Originally published In Press as doi:10.1074/jbc.M403712200 on April 12, 2004
J. Biol. Chem., Vol. 279, Issue 26, 27709-27718, June 25, 2004
A Fluorescence Resonance Energy Transfer-based Sensor Indicates that Receptor Access to a G Protein Is Unrestricted in a Living Mammalian Cell*
Inaki Azpiazu
and
N. Gautam
¶
From the
Departments of Anesthesiology and Genetics, Washington University School of Medicine, St. Louis, Missouri 63110
Received for publication, April 2, 2004
 |
ABSTRACT
|
|---|
Fluorescence recovery after photobleaching of muscarinic receptors and G protein subunits tagged with cyan or yellow fluorescent protein showed that receptors and G proteins were mobile and not immobilized on the cell membrane. The cyan fluorescent protein-tagged G
and yellow fluorescent protein-tagged G
subunits were used to develop sensors that coupled selectively with the M2 and M3 muscarinic receptors. In living Chinese hamster ovary cells, imaging showed that sensors emitted a fluorescence resonance energy transfer signal that was abrogated on receptor activation. When sequentially activated with highly expressed muscarinic receptors and endogenous receptors expressed at low levels, sensor molecules were sensitive to the sequence of activation and the receptor numbers. The results distinguish between models proposing that receptor and G protein types interact freely with each other on the cell membrane or that they function as mutually exclusive multimolecular complexes by providing direct support for the former model in these cells.
|
INTRODUCTION
|
|---|
G proteins and their receptors have been studied extensively, but it is still not known whether they are immobilized at certain locations on the cell membrane or are capable of diffusing to various regions on the cell membrane. By tagging a muscarinic receptor, a G subunit, and the G
complex with CFP1 or YFP and examining living cells with fluorescence recovery after photobleaching, we have determined that the receptor and G protein subunits are mobile. The mobility of the signaling molecules, however, did not indicate whether particular receptors and G proteins occur in mutually exclusive multimolecular complexes or freely diffuse to collide with each other. A longstanding model proposes that specificity is achieved by mechanisms that allow certain receptor types and G protein types to be associated in exclusive multimolecular complexes (1, 2). It has been proposed that receptors and G proteins exist as complexes even in the absence of an agonist stimulus and that G protein activation occurs in isolation from other such molecular assemblies (2, 3). In a further extension of this model, it has been suggested that receptor-G protein complexes persist during signaling activity (4). These models would explain how mammalian cells respond to extracellular signals with specificity, although they often express a variety of different receptor and G protein subunit types (5). Such a mechanism would aid the rapid response to the initial signal because of the association of appropriate receptor and G protein types before activation. Evidence for the presence of assemblies of receptors and G proteins in cell membranes and their potential interaction with adaptor or scaffolding proteins have also been thought to promote such receptor-G protein molecular assemblies (4, 6-9).
An alternative model predicts that receptor types and G protein types are mobile on the cell membrane, colliding with each other freely (1, 10). For optimal determination of which of these mechanisms regulate receptor-G protein coupling, living cells need to be used. It has not been possible, however, to examine these models in intact mammalian cells in the absence of appropriate methods. The biochemical approaches that have been used so far to identify the molecular mechanisms underlying G protein signaling have involved cell disruptive methods (11). In this study, we used non-invasive imaging methods to examine whether receptors and G proteins exist as multimolecular complexes. We developed a G protein sensor that emits a FRET signal that is sensitive to its activation state. Using this sensor, we directly observed the proportion of active and inactive G proteins by imaging living mammalian cells after the stimulation of particular receptors. We used a cell line expressing significantly different levels of Go coupling receptors to test whether a highly expressed receptor could act in a dominant-negative manner by acting as a sink for all Go molecules in the cell. To examine whether findings from the Go system applied to a different G protein signaling pathway, we mutated the o protein such that it coupled selectively with Gq coupling receptors. Examining the activation of the Go-q sensor by Gq coupling receptors using a similar experimental approach confirmed the results from the Go system.
|
MATERIALS AND METHODS
|
|---|
Chemicals and Expression Constructs—Chemicals were from Sigma unless listed otherwise. Cells were grown in CHO IIIa medium (Invitrogen) supplemented with dialyzed fetal bovine serum (Atlanta Biologicals), glutamine, penicillin/streptomycin, Fungizone, and methotrexate. CHO cells stably expressing muscarinic receptors (M2-CHO) have been described (12). M2-CHO cells were transfected with pDEST12.2 carrying appropriate combinations of o, o-CFP, 1, and YFP-1 (supplementary methods). Cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) from R. Y. Tsien (University of California San Diego) were mutated to obtain non-oligomerizing CFP (CFPm1) and non-oligomerizing Citrine (YFPcm1) (14, 29). Experiments with the Go sensor were performed with CFP/YFP and the experiments with the Go-q sensor with CFPm1/YFPcm1. Stable transfectants were selected with 0.5 mg/ml G418, and coexpressors were isolated by flow cytometry and maintained in the above media containing 0.3 mg/ml G418. Expression of fusion proteins was confirmed by fluorescence microscopy and by immunoblotting using antibodies to the expressed proteins. The M2-CHO expressing the Go sensor demonstrated growth properties similar to untransfected CHO cells (data not shown).
Estimating the Number of Molecules of o-CFP, YFP-1, and Endogenous 1 Subunits—Known number of M2-CHO cells stably expressing o-CFP + YFP-1 were lysed in SDS sample buffer and examined by immunoblotting after SDS-PAGE with antibodies to o, 1, or GFP. Densitometry (Image J software) of pure recombinant 1 protein was used to estimate number of molecules per cell. The level of o-CFP expression was determined similarly by comparing the expression level of o-CFP with YFP-1 using GFP-specific antibodies. o expression was also detected using antibodies specific to o.
Radioligand Binding to Membrane Receptors—Membranes were purified from M2-CHO cells stably expressing o-CFP + YFP-1 or transiently expressing M2-YFP and 10 nM [3H]N-methyl-scopolamine (Amersham Biosciences) or 10-50 nM [3H]5-hydroxytryptamine (Amersham Biosciences) binding was assayed essentially as described previously (24). All operations were conducted at 4 °C and the incubation buffer was supplemented with 10 µM pargyline and 0.1% sodium ascorbate. After 30 min of incubation, samples were filter-washed on glass membranes (Whatman) with chilled buffer. Radioactivity in glass membranes was measured using standard procedures.
