Research ArticleBiophysics

Coupling Mechanism of a GPCR and a Heterotrimeric G Protein During Chemoattractant Gradient Sensing in Dictyostelium

See allHide authors and affiliations

Science Signaling  28 Sep 2010:
Vol. 3, Issue 141, pp. ra71
DOI: 10.1126/scisignal.2000980

Abstract

The coupling of heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs) with G proteins is fundamental for GPCR signaling; however, the mechanism of coupling is still debated. Moreover, how the proposed mechanisms affect the dynamics of downstream signaling remains unclear. Here, through experiments involving fluorescence recovery after photobleaching and single-molecule imaging, we directly measured the mobilities of cyclic adenosine monophosphate (cAMP) receptor 1 (cAR1), a chemoattractant receptor, and a G protein βγ subunit in live cells. We found that cAR1 diffused more slowly in the plasma membrane than did Gβγ. Upon binding of ligand to the receptor, the mobility of cAR1 was unchanged, whereas the speed of a fraction of the faster-moving Gβγ subunits decreased. Our measurements showed that cAR1 was relatively immobile and Gβγ diffused freely, suggesting that chemoattractant-bound cAR1 transiently interacted with G proteins. Using models of possible coupling mechanisms, we computed the temporal kinetics of G protein activation. Our fluorescence resonance energy transfer imaging data showed that fully activated cAR1 induced the sustained dissociation of G protein α and βγ subunits, which indicated that ligand-bound cAR1 activated G proteins continuously. Finally, simulations indicated that a high-affinity coupling of ligand-bound receptors and G proteins was essential for cAR1 to translate extracellular gradient signals into directional cellular responses. We suggest that chemoattractant receptors use a ligand-induced coupling rather than a precoupled mechanism to control the activation of G proteins during chemotaxis.

Introduction

Heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs) constitute the largest family of membrane-spanning receptors in mammals, encompassing 1000 subtypes whose genes account for nearly 5% of the human genome. Although the primary sequence is highly divergent among GPCR subtypes, all receptors share a superstructure that consists of a seven-transmembrane α-helical region and extracellular and intracellular domains, both of which can vary greatly in size (13). GPCRs detect various extracellular signals, such as light, odorants, hormones, neurotransmitters, and chemokines, and mediate a large range of physiological responses. Despite the structural diversities of GPCRs and their ligands, the mechanism that controls the first step in transducing the signal by the extracellular ligand to the activation of intracellular signaling pathways is conserved at the molecular level. The binding of extracellular ligand activates GPCRs, enabling them to catalyze the exchange of guanosine diphosphate (GDP) for guanosine triphosphate (GTP) on the α subunit of the heterotrimeric G protein, which causes the dissociation of GTP-Gα from the Gβγ subunit, leading to activation or inhibition of various downstream effectors (1, 2, 4). The biological importance of GPCRs has inspired intensive research on many fundamental aspects that concern the function and regulation of these receptors. One important aspect that has become a subject for debate is whether a GPCR is precoupled with heterotrimeric G proteins to form higher-order signaling complexes before binding of the ligand.

There are currently two opposing models that attempt to explain the interactions between GPCRs and heterotrimeric G proteins. Studies of the β-adrenoceptor system in erythrocyte membranes in the late 1970s (5) led to a “collision coupling” model, which posits that G proteins freely diffuse on the inner leaflet of the plasma membrane and become activated when they “collide” and transiently interact with ligand-bound receptors. In support of this classic model, the photoreceptor rhodopsin activates many heterotrimeric G proteins in response to a single photon (6). However, studies involving the use of fluorescence resonance energy transfer (FRET) and bioluminescence resonance energy transfer (BRET) methods have suggested a “precoupling” model (710). This model suggests that GPCRs and their heterotrimeric G proteins are present in preformed complexes in the resting state. It has been argued that these preexisting GPCR–G protein complexes enable the receptors to generate fast activation, fast signal termination, and specificity (8, 9). FRET and BRET are powerful techniques that can be used to determine small changes (within 10 nm) in the distance between a GPCR and a G protein when each is tagged with a fluorescent protein (7, 11, 12). Because these measurements monitor spatial proximity between two fluorescently tagged proteins and indicate average distances of the populations of receptors and G proteins (11, 13, 14), random collisions could also result in detectable FRET or BRET signals. Although ligand-induced changes in FRET or BRET either between a GPCR and a G protein subunit or between Gα and Gβγ subunits demonstrate functional coupling between populations of GPCRs and G proteins, measurements of the mobilities of single GPCRs and G protein subunits would enable us to verify their association at a single-molecule level.

Interactions between GPCRs and the G protein machinery have been particularly well characterized in the genetically amenable social amoeba Dictyostelium discoideum (15). Individual amoebae use the GPCR cyclic adenosine monophosphate (cAMP) receptor 1 (cAR1) to detect the chemoattractant cAMP, which is secreted by neighboring cells to trigger robust chemotaxis, a process that is essential for the developmental program of the organism (1620). cAR1 couples with the heterotrimeric G protein Gα2Gβγ, and different coupling mechanisms of cAR1 and G proteins have been proposed (21, 22). Binding of cAMP to cAR1 induces the dissociation of Gα2 from the Gβγ subunits to activate multiple signal transduction pathways that mediate cell migration (15, 19, 20, 23, 24). Functional, fluorescently tagged cAR1 and Gβ proteins have been used to determine the dynamic properties of GPCR–G protein signaling in the membranes of live cells in real time (2527). FRET imaging methods have been developed to observe cAR1 activation–induced dissociation of Gα2 and Gβγ in live cells (12, 26, 28). In addition, we have developed computational models that enable us to examine the consequences for signaling dynamics that arise from different coupling mechanisms in cAR1- and G protein–regulated chemosensing (29, 30).

