Exploring the Activation Process of the β2AR‑G_s Complex

Abstract

G-Protein-coupled receptors (GPCRs) belong to an important family of integral membrane receptor proteins that are essential for a variety of transmembrane signaling process, such as vision, olfaction, and hormone responses. They are also involved in many human diseases (Alzheimer’s, heart diseases, etc.) and are therefore common drug targets. Thus, understanding the details of the GPCR activation process is a task of major importance. Various types of crystal structures of GPCRs have been solved either at stable end-point states or at possible intermediate states. However, the detailed mechanism of the activation process is still poorly understood. For example, it is not completely clear when the nucleotide release from the G protein occurs and how the key residues on α5 contribute to the coupling process and further affect the binding specificity. In this work we show by free energy analysis that the guanosine diphosphate (GDP) molecule could be released from the G_s protein when the binding cavity is half open. This occurs during the transition to the G_s open state, which is the rate-determining step in the system conformational change. We also account for the experimentally observed slow-down effects by the change of the reaction barriers after mutations. Furthermore, we identify potential key residues on α5 and validated their significance by site-directed mutagenesis, which illustrates that computational works have predictive value even for complex biophysical systems. The methodology of the current work may be applied to other biophysical systems of interest.

Graphical Abstract

I. INTRODUCTION

G-Protein-coupled receptors (GPCRs) constitute one of the most important receptor families in the human body.^1,2 GPCRs respond to extracellular stimuli such as stimulation by hormones and neurotransmitters that subsequently activate cellular signals by coupling to G proteins. Malfunction of the signal transduction processes is relevant to many human pathologies including Alzheimer’s and heart diseases.^3,4 Consequently GPCRs are also important targets for drug design, where more than 30% of the drugs approved by the Food and Drug Administration (FDA) are within this family.⁵ Thus, it is of utmost importance to understand the GPCR coupled signal transduction processes.

Depending on the type of α subunit, G proteins can be classified into four families: G_s, G_i/o, G_q/11, and G_12/13. One unresolved issue is the nature behind the binding preference in the GPCRs/G protein complexes. Usually, each receptor has its preference for one type of G protein. However, promiscuous coupling to one G protein from multiple receptor families is also observed.^6,7

The C-terminal of the α subunit (α5) has been considered to play an important role in determining the coupling specificity.^8–10 However, the coupling determinants vary from one G protein family to another, and there is no simple universal motif that explains the behavior of all families.¹¹ Based on structural analysis, some residues on G_s α5 were argued to serve as the “barcode” residues that account for the coupling to the β2 adrenergic receptor (β2AR).^11,12 Nevertheless, the structural/energetic basis of the coupling between receptors and G proteins is still poorly understood.

The structure of the β2AR-G_s complex was the first structure of the GPCR family to be solved by X-ray crystallography.¹³ The overall activation mechanism of β2AR-G_s is well-defined, as shown in Figure 1. The activation process starts by binding of the agonist adrenaline to the receptor β2AR. This shifts the receptor’s allosteric equilibrium to the active form and opens a cavity for G_s binding on the cytoplasmic side of the membrane. Next, the C-terminal of the α helix of the G_s α domain (α5) undergoes a conformational rearrangement and starts to intrude into the cavity of the receptor.

Figure 1.

Schematic description of the GPCR activation pathway. (A) The receptor is activated upon binding of the agonist adrenaline, while the G_s protein is in the GDP bound state. (B) The α5 helix then undergoes a rearrangement and intrudes halfway into the β2AR binding site (the experimental intermediate structure, PDB ID: 6E67). (C) Next, the α5 helix forms a tighter binding with the receptor, and the G protein experiences a large conformational change that facilitates the binding/unbinding of the nucleotides. (D) G protein decouples from GPCR and elicits downstream signal transduction within the cell.

