Analysis of β₂AR-G_s and β₂AR-G_i complex formation by NMR spectroscopy

Significance

Recent structures of GPCRs in complex with G proteins provide important insights into G protein activation by family A and family B GPCRs; however, important questions remain. We don’t fully understand the mechanism of G protein coupling specificity or coupling promiscuity of some GPCRs. The β₂AR preferentially couples to G_s and less efficiently to G_i, yet β₂AR-G_i coupling has been shown to play important roles in cardiac physiology. To better understand the structural basis for the preferential coupling of the β₂AR to G_s over G_i, we used NMR spectroscopy and supporting MD simulations to study the conformational changes in the intracellular surface of the β₂AR. These studies reveal a distinct difference in intracellular loop 2 interactions with G_s and G_i1.

Abstract

The β₂-adrenergic receptor (β₂AR) is a prototypical G protein-coupled receptor (GPCR) that preferentially couples to the stimulatory G protein G_s and stimulates cAMP formation. Functional studies have shown that the β₂AR also couples to inhibitory G protein G_i, activation of which inhibits cAMP formation [R. P. Xiao, Sci. STKE 2001, re15 (2001)]. A crystal structure of the β₂AR-G_s complex revealed the interaction interface of β₂AR-G_s and structural changes upon complex formation [S. G. Rasmussen et al., Nature 477, 549–555 (2011)], yet, the dynamic process of the β₂AR signaling through G_s and its preferential coupling to G_s over G_i is still not fully understood. Here, we utilize solution nuclear magnetic resonance (NMR) spectroscopy and supporting molecular dynamics (MD) simulations to monitor the conformational changes in the G protein coupling interface of the β₂AR in response to the full agonist BI-167107 and G_s and G_i1. These results show that BI-167107 stabilizes conformational changes in four transmembrane segments (TM4, TM5, TM6, and TM7) prior to coupling to a G protein, and that the agonist-bound receptor conformation is different from the G protein coupled state. While most of the conformational changes observed in the β₂AR are qualitatively the same for G_s and G_i1, we detected distinct differences between the β₂AR-G_s and the β₂AR-G_i1 complex in intracellular loop 2 (ICL2). Interactions with ICL2 are essential for activation of G_s. These differences between the β₂AR-G_s and β₂AR-G_i1 complexes in ICL2 may be key determinants for G protein coupling selectivity.

G protein-coupled receptors (GPCRs) are the largest family of membrane proteins responsible for most of the transmembrane signal transduction processes. The activation of GPCRs is mediated by diverse stimuli, including neurotransmitters, hormones, ions, and photons. It is now well known that GPCRs can activate different signaling pathways through coupling with distinct downstream transducers, for example G proteins and arrestins (1). It is generally accepted that GPCRs are highly dynamic proteins that can adopt multiple distinct conformations depending on the ligands, lipids, environments, and signaling transducers through a conformational selection mechanism (1, 2). The β₂ adrenergic receptor (β₂AR) has been an important model system for studying aspects of GPCR biology and pharmacology for more than 40 y. The β₂AR can activate more than one G protein isoform, while preferentially coupling to the stimulatory G protein (G_s), which activates adenylyl cyclase, it also couples to the inhibitory G protein family (G_i/o) (3), which inhibits adenylyl cyclase (Fig. 1A).

Fig. 1.

The β₂AR-G protein signaling pathways. (A) The agonist-bound β₂AR activates either G_s or G_i heterotrimer, which stimulates or inhibits the adenylyl cyclase activity, respectively. (B) Structure of β₂AR-Gs complex (PDB ID code: 3SN6), the lysine residues that are chosen as NMR probes are shown as solid spheres, other lysine residues were mutated to arginine as described in the text. (C–E) The lysine probes undergo conformational changes during the activation of the β₂AR, as shown by cytoplasmic views of active β₂AR (PDB ID code: 3SN6) (C), inactive β₂AR (PDB ID code: 2RH1) (D), and the overlap of active and inactive β₂AR (E).

