Live Cell Imaging of Gs and the β2-Adrenergic Receptor Demonstrates That Both αs and β1γ7 Internalize upon Stimulation and Exhibit Similar Trafficking Patterns That Differ from That of the β2-Adrenergic Receptor*[boxs]

To visualize and investigate the regulation of the localization patterns of Gs and an associated receptor during cell signaling, we produced functional fluorescent fusion proteins and imaged them in HEK-293 cells. αs-CFP, with cyan fluorescent protein (CFP) inserted into an internal loop of αs, localized to the plasma membrane and exhibited similar receptor-mediated activity to that of αs. Functional fluorescent β1γ7 dimers were produced by fusing an amino-terminal yellow fluorescent protein (YFP) fragment to β1 (YFP-N-β1) and a carboxyl-terminal YFP fragment to γ7 (YFP-C-γ7). When expressed together, YFP-N-β1 and YFP-C-γ7 produced fluorescent signals in the plasma membrane that were not seen when the subunits were expressed separately. Isoproterenol stimulation of cells co-expressing αs-CFP, YFP-N-β1/YFP-C-γ7, and the β2-adrenergic receptor (β2AR) resulted in internalization of both fluorescent signals from the plasma membrane. Initially, αs-CFP and YFP-N-β1/YFP-C-γ7 stained the cytoplasm diffusely, and subsequently they co-localized on vesicles that exhibited minimal overlap with β2AR-labeled vesicles. Moreover, internalization of β2AR-GFP, but not αs-CFP or YFP-N-β1/YFP-C-γ7, was inhibited by a fluorescent dominant negative dynamin 1 mutant, Dyn1(K44A)-mRFP, indicating that the Gs subunits and β2AR utilize different internalization mechanisms. Subsequent trafficking of the Gs subunits and β2AR also differed in that vesicles labeled with the Gs subunits exhibited less overlap with RhoB-labeled endosomes and greater overlap with Rab11-labeled endosomes. Because Rab11 regulates traffic through recycling endosomes, co-localization of αs and β1γ7 on these endosomes may indicate a means of recycling specific αsβγ combinations to the plasma membrane.

To visualize and investigate the regulation of the localization patterns of G s and an associated receptor during cell signaling, we produced functional fluorescent fusion proteins and imaged them in HEK-293 cells. ␣ s -CFP, with cyan fluorescent protein (CFP) inserted into an internal loop of ␣ s , localized to the plasma membrane and exhibited similar receptor-mediated activity to that of ␣ s . Functional fluorescent ␤ 1 ␥ 7 dimers were produced by fusing an amino-terminal yellow fluorescent protein (YFP) fragment to ␤ 1 (YFP-N-␤ 1 ) and a carboxyl-terminal YFP fragment to ␥ 7 (YFP-C-␥ 7 ). When expressed together, YFP-N-␤ 1 and YFP-C-␥ 7 produced fluorescent signals in the plasma membrane that were not seen when the subunits were expressed separately. Isoproterenol stimulation of cells co-expressing ␣ s -CFP, YFP-N-␤ 1 /YFP-C-␥ 7 , and the ␤ 2 -adrenergic receptor (␤ 2 AR) resulted in internalization of both fluorescent signals from the plasma membrane. Initially, ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 stained the cytoplasm diffusely, and subsequently they co-localized on vesicles that exhibited minimal overlap with ␤ 2 AR-labeled vesicles. Moreover, internalization of ␤ 2 AR-GFP, but not ␣ s -CFP or YFP-N-␤ 1 /YFP-C-␥ 7 , was inhibited by a fluorescent dominant negative dynamin 1 mutant, Dyn1(K44A)-mRFP, indicating that the G s subunits and ␤ 2 AR utilize different internalization mechanisms. Subsequent trafficking of the G s subunits and ␤ 2 AR also differed in that vesicles labeled with the G s subunits exhibited less overlap with RhoB-labeled endosomes and greater overlap with Rab11-labeled endosomes. Because Rab11 regulates traffic through recycling endosomes, colocalization of ␣ s and ␤ 1 ␥ 7 on these endosomes may indicate a means of recycling specific ␣ s ␤␥ combinations to the plasma membrane.
Multiple G protein signaling pathways operate in individual cells to maintain homeostasis and to bring about responses to external stimuli such as growth and differentiation. An important, but unresolved issue is how the specificity of these pathways is maintained among so much complexity. 23 G protein ␣ subunits, 5 ␤ subunits, and 12 ␥ subunits have been identified in mammals (1), which could give rise to more than 1300 combinations. However, inactivation of specific G protein subunits in vivo using antisense (2)(3)(4)(5)(6) and ribozyme (7,8) strategies has demonstrated a remarkable specificity of interaction between receptors, ␣␤␥ combinations, and effectors. For instance, ribozyme-mediated suppression of ␥ 7 in HEK-293 cells specifically reduces expression of ␤ 1 and disrupts activation of G s by ␤-adrenergic and D 1 dopamine receptors, but not by prostaglandin E 1 and D 5 dopamine receptors (8,9). Moreover, knockout of ␥ 7 in the mouse results in behavioral changes and reductions in the level of ␣ olf in the striatum (10).