A recent X-ray powder diffraction (XRD) study¹² captured an intermediate structure where the α5 is half intruded into the cavity as indicated by panel B of Figure 1. From here, while α5 forms a tighter binding with the receptor, the Ras-like (RS) domain and the α-helical domain (AHD) of the α domain separate from each other. This facilitates the release of the guanosine diphosphate (GDP) molecule and a binding of the GTP molecule. This step is considered to be the rate-determining step of the GPCR activation.¹⁴ The G protein then detaches from the receptor, dissociates into the α and the βγ domain, and leads to downstream cellular signals. How does the conformational change of the system couple to GDP binding and when is GDP released are two important questions. A recent crystallographic study has shown that upon mixing of β2AR and G_s, GDP release happens much faster than the formation of the stable β2AR-G_s^empty complex.¹⁵

With the fast development of cryogenic electron microscopy (cryo-EM) and XRD techniques, various structures of the GPCR-G protein complex have been solved.^16–24 This provides us the opportunity to use advanced modeling and simulation tools to investigate these systems. However, all earlier structural studies captured stable end-point states when the G protein is closed and α5 is outside the receptor (closed-out) with GDP bound (Figure 1A) or when the G protein is fully open and the α5 is coupled to the receptor (open-in) after the GDP is released (Figure 1C). There has been very limited information about the transition process between the two states. For example, it has been unknown at what stage the GDP molecule is released form the G protein. Very recently, Liu et al. stabilized and captured a complex structure of the β2AR receptor and the last 14 amino acids of the C-terminal of G_sα5.¹² This structure showed a half-intruded interaction pattern between β2AR and G_s and is believed to be a possible intermediate structure of the transition between the closed-out G_s^GDP and the open-in G_s^empty state. Hence the reported intermediate structure provides a foundation for our computational study of the transition process.

One major difficulty in modeling biophysical systems is the size and complexity of the system (usually with thousands of amino acids). This put a serious computational limit on the use of all-atom (AA) models. Here we seek to investigate the activation process from the G_s^GDP state to the G_s^empty state of the β2AR-Gs system using our consistently developed coarse-grained (CG) model.^25–27 It is worth noting here that in biophysical systems the electrostatic term usually contributes the largest contribution from the different types of interactions; this is clearly the case in other systems, such as ATPase,^28,29 myosin,³⁰ and kinase.³¹ To correctly represent the free energy surface of the simulated system, we need to choose models with a proper description of the electrostatic term.³² Our electrostatic-based CG model for membrane proteins combined with other techniques has been proved to be very effective in investigating GPCRs^33,34 and other biophysical systems, as mentioned above. Here we will also use it to investigate the β2AR-G_s activation process.

Our coarse-grained profile located the position of the conformational barrier and the rate-determining step of the activation process for the transition from the intermediate structure to the GDP free state. Furthermore, during the receptor conformational change and GDP release, the free energy map shows a low-barrier GDP release pathway, indicating that the GDP could be released after the nucleotide binding pocket is half-opened. The mape of the landscape also suggests that the GDP could also stay inside the binding pocket after the G_s is fully open, which is consistent with experimental observations.¹² On the contrary, it is not likely that the GDP will be released right after the formation of the intermediate structure due to the corresponding high barrier. By the calculations of the effect of mutating the last 21 residues of the α5 helix and using transition state theory, we explain the measured effect of slowing down mutations.

It was found that the mutational effect of each residue depends on its overall electrostatic energy, in contrast to the conclusions of studies that only focused only on the interactions with the nearest residues. Furthermore, this work identified potential residues that might have similar mutational effects and validated them in functional experiments. The current methodology may also apply to other systems of interest such as viruses and antibodies.

II. RESULTS AND DISCUSSION

II.1. CG Free Energy Profile for Conformational Change.

The crystal structure of the intermediate of the β2AR receptor and the last 14 residues of the G_s C-terminal¹² (PDB ID: 6E67) provide us with a junction point to connect the two end-point states (states A and C in Figure 1). We also utilized another two XRD structures (PDB ID: 6EG8, 3SN6) to construct the three major states of the receptor–Gs complex in this study: the closed-out G_s^G+DP state, the experimental intermediate state (with GDP), and the GDP free open-in Gs^empty state obtained by homology modelings.^35,36 Next, we generated a series of intermediate structures between each major state by targeted molecular dynamics (TMD). For each transition we picked 10 structures at equal intervals to reproduce the free energy profiles in Figure 2 (see the SI for details). Figure 2A shows the structural differences of the α5 between the three major states. From G_s^GDP (red) to the intermediate state (purple), the α5 undergoes a straightening, rotation, and intrusion movements¹² (Movie 1). On the other hand, from the intermediate state (purple) to the G_s^empty state (yellow) the α5 forms a tighter coupling with the receptor while G_s undergoes a large conformational change, during which the nucleotide binding pocket opens (Movie 2A and Movie 2B).