The structure of the β₂AR-G_s complex revealed the interaction between β₂AR and G_s, and the structural changes that take place upon complex formation (4), yet, the dynamic process of β₂AR signaling through G_s is still not fully understood. The structural features of the β₂AR-G_s complex have been observed in GPCR-G protein complexes for G_i, G_o, and G₁₁ proteins determined by cryo-electron microscopy (cryo-EM) (5–7); however, the molecular mechanism of G protein coupling specificity is still unclear. This may in part be due to the fact that all of the GPCR-G protein complex structures solved so far are captured in a biochemically stable nucleotide-free state, while specificity determination may happen at an earlier stage of complex formation. Single-molecule studies of the β₂AR coupling to G_s provide evidence for at least one intermediate state (8). The existence of intermediate states is further supported by monitoring β₂AR-G_s complex formation using time-resolved radiolytic footprinting and hydrogen-deuterium exchange (9).

NMR spectroscopy has previously been used to study GPCR dynamics (10–16). Several site-specific isotopic labeling strategies have been shown to be feasible for studying conformational dynamics of GPCRs, such as ¹³C-dimethylated lysine, ¹³CH₃-ε-methionine, and ¹⁹F-labeled cysteine (17). Previous NMR studies of the β₂AR using these methods have revealed weak allosteric coupling between ligand binding domain and G protein coupling domain, and that the β₂AR can adopt multiple conformational states during activation (11–13). However, due to the limitation of large molecular weight and difficulty in sample preparation, little is known about the conformational dynamics of the β₂AR in response to different G protein subtypes. In this study, we utilized ¹³C-dimethylated lysine as an NMR probe to investigate the conformational changes of several sites in the intracellular G protein binding interface of the β₂AR. ¹³C-dimethylated lysine is an ideal probe for this purpose because of its large, flexible side chain, enabling studies of larger protein complexes. Our studies show that in addition to TM6, agonist binding can also promote conformational changes in TM4, TM5, and TM7. We observe further structural changes in β₂AR coupled to G_s and G_i1. While the structural changes stabilized by G_s and G_i1 are qualitatively similar in most cytoplasmic domains probed, we detected distinct differences between the β₂AR-G_s and the β₂AR-G_i1 complex in intracellular loop 2 (ICL2). In the inactive-state structure of the β₂AR, ICL2 is a loop, while in the β₂AR-G_s complex, ICL2 forms an α-helix that positions F139^34.51 in ICL2 to engage a hydrophobic pocket in G_αs, thereby triggering nucleotide release (4).

Results

Design of Modified β₂ARs for ¹³C-Dimethylated Lysine NMR Studies.

¹³C-dimethylated lysine probes have long served as excellent probes for NMR experiments due to high sensitivity and relatively long transverse relaxation times (18). The WT β₂AR contains 16 lysine residues, most of which are located at the G protein coupling interface (Fig. 1 B–E and SI Appendix, Fig. S1A). Therefore, these lysine residues can be used as NMR reporters to monitor structural changes in response to agonists and G proteins.