One mechanism by which signaling specificity appears to be regulated is at the level of subcellular compartmentalization, which can facilitate or impair interactions between proteins expressed in the same cell (11,12). However, in the case of protein complexes such as G s , for which the localization patterns of the ␣ and ␤␥ subunits have been reported to change upon activation, it is not clear how specificity can be maintained. The G s subunits associate with the plasma membrane as a result of fatty acid modifications and association with each other. Targeting of ␤ subunits to the plasma membrane requires association with prenylated ␥ subunits (13) and is facilitated by association with ␣ subunits (14). Similarly, ␣ s attaches to the plasma membrane as a result of amino-terminal palmitoylation (15,16) and association with ␤␥ (17). Activation of G s results in depalmitoylation of ␣ s (18), and studies using immunohistochemistry (19) and an ␣ s -GFP 1 fusion protein (20) have demonstrated activation-dependent movement of ␣ s from the plasma membrane to the cytoplasm. Activation-dependent changes in ␤␥ localization have not been imaged in cells, but subcellular fractionation indicated that ␤␥ redistributed from the plasma membrane to low density microsomes upon stimulation of ␤-adrenergic receptors (21). In the face of these localization changes, it is not clear how specific ␣ s ␤␥ combinations can be preserved throughout multiple signaling cycles.
To address this issue, we have performed real time imaging of a G s heterotrimer, ␣ s ␤ 1 ␥ 7 , which mediates signaling from the ␤ 2 AR to adenylyl cyclase (7,8), in isoproterenol-stimulated HEK-293 cells. ␣ s was visualized using an internally tagged ␣ s -CFP fusion protein that has comparable activity to that of ␣ s , whereas ␤ 1 and ␥ 7 were imaged exclusively in the form of ␤ 1 ␥ 7 complexes using the strategy of BiFC (22). BiFC involves the production of a fluorescent signal by two nonfluorescent fragments of YFP when they are brought together by interactions between proteins fused to each fragment. When expressed * This work was supported by National Institutes of Health Grant GM50369. 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. together, fusion proteins consisting of an amino-terminal YFP fragment fused to ␤ 1 and a carboxyl-terminal YFP fragment fused to ␥ 7 produce a fluorescent signal that is not obtained with either subunit alone (23). Our results indicate that both ␣ s and ␤ 1 ␥ 7 internalize upon activation, showing significant colocalization on intracellular vesicles that are distinct from ␤ 2 AR-labeled vesicles. Moreover, internalization of the ␤ 2 AR, but not ␣ s or ␤ 1 ␥ 7 , was inhibited by a dominant negative dynamin 1 mutant, indicating that the G protein subunits dissociate from the receptor upon activation, and utilize a different mechanism for internalization from that of the receptor. Compared with vesicles labeled with the ␤ 2 AR, vesicles labeled with ␣ s and ␤ 1 ␥ 7 exhibited less overlap with RhoB-labeled endosomes and greater overlap with Rab11-labeled endosomes. Overlap of ␣ s and ␤ 1 ␥ 7 with each other on these Rab11-labeled endosomes may indicate a means of recycling specific ␣ s ␤␥ combinations to the plasma membrane, because Rab11 regulates trafficking through the pericentriolar recycling endosome (24). This approach of simultaneously imaging the ␣ and ␤␥ components of heterotrimeric G proteins in live cells will have many applications in elucidating the roles of specific ␣␤␥ combinations in particular signaling pathways.

EXPERIMENTAL PROCEDURES
Production of Fluorescent Fusion Proteins-␣ s -CFP in the expression vector pcDNAI/Amp (Invitrogen) was generated from the long splice variant of rat ␣ s cDNA (25) containing the EE epitope (26,27) and ECFP (Clontech) containing a substitution of His for Asn 164 (28). BamHI sites in the ␣ s cDNA were removed and ␣ s residues 73-84 were then replaced with Gly-Ser, adding a unique BamHI site. ECFP was amplified by a PCR in which BamHI sites and linker sequences (Ser-Gly-Gly-Gly-Gly-Ser) were added at each end and then subcloned into the BamHI site in ␣ s , so that ECFP was placed between Gly 72 and Asp 85 (Fig. 1A). ␣ s -GFP (20) in the expression vector pcDNA3.1 (Invitrogen) was kindly provided by Mark Rasenick. Production of YFP-N-␤ 1 and YFP-C-␥ 7 in pcDNAI/Amp was described previously (23). ␤ 2 AR-GFP, a fusion of EGFP to the carboxyl terminus of the ␤ 2 AR in the Clontech vector, pEGFP-N1, was kindly provided by Gerda Breitwieser. ␤ 2 AR-CFP was produced from ␤ 2 AR-GFP and pECFP-N1 (Clontech). ␤ 2 AR-GFP was digested with BamHI and NotI to remove EGFP, which was replaced with a BamHI/NotI fragment from pECFP-N1 encoding ECFP. Dyn1(K44A)mRFP was produced from dyn1(K44A)GFP (29) (kindly provided by Pietro De Camilli) and mRFP (30) (kindly provided by Roger Tsien). Dyn1(K44A)GFP (in the Clontech vector pEGFP-N1) was digested with BamHI and NotI to remove EGFP, which was replaced by a BamHI/NotI fragment encoding mRFP. CFP-Rab11 and YFP-Rab11 (31), fusions of Rab11 to the carboxyl termini of ECFP and EYFP, in the Clontech vectors, pECFP-C1 and pEYFP-C1, respectively, were kindly provided by Marino Zerial.