Localized Photobleaching and Fluorescence Recovery—CHO cells cultured on coverslips were transfected with appropriate constructs using LipofectAMINE. After 24-48 h, the cells were mounted on a 0.25-ml imaging chamber (Warner Instruments) and perfused at a flow rate of 0.1 ml/min with Hanks' buffered saline solution supplemented with 10 mM HEPES, pH 7.4, and 1 mg/ml glucose. Fluid delivery was controlled by a fluid delivery system (Automate Scientific). Cells were visualized using a Zeiss 63x oil objective (1.4 numerical aperture) in a Zeiss Axioscope with 20% neutral density filter and 50-watt mercury lamp. The filter wheels were run by a Sutter Lambda 10-2 device to select the following filters in combination with appropriate beam splitters in the filter cube. For CC images: D436/20 excitation, D480/40 emission, and 455DCLP beam splitter; for CY images: D436/20 excitation, D535/30 emission, and 455DCLP beam splitter; and for YY images: D500/20 excitation, D535/30 emission, and 515LP beam splitter. All filters were from Chroma. Images were acquired using a Hamamatsu charge-coupled device Orca-ER Camera. Both camera and filter wheels were controlled peripherally using MetaVue software (Universal Imaging). Cell(s) with similar CC and YY intensities were selected and initial CC, CY, and YY images were recorded. A portion of the cell was bleached with white light for 30 s by positioning that portion of the cell alone in the illuminated field controlled by the diaphragm. Thereafter, the entire cell was moved back into the illuminated field, and CC, CY, or YY images were recorded at various times. Both whole-cell and photo-bleached regions of the plasma membrane were delineated using MetaVue. All images were background-subtracted. Data was plotted as the relative intensity = 100 x (Ixb x Iiw)/(Iib x Ixw) (Ref. 13). Ixb is the average intensity in the bleached membrane region at time x. Iiw is the average intensity in the whole-cell membrane region at time 0. Iib is the average intensity in the bleached membrane at time 0 s. Ixw is the average intensity in the whole cell membrane region at time x. The introduction of the correction factor (Iib/Iiw) in the formula compensates for any initial uneven distribution of the fluorophore in the cell membrane. Data were fitted to second order polynomial equations.
Fluorescence Spectroscopy of Cells in Suspension—M2-CHO cells stably expressing various combinations of the tagged G protein subunits were cultured overnight and analyzed at 2.5 million cells per milliliter. Spectra of the cells were recorded using a Spex Fluoromax-3 spectrofluorometer. Spectra from M2-CHO cells were subtracted from cells expressing G protein sensor or tagged subunits ensuring that emission at 580-600 nm baseline was about the same. For emission spectra resulting from CFP excitation, cell suspensions were excited at 422 nm (5-nm bandpass) and recorded between 460 and 580 nm (5-nm bandpass). For the YFP emission spectra, samples were excited at 488 nm (5 nm bandpass) and recorded between 500 and 600 nm. The excitation wavelengths are not the maximums and were picked to facilitate background subtraction. To examine agonist activation, the emission spectra resulting from CFP excitation of a cell sample were recorded first. Then the same sample was treated with 100 µM carbachol, and after 1 min of incubation, the same spectra were recorded. 10 µM atropine was added immediately; 1 min later, the same emission spectra were recorded. In determining the mean values of the responses to agonist and antagonist, the emission intensities at 478 nm and at 526 nm in the spectra from different experiments were used to determine 526/478 nm ratios.
Fluorescence Spectroscopy of the G Protein Sensor in Vitro—M2 CHO cells with or without the sensor o-CFP+YFP-1 were cultured overnight, rinsed with phosphate-buffered saline solution without calcium/magnesium and lifted in phosphate-buffered saline containing 0.02% EDTA. The cell suspension was briefly centrifuged at low speed, and the pellet was resuspended in phosphate-buffered saline supplemented with 3 mM MgSO4, 0.05 mM EGTA, and protease inhibitors. The density of the cell suspension was adjusted to 2.5 million cells per milliliter. Cyan emission spectra after excitation at 422 nm were recorded for both cell suspensions, and difference spectra were generated as described in the previous paragraph. The presence of FRET at 526 nm shoulder was confirmed visually and by measuring the ratio of E526/E480. Next, 1 ml of cell suspension was supplemented with 0.5 mM dithiothreitol and 200 µM GDP, guanosine 5'-O-(2-thio)diphosphate, or GTP--S and lysed in a nitrogen Parr bomb. Homogenates were incubated on ice for 20 min, and the spectra of 250-µl suspensions excited at 422 nm as described earlier were recorded, ensuring the baselines were the same at 600 nm with the addition of homogenate buffer. The spectra of M2-CHO wild-type cell homogenates were subtracted from that of the sensor cell homogenates to obtain the difference spectra.
Microscopy, Image Acquisition, and FRET Analysis—Cells were imaged as above. Agonists, antagonists, or vehicle were delivered at a rate of 
1 ml/min. Images were acquired at the indicated times. Cells were observed and images acquired with an Olympus 60x oil immersion objective (1.3 numerical aperture) or 63x oil immersion Zeiss Plan Apochromat objective (1.4 numerical aperture), 10-20% neutral density filter, and 50- or 100-watt mercury lamp. Appropriate combinations of 430/25 excitation, 470/30 emission, 500/20 excitation, 535/30 emission, and a polychroic beam splitter (Chroma) were used. Exposure times were 1 s with 4 x 4 binning. Images were acquired every 15 s for a total of 4 or 6 min and stored as 12-bit grayscale image stacks using MetaVue. Images were processed using Metamorph (Universal Imaging). Images were background-subtracted and aligned, and plasma membrane regions of entire cells (or most of the cell) were selected after we determined that CC (o-CFP) and YY (YFP-1) signals were co-localized. Average intensities in these regions were measured. Ratios of the CY over CC intensities from these regions were plotted as shown in the figures. Images were pseudocolored. Colors were associated with specific ranges of grayscale intensities.