However, previous studies have shown that the ¹³C-dimethylated lysine spectra of C-terminally truncated WT β₂AR (β₂AR-365N) was too crowded to make assignments (11), probably due to the similar chemical environments of these intracellular lysines. To avoid the NMR signals overlapping, we made a truncated lysine-free β₂AR construct (β₂AR-365N-zero-K) by mutating all lysines to arginines as previously reported (19) and removing the C-terminal 48 residues (SI Appendix, Fig. S1 A and D). Radio-ligand binding studies show that these modifications have only small effects on the affinity of antagonists and agonists (SI Appendix, Fig. S1C). A previous study showed that the lysine-free β₂AR has functional properties that are comparable to WT β₂AR in stimulation of adenylyl cyclase activity (19) and the recruitment of β-arrestin (20). The only reported effect of removing lysine is loss of ubiquitination and subsequent targeting receptor to lysosome for degradation (20). We then individually introduced eight lysines to different transmembrane segments in the intracellular domain of the β₂AR based on the lysine-free construct (Fig. 1 B–E and SI Appendix, Fig. S1A). These eight residues were K60^1.59, K140^34.52, K147^4.39, K227^5.56, K267^6.29, K273^6.35, R328K^7.55, R333K^8.51 (superscripts are Ballesteros and Weinstein numbering; ref. 21), located at the intracellular end of TM1, TM4, TM5, TM6, TM7, ICL2, and H8. The resulting eight constructs were expressed, purified, and labeled through reductive methylation (SI Appendix, Figs. S1B and S2A). Methylation efficiency was high, as determined by loss of labeling with fluorescamine, an amine reactive fluorophore (SI Appendix, Fig. S1B). We applied ¹H-¹³C heteronuclear single quantum coherence spectroscopy (¹H-¹³C HSQC) to each construct. The signal of each lysine was found at around 2.85 ppm in the ¹H dimension and 45 ppm in the ¹³C dimension as compared to the spectrum of lysine-free β₂AR (SI Appendix, Fig. S3 A–H). The signals are unlikely from the N terminus because the N-terminal Flag tag was removed after methylation. In order to make sure they are not residual N terminus signals due to incomplete protein digestion, we measured the signals from the methylated N terminus and they are in different positions from the lysine signals (SI Appendix, Fig. S3I). For each β₂AR-365N-one-K construct, we obtained the ¹H-¹³C HSQC spectra under the following conditions: unliganded (apo-state) β₂AR, BI-167107-bound, carazolol-bound, BI-167107-bound β₂AR in complex with Gs or Gi1 in the presence of 300 µM GDP (G_s^GDP or G_i1^GDP), and BI-167107-bound β₂AR in complex with nucleotide-free Gs or Gi1 (G_s^EMPTY or G_i1^EMPTY). For these studies we used saturating concentrations of BI-167107, a potent full agonist with a dissociation half-life of 400 min and an association half-life of less than 4.4 min (12). For inverse agonist, we used a saturating concentration of carazolol. The very slow dissociation half-lives for these ligands ensures that they are bound throughout the NMR experiment.

There are crystal structures of carazolol-bound β₂AR (22) and the BI-167107–bound β₂AR-G_s complex (4); however, it has not been possible to crystalize native β₂AR in the apo-state or bound to agonist alone, most likely due to the inherent instability of the receptor under these conditions. While β₂AR also couples to G_i, there are no reported structures of a β₂AR-G_i complex. Therefore, the NMR experiments provide structural insights into the apo-state, the agonist-bound β₂AR, and the β₂AR-G_i1 complex relative to available crystal structures. Figs. 2–4 show the comparison of ¹H-¹³C HSQC spectra of the eight ¹³C-dimethylated lysine probes in TM1, ICL2, TM4, TM5, TM6, TM7, and Helix-8 in four different conditions: apo-state, inverse agonist-bound, agonist-bound, and agonist + G_s^EMPTY.

Fig. 2.

Conformational changes of TM1 and H8 during β₂AR activation. (A and B) The ¹H-¹³C HSQC spectra of K60^1.59 from TM1 (A) and K333^8.51 from H8 (B). The one-dimensional (1D) slices of corresponding ¹H-¹³C HSQC spectra in H dimension are shown at the bottom. (C–E) The local environments of the two probes are shown by side views in inactive (gray, C) and active conformations (green, D), as well as cytoplasmic views in inactive (gray, E) or active conformations (green, F). The agonist BI-167107 is abbreviated by BI.

Fig. 4.

Conformational changes of TM5, TM6, and TM7 during β₂AR activation. (A–D) The ¹H-¹³C HSQC spectra of K227^5.71 (A), K267^6.29 (B), K273^6.35 (C), and K328^7.55 (D). The 1D slices of corresponding ¹H-¹³C HSQC spectra in H dimension are shown at the bottom. The local environments of the four probes are shown in inactive (gray, E) and active (green, F) conformations. The α5 helix from Gsα is shown as purple cartoon. The agonist BI-167107 is abbreviated by BI.

Apo-β₂AR Compared to Inverse Agonist-Bound States.