pECFP-Mem and pEYFP-Mem, encoding fusions to the amino termini of ECFP and EYFP, respectively, of the amino-terminal 20 residues of GAP-43, which leads to targeting primarily to the plasma membrane (32,33), were obtained from Clontech. The fusion proteins are referred to as CFP-Mem and YFP-Mem. pECFP-Endo and pEYFP-Endo, encoding fusions of ECFP and EYFP, respectively, to the amino terminus of the human RhoB GTPase, which localizes to early endosomes (34) and multivesicular bodies, a prelysosomal compartment (35), were obtained from Clontech. The fusion proteins are referred to as CFP-RhoB and YFP-RhoB and vesicles labeled with these markers are referred to as RhoB-labeled endosomes.
Transient Expression and Assay for cAMP Accumulation-HEK-293 cells (ATCC, CRL-1573) (10 6 per 60-mm dish) were transfected with plasmids as described in the legend to Fig. 1 using 10 l of Lipo-fectAMINE 2000 reagent (Invitrogen). The cells were labeled with [ 3 H]adenine and intracellular cAMP levels were determined as described previously (36) in the presence of 1 mM of the phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine in the presence or absence of 10 M UK-14,304.
Membrane Preparations and Immunoblots-HEK-293 cells (12.5 ϫ 10 6 per 150-mm dish) were transfected with 75 g of plasmid using DEAE dextran (37). 48 h after transfection, membranes were prepared as described (36). 25 g of membrane proteins were resolved by SDSpolyacrylamide electrophoresis (10%), transferred to nitrocellulose, and probed with a monoclonal antibody to the EE epitope (26) to compare ␣ s and ␣ s -CFP or with a polyclonal antibody to GFP (A.v. Peptide Antibody, Clontech) to compare ␣ s -CFP and ␣ s -GFP. The antigen-antibody complexes were detected according to the ECL Western blotting protocol (Amersham Biosciences). Chemiluminescence was imaged using a Lumi-Imager (Roche Applied Science).
Imaging of Fluorescent Fusion Proteins-HEK-293 cells were plated at a density of 10 5 cells per well on Lab-Tek II, 4-well chambered coverslips and transiently transfected using 0.25 l of LipofectAMINE 2000 reagent. Plasmids were transfected using the following amounts: ␣ s and ␣ s -CFP, 0.15 g; ␤ 1 , YFP-N-␤ 1 , ␥ 7 , and YFP-C-␥ 7 , 0.075 g; ␤ 2 AR, ␤ 2 AR-GFP, and ␤ 2 AR-CFP, 0.05 g; mRFP, CFP-Mem, YFP-Mem, CFP-RhoB, YFP-RhoB, CFP-Rab11, and YFP-Rab11, 0.0125 g; Dynamin-1-K44A-mRFP, 0.3 g. Cells were imaged 2 days after transfection at ϫ63 using a Zeiss Axiovert 200 fluorescence microscope under the control of IPLab software (Scanalytics) as described (23). Using the motorized x-y-z stage, time course images of cells located at 5-6 positions in the well were collected simultaneously. Cells selected for imaging expressed all of the labeled proteins, had a clearly delineated plasma membrane, and had a region of cytoplasm that was generally free of vesicles or other structures. Individual exposure times were optimized for each cell and color channel. Images for each color channel and DIC were collected at each position in the well every 60 s. Following the second time point, a stimulus of 200 l of 40 M isoproterenol was added to the well, resulting in a final concentration of 10 M, and images were collected for 30 min. For each experimental condition, cells were imaged from plates transfected on 3 different days.
Image Analysis-Time course images were analyzed using IPLab software (Scanalytics Inc.). Before analysis, instrument background was subtracted from the images and corrections were made for bleaching so that the average intensity of the image was constant throughout the time course (see Supplemental Material). The in-focus membrane and vesicle features visible in the images were superimposed on a varying background of diffuse intensity because of out of focus, as well as soluble labeled protein. The images were processed to remove the low resolution diffuse background to produce a high resolution image that was used to identify groups of adjacent pixels or "segments" that corresponded to cellular features such as the plasma membrane and intracellular vesicles (see Supplemental Material). To distinguish between signals in the plasma membrane and in the cytoplasm, a "border" centered on the plasma membrane, 6 -10 pixels wide (0.6 -1.0 m), was drawn around the edge of the cell using a Cintiq pen based display screen (Wacom).
