Dissociated GαGTP and Gβγ Protein Subunits Are the Major Activated Form of Heterotrimeric Gi/o Proteins*

Background: The nature of the activated form of heterotrimeric Gi/o proteins is unclear. Results: Non-modified Gi/o heterotrimers dissociate upon activation, whereas fluorescently labeled Gi/o heterotrimers dissociate or rearrange depending on fluorescent protein localization. Conclusion: Dissociated GαGTP and Gβγ subunits represent the major activated form of Gi/o proteins. Significance: The identification of the Gi/o protein activated form is crucial for the elucidation of signaling mechanisms and rational drug design. Although most heterotrimeric G proteins are thought to dissociate into Gα and Gβγ subunits upon activation, the evidence in the Gi/o family has long been inconsistent and contradictory. The Gi/o protein family mediates inhibition of cAMP production and regulates the activity of ion channels. On the basis of experimental evidence, both heterotrimer dissociation and rearrangement have been postulated as crucial steps of Gi/o protein activation and signal transduction. We have now investigated the process of Gi/o activation in living cells directly by two-photon polarization microscopy and indirectly by observations of G protein-coupled receptor kinase-derived polypeptides. Our observations of existing fluorescently labeled and non-modified Gαi/o constructs indicate that the molecular mechanism of Gαi/o activation is affected by the presence and localization of the fluorescent label. All investigated non-labeled, non-modified Gi/o complexes dissociate extensively upon activation. The dissociated subunits can activate downstream effectors and are thus likely to be the major activated Gi/o form. Constructs of Gαi/o subunits fluorescently labeled at the N terminus (GAP43-CFP-Gαi/o) seem to faithfully reproduce the behavior of the non-modified Gαi/o subunits. Gαi constructs labeled within the helical domain (Gαi-L91-YFP) largely do not dissociate upon activation, yet still activate downstream effectors, suggesting that the dissociation seen in non-modified Gαi/o proteins is not required for downstream signaling. Our results appear to reconcile disparate published data and settle a long running dispute.


Although most heterotrimeric G proteins are thought to dissociate into G␣ and G␤␥ subunits upon activation, the evidence in the Gi/o family has long been inconsistent and contradictory. The Gi/o protein family mediates inhibition of cAMP production and regulates the activity of ion channels. On the basis of experimental evidence, both heterotrimer dissociation and rearrangement have been postulated as crucial steps of Gi/o protein activation and signal transduction. We have now investigated the process of Gi/o activation in living cells directly by two-photon polarization microscopy and indirectly by observa
Heterotrimeric G proteins (consisting of G␣, G␤, and G␥ subunits) are an important part of the cellular signal transduction system. They serve as transducers and amplifiers of signals from G protein-coupled receptors to intracellular effectors. Activation of G protein-coupled receptors leads to an exchange of a molecule of GDP bound to an inactive G␣ subunit for a molecule of GTP, causing changes in the G protein complex, termed "activation." Both the activated GTP-bound G␣ subunits and the G␤␥ dimers can modulate the activity of effectors such as adenylate cyclase, phospholipase C, or G protein-regulated inward rectifying potassium (GIRK) 2 channels. Hydrolysis of GTP to GDP, because of a GTPase activity of the G␣ subunit, leads to a return of the G protein molecule into the inactive state.
The Gi/o family of G proteins inhibits adenylate cyclase activity (decreasing the intracellular cAMP concentration) and activates GIRK channels (hyperpolarizing the cell membrane). However, it is not clear how activation of Gi/o proteins leads to these downstream signaling events. Despite a number of published studies, even the molecular identity of the activated state of the Gi/o proteins remains disputed (1)(2)(3)(4). One model (5) postulates that the G protein heterotrimer dissociates into a GTP-bound G␣ subunit (G␣ GTP ) and a G␤␥ dimer, whereas another model (6) proposes that the G protein heterotrimer becomes rearranged but does not dissociate upon activation. A combined model (2,7) asserts an equilibrium between dissociated and intact G protein heterotrimers in the active state. It is, however, unclear whether interactions with effectors are mediated by a rearranged, GTP-bound heterotrimer (G␣ GTP G␤␥) or by the dissociated components (G␣ GTP , G␤␥).