Here, we addressed the coupling mechanism between cAR1 and the Gα2Gβγ heterotrimer by measuring the mobility of yellow fluorescent protein (YFP)–tagged cAR1 and Gβ in live cells in response to stimulation with chemoattractant. Our results suggest that cAR1 and Gβγ diffuse throughout the plasma membrane independently at the resting stage and that it is only upon stimulation with cAMP that activated cAR1 transiently interacts with Gα2Gβγ to trigger the release of the Gβγ subunit for downstream signaling. We developed computational models to describe the possible coupling mechanisms between GPCR and a G protein. We computed the expected temporal kinetics in the abundance of GαGβγ heterotrimers and discovered that each coupling mechanism showed a distinct kinetic pattern. We determined the kinetics of cAR1-induced dissociation of Gα2 from Gβγ in live cells by FRET imaging and showed a functional coupling between activated cAR1 and G proteins. Furthermore, we used a detailed cAMP-sensing model to examine dynamic cell responses by incorporating different coupling mechanisms between cAR1 and G proteins. Our computational simulations indicated that cAR1-type GPCRs use a ligand-induced, rather than a precoupled, coupling mechanism to control the spatiotemporal activation of G proteins, which is required for proper chemosensing in D. discoideum.

Results

Different mobilities of cAR1 and Gβγ in the plasma membrane

To measure the apparent mobilities of cAR1 and G proteins, we first performed fluorescence recovery after photobleaching (FRAP) analysis (31) of YFP-tagged cAR1 (cAR1-YFP) and Gβγ (Gβ-YFP). cAR1-YFP and Gβ-YFP were functional and present in both car1 and gβ cell lines in which endogenous genes encoding cAR1 or the Gβ subunit were deleted (25, 27). Thus, signaling in the car1 and gβ cell lines was transduced by cAR1-YFP and Gα2Gβγ or by cAR1 and Gα2YFP-Gβγ, respectively (25, 27). Photobleaching experiments were performed by bleaching two small regions (~0.8 μm2) of the cell membrane. Cells were treated with latrunculin B to obtain reliable, quantified data for fluorescence recovery in bleached regions without any interference by cell movement (Fig. 1A).

Fig. 1

Different mobilities of cAR1 proteins and G protein subunits as measured by FRAP. (A) The montages show the fluorescent recovery of Gβ-YFP in two photobleaching areas (R1 and R2) of a single gβ cell containing Gβ-YFP. The area of the whole cell for photobleaching measurement in the calculation of recovery efficiency of fluorescent intensity is shown (R3). Scale bar, 5 μm. (B to E) Recovery curves of cAR1-YFP in car1 cells (n = 32 for 0 μM cAMP and n = 18 for 10 μM cAMP), Gβ-YFP in gβ cells (n = 24 for 0 μM cAMP and n = 24 for 10 μM cAMP), cAR1-YFP in gβ cells (n = 24 for 0 μM cAMP and n = 38 for 10 μM cAMP), and Gβ-YFP in car1 cells (n = 16 for 0 μM cAMP and n = 28 for 10 μM cAMP) that were untreated (0 μM cAMP, black) or were treated with cAMP (10 μM, red) were normalized and fitted to single-exponential function curves. Curves show the mean intensity ± SD at each time point.

Fluorescence recovery curves were normalized and fitted to a single-exponential function (Fig. 1, B to E) to calculate the recovery half-time (τ1/2) and the percentages of the tagged proteins that were either mobile (α) or immobile (1 − α), which we refer to as the mobile and immobile fractions of these proteins (31). Normalized recovery curves of cAR1-YFP in transfected car1 cells and of Gβ-YFP in transfected gβ cells photobleached in the presence or absence of cAMP (10 μM, a saturating concentration that fully activates cAR1) were prepared (Fig. 1, B and C), as were recovery curves of cAR1-YFP in gβ cells and Gβ-YFP in car1 cells (Fig. 1, D and E). The measured parameters of cAR1-YFP and Gβ-YFP in each cell line are summarized in Table 1. In the absence of cAMP, the recovery τ1/2 of cAR1 was 16.6 s, with mobile and immobile fractions of ~67 and 33%, respectively. In contrast, the τ1/2 of the Gβ subunit was shorter (5.5 s), with a mobile fraction of 90% and an immobile fraction of 10%. The difference in the τ1/2 values of cAR1 and Gβ suggested that the mobile fractions of ligand-free cAR1and Gβ proteins moved independently. When cells were stimulated with cAMP (10 μM), the τ1/2 of cAR1 was 13.5 s, with mobile and immobile fractions of 68 and 32%, respectively, whereas the τ1/2 of Gβ was 5.6 s, with mobile and immobile fraction of 74 and 26%, respectively, indicating that the binding of cAMP to cAR1 did not cause marked changes in the mobility of the receptor but resulted in a decrease in the number of Gβ molecules that were mobile. Without functional G proteins and in the absence of cAMP, 78% of cAR1 molecules were mobile, whereas 22% were immobile, and these percentages changed little when the cells were stimulated (Fig. 1D). Similar results were obtained for Gβ in cells lacking cAR1, in which the mobile and immobile fractions of Gβ in the absence of cAMP were 78 and 22%, respectively. Upon stimulation with cAMP (10 μM), neither the τ1/2 nor the percentages of mobile and immobile subunits changed substantially (Fig. 1E). Together, our data suggest that molecules in the mobile fractions of ligand-free cAR1 or cAMP-bound cAR1 moved independently of Gβγ subunits and that the binding of cAMP to cAR1 proteins led to a decrease in the fraction of mobile Gβγ subunits. It is unlikely that the immobile fractions of cAR1 and Gβγ were caused by the formation of precoupled complexes of cAR1 and G protein, because each of the fractions was detected without the other.

Table 1

Summary of the mobilities of cAR1-YFP and Gβ-YFP, as measured by FRAP (Fig. 1), and of the diffusion constants of cAR1-YFP and Gβ-YFP as determined by single-molecule imaging by PICS (Fig. 4). The mobilities of cAR1-YFP and Gβ-YFP molecules were measured in car1 and gβ cells. Two parameters were obtained by FRAP: the percentage of molecules in the mobile (Mo) fraction [immobile fraction (Im) = 1 − Mo] and the half-life of mobility (τ1/2). Two parameters were obtained by single-molecule analyses: the percentage of the mobile fast fraction [slow fraction (Im) = 1 − Mo], and the diffusion constant (D, μm2/s). The number of cells (n), SD (%), and SD (τ1/2) for FRAP and the n, SD (%), and SD (D) for single-molecule analysis are shown. Our measured Ds for cAR1 and Gβγ in the mobile fraction are consistent with diffusion coefficients that were previously reported (43, 44).