The β₂AR exhibits basal activity for G_s activation in the apo-state. This activity can be suppressed by inverse agonists such as carazolol. Previous ¹³C-methionine and ¹⁹F NMR studies revealed relatively subtle conformational changes when comparing inverse-agonist stabilized inactive-state and the apo-state β₂AR (12, 13). Consistent with these findings, our results show relatively small spectral changes in several domains of the intracellular surface in apo-state and the carazolol-stabilized inactive state. We observe the largest difference between the apo-state and carazolol-bound inactive state is in K60^1.59 (Fig. 2A), suggesting a change in its chemical environment possibly due to a change in polar interactions with the end of Helix 8 (Fig. 2 C–F). In the apo-state, K60^1.59 is represented by strong Peak 1 with a weak shoulder Peak 2. In carazolol-bound receptor we observe an up-field shift in Peak 1 and no change in Peak 2.

In the apo-state we observe two peaks for K147^4.39 at the end of TM4 (Fig. 3), K273^6.35 in TM6, and R328K^7.55 in TM7 (Fig. 4). This suggests the existence of conformational heterogeneity in these receptor domains with exchange between two states on slow time scale (seconds or slower). In K273^6.35 and R328K^7.55 we observe very small up-field shifts in the position of Peak 1 in carazolol-bound receptor compared to the apo-state that are in the opposite direction observed for agonist (Fig. 4 C and D).

Fig. 3.

Conformational changes of ICL2 from inactive to active state. (A and B) ¹H-¹³C HSQC spectra of K140^34.52 (A) and K147^4.39 (B). The 1D slices of corresponding ¹H-¹³C HSQC spectra in H dimension are shown at the bottom. The local environment of the two probes are shown by side views in inactive (gray, C, E) and active (green, D, F) conformations. G_αs is shown in purple (D). The agonist BI-167107 is abbreviated by BI.

Conformational Changes following Agonist Binding and G_s Coupling.

To correlate structural changes observed by NMR with the β₂AR-G_s crystal structure, we obtained spectra of the β₂AR bound to nucleotide-free G_s (β₂AR-G_s^EMPTY). β₂AR-G_s^EMPTY was formed by adding the nucleotidase apyrase after complex formation to degrade any released GDP. In the β₂AR-G_s^EMPTY crystal structure, the largest structural changes are the outward movement of TM5 and TM6 and inward movement of TM7, which allows the insertion of the α5 helix of G_αs (Fig. 1 C–E and SI Appendix, Fig. S4). We observed obvious spectral changes when comparing β₂AR bound to agonist alone with the β₂AR-G_s^EMPTY complex in all probes except K333^8.51 in Helix 8 (Figs. 2–4). This may be due to a combined effect of direct G_s protein binding, as well as a fully active conformation stabilized only by agonist and G_s protein, but not agonist alone (13).

ICL1 and helix 8.

The spectrum of K60^1.59 at the end of TM1 is nearly identical for apo-state and agonist-bound state (Fig. 2A), suggesting little perturbation of conformational dynamics in the junction of TM1 and ICL1. ICL1 and Helix 8 have no direct interaction with G_s in the β₂AR-G_s^EMPTY complex. The observation of intensity reduction upon coupling with G_s for K60^1.59 (Fig. 2A) and K333^8.51 (Fig. 2B) may result from the slower tumbling due to the formation of complex with a much larger molecular weight. We do observe a change in the spectra of K60^1.59 upon coupling to G_s. In the inactive state K60^1.59 forms a polar interaction with E338^8.56 in Helix 8 (Fig. 2 E and F). This is lost in the β₂AR-G_s^EMPTY complex due to a small ∼1.5 Å movement of TM1 away from Helix 8 (Fig. 2 C and D). We also observe the appearance of a weak downfield peak in R333K^8.51 in H8.

ICL2.

In the inactive state structure of the β₂AR, ICL2 has no secondary structure while it is an alpha-helix in the β₂AR-G_s^EMPTY complex (Fig. 3 C and D). The helix positions F139^34.51 in ICL2 to engage a hydrophobic pocket formed by H41, V217, F376, C379, R380, and I383 in the G_αs subunit (Fig. 3D). This interaction is essential for G_s activation (4, 23, 24). The conformational changes detected in the NMR spectra of K147^4.39 and K140^34.52 may represent the chemical environment change during the transition of ICL2 from a loop to a helix (Fig. 3 A and B).