The published evidence in support of the distinct models of Gi/o activation is often contradictory. In vitro studies indicate that, similar to other heterotrimeric G proteins, the Gi/o pro-teins dissociate upon activation by the non-hydrolyzable GTP analog GTP␥S (5,8,9). Furthermore, individual purified G protein subunits (including purified G␣i/o subunits) have been shown to interact with effectors associated with Gi/o signaling (5,8). Thus, in vitro evidence suggests that Gi/o proteins dissociate upon activation into a free G␣ GTP subunit and a G␤␥ dimer and that the dissociated subunits mediate downstream signaling (5,10). In contrast, in vivo experiments performed in cell cultures have produced conflicting results. Observations of increases in FRET (11,12) and bioluminescence resonance energy transfer (13) between modified G protein subunits upon G protein activation support heterotrimer rearrangement (rather than dissociation) upon activation. A decrease in FRET between G protein subunits labeled with fluorescent proteins (FPs) (14), internalization of G␤1␥11 dimers (15), and an increase in lateral mobility of FP-labeled G␤␥ dimers in the plasma membrane (16) upon Gi/o protein activation are consistent with heterotrimer dissociation upon activation. Because the in vitro results were obtained under conditions far from natural and the in vivo evidence is inconsistent and has mostly been obtained using G protein heterotrimers modified with two large protein tags (17), better evidence is needed to elucidate the process of Gi/o activation under natural conditions. Inspired by recent advances in imaging of G protein activation, we decided to investigate the molecular nature of the activated state of Gi/o proteins under conditions as close to natural as possible.
To detect interactions between G␣ and G␤␥ subunits, we employed the technique of two-photon polarization microscopy (2PPM), developed recently in our laboratory and described in detail in Ref. 18. The 2PPM technique allows measurements of two-photon linear dichroism (LD) (the differences in two-photon absorption of light of distinct (perpendicular) linear polarizations in fluorescently labeled samples). LD is present in all assemblies of non-randomly oriented fluorophores. We have shown recently that, using 2PPM, LD can be detected in ϳ80% of FP-labeled membrane proteins expressed in mammalian cells, including several G proteins (18). Changes in LD can be used for sensitive observations of changes in protein-protein interactions and conformational changes in proteins. Our mathematical and software tools allow reliable quantification of LD and derivation of quantitative descriptions of molecular processes taking place in living cells (18). Because of the requirement of 2PPM for only a single FP label to observe protein-protein interactions, 2PPM is a promising tool for investigating processes such as G protein activation under conditions close to natural.
In 2PPM, two images of a fluorescent sample are acquired, one with excitation light polarized horizontally in the image and one with excitation light polarized vertically. Differences between the two images signify the presence of LD, caused by presence of fluorophores that are orientationally distributed non-randomly, for example because of their tethering to a cell membrane.
The extent of LD in membrane-tethered fluorophores depends on two factors: the orientation of the observed membrane with respect to the excitation light polarization and the orientation of the fluorophore assembly with respect to the cell membrane. No LD is present in membranes oriented diagonally in 2PPM images because the orientation of the membrane (and fluorophores) with respect to both excitation polarizations is identical. Maximum LD occurs in membranes oriented parallel and perpendicular to the used polarizations (vertically or horizontally in the image). Therefore, we measured LD in sections of cell outline that are oriented almost horizontally or vertically.
For a particular section of the cell outline, the extent and "sign" of LD depends on fluorophore orientation with respect to the membrane. The closer a fluorophore is to being parallel to the cell membrane, the larger the excess of fluorescence it will exhibit with light polarized parallel to the cell membrane. Conversely, the closer a fluorophore is to being perpendicular to the cell membrane, the larger the excess of fluorescence it will exhibit with light polarized perpendicular to the cell membrane. Fluorophores with a random distribution of orientations, or fluorophores at an angle of 52.0°(the so-called "magic angle" for two-photon excitation) with respect to the cell membrane will produce no LD.