View this table:

Single-molecule imaging of cAR1-YFP and Gβ-YFP by total internal reflection fluorescence microscopy

To test whether our FRAP-based observations could be confirmed with a different experimental technique, we used total internal reflection fluorescence (TIRF) microscopy to study the dynamic properties of cAR1 and Gβγ subunits at the single-molecule level in the plasma membranes of live cells (14, 32). To validate our imaging system (33), we attached monomeric, purified YFP proteins to glass coverslips at low concentration to visualize single YFP molecules. Images were recorded on video at a rate of 30 frames per second by TIRF illumination with a cooled, intensified charge-coupled device (CCD) camera (fig. S1A and movie S1). As expected, single YFP molecules showed individual and diffraction-limited fluorescence spots, which disappeared in a single step with occasional on-and-off blinking (fig. S1, C and D). Under the same imaging conditions, we then visualized single cAR1-YFP and Gβ-YFP molecules in live cells. After cells transfected with plasmids expressing cAR1-YFP or Gβ-YFP were attached to the coverslips, the plasma membrane most proximal to the coverslip was imaged by TIRF illumination (Fig. 2A and movies S2 and S3). Fluorescence spots observed in live cells transfected with plasmid encoding either cAR1-YFP or Gβ-YFP displayed the characteristics of single molecules having the similar fluorescence intensity and photobleaching characteristics of purified YFP proteins (Fig. 2C).

Fig. 2

Single-molecule imaging of cAR1-YFP and Gβ-YFP in live cells by TIRF microscopy. (A) Images of single cAR1-YFP and Gβ-YFP molecules in living cells. (B) Tracks of single molecules of cAR1-YFP or cAR1 and Gβ-YFP in the membranes of live cells. (C) Images of a single cAR1-YFP molecule, which exhibits single-step blinking, indicate the characteristics of single YFP molecules. The measured total intensity of the spot in each frame in one time lapse experiment is shown as an example.

Mobilities of cAR1-YFP and Gβ-YFP analyzed by single-particle tracking

To quantify the mobilities of single molecules of cAR1-YFP and Gβ-YFP in cells, we identified fluorescent particles in each frame by two-dimensional Gaussian fitting with MATLAB (MathWorks) scripts developed in-house (see Materials and Methods). The fitting analyses provided positions (x, y) for each single-molecule peak. Individual peak positions were automatically tracked over time in an image series with a MATLAB-based single-particle–tracking algorithm to build trajectories (Fig. 2 and movie S4; Materials and Methods). In short, the algorithm required that the maximum range of displacement of a particle between consecutive frames be assumed a priori (the maximum displacement was set to 0.45 μm). If a particle was displaced between two frames of a sequence by a distance less than or equal to the maximum range, it was considered to be the same particle, and the positions between frames were connected to form a step of the trajectory (Fig. 2B).

The movements of cAR1-YFP and Gβ-YFP were calculated as the square of the distance between the sequential positions, which is referred to as the square displacement (SD). Subsequently, all SDs found between consecutive frames (30 ms between frames) for all particles in an image sequence were expressed as a cumulative distribution function (CDF) of the SDs (21). The distribution of SDs for cAR1-YFP was similar in the presence or absence of cAMP (Fig. 3A), which suggested that the binding of cAMP to cAR1 did not substantially change its mobility in live cells. In contrast, the distribution of the SDs of Gβ-YFP shifted to the left in the presence of cAMP (10 μM) compared to those in the absence of cAMP (Fig. 3B), which indicated that the activation of cAR1 triggered the reduced mobility of Gβ subunits in live cells. The mobility of cAR1-YFP in gβ cells was similar in the presence and absence of cAMP (Fig. 3C). In addition, the mobility of Gβ-YFP in car1 cells did not change upon stimulation with cAMP (Fig. 3D), indicating that the cAMP-induced reduction in the mobility of Gβ required functional cAR1. The mobilities of cAR1-YFP and Gβ-YFP in live cells were different (Fig. 3E). In the absence of cAMP, when cAR1 and G proteins were at rest, cAR1-YFP was less mobile than was Gβ-YFP. In the presence of cAMP, when cAR1 and G proteins were fully activated, we detected little change in the mobility of cAR1-YFP, but observed a substantial reduction in the mobility of Gβ-YFP in live cells (Fig. 3E).

Fig. 3

Mobilities of cAR1-YFP and Gβ-YFP in the membranes of live cells. (A to D) Cumulative probability of the square displacements (SDs, in μm2) of cAR1-YFP in car1 cells (n = 395,366 molecules for 0 μM cAMP and n = 250,753 molecules for 10 μM cAMP), Gβ-YFP in gβ cells (n = 514,405 molecules for 0 μM cAMP and n = 160,832 molecules for 10 μM cAMP), cAR1-YFP in gβ cells (n = 317,192 molecules for 0 μM cAMP and n = 202,016 molecules for 10 μM cAMP), and Gβ-YFP in car1 cells (n = 39,924 molecules for 0 μM cAMP and n = 37,681 molecules for 10 μM cAMP) that were untreated (black) or treated with cAMP (10 μM, red). (E) Cumulative probability of the SDs of cAR1-YFP in car1 cells (green) and Gβ-YFP in gβ cells (blue) that were untreated (solid lines) or were treated with cAMP (10 μM, dashed lines).