K147^4.39 is located at the junction of TM4 and ICL2 (Fig. 1 C and D). In the inactive state, K147^4.39 forms a hydrogen bond with the backbone carbonyl of Q65^12.51 in ICL1 (Fig. 3E). This interaction is lost in the structure of β₂AR-G_s^EMPTY (Fig. 3F). In the apo state, the two peaks likely represent an equilibrium between the formation (Peak 2) and disruption (Peak 1) of the hydrogen bond in slow exchange (Fig. 3B). The observation of one dominant peak for K147^4.39 in the agonist-bound state compared to the two peaks in the apo-state most likely suggests faster exchange rate between the different conformations observed in the apo state (Fig. 3B). In β₂AR-G_s complex we observe predominantly Peak 1, consistent with the loss of the hydrogen bond (Fig. 3B).

In contrast to the dynamics in the junction of TM4 and ICL2 observed in the apo and agonist-bound receptor, we observe a strong single peak for K140^34.52 at the junction of ICL2 and TM3 (Fig. 3A). In the inactive state, K140^34.52 forms a hydrogen bond with Q229^5.58 in TM5, while in the β₂AR-G_s complex it is adjacent to the β₂AR-G_s interface (Fig. 3 C and D). Coupling with G_s dramatically alters the chemical environment of K140^34.52, as evidenced by the appearance of a new downfield peak (labeled as peak2). Only a small fraction of Peak 1 remains following the formation of β₂AR-G_s^EMPTY, consistent with a minor population of uncoupled β₂AR (Fig. 3A).

TM5, 6, and 7.

While we observe spectral changes in K227^5.56, K267^6.29, K273^6.35, and R328K^7.55 upon agonist binding, further changes are observed following G protein coupling (Fig. 4 A–D), consistent with changes observed in TM5, TM6, and TM7 in the β₂AR-G_s structure (Fig. 4 E and F). The slight broadening of the peak representing K227^5.56 upon agonist binding suggests the appearance of two or more conformations in intermediate exchange at the intracellular side of TM5 (Fig. 4A). Upon coupling to G_s we see the appearance of a new downfield peak (Peak 2) and a weaker Peak 1 (Fig. 4A).

The two TM6 probes, K267^6.29 and K273^6.35, show significant differences in terms of peak patterns. In the apo and agonist-bound state K273^6.35 is represented by two weak peaks, indicating that there are at least two conformations with a slow exchange rate, while K267^6.29 is represented by a stronger single peak (Fig. 4 B and C). This is because K267^6.29 is located at the end of TM6 and its side chain is expected to exhibit relatively high conformational flexibility, whereas K273^6.35 is located in the transmembrane core and its side chain can interact with neighboring residues I277^6.39 and R328^7.55 (Fig. 4 E and F). Upon agonist binding, there is a downfield shift in the peak representing K267^6.29. There are two peaks for K267^6.29 in the β₂AR-G_s complex (Fig. 4B). Peak 2 has the same chemical shift seen with agonist alone but with marked reduction in intensity. Of note, the dynamics of the cytoplasmic end of TM6 may impact the cytoplasmic end of TM5, given their close proximity (Fig. 4F). K227^5.56 in TM5 is close enough to have a polar interaction with E268^6.28 (Fig. 4F). Thus, the minor Peak 1 in K227^5.56 and in K267^6.29 may reflect different TM5–TM6 interactions in a small population of nucleotide-free β₂AR-G_s complex.

It is interesting to note that the spectra of K273^6.35 and R328K^7.55 are similar (Fig. 4 C and D). This may be due to the close proximity of K273^6.35 and R328^7.55 observed in the inactive state crystal structures (Figs. 1 C and D and 4E) where R328K^7.55 could form a hydrogen bond with the backbone carbonyl of K270^6.32. This interaction would be lost upon coupling with G_s. For both K273^6.35 and R328K^7.55, agonist binding leads to a downfield shift in Peak 1, and Peak 1 is lost in the β₂AR-G_s complex. Peak 2 may represent a conformation that is more exposed to the solvent, therefore Peak 2 is less affected by agonist-binding or G protein coupling than Peak 1.

Conformational Differences in G_s and G_i1 Coupling.