Changes in the plasma membrane intensity of labeled proteins were measured in cells co-expressing a membrane marker (YFP-Mem or CFP-Mem) that was used to segment membrane pixels and correct for intensity changes due to changes in cell shape. A segment of pixels covering a length of the plasma membrane was identified from the high resolution image of the membrane marker, using a portion of the border as a mask (Fig. 5). The average intensities of these pixels in the background-and bleach-corrected images of the labeled protein and membrane marker were determined. The membrane marker intensity values were normalized to a starting value of one and the labeled protein intensity values were divided by the normalized membrane marker values. This corrected for changes in cell shape during the time course, because the distribution of the membrane marker did not change in response to agonist stimulation. The corrected labeled protein intensities were normalized to a starting value of one and averaged with values from multiple cells. Images of unstimulated cells analyzed in this manner showed no response or drift in signal (Fig. 3A). Cells were designated as non-responders if the plasma membrane intensity of the labeled protein dropped less than 3% during the 30 min time course.
Labeled intracellular vesicles within the inside edge of the cell border were segmented and the areas of these segments were calculated using the high resolution images. The area of overlap between vesicles labeled with two different proteins was defined as the area of pixels that was segmented for both proteins. Because the images were acquired sequentially, vesicle movement reduced the measured overlap for vesicles that contained both proteins. Consequently, overlap values of 50 -70% were obtained for images that appeared to have very similar vesicle labeling, such as those labeled with ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 (Figs. 2 and 3), or with ␤ 2 AR-GFP and CFP-RhoB (Fig. 6).
Changes in the intensities of labeled proteins in the cytoplasm were made in cells co-expressing mRFP as a cytoplasm marker to correct for intensity changes due to changes in cell shape. In addition to drawing the cell border, the nucleus was traced using the DIC image. Vesicle pixels for the labeled proteins were segmented as described above. The cytoplasm segment was defined as pixels within the inside edge of the cell border, outside of the nucleus, and not part of a vesicle segment. The average intensities of the cytoplasm segments in the backgroundand bleach-corrected images of the labeled protein and the cytoplasm marker were determined. The cytoplasm marker intensity values were normalized to a starting value of one. The labeled protein intensity values were divided by the normalized cytoplasm marker values to correct for changes in cell shape during the time course, because the distribution of the cytoplasm marker did not change in response to agonist stimulation. The corrected labeled protein intensities were normalized to a starting value of one and then averaged with values from multiple cells.

RESULTS
Production of a Functional Fluorescent G Protein ␣ s Subunit-Because the amino and carboxyl termini of G protein ␣ subunits are important for interactions with receptors, effectors, the G protein ␤␥ subunits, and the plasma membrane (38 -40), we attempted to produce a functional ␣ s -GFP by inserting GFP into ␣ s at the position corresponding to an insertion site that produced a functional ␣ q -GFP (41), but this fusion protein did not have activity (data not shown). We then used a random insertion approach that utilized a synthetic transposon containing GFP within the Tn5 transposon. Surprisingly, the most functional fluorescent fusion protein obtained contained the GFP within the ␣A helix of ␣ s , whereas several insertions in exposed loops were not functional. This functional fusion protein was not expressed as well as ␣ s was, but, when corrected for expression level, had comparable activity to that of ␣ s (42).
In parallel with the random insertion approach, we produced an ␣ s -CFP fusion (Fig. 1A), based on a report that inserting GFP in the ␣1/␣A loop, the site of alternative splicing in ␣ s , leaves function intact (20). ␣ s -CFP differs from the previously published ␣ s -GFP construct in that it contains a Ser-Gly-Gly-Gly-Gly-Ser linker on each side of the inserted CFP. We found that such a linker was essential for producing an ␣ q -GFP fusion that had the same expression and activity levels as those of ␣ q (41). The expression levels (Fig. 1B) and activities (Fig. 1C) of ␣ s and ␣ s -CFP were similar, and ␣ s -CFP produced a fluorescent signal in the plasma membranes of HEK-293 cells (Fig. 2). As with ␣ q -GFP, the Ser-Gly-Gly-Gly-Gly-Ser linkers appear to be important for activity, because the activity of ␣ s -CFP was greater than that of a comparable ␣ s -GFP fusion lacking these linkers (20) (Fig. 1D), although the expression levels of the two fusion proteins were similar (data not shown).
Production of a Functional Fluorescent G Protein ␤ 1 ␥ 7 Complex-We applied the strategy of BiFC (22) to visualize ␤ 1 ␥ 7 dimers (23). BiFC involves the production of a fluorescent signal by two nonfluorescent fragments of YFP when they are brought together by interactions between proteins fused to each fragment. Briefly, a functional fluorescent ␤ 1 ␥ 7 dimer was produced by fusing an amino-terminal YFP fragment (residues 1-158, referred to as YFP-N) to the amino terminus of ␤ 1 and a carboxyl-terminal YFP fragment (residues 159 -238, referred to as YFP-C) to the amino terminus of ␥ 7 . When expressed together, YFP-N-␤ 1 and YFP-C-␥ 7 produced a fluorescent signal in the plasma membrane (Fig. 2) that was not seen when either subunit was expressed alone (23). Functionality of YFP-N-␤/YFP-C-␥ complexes was demonstrated by the ability to potentiate activation of adenylyl cyclase by ␣ s in COS-7 cells (23). The BiFC method assures exclusive visualization of ␤ 1 ␥ 7 dimers rather than individual subunits and leads to more selective labeling of the plasma membrane than is obtained with individually tagged subunits, which when co-expressed, are not functional (23).