Calculation of diffusion constants and domain size through particle image correlation spectroscopy

Single-particle tracking (SPT) methods are best suited for conditions in which the particle of interest can be tracked for numerous consecutive image frames. The disadvantage of using common biological fluorescent tags such as green fluorescent protein (GFP) or YFP is their inherent tendency to photoblink and photobleach (reversible and irreversible transitions to nonemissive states, respectively), which substantially decreases the number of steps over which a tagged molecule can be tracked. In addition, the generation of fluorescently tagged proteins in transfected cells under most circumstances leads to crowded systems where tracking algorithms are unable to robustly determine the origin of a particle from the previous frame. Particle image correlation spectroscopy (PICS) is a robust statistical method to circumvent the potential problems with fluorescence proteins (fig. S3) (34). Briefly, PICS does not require an a priori assumption of the expected mobility of a particle (a search radius is necessary for SPT algorithms) and, moreover, PICS has the ability to calculate the diffusion of a particle without the need to establish a trajectory. This particular feature removes complications such as tracking errors because of overlapping trajectories in a crowded system, the loss of particles because of photobleaching, and the missing of steps due to loss of a particle because of photoblinking.

With PICS, we analyzed the movement of ~106 particles in cells containing cAR1-YFP and Gβ-YFP in the presence or absence of cAMP (Fig. 4). For all of the data obtained from each cell type, the respective CDFs (fig. S3) of the SDs were best fit to a biexponential model, which indicated the presence of two populations, one fast and one slow (which may reflect the mobile and immobile fractions observed by FRAP). From these fits, we calculated the mean square displacement (MSD) for the percentages of particles in the fast (MSDfast, α) and slow (MSDslow, 1 − α) populations, respectively, which were performed for 10 consecutive frames (Fig. 4 and Table 1; see details in Materials and Methods).

Fig. 4

Calculation of diffusion constants and diffusion area through analysis by PICS. (A to D) Single-molecule positions of cAR1-YFP in car1 cells (A), Gβ-YFP in gβ cells (B), cAR1-YFP in gβ cells (C), and Gβ-YFP in car1 cells (D) were recorded in the absence (black) in 15, 16, 16, and 16 independent experiments, respectively, or in the presence of cAMP (10 μM, red) in 16, 18, 15, and 14 independent experiments, respectively, and analyzed by PICS. The PICS analysis yielded three parameters for each of the 10 time lags that were analyzed, which are plotted against the time lags: the fractions of the fast population of molecules (α, left panels), the MSDs of the fast population (MSDfast, center panels), and the MSDs of the slow population (MSDslow, right panels). Models for free diffusion (straight curves; see Eq. 8) or confined diffusion (dashed curves; see Eq. 9) were fitted to each MSD plot to obtain diffusion constants and diffusion areas. The results of theses fits are summarized in Table 1. Only the mobility of Gβ-YFP in gβ cells (B) substantially changed upon addition of cAMP. The diffusion coefficient D decreased for both the fast [(B), center panel] and the slow fraction [(B), right panel], as indicated by the decrease in the slopes; however, the fraction of the fast population [(B), left panel] decreased, demonstrating that a small fraction of Gβ-YFP (7%) showed a change in mobility as it transitioned from the fast to the slow fraction. The mobilities of the fast populations of cAR1-YFP [center panels in (A) and (C)] are best described by a model for confined diffusion (straight curves; see Eq. 9), whereas the mobility of Gβ-YFP [center panels in (B) and (C)] did not show a marked curvature and were well described by a model for free diffusion (dashed curves; see Eq. 8). Because of the limited positional accuracy in our recordings, the minimal detectable mobility (dotted horizontal lines) was 0.0027 μm2. The slow fractions of cAR1-YFP [right panels in (A) and (C)] did not exceed this minimum and, hence, appeared immobile in our analysis. Only the fast fractions [center panels, (A) to (D)] and the slow fraction of Gβ-YFP in gβ cells [right panel, (B)] substantially exceeded this minimum.

We found that when car1 cells expressing cAR1-YFP were stimulated with chemoattractant, the fraction of cAR1-YFP particles that belonged to the fast population (~63%) was not altered (Fig. 4A and Table 1), and we did not observe a substantial change in the speed of the fast fraction of cAR1-YFP particles upon stimulation. We considered the slow fraction to be immobile because each MSD value barely exceeded the minimum SD of 0.0027 μm2 (horizontal dotted line in all MSD plots) that was determined for our imaging conditions (Fig. 4A and Table 1). In contrast to what we observed with cAR1-YFP, we saw a substantial decrease in the number of Gβ-YFP particles in the fast fraction and a decrease in the speed of both the fast and the slow fraction upon stimulation (Fig. 4B and Table 1). As expected, we did not observe any substantial differences in diffusion coefficients upon stimulation in control experiments with gβ cells expressing cAR1-YFP or in car1 expressing Gβ-YFP, except that Gβ-YFP diffused more slowly in the absence than in the presence of cAR1 (Table 1). We measured the mobilities of cAR1 and Gβγ subunits in both the presence (FRAP measurement) and the absence (single-molecule analysis) of latrunculin B. Both measurements yielded consistent results, suggesting that the actin cytoskeleton did not substantially affect the mobilities of cAR1 or Gβγ.

To estimate whether cAR1 and Gβ diffused in a confined area in the cell membrane, we fitted all of the MSD plots with either a free or a confined model (Fig. 4). The free diffusion model yielded a constant diffusion coefficient (D in Table 1), whereas the confined model yielded an initial diffusion coefficient and an estimation of the domain size to which the diffusion was confined, D0 and L, respectively (35). We found that the mobility of the fast population of cAR1-YFP in car1 cells was confined to an area of ~250 nm in diameter, which was unchanged in the absence of Gβ and did not change upon stimulation (Fig. 4, A and B, and Table 1). We could not calculate a confinement zone for the immobile (slow) population of cAR1-YFP. In contrast, the fast population of Gβ-YFP in gβ cells was only slightly confined to an area of ~1 μm in diameter (Fig. 4B); therefore, its mobility was more accurately described by the free diffusion model (Fig. 4B).

Modeling and validation of the functional coupling between ligand-bound cAR1 and G proteins

Coupling between GPCRs and heterotrimeric G proteins is described by three possible mechanisms. The first, the R/G model, suggests that only ligand-free GPCRs couple to G proteins (Fig. 5A) and that ligand-bound GPCRs do not recruit G proteins, but activate only those that were recruited before ligand binding. In the second proposed mechanism, the L/R/G model, only ligand-bound GPCRs interact with G proteins (Fig. 5A), and GPCRs do not couple to G proteins in the absence of ligand. Activation of G proteins occurs in two steps after receptor-ligand binding: G proteins are first recruited and then activated. In the third mechanism, which combines the R/G and L/R/G models, ligand-free and ligand-bound GPCRs can interact with G proteins, but only ligand-bound receptors can activate G proteins. A previous study showed that the binding of cAMP did not induce internalization of cAR1 (25), and our results also support this (fig. S4). Thus, ligand-induced internalization of a GPCR was not considered in our model.