The extent of LD can be described by a dichroic ratio (r): the ratio of fluorescence intensities excited by light polarized horizontally (F h ) and vertically (F v ), respectively, in the microscopy image. Because r depends on the orientation of the cell membrane, we characterize constructs by using a maximum dichroic ratio (r max ), calculated as F h /F v in sections of the cell outline oriented close to horizontal in the image and as the inverse ratio in sections of the cell outline oriented close to vertical in the image. This way, fluorophore distributions close to parallel to the cell membrane exhibit r max Ͼ 1, and fluorophore distributions close to perpendicular to the cell membrane show r max Ͻ 1. Wide distributions of fluorophore tilt angles and/or closeness of the mean fluorophore tilt angle to the magic angle show little or no LD (r max ϭ ϳ1). In this work, we rely on observations of LD by the 2PPM technique to infer information on dissociation and rearrangement of G protein complexes.
To verify our 2PPM results, and to extend them to non-modified G proteins, we built on a published observation (19) that fluorescently labeled peptides derived from a G protein-coupled receptor kinase 3 (GRK3) selectively bind free G␤␥ dimers and can, thus, serve as indicators of G protein dissociation. The binding site of the GRK3 C-terminal domain (GRK3ct) on G␤1␥2 overlaps extensively with the binding site of G␣ on G␤1␥2 (17,20,21). Thus, binding of GRK3ct to G␤1␥2 likely occurs only upon dissociation of the heterotrimeric G protein complex (2,20). Interactions of GRK3ct with G␤1␥2 have been observed by resonant energy transfer (FRET and bioluminescence resonance energy transfer) between suitable labels attached to both GRK3ct and G␤1␥2. For G␣o, the GRK3ct -G␤1␥2 interaction has also been observed by monitoring cellular localization of GRK3ct-Venus, allowing the use of nonlabeled G␤1␥2. Observations of GRK3ct-FP translocation, in combination with our imaging and image processing tools, promised to yield evidence of the mechanism of Gi/o protein activation not only in G␣i/o constructs labeled with an FP but also in non-modified G␣i/o subunits.
Our 2PPM and GRK3ct translocation results indicate that non-modified G␣i/o constructs extensively dissociate from G␤␥ partners upon activation. In contrast, some of the FP-tagged G␣i/o constructs investigated remain in the form of a heterotrimer. Interestingly, our results show that even FPtagged constructs that do not dissociate upon activation can activate downstream signaling pathways. Thus, at least in FPmodified constructs, G␣i/o dissociation is not required for downstream signaling. However, the extent of dissociation in non-modified Gi/o proteins indicates that the dissociated G␣ and G␤␥ subunits are the major activated form that mediates downstream signaling in a natural setting.
Cell Culture-HEK293 cells were cultured at 37°C under an atmosphere of 95% air, 5% CO 2 in Dulbecco's modified Eagle's medium with Glutamax I and high glucose (Life Technologies) supplemented with 10% fetal bovine serum. Cells were plated on 8-chamber microscopy slides (-Slides, Ibidi GmbH, Germany) or 6-chamber perfusion microscopy slides (-Slide VI Luer slides, Ibidi GmbH). Transfections were carried out using Lipofectamine 2000 (Life Technologies) according to the protocol of the manufacturer. In cotransfections, we used equimolar amounts of plasmids encoding G␣i/o, G␤1, G␥2, and ␣2aAR. The GRK3ct-encoding constructs were cotransfected at 0.8 equivalents. Microscopy experiments were carried out 48 h after transfection. In experiments involving GRK3ct constructs, we inhibited endogenous Gi/o signaling by treatment with 100 ng/ml of pertussis toxin for 16 h prior to microscopy observations. G protein activation experiments were performed in perfusion slides at room temperature under a continuous flow of peristaltically pumped (Minipuls 3, Gilson, UK), HEPES-buffered Hanks' balanced salt solution (pH 7.4). Norepinephrine ((Ϯ)-norepinephrine (ϩ)-bitartrate salt (NE), Sigma) at a final concentration of 1 M was applied for 30 s.
Fluorescence Microscopy-We conducted all microscopy experiments on a customized laser scanning microscope (iMic2, Till Photonics, Germany) equipped with a titanium: sapphire laser (Chameleon Ultra II with group velocity dispersion compensation, Coherent) using a UApoPlan/IR ϫ60 numerical aperture 1.2 water immersion objective lens (Olympus, Japan). A long-pass dichroic mirror and an emission filter (Q565LP, Chroma) and Brightline 479/40 (Semrock) for CFP and 740DCXR (Chroma) and Brightline 542/27 (Semrock) for YFP separated fluorescence from the excitation laser beam. Fluorescence was detected by a photomultiplier (R6357, Hamamatsu Photonics) equipped with an IR-blocking filter (HQ700SP-2P, Chroma). For each imaging experiment, we quantitatively analyzed at least 10 representative cells.