Figure 2.

(A) Protein structure of the β2AR-G_s complex. The orientation of α5 in the three major states is indicated: G_s^GDP (red), intermediate (purple), and G_s^empty (yellow). Receptor is in blue and G_s is in green. For clear visualization, the whole protein is only shown for the G_s^empty state. (B) The CG free energy profile for the conversion between the three major states. These states are labeled as intermediate (I), transition state (T), and product (P); structures are shown in Figure S6. Note that the modeled intermediate structure is on the left of the experimental structure.

The CG free energy profile of the conformational transition between the three states is depicted in Figure 2B. Once the α5 system starts to bind the receptor, the free energy readily goes down. Note that the modeled intermediate structure is more stable compared to the experimental structure (the energy of the I state is lower than the energy of the exp(inter) state on the energy curve), and this might reflect the fact that the receptor at the experimental structure is stabilized by the linker units without the presence of the G_s− unit, while in the modeled structure the G_s unit is complete. The highest barrier for the whole process occurs during the transition from the intermediate state to the open-in G_s^empty state. This is the rate-determining step of the conformational transition process in the present study. We also labeled the intermediate structure (I), the barrier structure (T), and the open-in state (P) in the curve for the subsequent discussion.

II.2. Coupling between the Conformational Change and the GDP Release.

One important issue is the identification of the stage where the nucleotide is released from G_s. A recent XRD study suggests that this event should happen during the transition between the experimental recorded intermediate state and the G_s^empty state, but the exact point along the reaction coordinate is unclear.¹² On the other hand, some anticipated movements of the α5 helix and the β3 strand were not observed long after the GDP was released.¹⁵ This indicates that there is a gap between the nucleotide release event and the stable G_s^empty formation. To investigate the details of the changes of each structure from the experimental intermediate state to the G_s^empty state, we generated the GDP release free energy profile by docking a GDP and a Mg²⁺ ion at different distances away from the nucleotide binding pocket to the bulk. For each combination of system conformation and the nucleotide position we generated an individual intermediate structure for the free energy map. We first perform extensive MD relaxation of those structures and then used PDLD/s-LRA/β binding energy calculations to get the free energies (see the SI for details).

The free energy map that couples the conformational change and the GDP release is shown in Figure 3, a process that corresponds to the transition from panel B to C of Figure 1. The X axis is the conformational change coordinate, while the Y axis denotes the GDP/Mg²⁺ distance from the nucleotide binding site. Very encouragingly, we identified a least energy pathway from A to B, indicated by a black dashed line in the figure. The barrier for this pathway is 12.2 kcal/mol. Once the system reaches point B, there is no apparent barrier that can block the nucleotide from releasing from the G_s protein (moving to the upper side of the free energy map). The structure of point A corresponds to the first frame of Movies 2A and 2B. Point B is a middle state from panel 1B to 1C and corresponds to the sixth frame of Movies 2A and 2B. After reaching point B, the GDP molecule could be released from the G_s protein at any time, but it could also stay at the binding site until the stable open-in state is formed. This is consistent with the experimental findings that part of the GDP molecules stays bound to the G proteins after mixing β2AR and G_s^GDP for a sufficient time.¹² We have examined the possibility for GDP to be directly released right after the intermediate state is formed while before any major conformational change happens, as indicated by the route from A to C in Figure 3. However, the barrier of the A–C route is 20.2 kcal/mol, which is much higher than that for the A–B pathway. The current result suggests that GDP release could happen halfway during the conformational change from the experimental intermediate state to the stable G_s^empty open-in state (Figure 1B–2C). Before reaching point B on the free energy map, there is one possible least energy pathway with relatively small degrees of freedom (DOFs) on the free energy map, as indicated by the narrow tunnel before reaching point B, but after B point is reached the GDP can visit most areas on the right part of the free energy map.

Figure 3.