Recently, a number of GPCR-G_i/o complex structures have been determined by cryo-EM (5–7, 25–28). The G_s and β₂AR interaction interface is composed by TM3, TM5, TM6, and ICL2. While G_αi/o interactions with TM3, TM5, and TM6 are similar in GPCR-G_i/o protein complexes, TM2, TM7, and H8 are also involved in formation of the A1 adenosine receptor (A1R)-G_i complex, and ICL2 and the junction between TM7 and H8 contribute to formation of the rhodopsin-G_i complex (SI Appendix, Fig. S5). For GPCR-G_i/o complexes, the outward displacement of TM6 is smaller than observed for the β₂AR-G_s complex. The differences in the interaction interface for these complexes might contribute to G protein coupling preferences. As noted above, the β₂AR can couple to both G_s and G_i. However, efforts toward obtaining the structure of β₂AR-G_i complex have failed due to the relatively weak interaction and instability of the complex. As a consequence, little is known about differences in the interactions between β₂AR and G_i compared with the β₂AR-G_s complex.

To provide more structural insights into the basis of β₂AR coupling selectivity for G_s over G_i, we next sought to compare β₂AR binding with G_s and G_i by using the ¹H-¹³C HSQC spectra of the β₂AR-BI-167017-G_i1^EMPTY complex (Fig. 5). The β₂AR-G_i1^EMPTY complex was formed by adding the nucleotidase apyrase after complex formation to degrade any released GDP. Fig. 5 compares the spectra of β₂AR-G_s^EMPTY and β₂AR-G_i1^EMPTY complexes. Consistent with the previous observation that the β₂AR-G_i1 complex is less stable than the β₂AR-G_s complex (29), we observed smaller spectral changes in the β₂AR-G_i1 complex sample. The NMR spectral changes of K227^5.56, K267^6.29, K273^6.35, and R328K^7.55 corresponding to the conformational changes of TM5, 6, and 7 are similar to those obtained from the β₂AR-G_s complex; however, peak 2 of K227^5.56 and K273^6.35 and peak 1 of K267^6.29 are notably weaker in the β₂AR-G_i1 complex compared to the β₂AR-G_s complex. However, it’s hard to discriminate whether the TM6 outward displacement of β2AR−G_i1 is smaller than that of β2AR−G_s based on the signal intensity difference observed by NMR.

Fig. 5.

Comparison of conformational changes in β₂AR coupled to Gs or Gi1. BI is short for BI-167107. (A–H) The ¹H-¹³C HSQC spectra of the eight probes from TM1 to H8. The eight probes are K60^1.59 (A), K140^34.52 (B), K147^4.39 (C), K227^5.56 (D), K267^6.29 (E), K273^6.35 (F), R328K^7.55 (G), and R333K^8.51 (H). For each probe, the overlay of the ¹H-¹³C HSQC spectra of agonist-bound β₂AR (green) and β₂AR-Gs^EMPTY (blue) are shown at the top, followed by the comparison of the ¹H-¹³C HSQC spectra of agonist-bound β₂AR (green) and β₂AR-Gi1^EMPTY (pink), then by the overlay of β₂AR-Gs^EMPTY (blue) and β₂AR-Gi1^EMPTY (pink). The 1D slices of corresponding ¹H-¹³C HSQC spectra in H dimension are shown at the bottom.

Remarkably, there are nearly no chemical shift changes for K140^34.52 in ICL2 and K147^4.39 in TM4 in the β₂AR-G_i1 complex, which is in contrast to what is observed for β₂AR-G_s complex (Fig. 5 B and C). The observed decrease of peak intensity is likely due to the increase in mass. The lack of spectral changes for K140^34.52 and K147^4.39 suggest that ICL2 does not form an alpha helix when coupled to G_i1 and may have only weak interactions with G_i1, possibly explaining the less efficient coupling and the overall instability of the β₂AR-G_i1 complex.