Agonist Concentration Dependence of the Decrease in the FRET Signal—Images of live cells were recorded as above for 6 min. At the 90-s time point, cells were stimulated with increasing concentrations of carbachol (10 nM-10 µM) and at the 210-s time point, cells were stimulated maximally (with 1 mM). The regression lines were drawn from the baselines obtained using Prism software (GraphPad Software). The percentage of activation by submaximal doses was determined by the formula 100(Rx-Ro)/(Rm-Ro). Rx, Ro, and Rm are the values of the respective regression lines at the 210-s time point, where Ro was the initial baseline, Rx was the submaximal response, and Rm was the maximal response.
Image Processing—All CC and CY images corresponding to the data presented were stored in stacks. The CC and CY images were separated and were background-subtracted. Masks were generated using MetaVue from the images in the CC stack. First, the CC images were sharpened and a grayscale intensity value constant was subtracted from these images. The sharpening and substraction of grayscale intensity values results in highlighting the regions where o-CFP is preponderant (cell membranes and some intracellular particles). Pixels corresponding to the intracellular particles were assigned an intensity value of 0. These images were converted into binary images by assigning a value of 0 or 1 to pixels with 0 or >1 grayscale value. The CY and CC images recorded at each time point were used to obtain a CY/CC ratio image, and the resulting image was multiplied by the corresponding binary image. The intensities in the CY/CC images were pseudocolored where warmer hues represent higher ratios. Stacks were converted to QuickTime (.mov) files. The hues in the initial and final images were set at about the same for the two stacks presented. Legends on the images indicate time points at which agonists were introduced.
Calculating the Actual FRET Signal Intensity—To calculate the FRET signal component in the CY signal, the bleed through of the CFP emission into the CY channel and the cross-excitation of YFP by the CFP excitation wavelengths have to be determined and subtracted. M2-CHO cells stably expressing o-CFP or YFP-1 were used to identify these components. A, the factor for the cross-excitation of YFP by the CFP excitation wavelengths was the ratio of the emission intensity from YFP-1 cells using 430/25 excitation and 535/30 emission filters and the emission intensity from the same cells using 500/20 excitation and 535/30 emission filters. B, the factor for the bleed through of CFP excitation wavelengths was the ratio of the emission intensity from o-CFP cells using 430/25 excitation and 535/30 emission filters and the emission intensity from the same cells using 430/25 excitation and a 470/30 emission. When analyzing cells of different genotypes, we ensured that the CC emission intensity from o-CFP cells and YY emission intensity from YFP-1 cells were similar to the corresponding intensities of CC and YY channels of o-CFP YFP-1 cells. Ratios were determined from at least five cells, and the mean was used in the calculations below. The FRET signal intensity was calculated using the formula FRET = CY - [A x YY] - [B x CC], where A and B are conversion factors for YFP cross-excitation (0.369) and CFP bleed through (0.709), respectively. The actual FRET signal that occurs between o-CFP and YFP-1 was determined by acquiring and analyzing images of CHO cells transiently expressing these proteins as above. Likewise, to determine any potential FRET signal from two non-inter-acting fluorescent proteins targeted to the plasma membrane, o-CFP and a YFP construct containing a CAAX box at the C terminus to facilitate prenylation and membrane targeting were transiently co-expressed in CHO cells, and the images of these cells were acquired and analyzed as above.
|
RESULTS
|
|---|
Expression of the G Protein Sensor—To obtain cells expressing defined levels of the G protein sensor, we stably integrated the cDNAs for Go tagged with CFP (14) and G1 tagged with YFP into the genome of M2-CHO cells (Chinese hamster ovary cells expressing the M2 muscarinic receptor) (12, 15, 16). The M2-CHO cells do not contain endogenous muscarinic receptors but express the stably integrated M2 receptor cDNA (12). The CFP molecule was inserted downstream of Gly92 in Go because this region forms a loop that projects away from the complex in the crystal structure of the G protein, and the Dictyostelium discoideum G subunit is active when it is fused to CFP in a similar fashion (17, 18). The YFP molecule was fused to the N terminus of G1. It was anticipated that fusion to the N terminus would not affect receptor interaction based on the model for receptor interaction of the G protein (15) and the innocuous effect of YFP-tagged D. discoideum subunit on interaction with a D. discoideum receptor (18). Endogenous G subunits in the CHO cells were expected to form viable YFP-1 complexes. Association of -CFP subunit and YFP- complex was expected to bring the YFP and CFP molecules sufficiently close for FRET to occur, and receptor activation was expected to abrogate the FRET signal as a result of subunit dissociation (Fig. 1A) (18). Images of M2-CHO cells expressing the G protein sensor were captured at different excitation and emission wavelengths (Fig. 1, B-D) (see "Materials and Methods"). The FRET emission is observed in the CY channel (along with bleed through of CFP emission and YFP emission caused by excitation by the CFP excitation wavelengths; discussed below). When images of cells expressing o-CFP + YFP-1 were captured in the different channels, o-CFP and YFP-1 were localized predominantly in the plasma membrane as anticipated (Fig. 1B). The presence of the 1-YFP fusion on the plasma membrane indicated that it was in complex with a lipid-modified endogenous subunit. Furthermore, the distribution of o-CFP is similar to the distribution of the YFP-1 complex, suggesting that the subunits are mostly in the heterotrimer form. This result also indicates that the o-CFP + YFP-1 complex is representative of the heterotrimer population in the cell, which will also consist of o-CFP or YFP-1 in complex with endogenous untagged and subunits. We determined using immunoblotting with specific antibodies that the introduced YFP- subunit formed about 45% of the total 1 subunit in the cell (o is not detectable in these cells) and that the levels of expressed o-CFP and YFP-1 were approximately equal (see "Materials and Methods"). The plasma membrane region in which the G protein sensor was localized was selected when analyzing images of cells to determine the effect of agonists or antagonists on the activation state of the G protein sensor (Fig. 1C). We confirmed using cells expressing o-CFP or YFP-1 that CFP emission is observed in the CC channel but not in the YY channel, and YFP emission is observed in the YY channel but not in the CC channel (Fig. 1D). Bleed through of CFP emission from o-CFP and cross-excitation of YFP- are seen in the corresponding CY channels.