Coupled free energy map of the conformational change of the system and the GDP release. The possible pathways are indicated by black dashed lines. Point A is the experimental intermediate structure, while point B indicates the time point when GDP is free to release or stay in the binding pocket. The barrier along route A–B is 12.2 kcal/mol, while along A–C it is 20.2 kcal/mol. Structures of point A and B are shown in Figure S5.

II.3. Mutation Effects of Key Residues of α5.

The binding specificity between GPCRs and G proteins may be promiscuous,^6,7 where G_s-coupled receptors are found to couple to G_i proteins to some extent.⁷ However, Liu et al. mutated “barcode” residues^11,12 and found that upon mutating G_s-R389/E392 or G_s-H387/Y391 to alanine, the rate of coupling between β2AR and G_s is reduced. They also found that such mutation would decrease the rate of GDP release. These investigators argued that the mutation could disrupt the salt bridge between those key residues and the counterpart residues on the receptor, either in the intermediate or in the open-in G_s^empty state, that might be related to the GDP release.

Our initial aim was to use simulations to investigate if there are other residues on α5 that could lead to the destabilization of the structure by mutating different residues to ALA and slowing down the rate of the receptor coupling. The results of our PDLD/s-LRA/β calculations for mutational effects for the intermediate state (I), the transition state (T), and the product state (P, in Figure 2) are shown in Figure 4A. To our surprise, only E392A shows a large free energy increase (destabilization) after mutating all three states. In contrast, all the other three residues (R389A, Y391A, and H387A) show a decrease in free energy (stabilization) after mutations. These results contradict the explanation proposed by Liu et al.¹² as mentioned in the paragraph above.

Figure 4.

(A) Mutational effects after substitutions of the last 21 residues on α5 of the three states I, T, and P for the coupling process that was defined in Figure 2. (B) The barrier change calculated by forward barrier = G(T) – G(I) and backward barrier = G(T) – G(P).

In general, the decrease in the rate of a reaction at a constant temperature reflects the increase in the reaction barrier. Taking the coupling reaction as defined in Figure 2 into account, the forward reaction barrier is equal to G(T) – G(I), while the backward barrier is G(T) – G(P). Both the increase in the forward barrier and the decrease in the backward barrier can slow down the binding rate between β2AR and G_s. It is worth noting here that even though the backward barrier also affects the reaction rate, in this nonequilibrium system when the reaction starts to proceed, the reactant (state I) population is much higher than the population of the P state and the energy of P is relatively low compared to that of I. Thus, the forward barrier plays a dominant role.

The contributions of the last 21 residues of α5 to the barrier change are shown in Figure 4B. All four residues (E392A, R389A, Y391A, H387A), whose mutations caused a slow-down of the binding rate and the GDP release rate, contribute to the calculated increase in the reaction barrier. Structurally (see Figure S1), both E392 and R389 are surrounded by multiple positively charged residues (colored in blue) with respect to the electrostatic environment around the key residues in the intermediate state (PDB ID: 6E67). The mutational effects of the key residues are determined by the overall electrostatic energy and not just one salt-bridge interaction, thus explaining why R387 presents a large stabilization effect after alanine substitution (Figure 4A). This energy-based explanation accounts well for the slow-down mutational effect that is observed experimentally.¹² We note in this respect that some of the PDLD calculations clearly present an overestimate due to insufficient relaxation (which would be reflected by higher dielectric constant), but the trend is correct.

On the basis of the current analysis, we can further identify other residues on α5 that might have the same slow-down mutational effect on the receptor coupling other than the reported barcode residues E392, R389, Y391, and H387. By examining the barrier change of the last 21 residues on α5, we found that D381A, M386, and C379 might show similar mutational effects to those of E392A/R389A and Y391A/ H387A. D381A should behave similarly to E392A based on the change in negative charge. On the other hand, there is no adjacent residue that can form a disulfide bond with either M386 or C379. The mutational effect for these two residues might show a response of the amino acid to the electrostatic environment.

II.4. Experimental Validation of Computational Predicted Mutational Effects.