Fig. 5

Modeling and validation of a functional coupling mechanism between a ligand-bound cAR1 and its G proteins. (A) Schemes show two coupling models by which heterotrimeric G proteins are thought to be activated. (B to D) Results of the simulations with the three possible coupling mechanisms between a GPCR and heterotrimeric G proteins. The total number of heterotrimeric G proteins (receptor-bound and free) (y) is a function of the time (x) after the cells were stimulated with a saturating concentration of ligand. (B) R/G coupling. Heterotrimeric G proteins that are coupled to receptors before the cAMP stimulus are rapidly activated and dissociate into their Gα and Gβγ subunits, leading to a sudden decrease in the concentration of the heterotrimer. After their dissociation, the abundance of heterotrimer recovers at a rate that is determined by the rate of hydrolysis of GTP by the G protein α subunit. Because the receptor does not recruit new heterotrimeric G proteins while ligand-bound, the recovery in the amount of the heterotrimer is not hindered by reassociation with the receptor followed by reactivation. (C) L/R/G coupling. After stimulation of cells with cAMP and activation of cAR1, heterotrimeric G proteins are recruited to the receptor at a rate that is limited by diffusion. Once the G protein is associated with the receptor, the α subunit becomes rapidly activated, leading to release of Gα and Gβγ from the receptor, thereby reducing the amount of available heterotrimeric G proteins. Hydrolysis of GTP makes Gα available for reassociation with Gβγ and another round of interaction with cAR1 and subsequent activation. Within a few seconds, these two processes reach a dynamic equilibrium, which is reflected in the stable extent of dissociation of the heterotrimer. (D) R/G + L/R/G coupling. In this scenario, the heterotrimeric G protein is coupled to the receptor before the application of cAMP. Upon stimulation, these G proteins are rapidly activated and dissociate, leading to a rapid reduction in the amount of heterotrimer. Additionally, the activated receptors recruit new G proteins and activate them in a cycle identical to the one described in (C). After reassociation of the G proteins that had been precoupled to the receptors, we observe the same G protein dynamics as those that are observed in the pure L/R/G model. The initial dip in G protein dissociation is a result of a high receptor-ligand on-rate. For slower reaction kinetics, the difference between (C) and (D) would be less pronounced. The limited time resolution of the experimental data does not enable us to distinguish whether precoupling (R/G) acts in addition to ligand-induced recruitment of G proteins (L/R/G). (E) Activation of heterotrimeric G proteins (dissociation) measured experimentally as the loss of FRET between Gβγ-CFP and Gα-YFP subunits when cells were stimulated with cAMP (10 μM). Means and SDs are shown (n = 5 experiments). Gα2-CFP and YFP-Gβ were coexpressed in gα2 cells and signaling was largely transduced by cAR1, Gα2-CFP, and YFP-Gβγ [for details, see (26)].

To investigate the dynamics of G protein activation that would be expected from the various coupling mechanisms, we developed simple computational models that described dynamic interactions between GPCRs and G proteins and simulated temporal changes in the abundances of heterotrimeric G proteins (Fig. 5, B to D). When the GPCR–G protein systems were stimulated with a saturating dose of a ligand, the R/G and L/R/G coupling mechanisms showed different kinetics for the dissociation of G proteins. In the R/G coupling mechanism (Fig. 5B), stimulation with ligand triggered the immediate dissociation of G proteins into GTP-Gα and Gβγ subunits (resulting in a fast decrease in the abundance of heterotrimeric G proteins). Because the saturating dose of ligand would have converted all of the receptors into a ligand-bound form (L/R) that does not recruit reassociated G proteins, one would thus observe a gradual recovery in the amount of heterotrimeric G proteins while GTP on the G protein α subunit was hydrolyzed to generate GDP and Gα reassociated with Gβγ. In contrast, in the L/R/G coupling mechanism (Fig. 5C), a gradual, but persistent, decrease in the amount of heterotrimeric G proteins was induced, because the ligand-bound GPCR would be expected to continually activate reassociated G proteins. In the “R/G + L/R/G” coupling mechanism (Fig. 5D), ligand stimulation also caused the persistent dissociation of G proteins, which was similar to that predicted by the L/R/G mechanism, but showed a rapid initial dissociation of G proteins that was a result of the activation of precoupled G proteins.

Different strengths of guanosine triphosphatase (GTPase) activity, including the intrinsic GTPase activity of the α subunit and those of regulator of G protein signaling (RGS) proteins, would affect the kinetics of G protein activation and deactivation (36). We used simulations to investigate the effect of RGS activity on the kinetics of G protein activation generated by the R/G and L/R/G coupling mechanisms (fig. S5) by varying Gα deactivation (GTPase activity) rates over a wide range. In the case of the R/G coupling mechanism, the simulations showed that an increased RGS activity would lead to a faster return to the prestimulus amounts of heterotrimeric G proteins and a smaller amount of active G proteins. In the case of the L/R/G coupling mechanism, a lower extent of persistent activation of G proteins was observed for high GTPase activity. Altering the GTPase activity did not change the characteristics of the G protein activation curves: G proteins were activated transiently in the R/G model and persistently in the L/R/G model. Thus, different coupling mechanisms can be distinguished by measuring the kinetics of G protein activation in live cells.

To determine the coupling mechanism between cAR1 and its heterotrimeric G proteins, we measured the kinetics of cAR1-triggered dissociation of G proteins in response to a saturating dose of cAMP in live cells with our developed FRET imaging method (12). The cAMP-induced dissociation of G proteins was monitored by a change in the FRET signal (YFP intensity/CFP intensity as a function of time) between Gα2-CFP (the FRET donor) and Gβ-YFP (the FRET acceptor) (Fig. 5E). Stimulation of cAR1 induced a gradual and persistent dissociation of Gα2 from Gβγ, which was different from the dynamics predicted by the R/G coupling model, indicating that cAMP-bound cAR1 proteins interacted with and activated G proteins.