View larger version (42K):
[in this window]
[in a new window]
|
FIG. 1.
Imaging receptor activation of G proteins. A, diagrammatic representation of o-CFP and YFP-15. Association of o-CFP with YFP-1 is expected to bring CFP and YFP within 100 Å and produce a YFP emission signal resulting from FRET (red arrow) when CFP is excited (black arrow). Receptor mediated dissociation is expected to eliminate the FRET signal. Excitation and emission wavelengths for CFP and YFP are shown in gray. B, fluorescence from M2-CHO cells stably expressing o-CFP and YFP-1 was imaged. CC, CY, and YY are images acquired using the following excitation filters: CC, 436/20 excitation and 480/40 emission; CY, 436/20 excitation and 535/30 emission; and YY, 500/20 excitation and 535/30 emission. C, in experiments to visualize G protein sensor activation and deactivation, the membrane region where the G protein sensor was localized was selected typically as shown, and the intensity of the signal was followed over time. D, fluorescence from M2-CHO cells stably expressing o-CFP or YFP-1 alone was imaged as above and obtained as described under "Materials and Methods."
|
|
Mobility of Receptor and G Protein on the Cell Membrane—At first, to examine whether receptors and G proteins are immobilized at specific regions of the cell membrane or whether they are capable of diffusing to various parts of the cell membrane, we performed fluorescence recovery after photobleaching experiments (19) on CHO cells transiently expressing YFP-tagged M2 or o-CFP + YFP-1. Expression of M2-YFP was independently confirmed by the presence of [3H]-N-methyl scopolamine binding sites in M2-YFP transfected cells (see "Materials and Methods"). A randomly selected region of a cell was bleached. Fluorescence was recovered in the bleached area over a period of time in the case of M2-YFP cells and cells expressing o-CFP YFP-1 (Fig. 2, A and B). The average intensity in the whole-cell membrane region at different times was used as a correction factor as described under "Materials and Methods" to control for any changes in fluorescence emission that resulted from causes other than the mobility of the fluorescent protein into the bleached region of the cell membrane. Plotting the time course of recovery indicated that within 5 min, about 70 to 85% of M2-YFP (receptor YY), o-CFP (G protein CC), YFP- (G protein YY), and G protein heterotrimer (G protein CY (see below)) were recovered in the bleached region (Fig. 2C). The time course also indicated that recovery was not complete, especially in the case of the receptor (Fig. 2C). It is thus likely that most of the receptor proteins will be capable of diffusion into the bleached region over a longer period. This slower diffusion of the receptor in contrast to the G protein subunits is consistent with the slower diffusion of transmembrane proteins compared with peripherally associated proteins (20). Although it cannot be ruled out that a relatively small proportion of each of these molecules is immobile, the result indicates that receptors and G proteins are unlikely to be immobilized on the cell membrane as in Fig. 3A. In these cells, the predominant proportions of the receptor, G protein subunits, and G protein heterotrimer are capable of free movement to any part of the cell membrane, as shown in Fig. 3, B and C. These results, however, do not indicate whether certain receptor and G protein types are present as pre-existing complexes that diffuse as multimolecular assemblies or whether they freely diffuse to collide with each other (Fig. 3, B or C). To distinguish between these models, we developed a G protein sensor as described below.
View larger version (36K):
[in this window]
[in a new window]
|
FIG. 3.
Competing models that propose mechanisms at the basis of receptor-G protein association. A, particular receptor types and G protein types may be present as exclusive complexes (within circle) ensuring rapid responses to a signal and selective activation of pathways. Such complexes may be immobile and localized to specific regions on the cell membrane. B, on the other hand, such complexes may be mobile but receptor and G protein types may still reside and function within these multimolecular complexes. C, finally, G proteins and receptors may freely diffuse along the cell membrane to interact with each other randomly and stochastically.
|
|
The G Protein Sensor in Living Cells Emits a FRET Signal Measurable Using Spectroscopy—We first examined whether the G protein sensor (o-CFP YFP-) emitted a FRET signal and whether this signal was sensitive to the activation state of the G protein using spectrofluorometric and imaging experiments (21-23). Spectra from M2-CHO cells expressing o-CFP + YFP-1 were compared at the excitation wavelength for CFP (422 nm) with cells expressing -CFP alone. In o-CFP + YFP-1 cells, the appearance of an emission peak at 526 nm was observed along with a corresponding decrease in the emission peak of CFP (478 nm) (Fig. 4A). This peak corresponded with the emission peak observed when the same cells were excited at 488 nm; in addition, no emission was observed when o-CFP cells were excited at 488 nm or YFP-1 cells at 422 nm (supplementary Fig. 4). These results are consistent with a FRET signal originating from YFP because of CFP excitation, caused by G protein subunits forming a heterotrimer (Fig. 1A). Abrogation of FRET will result in an increase in the emission intensity at the 478-nm peak with a corresponding decrease in the 526-nm peak. Changes in the FRET signal can thus be determined by following the 526/478-nm ratio after receptor stimulation or inactivation. When the M2 receptors in these cells were activated with an agonist, carbachol, the ratio of emission at the 526-nm peak and the 478-nm peak decreased and then recovered on subsequent deactivation with an antagonist, atropine (Fig. 4B). The sensor thus indicated the activation state of the population of Go molecules inside the cell.
View larger version (24K):
[in this window]
[in a new window]
|
FIG. 4.