To validate our predictions, we performed mutational experiments for D381A, M386A, and C379A. After binding with agonists, the G_s can couple to beta-2 adrenergic receptor (β2AR) and then induce production of cAMP.^37,38 In the present study, HeLa cells were cotransfected with β2AR and the G_s mutants, and the cAMP levels were detected by a time-resolved fluorescence resonance energy transfer (TR-FRET) after stimulation with the β2AR agonist isoproterenol (ISO).^39–41 The coupling ability of the G_s mutants can be reflected by the cAMP levels and the EC₅₀ values (half-maximal effective concentration of the drug ISO). The concentration of cAMP was calculated against the cAMP standard curve, as shown in Figure 5. The results show that the initial cAMP concentration in the G_s wild-type (G_s WT) group is higher than in the G_s mutants group (D381A < C379A < M386A < WT) (Table 1). In the resting state, the cAMP level in the G_s WT group is also higher than in the G_s mutants (Figure 5A). This means that self-activation of cAMP is decreased with the G_s mutations. The results indicate that the receptor coupling ability of the G_s mutants decreases compared to the corresponding coupling ability in the wild-type the G_s. After stimulation of 10⁻¹¹ M ISO, the G_s mutants are also less effective at triggering cAMP production than G_s WT (Figure 5B). The G_s WT and G_s mutants show no significant difference in response to the 10⁻⁸ M ISO in cAMP level (Figure 5C). In addition, the EC₅₀ value that induces a response halfway between the baseline and maximum was increased in the G_s mutants group (D381A > C379A > M386A > WT, Figure 5D and Table 1). In other words, a higher concentration of ISO is needed to obtain the half-maximal effect after mutations of the G_s. These results validate the predictions that the coupling ability between the G_s mutants and the receptor was decreased compared to the wild-type G_s.

Figure 5.

cAMP assay. HeLa cells (3000 cells/well) were cotransfected with pcDNA3.1 β2 AR and either Gs WT or one of the Gs mutants (D381A, M386A, and C379A). The transfectants were treated with phosphate-buffered saline (A), 10⁻¹¹ M ISO (B), or10⁻⁸ M ISO (C), and the cAMP level was detected by an Envision multilabel plate reader; (D) cAMP dose curves of different Gs mutants. The transfectants were treated with different concentrations of ISO. The data are expressed as means ± SEM from three experiments.

Table 1.

Parameters of Dose-Dependent Curves

	initial cAMP concentration (nM)	EC50 (M)
Gs WT	62.84	8.87 × 10−12
Gs C379A	44.16	2.30 × 10−11
Gs D381A	31.96	6.13 × 10−11
Gs M386A	55.13	1.80 × 10−11

The above discussion describes a systemic way for investigating mutational effects of key residues that might play essential roles at the interface between the receptor and the functional proteins. This is done by examining the change of the free energy barriers upon mutations. This methodology has a major advantage over structural analysis that relies on a distance comparison of key interacting residues of a single or several structures at stable states, which lacks the fundamental information on the relevant transition states. The current methodology could also be applied to other systems in addition to GPCRs.

II.5. Discussion.

Crystallographic experiments captured the high-resolution structures of β2AR-G_s complexes at key intermediate states during the coupling process,¹² thereby paving the way for theoretical and modeling studies that explore the transition reaction at the molecular levels. The intermediate structure of α5, which is a half-intruded β2AR-Gs complex, bridges the gap between the β2AR + G_s^GDP state and the well-coupled β2AR-G_s^empty open-in state. By utilizing this structure, we have generated the free energy profile for the conformational change and to identify the rate-determining step of the coupling process. To investigate the accurate GDP release time, we coupled the conformational change of the system with the GDP release profile in order to generate a full free energy map. On the landscape map we identified a specific pathway for the GDP release with a relatively low barrier of 12.2 kcal/mol. According to the free energy map, the GDP is free to release to the bulk after G_s is half open, but it may also stay in the pocket until the formation of the stable β2AR-G_s^empty complex.

To date, most structural analyses of the mutational effects have been based on assumed change of “key interactions”. However, suchan analysis missed the essential information on the structure of transition states and reaction barriers, which plays a dominating role in determining reaction rates. On the other hand, the mutational effect is affected by the overall electrostatic energy of the substitute site and not a single interaction. Our kinetic analysis of the experimentally observed slow-down mutational effect of key residues (E392A/R389A, Y391A/H387A) establishes a general methodology. The mutations to ALA of the four barcode residues were found to lead to an increase in the barriers for coupling. By examining the local electrostatic environment in the binding pocket, we found that α5 is surrounded by more positively charged residues than negatively charged residues; this caused a stabilization of positively charged residues after mutations such as R389A. Our analysis also identified potential residues (D381, M386, C379) that could behave in a similar way. These predictions were validated by HeLa cell experiments, which further demonstrates the strength of in-silico computations and predictions for large biophysical systems.

