 as above) and imaged cells after agonist stimulation using confocal microscopy. We observed clear colocalization of SNAP-mGluR8 with both Gαi1-Halo and β-arr1-CLIP in intracellular compartments with endosomal morphology (Fig. 8A and Supplementary Fig. 9A). Importantly, no intracellular localization of Gαi1-Halo and β-arr1-CLIP was seen in the absence of mGluR8 (Supplementary Fig. 9B).
Fig. 8: Single molecule analysis of formation of mGluR8/b-arr/G protein megacomplexes.
A Representative confocal image and line scan showing colocalization of intracellular mGluR8, β-arr1, and Gαi1 based on at least 3 separate biological replicates. Scale bars are 5 μm and 1 μm for zoom-ins. Schematic (B) and representative image (C) showing co-localization of β-arr1 and Gαi1 upon pulldown via mGluR8. (D) Summary graph of β-arr1 and Gαi1 colocalization efficiency across conditions. E Summary graph of colocalized β-arr1 and Gαi1 photobleaching step distributions. Points represent individual movies from at least 2 biological replicates (D, E). GRK2-CaaX, Gβ1, and Gγ2 were co-expressed for all experiments shown in this figure. All data shown as mean ± SEM.
While this observation is suggestive of mGluR8/β-arr/G protein co-assembly it does not directly address whether mGluR8 can simultaneously interact with both transducers in a single megacomplex. We, thus, turned to SiMPull to assess complex formation. We immobilized HA-mGluR8 from cells co-transfected with Gαi1-Halo (WT or Ins4A) and β-arr1-SNAP (WT or ∆CTD) (Fig. 8B). We observed substantial pulldown of both Gαi1-Halo and β-arr1-SNAP in the presence of HA-mGluR8 with more efficient pulldown for conditions containing Gαi1-Ins4A or β-arr1-∆CTD (Fig. 8C and Supplementary Fig. 9C). Importantly, little-to-no pulldown was seen in the absence of HA-mGluR8 (Fig. S9D). 20–30% of Gαi1-Halo spots and 5–10% β-arr1-SNAP spots were colocalized (Fig. 8D and Supplementary Fig. 9E), confirming the presence of mGluR8/β-arr/G protein megacomplexes in cells. Analysis of photobleaching steps for colocalized spots revealed primarily 1-step bleaching events for both transducers, but a slightly above-background level of 2-step bleaching for β-arr1-SNAP (Fig. 8E), indicating that most complexes consist of 1 copy of each transducer. Strict 1-step bleaching for Gαi1-Halo was unsurprising given our prior SiMPull data (Fig. 2) and that only 1 G protein heterotrimer is seen in our cryo-EM structure in our companion paper81.