Fluorescence spectroscopy of M2-CHO cells expressing tagged G protein subunits. A, representative emission spectra of M2-CHO cells stably expressing o-CFP + YFP-1 superimposed upon similar spectra from -CFP alone. Excitation was at 422 nm. Inset, immunoblot of purified recombinant o protein (50 ng) and membrane fraction from M2-CHO cells expressing o-CFP + YFP-1 (lane 2, 5 x 106 cells; lane 3, 15 x 106 cells) and o-CFP (lane 4, 5 x 106 cells; lane 5, 15 x 106 cells). B, effect of agonist addition followed by antagonist to M2-CHO cells expressing the G protein sensor. Bars are the mean values of the 526/478 nm ratios ± S.E. (n = 6). Spectra in A and B were obtained after subtraction of the spectra from untransfected M2-CHO cells. Cell numbers assayed were similar, and base lines were the same. Asterisks denote that the ratio after agonist treatment is significantly different from the ratio before treatment or antagonist treatment (p < 0.01). Fluorescence spectroscopy of G protein sensor in vitro. C, representative spectra of cell homogenates expressing o-CFP and 1-YFP prepared in the presence of GTP--S, guanosine 5'-O-(2-thio)diphosphate (GDP--S). D, means of the 526 nm/478 nm emission ratio from spectra of homogenates treated with different guanine nucleotides (± S.E. of four experiments with samples treated with GTP--S or GDP--S and two experiments with GDP-treated samples). In all spectroscopic measurements, the bandpass was 5 nm, as mentioned under "Experimental Procedures." Asterisks denote that the difference between the GTP--S- and GDB--S-treated samples is significant at p < 0.01.
|
|
To examine the properties of the sensor protein directly in vitro, cells expressing the G protein sensor were homogenized in the presence GTP--S, guanosine 5'-O-(2-thio)diphosphate, or GDP, and each of these samples was assayed by fluorescence spectroscopy. Emission spectra were recorded after CFP excitation. Comparison of the spectra after GTP--S treatment with spectra after guanosine 5'-O-(2-thio)diphosphate or GDP treatment showed a significant decrease in the emission at 526 nm (Fig. 4, C and D). GTP--S was used because the fluoride ion in aluminum fluoride quenches the YFP emission similar to Cl-. This result indicated that the non-hydrolyzable GTP analog activates the subunit and leads to its dissociation from the complex, resulting in a loss in FRET signal intensity. This result also indicated that the FRET signal changes seen in cells expressing the G protein sensor as a result of receptor action are caused by G protein activation.
Imaging Methods Can Measure the FRET Signal Changes from the G Protein Sensor in Response to an Agonist or an Antagonist—We then acquired images of single M2-CHO cells stably expressing -CFP + YFP- sequentially before and after the addition of the agonist. The plasma membrane region containing the Go sensor was selected for image analysis (Fig. 1C). We ensured that cells chosen for analysis expressed similar levels of -CFP and YFP- fluorescence intensity levels. Immediately after the cells were exposed to the agonist, the CY/CC ratio decreased significantly (Fig. 5A). The response was sensitive to the concentration of carbachol (Fig. 5B). The EC50 for carbachol activation of the G protein was 600 nM, consistent with results from a reconstituted system (16, 24). This indicates that the G protein sensor used here in living cells behaves similarly to the native Go protein. The CY/CC ratio was unaffected by buffer alone or an unrelated agonist; in addition, agonist activation was blocked by an antagonist (supplementary Fig. 5). Thus, the decrease in the FRET signal in response to an extracellular signal is both specific and sensitive to its strength. A YY emission acquired before and after exposure to agonist indicated that the decrease in the FRET signal was not caused by a decrease in the YFP emission (supplementary Fig. 5).
Furthermore, when M2-CHO cells expressing the G protein sensor were first exposed to the agonist and subsequently to the antagonist, the FRET signal decreased after addition of the agonist but reappeared immediately after exposure to the antagonist (Fig. 5C). This result established that the sensor responds to different states of activation with measurable changes in the FRET signal intensity. It also showed that to maintain the G protein in the activated state, the receptor needs to be continually stimulated. The deactivation of a receptor by agonist removal and the resultant rapid deactivation of a G protein may be a mechanism that allows a mammalian cell to respond quickly to subsequent signals. Typical plots of the CC and CY emission intensities corresponding to one of the experiments above are shown (Fig. 5D). The changes in CC emission intensity in response to agonist and antagonist addition are inversely correlated with changes in CY emission intensity. This result demonstrates unambiguously that the decrease in CY emission intensity in response to an extracellular signal represents a decrease in the FRET signal resulting from the activation of the G protein sensor.
Photobleaching of the Acceptor YFP of the Go Sensor and Computation Indicate the Presence of a FRET Signal—To further confirm that the Go sensor emits a FRET signal, we examined the increase in the CC emission intensity as a result of bleaching the FRET acceptor, YFP (Fig. 6A). The Go sensor expressed in M2-CHO cells showed a significant increase in the CC emission, when YFP was bleached selectively. o-CFP alone or o-CFP plus a form of YFP containing a CAAX sequence at the C terminus that was prenylated and directed to the cell membrane showed no significant increase in CC emission after photobleaching, indicating that the FRET signal was specific to the Go heterotrimer. To obtain evidence for FRET using another method, we computed the FRET signal emitted by the G protein sensor by determining the extent of CFP emission bleed through into the CY channel and the extent of YFP cross-excitation as described under "Materials and Methods." A significant FRET signal was emitted by CHO cells expressing o-CFP + YFP-15 in contrast to o-CFP plus the prenylated YFP (Fig. 6B). The FRET signal identified here provided a direct measure of the proportion of inactive Go heterotrimers in a cell.
View larger version (20K):
[in this window]
[in a new window]
|
FIG. 6.
FRET measurement by acceptor photobleaching and computation. A, bar diagram of averaged grayscale intensities from membrane regions of M2-CHO cells transiently expressing o-CFP and prenylated-YFP and M2-CHO cells stably expressing o-CFP alone or o-CFP + YFP-1. Cells were exposed to buffer with or without the appropriate agonist, 20 µM serotonin or 1 mM carbachol. CC, CY, and YY images were captured as reference. Cells with the appropriate expression levels of CFP and YFP were selected, and images were processed with Metamorph. Subsequently, a CC image was acquired and the cells were photobleached for 2 min using yellow excitation without a neutral density filter. A CC image was captured immediately. A YY image confirmed complete photobleaching of YFP. Average grayscale intensities were measured, and the FRET signal was determined using the formula % change in CC intensity = 100 x (Efinal - Einitial)/Einitial, where E is CFP emission. Differences between +serotonin and + or -carbachol are significant at p < 0.001. Differences were not statistically significant between o-CFP and prenylated-YFP, o-CFP alone, and +carbachol treatments. Data presented has been corrected such that o-CFP+YFP-prenyl value was zero. B, intensities of the calculated FRET signal emission from CHO cells expressing o-CFP + prenylated-YFP (n = 10) (a) and o-CFP + YFP-1 (n = 21). Mean intensities (± S.E.) (b) Asterisk indicates absence of a detectable FRET signal.