As noted above, the insertion of F139^34.51 in ICL2 of the β₂AR into a hydrophobic pocket formed by H41, V217, F376, C379, and R380 in the G_αs is essential for GDP release (SI Appendix, Figs. S6A and S7A). Of these, only F376 is conserved as F336 in G_i (SI Appendix, Fig. S6 B and C). When comparing sequences of ICL2 from GPCRs that couple to G_s, G_i/o, G_q/11, and G₁₂, position 34.51 is most often a L, F, I, or M for G_s coupled receptors, while a broader range of amino acids are found in G_i/o coupled receptors (SI Appendix, Fig. S6D). It is possible that interactions with 34.51 in ICL2 may be less important for coupling with G_i1 than G_s. This is in agreement with the recent structures of G_i in complex with the cannabinoid receptor subtype 1 (CB1) (27), the µ-opioid receptor (µOR) (6), and the A1R (7). For these receptors, the interactions between the amino acid at position 34.51 in ICL2 (L in CB1 and A1R, and V in µOR) and G_i are much weaker, interacting with only one or two side chains of G_i (SI Appendix, Fig. S7 B–D). However, this weaker interaction between ICL2 and G_i is not universal. For the neurotensin receptor subtype 1 (NTSR1), which couples promiscuously to G_i/o, G_s, and G_q/11 (30), F174^34.51 in ICL2 packs into a pocket formed by four amino acids in G_i (SI Appendix, Fig. S7E). The importance of the amino acid in position 34.51 was initially demonstrated for the G_q/11 coupled M1 muscarinic receptor (M1R) where the substitution of L131^34.51 for Ala in ICL2 led to a loss of G_q/11 coupling (24). In the recent cryo-EM structure of the M1R-G11 complex, L131^34.51 packs into a pocket formed by four amino acids in G11 (28) (SI Appendix, Fig. S7F). Like G_s coupled GPCRs, L, F, I, or M are more commonly found at position 34.51 in G_q/11 coupled GPCRs (SI Appendix, Fig. S6D). While the evidence suggests that weak interactions between F139^34.51 in ICL2 of the β₂AR and G_i1 may account for poor coupling efficiency, we cannot exclude the possibility that β2AR simply couples less efficiently to G_i and the lack of strong engagement with ICL2 is a manifestation of that.

Similar to G_s, G_i1 binding also has little effect on R333K^8.51 located at H8 (Fig. 5H). The decrease in intensity is likely due to the increased size of the complex. Notably, the decrease in intensity is greater for G_s^EMPTY compared to G_i1^EMPTY, consistent with the formation of a more stable complex. The low intensity downfield peak observed in R333K^8.51 with G_s is not observed with G_i1. The spectra of K60^1.59 in TM1 are similar for G_s and G_i1, with a reduction of intensity and the peak splitting into a slightly upfield-shifted peak and a smaller downfield peak (Fig. 5A). ICL1 and H8 are not observed to interact with Gs in the β₂AR-G_s complex; however, the orientation of ICL1 relative to H8 changes upon complex formation with the loss of an electrostatic interaction between K60^1.59 in ICL1 and R328K^7.55 in H8.

We have previously observed evidence for a transient intermediate state following the disruption of the β₂AR-G_s complex by the addition of GDP (8). In an effort to detect this state by NMR spectroscopy, we formed β₂AR-G_s and β₂AR-G_i1 complexes in the presence of 300 µM GDP. As can be seen in SI Appendix, Figs. S8 and S9, we don’t observe a distinct state in the presence of GDP. The spectra are consistent with a mixture of nucleotide-free β₂AR-G_s or β₂AR-G_i1, and β₂AR bound to agonist alone. Our inability to detect a distinct GDP-bound β₂AR-Gs complex could be due to the insensitivity of our probes or to the fact that the amount of the intermediate may be too small to detect.

Molecular Dynamics Simulations Showed Different ICL2 Behavior in the β₂AR-Gi^EMPTY and the β₂AR-Gs^EMPTY Complex.