Stimulation of the ␤ 2 -AR Leads to Internalization of Both ␣ s and ␤ 1 ␥ 7 and Substantial Co-localization of These G s Subunits on Vesicles-To visualize and investigate the regulation of the subcellular localization patterns of the subunits of G s during signaling, we imaged ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 after agonist stimulation in transfected HEK-293 cells. Stimulation of cells co-expressing ␣ s -CFP, YFP-N-␤ 1 /YFP-C-␥ 7 , and the ␤ 2 AR with the ␤-adrenergic agonist, isoproterenol, resulted in internalization of both ␣ s -CFP (red in Fig. 2) and YFP-N-␤ 1 / YFP-C-␥ 7 (green in Fig. 2) from the plasma membrane. Ini-tially, upon leaving the plasma membrane, ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 exhibited a diffuse staining pattern in the cytoplasm (Fig. 2, 3 min). Subsequently, the signals for these G s subunits exhibited significant co-localization (in yellow) on intracellular vesicles (Fig. 2, 24 min).
To quantify the distribution and movements of the labeled proteins following stimulation, we developed image analysis protocols to measure the intensity, area, and overlap of labeled cellular features. The pixels in each time course image that corresponded to a specific cellular feature such as the plasma membrane or intracellular vesicles were marked or "segmented" and analyzed as described under "Experimental Procedures." The amounts of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 signal that moved out of the plasma membrane upon stimulation were quantified in cells co-expressing ␣ s -CFP or YFP-N-␤ 1 / YFP-C-␥ 7 and a plasma membrane marker (YFP-Mem or CFP- Mem) to identify plasma membrane pixels and to control for changes in cell shape during the responses. The plasma membrane signals for ␣ s -CFP (Fig. 3A, filled circles) and YFP-N-␤ 1 / YFP-C-␥ 7 (Fig. 3A, filled squares) decreased with the same kinetics and the responses were essentially complete by 6 min. In the absence of stimulation, changes in the plasma membrane intensity of ␣ s -CFP (Fig. 3A, open circles) and YFP-N-␤ 1 / YFP-C-␥ 7 (Fig. 3A, open squares) did not occur.
The amounts of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 signal that stained the cytoplasm diffusely were quantified relative to that of a cytoplasm marker (mRFP) and were determined for intracellular pixels, excluding vesicles that became labeled and the nucleus (see "Experimental Procedures"). This analysis demonstrated that the cytoplasm signals for both ␣ s -CFP (Fig. 3B,  circles) and YFP-N-␤ 1 /YFP-C-␥ 7 (Fig. 3B, squares) increased upon stimulation at rates similar to that at which they decreased in the plasma membrane. Labeling of vesicles with ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 was quantified by determining the per cell areas of the labeled vesicles and the extent to which vesicles labeled with the two signals overlapped (Fig.  3C). The total areas of vesicles labeled with either ␣ s -CFP (Fig.  3C, circles) or YFP-N-␤ 1 /YFP-C-␥ 7 (Fig. 3C, squares) increased with the same kinetics and to the same extent. Approximately 50% of the vesicle areas contained both signals (Fig. 3C, triangles). This is an underestimate of the overlap of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 on vesicles, because vesicle movement in between acquisition of the sequential images reduced the measured overlap.
Stimulus-dependent decreases in the plasma membrane intensity of ␣ s -CFP were somewhat greater in magnitude than those of YFP-N-␤ 1 /YFP-C-␥ 7 (Fig. 3A) and the increases in cytoplasm signal were greater for ␣ s -CFP than for YFP-N-␤ 1 / YFP-C-␥ 7 (Fig. 3B). However, because the stimulus-dependent changes in the intensities of the two signals in these locations were determined relative to each of their starting values and do not represent absolute amounts of the proteins, these changes in the intensities of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 are not directly comparable. Nevertheless, the similar timing of the ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 internalization responses (Fig. 3) and the similar localization patterns of these G protein subunits in the diffuse cytoplasm and on vesicles (Fig. 2) suggest that the activated populations may traffic together.