|
|
Access of M2 and an Endogenous Gi/o Coupling Receptor to the Go Sensor—The Go sensor was used to examine whether particular receptors and G proteins existed as exclusive complexes in the cell membrane or whether receptors couple freely through molecular collisions with G proteins. We used M2-CHO cells stably expressing the G protein sensor for these studies because the cells express high levels of M2 receptors compared with an endogenous Go coupling serotonin receptor. A filter-binding assay using radiolabeled ligands indicated that the M2-CHO cells express 500-fold more introduced muscarinic receptors compared with endogenous serotonin receptors. Each cell expressed 400,000 M2 receptors, 800 serotonin receptors, and 45,000 molecules of YFP-1 (indicating that, because 1 is the predominant mammalian G, total G proteins capable of interaction with these receptors were not likely to be significantly more than 100,000 molecules per cell) (see "Materials and Methods"). We expected that the higher receptor numbers of M2 would act in a dominant-negative manner by acting as a sink for the G protein sensor if G proteins were in an exclusive multimolecular complex with a receptor type.
If exclusive complexes of M2 receptors with Go existed, other Go coupling receptors would not be able to activate the G protein sensor in these cells because M2 would act as a sink for all the Go molecules in the cell. When the cells were exposed to saturating concentrations of first serotonin and then carbachol, the sensor responded to both agonists (Fig. 7, A and C) (in results not shown, a serotonin receptor antagonist, cyanopindolol, blocked serotonin activation, independently confirming the presence of serotonin receptors in the M2-CHO cells; CHO cells have previously been shown to possess serotonin receptors coupled to Gi/Go (25)). In contrast, when the cells were exposed to first carbachol and then serotonin, the G protein sensor responded to carbachol but not to serotonin (Fig. 7, B and D). Movies of the captured images show that changes in the FRET signal occur on the cell membrane in response to serotonin and carbachol (supplementary Fig. 7). The result in Fig. 7A indicated that stimulation of the small population of serotonin receptors relative to G protein sensors activated almost half the sensor molecules even in the presence of a vast excess of M2 receptors in a cell membrane. This result demonstrates the large amplification of an external signal in an intact living cell (20-50 Go molecules per serotonin receptor). In contrast, when the larger population of M2 receptors was stimulated first, the entire population of G protein sensors was activated. No Go sensor molecules were available for activation by the serotonin receptors despite the ability of these receptors to activate half the sensor molecules when stimulated first (Fig. 7A).
The Proportion of Go Sensors Remaining in the Deactivated State after M2 Activation—To estimate the proportion of functional but inactive G protein sensors in a cell, the sensor was activated with carbachol and the acceptor photobleaching experiment was performed. No detectable increase in CC emission was detected (Fig. 6A). This result indicated that FRET from o-CFP to YFP- was abrogated because of the activation of the Go sensor protein. When the change in CFP emission of the Go sensor was examined by acceptor photobleaching after the addition of serotonin, approximately 50% of the FRET signal remained (Fig. 6A). Results in Fig. 7A and from five other similar experiments were consistent with this finding, indicating that a saturating concentration of serotonin activated about half the G protein sensors on the membrane of a single cell. If serotonin receptors were associated with Go in an exclusive multimolecular complex, the intensity of the FRET signal after M2 activation would be expected to be half the FRET signal intensity before M2 activation. When we determined the FRET signal before and after M2 receptor activation, using photobleaching no FRET signal was detected after activation of the muscarinic M2 receptor (Fig. 6A).
Access of M3 and Other Gq Coupling Receptors to a G Protein Sensor—To test whether the findings from the Gi/Go coupling receptors above were applicable to a distinctly different receptor system, we used CHO cells expressing the Gq coupling M3 receptor. These cells express approximately the same number of M3 receptors as M2 receptors described above (26, 27). To obtain a G protein sensor capable of measuring activation by the M3 receptor, we substituted the C-terminal residues of the o subunit with corresponding residues from the q subunit (Fig. 8A). Previous reports have shown that the C-terminal residues of the subunits govern specificity of interaction with Gi/o-and Gq-coupled receptors (28). The chimeric o-q subunit was introduced along with YFP-1 into M2-CHO cells. When stimulated with carbachol, in contrast to the Go sensor, the Go-q sensor was not activated (Fig. 8B) indicating that the C-terminal domain did determine receptor specificity in vivo. When the Go-q sensor was introduced into M3-CHO cells and stimulated with carbachol, there was a substantial change in the FRET signal (Fig. 8C). This result indicated that the Go-q sensor was highly selective in terms of receptor coupling and was activated strongly by the M3 receptor but not by M2. Control experiments with the Go sensor showed that it was poorly activated by the M3 receptor (Fig. 8C). To perform sequential receptor activation experiments as in the case of the Go sensor described above, we assayed M3-CHO cells expressing the Go-q sensor with a variety of agonists of known Gq-coupled receptors. LPA, ATP, and bradykinin elicited FRET signal decreases from this sensor, indicating the presence of these Gq-coupled receptors in CHO cells (Fig. 9A). The FRET signal in response to bradykinin and LPA recovered rapidly, presumably because the receptors were desensitized rapidly. When M3-CHO cells were exposed first to acetylcholine and then to a pool of LPA, ATP, and bradykinin, acetylcholine evoked a large decrease in FRET, but the subsequent addition of the pool of agonists had no effect (Fig. 9B). When the pool of agonists was introduced first, a FRET decrease was detected, followed by another decrease in the FRET signal when acetylcholine was introduced (Fig. 9B). Examination of the images of cells during the activation process shows the corresponding changes in the FRET signal on the cell membrane (Fig. 9C). We then performed photobleaching experiments to more rigorously determine the FRET signal after the addition of carbachol. Results in Fig 10 show that photobleaching the acceptor YFP increases the CFP emission in M3-CHO cells expressing the Go-q sensor before the addition of ACh. This result confirmed the presence of a FRET signal from this sensor. After the addition of ACh, no change in the CFP emission was detected after YFP photobleaching, indicating that the entire complement of detectable Go-q molecules was activated by the M3 receptor. Despite this ability of the M3 receptor to activate all detectable levels of Go-q heterotrimers, various endogenous Gq-coupled receptors were able to activate Go-q molecules in M3-expressing cells (Fig. 9, A-C). The ability of the pooled agonists to evoke a response when added first indicated that they were functional and capable of activating any available Go-q protein. The M3 receptors are thus not present in an exclusive multimolecular complex with Go. Together, these results are consistent with those obtained with the Go sensor, carbachol, and serotonin using the M2-CHO cells.