Requirement for coupling between cAMP-bound cAR1 and G proteins for sensing chemotactic gradients

To examine whether and how precoupling of cAR1 to G proteins might affect the sensing of chemotactic gradients (24, 37), we used a spatiotemporally resolved signaling network (29) that we previously constructed to simulate phosphatidylinositol 3,4,5-trisphosphate (PIP3) responses (12, 30, 38). In this model, we varied the affinity between ligand-free cAR1 and G proteins to change the fraction of cAR1 molecules that were precoupled to G proteins. The interactions between cAR1 and heterotrimeric G proteins were quantitatively described by a set of parameters that included the constants Kon1 and Koff1 for ligand-free cAR1, and Kon2 and Koff2 for cAMP-bound cAR1 (Fig. 6A). Our signaling network simulated the dynamics of PIP3 observed experimentally when cells were stimulated with uniform cAMP fields or gradients by incorporating a high-affinity coupling between cAMP-bound cAR1 and G proteins (Fig. 6, B to E, and figs. S6 and S7).

Fig. 6

High-affinity coupling between cAMP-bound cAR1 and G proteins is essential for sensing of cAMP gradients. (A) The binding constants for ligand-free cAR1 (Kon1 and Koff1) and ligand-bound cAR1 (Kon2 and Koff2) control the association and dissociation of G proteins. Activation of G proteins, in turn, regulates a signaling network that controls the production of PIP3 in the plasma membrane of the cells as they respond to cAMP. (B) Experimental data showing the PIP3 response of Dictyostelium cells exposed to a rapidly initiated, then constant, gradient of cAMP, released from a micropipette (12). (C to E) Simulation of PIP3 responses with a model described previously (29). Three conditions were simulated. In (C), only cAMP-bound cAR1s can bind to G proteins; in (D), cAMP-free cAR1s can bind to G proteins; whereas in (E), both forms of cAR1 can bind to the G proteins. In all three cases, only the cAMP-bound cAR1 proteins can activate Gα. “Front” and “Back” refer to the regions of the cells that experience the highest or lowest extracellular concentration of cAMP, respectively.

Our simulations indicated that provided that cAMP-bound cAR1 effectively activated the G protein (that is, Kon2 was high), altering Kon1 (and thereby the fraction of precoupled cAR1–G protein complexes) did not substantially change PIP3 dynamics. This includes the adaptation of the PIP3 response upon stimulation with a uniform concentration of cAMP (fig. S6), as well as the previously described biphasic PIP3 response to a gradient of cAMP (Fig. 6, B, C, and E, and fig. S7) (12, 37). However, when coupling between cAMP-bound cAR1 and G proteins became less effective (that is, Kon2 was small), the cAR1-controlled signaling network could no longer generate directional PIP3 responses to a cAMP gradient (Fig. 6D), indicating that a high-affinity coupling between cAMP-bound cAR1 and G proteins, but not the coupling ability of cAMP-free cAR1 and G proteins, was critical for the role that this GPCR plays in the ability of the cell to sense and respond to chemoattractant gradients.

Discussion

Here, we investigated the earliest step in GPCR-mediated signaling, the coupling between a GPCR and its heterotrimeric G protein. Upon ligand binding, activated GPCRs promote the exchange of GDP for GTP on the α subunit, leading to the dissociation of GTP-bound α subunits from Gβγ subunits, which activate downstream effectors. Hydrolysis of GTP to GDP on α subunits enables reassociation of GDP-Gα subunits and Gβγ subunits to form inactive GαGβγ heterotrimers, which can engage GPCRs for a new round of signaling. We developed methods to study the coupling mechanism of a chemoattractant GPCR, cAR1, by directly measuring the mobility of cAR1 and Gβγ in live cells and by modeling and verifying the kinetics of cAR1-induced dissociation of G proteins. Furthermore, through computer simulations, we explored the effects of various coupling mechanisms on cAR1-controlled gradient sensing.

FRAP experiments and single-molecule analysis of the mobility of cAR1 and Gβγ yielded consistent results (Table 1). Our data indicated that cAR1 proteins diffused more slowly than did Gβγ subunits in the absence of cAMP, indicating that a large percentage of cAR1 and Gβγ did not form complexes in unstimulated cells. When cAR1 was fully activated, its mobility remained unchanged, whereas the mobility of a fraction of the fast-moving Gβγ subunits decreased, which can be explained by the occurrence of interactions between cAMP-bound cAR1 proteins and G proteins and between free Gβγ subunits and effectors. In cells that did not have cAR1, the dynamics of the fast-moving Gβγ subunits remained unchanged upon exposure to cAMP. In addition, the mobility of cAR1 did not change in cells that did not have functional G proteins. The results suggested the following coupling mechanism between cAR1 and G proteins: In the absence of cAMP, heterotrimeric G proteins moved relatively quickly within the cell membrane, whereas cAR1 was relatively immobile. We found that diffusion collisions between G proteins and cAR1 rarely led to the formation of a stable complex.

A previous study indicated that cAR1 and its downstream effector, adenylyl cyclase, exclusively localize to detergent-resistant microdomains in the plasma membrane, which are about ~100 to 200 nm in diameter, whereas the Gα2 and Gβγ subunits largely reside in a fluid environment that is outside of these microdomains (39). Our measurements of the mobilities and confinement of cAR1-YFP and Gβ-YFP reflected their localization in the heterogeneous microenvironments of the plasma membrane. We speculate that upon stimulation of cells with cAMP, the rapidly diffusing heterotrimeric G proteins interact with cAMP-bound cAR1, which results in activation of the G proteins, after which the freed Gβγ subunits interact with downstream effectors in these microdomains. The interaction with effectors localized in different microdomains contributes to the change in the mobility of Gβγ subunits. Consistent with this notion, we found that when a cell was exposed to a gradient of cAMP, there was a larger fraction of slow-moving YFP-Gβγ subunits at the front of the cell where the extent of stimulation of the cell by cAMP was greatest (fig. S8). These data are consistent with a recent study that used single-molecule epifluorescence microscopy, which showed that there were more slow-moving Gβγ subunits at the front of cells undergoing chemotaxis (40).