Vesicles Labeled with ␣ s and ␤ 1 ␥ 7 Exhibit Minimal Overlap with ␤ 2 AR-labeled Vesicles-The trafficking patterns of the G s subunits were compared directly with that of the ␤ 2 AR in cells that co-expressed the labeled proteins. In cells co-expressing ␣ s -CFP, unlabeled ␤ 1 ␥ 7 , and ␤ 2 AR-GFP, there was minimal overlap between vesicles labeled with ␣ s -CFP (red in Fig. 4A) and those labeled with ␤ 2 AR-GFP (green in Fig. 4A). Image analysis of these cells demonstrated that similar amounts of vesicles, as defined by their per cell areas, became labeled with ␣ s -CFP (Fig. 4B, filled circles) and ␤ 2 AR-GFP (Fig. 4B, open  circles), despite the minimal overlap of the two vesicle populations (Fig. 4B, filled triangles). Similarly, in cells that co-ex-pressed unlabeled ␣ s , YFP-N-␤ 1 /YFP-C-␥ 7 , and ␤ 2 AR-CFP, there was minimal overlap between vesicles labeled with YFP-N-␤ 1 /YFP-C-␥ 7 (red in Fig. 4C) and those labeled with ␤ 2 AR-CFP (green in Fig. 4C). Again, image analysis showed that similar amounts of vesicles became labeled with YFP-N-␤ 1 / YFP-C-␥ 7 (Fig. 4D, filled circles) and ␤ 2 AR-CFP (Fig. 4D, open  circles), although the two vesicles populations did not colocalize (Fig. 4D, filled triangles). In addition, the fluorescent G s subunits (filled circles in Fig. 4, B and D) labeled vesicles more rapidly than the fluorescent ␤ 2 AR (open circles in Fig. 4,  B and D).
Receptor-stimulated Internalization of the ␤ 2 AR, but Not ␣ s or ␤␥, Is Inhibited by a Dominant Negative Dynamin 1 Mutant-The minimal overlap of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 vesicles with ␤ 2 AR-labeled vesicles suggested that the internalization pathways of these G s subunits might differ from that of the ␤ 2 AR. To test this hypothesis, we determined the effects on G s subunit and ␤ 2 AR internalization of a dominant negative dynamin 1 mutant fused to mRFP, Dyn1(K44A)-mRFP. The amounts of ␤ 2 AR-GFP, ␣ s -CFP, and YFP-N-␤ 1 /YFP-C-␥ 7 in the plasma membrane (relative to CFP-Mem or YFP-Mem) before and after stimulation with isoproterenol were determined in the presence and absence of Dyn1(K44A)-mRFP. Dyn1(K44A)-mRFP effectively blocked internalization of ␤ 2 AR-GFP (Fig. 5,  A, B, and G), confirming previous studies (43,44). However, Dyn1(K44A)-mRFP did not reduce either the magnitude or the frequency of the internalization responses of ␣ s -CFP (Fig. 5, C,  D, and H) or YFP-N-␤ 1 /YFP-C-␥ 7 (Fig. 5, E, F, and I).
Vesicles Labeled with the ␤ 2 AR Exhibit Greater Co-localization with RhoB-labeled Endosomes Than Do Vesicles Labeled with ␣ s or ␤ 1 ␥ 7 -The initial diffuse labeling of the cytosol by both ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 after stimulation and the inability of Dyn1(K44A)-mRFP to block their internalization from the plasma membrane suggested that these G s subunits may not directly internalize on vesicles. However, the subsequent labeling of vesicles by both ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 suggested that vesicular trafficking may play a role in effector modulation and/or the recycling of G s back to the plasma membrane. Although G s subunit-labeled vesicles did not co-localize well with ␤ 2 AR-labeled vesicles (Fig. 4), we investigated the possibility that the G s subunit-labeled vesicles might be endosomes, because diverse internalization pathways can converge on endosomes (45).
To explore the nature of the vesicles labeled by the G s subunits, stimulus-dependent overlap of the G s subunits and the ␤ 2 AR with RhoB-labeled endosomes was measured. RhoB has been reported to associate with early endosomes (34) and multivesicular bodies, a prelysosomal compartment thought to be involved in sorting internalized receptors for degradation (35). Vesicles labeled with ␤ 2 AR-GFP that formed after stimulation occupied similar per cell areas as did CFP-RhoB-labeled endosomes, which exhibited unchanged areas upon stimulation (Fig. 6, A and B). After stimulation, ϳ50% of the RhoB-labeled vesicle area overlapped with that of ␤ 2 AR-labeled vesicles (  6, A and B). As with ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 vesicles (Figs. 2 and 3), the actual overlap was underestimated, because of the sequential acquisition of images. In contrast, although vesicles labeled with ␣ s -CFP or YFP-N-␤ 1 /YFP-C-␥ 7 occupied similar per cell areas, respectively, as did YFP-RhoB-and CFP-RhoB-labeled endosomes, the G s subunit vesicles co-localized with only ϳ25% of the RhoB-labeled endosome area (Fig.  6, C-F).
Vesicles Labeled with ␣ s and ␤ 1 ␥ 7 Exhibit Greater Co-localization with Rab11-labeled Endosomes Than Do Vesicles Labeled with the ␤ 2 AR-Some of the internalized ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 localized to the Golgi apparatus region, but did not overlap directly with the trans-medial Golgi region, labeled with YFP or CFP, respectively, fused to the amino-terminal 81 residues of ␤1,4-galactosyltransferase (46, 47) (not shown). Similar results were obtained using a GFPtagged marker for the cis-Golgi region, GFP-GM130 (48, 49) (not shown). To further investigate the significance of the appearance of ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 in this cellular region, we tested for overlap with Rab11, which localizes to recycling endosomes and post-Golgi membranes, including the trans-Golgi network, and secretory vesicles (24,50), and plays a role in regulating recycling through the slow perinuclear recycling endosome route (24).