View larger version (30K):
[in this window]
[in a new window]
|
FIG. 9.
Response of the Go-q sensor to sequential stimulation of endogenous receptors followed by overexpressed M3 receptors. A, plots of CY/CC ratio from images of M3-CHO cells expressing o-q-CFP + YFP-1 were acquired before and after exposure to an agonist (10 µM bradykinin (Brdkn), 100 µM ATP or 4 µM lysophosphatidic acid (LPA)) and subsequently to carbachol (1 mM). The responses to these agonists are mostly weak and the stronger responses are shown. Application of the first agonist continued after the introduction of carbachol. The lack of response to 0.2 µM cholecystokinin (CCK) is shown as an indicator of the baseline. For Brdkn, 21 cells (six coverslips) were observed, 18 of which showed a positive response to this agonist. For ATP, 15 cells (five coverslips) were observed, 13 of which had a positive response to this agonist. For LPA, 12 cells (three coverslips) were observed, 11 of which had a positive response to this agonist. As for CCK, 11 cells (three coverslips) were observed; only 1 had a positive response to this agonist. B, plots of CY/CC ratio from images of the same cells before and after initial exposure to 1 mM acetylcholine (ACh) and a pool of agonists including 10 µM brdkn, 100 µM ATP, and 4 µM LPA. The response to the pooled agonist is not additive compared with the individual responses in A, because the traces in A represent the stronger responses. Application of ACh continued after the introduction of the pool of agonists. Seven cells (three coverlips) were studied, all of which showed responses similar to the plot shown. C, similar plots of the same cells before and after exposure to a pool of agonists (10 µM brdkn, 100 µM ATP, and 4 µM LPA) and subsequently to carbachol (1 mM). Application of the first agonist continued after the introduction of carbachol. Images correspond to the time points circled and numbered. 16 cells (five coverslips) were observed; all of them showed responses similar to the plot shown.
|
|
View larger version (16K):
[in this window]
[in a new window]
|
FIG. 10.
Bar diagram of averaged gray scale intensities from membrane regions of M3-CHO cells expressing the Go-q sensor. Cells were exposed to buffer with or without 1 mM ACh. CC, CY, and YY images were captured as reference. Photobleaching was performed as described in the legend to Fig. 6. Differences were statistically significant at p < 0.001. For each treatment, CC intensity changes from at least 30 cells from 8 different coverslips were observed.
|
|
|
DISCUSSION
|
|---|
In the design of the G protein sensors used here, the insertion of CFP in the subunit restricts mobility of the CFP partner relative to the YFP attached to the N terminus of 1. This potentially reduces the orientations that are optimal for efficient FRET. In combination with the binding of the tagged subunits with endogenous G protein subunits, this was expected to result in the emission of a relatively low FRET signal from the G protein sensors developed here. The calculated FRET signal is thus comparable with a similar D. discoideum G protein sensor (18). The responses to agonist and antagonist measured by ratiometry, by computation of FRET through independently determining the contaminating factors, and by acceptor photobleaching indicate unequivocally that the sensors respond to activation and deactivation with corresponding changes in the FRET signal.
Various mechanisms have been proposed to explain the specificity of response seen in G protein-mediated signaling. A longstanding model invokes the possibility of particular receptors and G proteins occurring as exclusive multimolecular complexes (1, 2). As an extension of this model, it has also been proposed that in such complexes, receptors and G proteins may be coupled even before the receptor is stimulated by an agonist and may function as a signaling complex (2-4). The absence of cross-talk among certain signaling pathways in the same cell, evidence for association of G protein subunits with cytoskeletal elements, and copurification of some receptors and G proteins have been thought to support this model (2, 4). These models are attractive because the potential mechanisms proposed can help achieve both specificity of signaling and rapid response to the initial signal. There has been a dearth of experimental methods available to test these models so far. The cell-disruptive methods used to obtain the extensive information about the molecular basis of G protein signaling are not appropriate for examining the association of receptors and G proteins on the cell membrane. Methods to observe receptors and G proteins in living cells will allow the rules for the association of receptors and G proteins to be identified. We developed sensors based on G proteins that can couple with two different receptor types, Gi/o and Gq. These sensors emit a FRET signal that responds to the activation state of the G protein. The sensors facilitated the observation of receptor activation of G proteins in live cells in real time.
When M2-CHO cells expressing the Go sensor were stimulated with saturating levels of serotonin, almost half of the Go sensor molecules were activated by the endogenous serotonin receptors (Figs. 6A and 7A). These results indicated that M2-CHO cells possess native Go coupling receptors apart from M2 that were capable of activating a significant proportion of the Go protein in the cell. According to a model that proposes the existence of specific and exclusive receptor-G protein complexes, these receptors should be associated with half the Go sensor molecules. An expectation of the model is that agonist stimulation of M2 preceding serotonin receptor stimulation should result in the activation of less than half the Go sensors. The Go associated with receptors other than M2 should remain in the inactive heterotrimer form even after the activation of the muscarinic receptors because of their inaccessibility to the muscarinic receptors (Fig. 3B). However, carbachol stimulation resulted in the loss of all detectable FRET from the Go sensor measured ratiometrically or by using photobleaching (Figs. 6A and 7B), indicating that M2 receptors have access to the entire pool of Go detectable in this assay. The serotonin receptors are thus not associated in an exclusive multimolecular complex with the Go protein in CHO cells.