How do the different potential coupling mechanisms of GPCRs and G proteins affect the temporal kinetics of G protein dissociation? To address this important question, we generated models to calculate temporal changes in the abundance of inactive heterotrimeric G proteins (Fig. 5). In all of the models in which a GPCR was fully activated, the amounts of heterotrimeric G proteins decreased because of the dissociation of their subunits; after GTP hydrolysis, the heterotrimer reformed. In a coupling mechanism in which ligand-bound receptor cannot interact with reassociated heterotrimers (the R/G model), the amount of heterotrimeric G proteins gradually returned to that found before any stimulus was applied (Fig. 5B). In contrast, if ligand-bound GPCRs continued to couple to G proteins that recycled back to the heterotrimeric complex (as in the L/R/G model), inactive G proteins would remain low in abundance (Fig. 5, C and D). Our simulations revealed that different coupling mechanisms could be distinguished by measuring the kinetics of G protein activation in live cells. Through FRET imaging, we found that stimulation with a saturating concentration of ligand induced a persistently low abundance of inactive heterotrimeric G proteins, indicating that cAMP-bound cAR1 continually promoted the dissociation of G proteins into Gα2 and Gβγ subunits in live cells (Fig. 5E). This is consistent with a previous study that supported the idea that cAMP-bound cAR1 could indeed trigger the dissociation of G protein subunits (26).

What is the essential coupling mechanism for a chemoattractant GPCR to sense a chemoattractant gradient? Through the use of a spatiotemporally resolved signaling network that incorporated the coupling between cAMP-bound cAR1 and G proteins, we simulated the dynamics of cAR1-controlled signaling events, including PIP3 responses (Fig. 6 and fig. S4). Our simulations demonstrated that provided the cAMP-bound cAR1 could effectively interact with and activate G proteins, the cAR1–G protein system could generate appropriate cellular responses to stimulation by either a uniform concentration or a gradient of cAMP, regardless of the portion of receptors and G proteins that were precoupled. Our results indicate that effective coupling between the ligand-bound GPCR and its G proteins generated persistent activation of G proteins. We suggest that a sustained difference in G protein activation generates an intracellular gradient that serves as a cue to direct a signaling network to sense and respond to spatiotemporal changes of extracellular chemoattractants, thereby controlling the sensing of the chemoattractant gradient.

GPCRs represent the largest family of cell surface receptors that regulate diverse physiological processes. Whereas our current study on cAR1 might shed light on the coupling mechanisms of chemokine receptors (4), it is possible that GPCRs from different classes will couple to G proteins through different mechanisms. Our model focuses on how a GPCR controls spatiotemporal responses during the sensing of a chemoattractant and does not contain all of the components that are required for cell migration, such as the actin cytoskeleton and its regulators. In addition, our single-molecule analysis was not able to determine the stoichiometries of the protein complexes involved, and this analysis still needs to be improved for imaging the possible oligomerization of cAR1 and the formation of cAR1–G protein complexes in vitro or in live cells. Despite these limitations, our study provides a new avenue for investigating the coupling mechanisms between a GPCR and its G proteins and evaluating the dynamics of GPCR signaling.

Materials and Methods

Cell lines and expression of fluorescently tagged proteins

D. discoideum cells expressing cAR1-YFP, Gβ-YFP, and Gα2-CFP were developed to the chemotactic stage as previously described (12). The proper expression of cAR1-YFP in car1 cells (25), Gβ-YFP in gβ cells (27), and Gα2-CFP and Gβ-YFP in gα2 cells (26) from plasmids enabled these fluorescently tagged proteins to mediate cAMP-dependent signaling in the respective mutant cells. Live cells were imaged for FRAP, FRET, or single molecules with a Zeiss Laser Scanning Microscope (28) or a TIRF microscope (14).

Fluorescence recovery after photobleaching

FRAP measurements were recorded on a Zeiss LSM 510 Meta with a 100× Zeiss PLAN Apo objective. Two circular regions on the plasma membrane on opposite ends of a D. discoideum cell were bleached within 4 s with the full power of the 488- and 514-nm lines of an argon-krypton (Ar-Kr) mixed gas laser. Each FRAP region had a radius of 0.5 μm. Recovery of the membrane-localized fluorescence was recorded every 3 s for a total of 160 s. D. discoideum cells were maintained and then treated with latrunculin B (1 μM) for 5 min to stop cell migration, as previously described (12), before FRAP was recorded. FRAP intensities were normalized according to the “double-normalization” procedure (41) as shown by Eq. 1. Normalized FRAP curves were then fitted to a single-exponential function (Eq. 2) to obtain the fitting parameters for the half-life τ1/2 (Eq. 3) and mobile fraction α (Eq. 4). Equation 5 summarizes these last three steps.

IFRAPnorm(t)=IFRAP(t)IBACK(t)IpreFRAPIpreBACK×IpreWHOLEIpreBACKIWHOLE(t)IBACK(t)(1)IFRAPnorm(t)=y0Aeτt(2)τ1/2=ln 2t(3)α=A1(y0A)(4)IFRAPnorm(t)=y0+α(y01)1αeln 2τ1/2t(5)

Two-sample t tests were performed with MATLAB on all pairs of FRAP curves, which were recorded in the presence or absence of cAMP (10 μM) and otherwise identical conditions (that is, in terms of cell line and expressed protein).

TIRF microscopy with single-molecule sensitivity

Individual YFP-tagged proteins (that is, cAR1 and Gβ) were imaged on a custom-built Olympus TIR setup. YFP was excited by the 512-nm line selected from an Ar gas laser by an acousto-optical tunable filter (AOTF, NEOS Technologies). The beam was further cleaned with an excitation filter (Chroma Technology Corp.) and a single-mode fiber and was coupled into the microscope through an Olympus TIR port. The evanescent field was obtained by objective type TIR with a 100× Plan Apochromat numerical aperture 1.45 lens (Zeiss) (33). Fluorescent emission was collected through a 540/50 emission filter onto an electron multiplying CCD (EMCCD; Cascade 512B CCD, Roper Scientific). Images were acquired every 35 ms (28.6 Hz).