The current work emphasizes the fact that the change of the reaction rates after mutation reflects the alternation of reaction barriers and cannot be accounted for just by distance measurements between key residues of stable structures.

III. METHODS

III.1. Complex Assembling.

We used Modeler^35,36 to perform homology modeling in order to generate the three major states in this study. The relevant PDB structures are 6EG8 (G_s^GDP state), 6E67 (exp intermediate), and 3SN6 (G_s^empty state). Then we used targeted molecular dynamics⁴² (TMD) to generate intermediate structures that connect them. For the GDP release coordinates, we performed GDP + Mg²⁺ docking calculations by Molaris-XG^43,44 on each intermediate structure from the nucleotide binding sites to bulk at different distances; totally 25 structures were generated for each intermediate structure. After obtaining the structures, we add membrane particles at the β2AR domain and perform extensive MD relaxation using Molarix-XG until the energy is converged. The binding energy and solvation energy are calculated by the PDLD/s-LRA/β method.^45–47 See the SI for details.

III.2. Total Energy of the CG Model.

We have been consistently developing a CG model^26,27 that emphasizes the electrostatic effects in proteins and consistently reflects the solvation of ionizable residues. In applying this model we first used Molaris-XG to trim the structures into CG presentation, in which the main chain of the amino acids is still in all-atom form, but the side chain is reduced into a simplified united atom. Then we perform another relaxation and use a Monte Carlo proton transfer algorithm to determine the optimistic charge distribution of the chargeable amino acids before free energy evaluation. The total energy of our CG model is

$$\begin{array}{l}\Delta G_{\text{fold}}\hfill \\ =\Delta G_{\text{main}}+\Delta G_{\text{side}}+\Delta G_{\text{main-side}}\hfill \\ =c_1\Delta G_{\text{side}}^{\text{vdw}}+c_2\Delta G_{\text{solv}}^{\text{CG}}+c_3\Delta G_{\text{HB}}^{\text{CG}}+\Delta G_{\text{side}}^{\text{elec}}+\Delta G_{\text{side}}^{\text{polar}}+\Delta G_{\text{side}}^{\text{hyd}}\hfill \\ +\Delta G_{\text{main-side}}^{\text{elec}}+\Delta G_{\text{main-side}}^{\text{vdw}}\hfill \end{array}$$

The terms on the right are the side chain van der Waals energy, main chain solvation energy, main chain hydrogen bond energy, side chain electrostatic energy, side chain polar energy, side chain hydrophobic energy, main chain/side chain electrostatic energy, and main chain/side chain van der Waals energy, respectively. The scaling coefficients $\text{c}1$, $\text{c}2$, and $\text{c}3$ take values of 0.10, 0.25, and 0.15 in this work.

III.3. cAMP Assay.

HeLa cells were cotransfected with pcDNA3.1 β2 AR and pcDNA3.1 Gs wild type or Gs mutants (D381A, M386A, and C379A) by using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA). After 24 h of incubation, 5 μL (total 3000 cells) of cell solution and 5 μL of different concentrations of ISO or phosphate-buffered saline (PBS) was added to a 384-well plate and incubated for 40 min at 37 °C. The cAMP levels were detected according the cAMP kit protocol (LANCE cAMP ultra assay kit, cat #TRF0263; PerkinElmer, Inc., Waltham, MA, USA). The cAMP production was detected by an Envision multilabel plate reader. The TR-FRET ratio (665 nm/615 nm) obtained from the Envision multilabel plate reader was converted to cAMP concentration by using the cAMP standard curve.

Supplementary Material

ja1c03696_si_004

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.1c03696.

Additional computational details and results (PDF)

ja1c03696_si_003

Movie 1: from G_s^GDP to the exp(inter) state (MP4)

ja1c03696_si_002

Movie 2A: from exp(inter) to the G_s^empty state (MP4)

ja1c03696_si_001

Movie 2B: from exp(inter) to the G_s^empty state from another viewing angle (MP4)