CFP-and YFP-Rab11 (31) selectively labeled a small perinuclear region (Fig. 7, A, C, and E). Segmentation analysis demonstrated that the cellular area occupied by Rab11-labeled vesicles (open circles in Fig. 7) was much smaller than that occupied by Rho-B-labeled vesicles (open circles in Fig. 6). However, we observed significant stimulus-dependent overlap in the perinuclear region between ␣ s -CFP and YFP-Rab11 (Fig. 7,  A and B). Similar overlap was obtained with YFP-N-␤ 1 /YFP-C-␥ 7 and CFP-Rab11 (Fig. 7, C and D). Approximately 40% of the Rab11-labeled vesicles co-localized with G s -subunit-labeled vesicles by the ends of the time courses. The total area of overlap between Rab11-labeled vesicles and those labeled with G s subunits (filled triangles in Fig. 7, B and D) was less than that between RhoB-and G s subunit-labeled vesicles (filled triangles in Fig. 6, D and F). However, a greater percentage of the Rab11-labeled vesicles than the RhoB-labeled vesicles colocalized with G s subunit-labeled vesicles. In contrast, there was minimal overlap between ␤ 2 AR-GFP and CFP-Rab11 (Fig.  7, E and F), confirming previous studies showing that the ␤ 2 AR recycles to the plasma membrane directly from early endosomes via a Rab4-dependent mechanism (51), rather than through the slow perinuclear recycling endosome route. DISCUSSION By imaging functional fluorescent fusion proteins in living cells, we have observed internalization of both the ␣ and ␤␥ components of the G s heterotrimer, ␣ s ␤ 1 ␥ 7 , upon stimulation of the ␤ 2 AR. ␣ s -CFP, in which CFP was inserted into ␣ s at an internal site, exhibited similar activity to that of ␣ s . ␤ 1 and ␥ 7 were imaged exclusively in the form of ␤ 1 ␥ 7 complexes by means of BiFC (22), using fusion proteins, YFP-N-␤ 1 and YFP-C-␥ 7 , that only produced fluorescence and functional activity when expressed together (23). Initially, after leaving the plasma membrane, ␣ s -CFP and YFP-N-␤ 1 /YFP-C-␥ 7 labeled the cytoplasm diffusely, with the same kinetics. Subsequently, they exhibited a high degree of co-localization on vesicles that exhibited minimal overlap with ␤ 2 AR-labeled vesicles. In gen-eral, our results confirm previous imaging studies on the trafficking of ␣ s . Moreover, to our knowledge, our study represents the first visualization of stimulus-dependent internalization of ␤␥. In addition, our recently developed methodologies enabled direct comparisons to be made between the trafficking patterns of ␣ s and ␤ 1 ␥ 7 and between these G s subunits and the ␤ 2 AR, to investigate potential regulatory mechanisms, and to address the issue of how the specificity of ␣ s and ␤␥ interactions is maintained during signaling cycles.
A previous study using immunohistochemistry demonstrated activation-dependent redistribution of ␣ s to the cytosol in HEK-293 cells stably transfected with the ␤ 2 AR and HAtagged ␣ s (19). The staining pattern for internalized ␣ s was punctate, but could be distinguished from that of the ␤ 2 AR. In a study using a different ␣ s -GFP construct, isoproterenol stimulation of COS-1 cells resulted in a partial decrease in association of ␣ s -GFP with some regions of the plasma membrane as well as insertion of ␣ s -GFP at other membrane sites (20). Differences between our results with ␣ s -CFP expressed in HEK-293 cells and these previous results with ␣ s -GFP could be because of the greater activity of the ␣ s -CFP construct and/or the different cell types examined. Stimulus-dependent ␤␥ internalization has not been imaged previously, but using an enzyme-linked immunosorbent assay, redistribution of ␤␥ from the plasma membrane to low density microsomes was observed in response to isoproterenol stimulation (21).
Our results indicate that G s and the ␤ 2 AR dissociate upon hormonal stimulation and utilize different mechanisms for internalization from the plasma membrane, because internalization of the ␤ 2 AR, but not ␣ s or ␤ 1 ␥ 7 , can be blocked by a dominant negative dynamin 1 mutant. The inability of Dyn1(K44A)-mRFP to block internalization of ␣ s extends a previous study in which hypertonic sucrose, which inhibits the formation of clathrin-coated pits, blocked internalization of the ␤ 2 AR, but not ␣ s (19). Because dynamin is involved in both clathrin-mediated endocytosis (52,53) and endocytosis of caveolae (54), our results demonstrate that internalization of ␣ s and ␤ 1 ␥ 7 is independent of both of these mechanisms. However, because ␣ s and ␤ 1 ␥ 7 subsequently exhibit partial co-localization with RhoB-and Rab11-labeled endosomes, these G s subunits appear to join the endosomal pathway by a different and as yet uncharacterized mechanism. There is precedent for alternative pathways from the plasma membrane to endosomes in that endocytosis of the M 2 -muscarinic receptor is independent of clathrin, but involves subsequent transfer to endosomal compartments of the clathrin-dependent pathway (45).