Analysis of single-molecule dynamics

Single molecules were identified with modified MATLAB routines, which are based on the publicly available codes based on the MATLAB adaptation (from D. Blair and E. Dufresne) of IDL code written by D. Grier, J. Crocker, and E. Weeks (http://physics.georgetown.edu/matlab/). Briefly, image sequences were filtered to suppress noise and background variations while at the same time enhancing signals with the spatial dimensions of single-molecule signals. Structures with peak intensities above a set threshold were then fitted to a two-dimensional Gaussian function to obtain the position and integrated brightness for each molecule. Positions were traced with the algorithm by Grier, Crocker, and Weeks. Subsequently, MSDs were calculated for each track that exceeded a track length of five steps. A model for free diffusion (MSD = 4Dtlag, where D represents the diffusion coefficient and tlag represents the time lag) was fit to the resulting MSD plot of each track to obtain its diffusion coefficient (D). In addition to tracing single molecules, peak positions were analyzed with PICS (34) to obtain the CDF of the SDs for each given image frame. In contrast to tracking algorithms, no previous assumptions of the mean or maximum expected diffusion are required for PICS. A thorough description is given in the main text. CDFs of the SDs were fit to a model, which was modified as described previously (42). The first, monoexponential model (Eq. 6) assumes that only one population of molecules causes the distribution of SDs, and consequently yields only one MSD value:

P(SD,tlag)=1e(SD/MSD)1e(SDmax/MSD),with MSD=4Dtlag(6)

The second, biexponential model (Eq. 7) accounts for the presence of two populations, one with a fast mobility and the other with a slow mobility, and therefore yields MSDs for the fast (MSDfast) and slow (MSDslow) components, as well as their relative sizes [that is, α and (1 − α), respectively].

P(SD,tlag)=1[αe(SD/MSDfast)+(1α)e(SD/MSDslow)]1[αe(SDmax/MSDfast)+(1α)e(SDmax/MSDslow)]with MSD fast=4Dfasttlag,MSDslow=4Dslowtlag(7)

Least-square fits of Eqs. 6 and 7 to the CDFs of the SDs were performed for curves obtained after correlation analysis of consecutive frames (that is, frames with a time distance of one time lag) and for curves obtained after correlation analysis between each 2nd, 3rd, and up to 10th frames. Plots of the parameters for MSDfast and MSDslow against the respective multiples of the time lags yielded the MSD plots. The latter were fit to models for free (Eq. 8) or confined diffusion (Eq. 9).

MSD(tlag)=4Dtlag(8)MSD(tlag)=L23(1e12D0tlagL2)(9)

Computational model of the interactions between cAR1 and G proteins

The model was developed and simulated with the Simmune software, as previously reported (29, 30). cAR1 was modeled as a transmembrane receptor having one intracellular binding site for Gα and one extracellular binding site for cAMP. Ligation of the extracellular binding site was assumed to instantly activate the receptor (the transition from Rec to Rec_act in the model), thereby changing the properties of its intracellular domain. Gα was modeled as a constitutively membrane-bound protein (Gα) with two binding sites, one for Gβγ (the model representative of Gβγ) and one for the cytosolic domain of cAR1. Ligation of Gα through Gβγ was assumed to instantly transform the Gα from a state in which it cannot bind to the receptor into Gα_bnd, which can bind to the cytosolic domain of cAR1. The complex consisting of cAMP, Rec_act, Gα_bnd, and Gβγ transforms into a complex wherein Gα_bnd is replaced by Gα_act_bnd (representing the GTP-bound state of Gα_bnd). Gα_act_bnd and its associated Gβγ are released from the receptor and Gα_act_bnd releases Gβγ to produce free Gα_act and Gβγ. Finally, Gα_act transforms back into Gα, representing the replacement of GTP by GDP. The resulting equations, automatically generated by Simmune, and a full description of the rate constants are provided in the Supplementary Materials. Because the simulations were performed as deterministic integrations of the equations that describe the concentration kinetics, each parameter set was simulated only once. A detailed description of the modeling approach is provided in the Supplementary Materials (30).

Acknowledgments

Acknowledgments: We thank M. Ueda and T. Yanagida for help in the single-molecule imaging study and X. Xiang and D. Hereld for critical reading of the manuscript. Funding: This work was supported by the intramural fund of National Institute of Allergy and Infectious Diseases and the NIH Intramural AIDS Targeted Antiviral Program. X.X. is supported by the American Heart Association (AHA0930127N). Author contributions: X.X. and T.J. designed the experiments; X.X., J.A.B., J.Y., and T.J. performed the experiments; T.M., X.X., and J.A.B. analyzed the data; M.M.-S. constructed the models and performed the simulations; and X.X., T.M., M.M.-S., and T.J. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. The simulation software used in this work is freely available from the NIH but requires a license agreement.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/3/141/ra71/DC1

Methods

Technical details for simulations

Fig. S1. Imaging of single molecules of YFP with TIRF microscopy.

Fig. S2. Track lengths obtained from tracking cAR1-YFP and Gβ-YFP by SPT.

Fig. S3. Individual steps in the analysis of the mobilities of single molecules.

Fig. S4. cAR1-YFP proteins remain on the membrane after stimulation with cAMP.

Fig. S5. Effects of GTPase activities on the activation kinetics of G proteins.

Fig. S6. Simulated adaptation of the PIP3 response of cells exposed to a homogeneous, constant concentration of cAMP.

Fig. S7. Simulated PIP3 responses of cells exposed to a rapidly induced, but then constant, extracellular 2:1 gradient of cAMP.

Fig. S8. Diffusion constants and region size of cAR1-YFP and Gβ-YFP at the front and back of a cell exposed to a gradient of cAMP.

References

GproteinCoupling.zip file

Movies S1 to S4

References and Notes

View Abstract

Navigate This Article