To investigate the dynamics at the receptor G protein interface, classical unbiased MD simulations of the β₂AR-G_i1^EMPTY and the β₂AR-G_s^EMPTY complex were carried out. The β₂AR-G_i1^EMPTY model was built based on the β₂AR-G_s^EMPTY complex (Protein Data Bank [PDB] ID code: 3SN6) (see Methods for more details) and as a result, ICL2 in the starting structures of both β₂AR-G_i1^EMPTY and β₂AR-G_s^EMPTY complex are in an alpha-helical conformation. Three independent, 3-µs-long molecular dynamics (MD) simulations were carried. While the simulations are too short to observe unfolding of the secondary structure, the significantly higher rmsd of the ICL2 region of the β₂AR-G_i1 complex compared to the β₂AR-G_s complex suggests the starting alpha-helical structure of ICL2 is less stable in the β₂AR-G_i1^EMPTY structure. Of note, such changes are not observed for ICL1, H8, nor G_α(α5) when both systems are compared (Fig. 6 A–D). The only other structural elements at the receptor G protein interface with increased rmsd values are TM5 and TM6 (Fig. 6 E and F), consistent with the high dynamics of their cytoplasmic ends observed by chemical shift analysis in β₂AR-G_i1^EMPTY and in β₂AR-G_s^EMPTY complexes. A close inspection of the MD simulations results suggests a higher flexibility of K140^34.52 and K147^4.39 from ICL2 for the β₂AR-G_i1^EMPTY compared to ICL2 from β₂AR-G_s^EMPTY as measured by the average rms fluctuation (RMSF) (SI Appendix, Fig. S10). The average RMSF value for K140^34.52 is two times higher and for K147^4.39 is six times higher in β₂AR-G_i1^EMPTY than β₂AR-G_s^EMPTY. The MD simulations results are in agreement with the NMR experimental observation that ICL2 plays different roles when β₂AR couples to G_i or G_s.

Fig. 6.

Comparison of conformational changes measured by MD simulations in β₂AR coupled to G_s or G_i. Backbone rmsd in Å extracted from 3 × 3-µs-long MD simulations of β₂AR-Gi (pink) and β₂AR-Gs (cyan) of structural elements at the receptor-G protein interface. (A) ICL1 of β₂AR: E62^12.48-Q65^12.51. (B) ICL2 of β₂AR: I135^3.54-A150^4.42. (C) H8 of β₂AR: S329^8.47-C341^8.59. (D) α5 helix of Gαi and Gαs. (E) TM5 of β₂AR: N196^5.35-R239^ICL3. (F) TM6 of β₂AR: C265^6.27-Q299^6.61. The rmsd values were calculated relative to the respective structural element in the starting structure of β₂AR-Gi^EMPTY and β₂AR-Gs^EMPTY.

Discussion

We applied NMR spectroscopy to monitor structural changes in the cytoplasmic surface of the unliganded β₂AR and in response to an inverse agonist, an agonist, G_s, and G_i. We have been unable to obtain crystal or cryo-EM structures of the apo-β₂AR and agonist-bound β₂AR due to its instability and possibly due to conformational heterogeneity. Our NMR studies provide further evidence for conformational heterogeneity as we observe more than one peak for K147^4.39, K273^6.35, and K328^7.55 in apo and agonist-bound β₂AR (Figs. 2–4). Agonist binding leads to changes in K147^4.39 in ICL2, K267^6.29, and K273^6.35 in TM6, and in K328^7.55 in TM7; however, additional changes are observed upon the addition of G_s and G_i1. This suggests that for the β₂AR agonist, binding alone cannot fully stabilize the G protein bound conformation, as has been observed in previous fluorescence and double electron electron resonance spectroscopy experiments (8, 13). The structural changes stabilized by G_i1 and G_s are qualitatively the same for TM5, TM6, and TM7; however, our NMR studies and supporting MD simulations suggest that G_i1 does not promote the formation of an alpha helix in ICL2.

Methods

The receptors were expressed in Sf9 insect cells, purified by affinity column and size exclusion chromatography, and labeled by reductive methylation. NMR data were collected at 25 °C on a Burker Avance 800-MHz spectrometer equipped with a cryoprobe. The ¹H-¹³C HSQC spectra were recorded with spectral width of 11,160.71428 Hz in the ¹H-dimension (w1) and 14,492.7536 Hz in the ¹³C-dimension (w2) centered at 46 ppm. For all spectra, 1024 × 256 complex points were recorded and a relaxation delay of 2 s was used to allow spin to relax back to equilibrium. MD simulations were performed using Gromacs simulation package. Further details are proved in SI Appendix.

Data Availability.

All study data are included in the article and SI Appendix.