Our observation that both ␣ s and ␤ 1 ␥ 7 leave the plasma membrane upon activation is consistent with previous reports that the combined effects of their plasma membrane targeting motifs are important for targeting each of them to the plasma membrane (14,55,56). Activation-dependent depalmitoylation of ␣ s is associated with its internalization from the plasma membrane to the cytoplasm (18,19), and mutant forms of ␣ s containing alternative plasma membrane-targeting motifs do not show activation-dependent internalization (57). It is possible that depalmitoylation of ␣ s is sufficient to result in movement of ␤␥ to the cytoplasm, based on the observation that efficient plasma membrane targeting of ␤ 1 ␥ 2 expressed in HEK-293 cells required either co-expression of ␣ s or introduction of a palmitoylation site into ␥ 2 (14).  tion of cytosolic ␣ s and ␤␥ with vesicles could be due to repalmitoylation of ␣ s upon its deactivation. Another potential contributing factor to the internalization of both ␣ s and ␤␥ could be their transient dissociation upon stimulation. Based on in vitro studies, the G protein ␣ and ␤␥ subunits have been thought to dissociate upon activation, but recent studies using fluorescence resonance energy transfer have produced differing results as to whether or not this actually takes place in vivo (58 -60). Because each of these studies utilized a different G protein heterotrimer, it is possible that some, but not all G proteins dissociate upon activation. Future fluorescence resonance energy transfer studies will address the issue of whether or not the G s subunits dissociate upon activation. In addition, it will also be interesting to determine whether ␤␥ internalization requires internalization of the associated ␣ subunit. For instance, when G q is activated by stimulation of the ␣ 2a -adrenergic receptor, ␣ q does not internalize (41).
Localization of ␣ s and ␤ 1 ␥ 7 on vesicles may be important for certain G s effector functions. For instance, there is evidence that both the ␣ and ␤␥ subunits of G s may play roles in regulating the fusion of early endosomes (61,62). In addition, the G s subunits have been implicated in the regulation of transcytosis of the polymeric immunoglobulin receptor (63,64), and activation of ␣ s stimulates transport of influenza hemagglutinin protein from the trans-Golgi network to the apical surface of Madin-Darby canine kidney cells (65). Additional characterization of the vesicles labeled by ␣ s and ␤ 1 ␥ 7 may help to elucidate further the roles of these subunits in vesicular trafficking. There is clearly a population of vesicles labeled with both ␣ s and ␤ 1 ␥ 7 that does not overlap with either RhoB-or Rab11labeled endosomes. Future experiments using other markers will attempt to identify these vesicles.
Localization of ␣ s and ␤ 1 ␥ 7 to Rab11-labeled vesicles may indicate that these G s subunits recycle to the plasma membrane via the slow perinuclear recycling endosome route, which Rab11 plays a role in regulating (24). Slow recycling of certain G protein-coupled receptors, such as the vasopressin V2, somatostin 3, and CXC chemokine 2 receptors, is blocked by overexpression of dominant negative Rab11 mutants (66 -68). In contrast, the ␤ 2 AR appears to recycle rapidly from early endosomes and can be blocked by a dominant negative Rab4 mutant (51). We observed little overlap between vesicles labeled with the ␤ 2 AR and Rab11, consistent with a previous study showing that the ␤ 2 AR exhibits relatively little co-localization with Rab11 unless cells are treated with the proton pump inhibitor, bafilomycin A 1 (69), which raises endosome and lysosome pH and blocks the formation of vesicles that traffic between endosomes to lysosomes.
The high degree of co-localization of ␣ s and ␤ 1 ␥ 7 on intracellular vesicles and the overlap of some of these vesicles with Rab11-labeled endosomes may indicate a means of recycling specific ␣ s ␤␥ combinations to the plasma membrane. Although different ␤␥ combinations generally exhibit similar abilities to modulate the activities of effectors such as adenylyl cyclase (70), phospholipase C (71), and GIRK channels (72), reconstitution experiments have indicated clear differences in the ␣␤␥ combinations that are preferred by particular receptors (73)(74)(75)(76)(77)(78). Moreover, inactivation of specific G protein subunits in vivo using antisense (2-6) and ribozyme (7,8) strategies has demonstrated a remarkable specificity of interaction between receptors, ␣␤␥ combinations, and effectors. Future studies will investigate whether ␣ s and ␤ 1 ␥ 7 recycle to the plasma membrane together by following their localization patterns upon removal of the stimulus and test for a potential regulatory role of Rab11 in this process. In addition, we will investigate whether other ␣ s ␤␥ combina-tions can respond to stimulation of the ␤ 2 AR and examine the ␣ s ␤␥ specificities of other G s -coupled receptors.