To test whether this rule applies to other receptor-G protein systems in these cells, we examined the association of Gq-coupled receptors by developing a Go-q sensor. The C-terminal receptor coupling domain of the o subunit in the sensor was substituted with the C-terminal domain of the q subunit. As anticipated, the Go-q sensor selectively coupled to the Gq-coupled M3 receptor. Several known Gq-coupled receptors endogenous to M3-CHO cells were able to activate the Go-q sensor. Stimulation of a set of these receptors first and then of the M3 receptor indicated that 35% of the Go-q sensors were activated by the endogenous receptors. If these receptors were associated in signaling complexes with the Go-q sensor molecules, activation of the sensor by the M3 receptor preceding activation by the endogenous receptors should result in the activation of only 65% of the Go-q sensors on the cell membrane. However, the measurement of FRET changes and also photobleaching experiments show that M3 stimulation resulted in the loss of the entire FRET signal on the cell membranes, indicating that all the detectable Go-q sensors were activated. These results imply that the Go and Gq proteins are freely accessible to the two classes of G protein coupled receptors that couple to them. It is unlikely that exclusive multimolecular complexes of particular receptors and G proteins exist in these mammalian cells before agonist stimulation.
The findings here indicate that cells similar to CHO cells are unlikely to contain multimolecular complexes of Go/Gi or Gq with specific receptors that exclude exchange between the complexes. Because these two classes of G proteins and their receptors mediate a variety of major signaling pathways and the properties of receptors and G proteins are highly conserved, these results may apply to G proteins in general. The relevance of these results to highly differentiated cells is unclear. In the case of one such cell type, retinal photoreceptor cells, only one type of G protein (GtR or GtC) and a single receptor type (rhodopsin or a color opsin) are expressed. Other differentiated cells do express multiple receptors and G proteins. This approach can potentially be used to examine whether these cells possess potential signaling complexes made up of receptors and G proteins or whether all receptors have free access to the population of G protein molecules on the cell membrane.
These results imply that specificity of G protein signaling is probably achieved through kinetic determinants intrinsic to the proteins and stochastic receptor-G protein collisions (1). A model that proposes specific exclusive receptor-G protein complexes may intuitively explain specificity and rapid responses. But free diffusion of receptors and G proteins and their collisions on the membrane combined with the known relatively low affinity of a G protein for a receptor may allow the progressive and sustained amplification of the initial signal that is common to many cells as predicted before (10). Finally, these results demonstrate the utility of sensors such as the ones used here to address long-standing questions about the mechanistic bases of signaling pathway function in intact living mammalian cells.
|
FOOTNOTES
|
|---|
* This work was supported by National Institutes of Health Grant GM46963. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 
The on-line version of this article (available at http://www.jbc.org) contains supplemental text and figures.
¶ To whom correspondence should be addressed: Box 8054, Washington University School of Medicine, St. Louis, MO 63110. Tel.: 314-362-8568; Fax: 314-362-8571; E-mail: gautam{at}morpheus.wustl.edu.
1 The abbreviations used are: CFP, cyan fluorescent protein; YFP, yellow fluorescent protein; FRET, fluorescence resonance energy transfer; CHO, Chinese hamster ovary; GTP--S, guanosine 5'-O-(3-thio)tri-phosphate; LPA, lysophosphatidic acid.
|
ACKNOWLEDGMENTS
|
|---|
We thank M. Akgoz and V. Kalyanaraman for experimental help and discussions.
|
REFERENCES
|
|---|
- Simon, M. I., Strathmann, M. P., and Gautam, N. (1991) Science 252, 802-808[Abstract/Free Full Text]
- Neubig, R. R. (1994) FASEB J. 8, 939-946[Abstract]
- Hur, E. M., and Kim, K. T. (2002) Cell Signal. 14, 397-405[CrossRef][Medline]
[Order article via Infotrieve]
- Rebois, R. V., and Hebert, T. E. (2003) Recept. Channels 9, 169-194[CrossRef][Medline]
[Order article via Infotrieve]
- Neves, S. R., Ram, P. T., and Iyengar, R. (2002) Science 296, 1636-1639[Abstract/Free Full Text]
- Okamoto, T., Schlegel, A., Scherer, P. E., and Lisanti, M. P. (1998) J. Biol. Chem. 273, 5419-5422[Free Full Text]
- Anderson, R. G. (1998) Annu. Rev. Biochem. 67, 199-225[CrossRef][Medline]
[Order article via Infotrieve]
- Oh, P., and Schnitzer, J. E. (2001) Mol. Biol. Cell 12, 685-698[Abstract/Free Full Text]
- Ostrom, R. S. (2002) Mol. Pharmacol. 61, 473-476[Free Full Text]
- Hille, B. (1992) Neuron 9, 187-195[Medline]
[Order article via Infotrieve]
- Ross, E. M., and Wilkie, T. M. (2000) Annu. Rev. Biochem. 69, 795-827[CrossRef][Medline]
[Order article via Infotrieve]
- Azpiazu, I., Cruzblanca, H., Li, P., Linder, M., Zhuo, M., and Gautam, N. (1999) J. Biol. Chem. 274, 35305-35308[Abstract/Free Full Text]
- Phair, R. D., and Misteli, T. (2000) Nature 404, 604-609[CrossRef][Medline]
[Order article via Infotrieve]
- Zhang, J., Campbell, R. E., Ting, A. Y., and Tsien, R. Y. (2002) Nat. Rev. Mol. Cell. Biol. 3, 906-918[CrossRef][Medline]
[Order article via Infotrieve]
- Azpiazu, I., and Gautam, N. (2001) J. Biol. Chem. 276, 41742-41747[Abstract/Free Full Text]
- Peralta, E. G., Winslow, J. W., Peterson, G. L., Smith, D. H., Ashkenazi, A., Ramachandran, J., Schimerlik, M. I., and Capon, D. J. (1987) Science 236, 600-605[Abstract/Free Full Text]
- Lambright, D. G.
TOP
ABSTRACT
INTRODUCTION
, 
, 
) before and after exposure to carbachol (1 mM). Timing of drug injections are indicated with arrows. The graph is representative of 10 experiments. B, plot of FRET decrease in response to various concentrations of carbachol. Points are means ± S.E. (n = 3). Curve was fitted using GraphPad Prism software. C, plots of CY/CC ratio derived from plasma membrane regions of different cells (