The method of 2PPM, including the experimental arrangement, is described in detail in Ref. 18. In 2PPM experiments, the direction of polarization of the excitation beam was alternated between horizontal and vertical by a polarization modulator (RPM-2P, Innovative Bioimaging) operated at 100 kHz and synchronized with acquisition of individual pixels by the microscope.
FRET measurements were performed on an iMic2 microscope equipped with a Polychrome IV light source (both Till Photonics) using an UApoPlan/IR ϫ60 numerical aperture 1.2 water immersion objective lens (Olympus). FRET was determined as a ratio of YFP to CFP emission (F YFP /F CFP ) using 438 Ϯ 12 nm of excitation. The CFP signal was recorded using a Brightline 438/24 excitation filter, Brightline 458 long-pass dichroic beam splitter, and Brightline 472/30 emission filter (all Semrock). The YFP signal was recorded with a Brightline 438/24 excitation filter for FRET images or a Brightline 479/40 for direct YFP excitation, a Brightline 509 long-pass dichroic beam splitter, and a Brightline 530/25 emission filter (all Semrock). Images were recorded sequentially with an ImagoQE camera (Till Photonics). The illumination time was set to 500 ms. CFP spillover into the FRET channel and YFP direct excitation upon 438 Ϯ 12 nm of illumination were calculated and subtracted from the FRET signal to give the corrected F YFP / F CFP ratio. For each imaging experiment, we quantitatively analyzed at least 10 representative cells.
Image Processing and Analysis-To display the ratiometric information contained in 2PPM images, we processed the raw images in ImageJ and Polarisϩ (Innovative Bioimaging) (Fig. 1). In short, each raw 2PPM image was deinterleaved into two images containing pixels obtained with horizontally and vertically polarized excitation, respectively. After background subtraction and adjustment for differences in laser beam intensity and pulse duration between the two polarizations, we applied a lookup table designed to express fluorescence intensity by brightness and values of the dichroic ratio by hue. On the color scale used, green indicates an excess of fluorescence excited by vertical polarization (F v ) and red an excess of fluorescence excited by horizontal polarization (F h ). Pure red or green color pixels indicate values of the dichroic ratio exceeding the maximum value set by the color scale.
Quantitative analysis of 2PPM images was carried out as described in Ref. 18 using Matlab scripts developed in-house. Briefly, we calculated the dichroic ratio (r ϭ F h /F v ) and log ratio (log 2 (r)) for each pixel corresponding to the cell outline. We then associated the values of r (and log 2 (r)) with information on cell membrane orientation (angle ). We determined the values of r max (and log 2 (r max )), characteristic of the construct being investigated, by calculating the mean of values of r (and log 2 (r)) for parts of the cell outline close to horizontal ( -/2 Ͻ 3°) and 1/r (and log 2 (1/r)) for parts of the cell outline close to vertical ( Ͻ 3°). Thus, values of r max Ͼ 1 (and log 2 (r max ) Ͼ 0) indicate that ␣ 0 , the mean fluorophore angle with respect to the membrane normal, is larger than the two-photon magic angle (␣ 0 Ͼ 52.0°).
To visualize and quantitate GRK3ct-FP translocation and localization (Figs. 5 and 7), we performed a background subtraction (using an image area outside of a cell), a correction for bleaching, and image thresholding in ImageJ. For visualization, we applied a lookup table designed to express values of fluorescence intensity ratios by hue. We quantitated GRK3ct-FP translocation (Fig. 5, D-G) by manually selecting a cytoplasmic region of a cell and taking a ratio of fluorescence intensities before and during stimulation of the cell with NE. The procedure of quantitating GRK3ct-FP membrane localization (Fig. 7) consisted of selecting the whole cell area by thresholding, measuring the fluorescence intensity (F membrane ) in a manually selected, 250-nm-wide cell outline, and comparing it to the fluorescence intensity in the cytoplasm (F cytoplasm ). To account for low fluorescence intensities at the outer edge of the cell outline, the F membrane /F cytoplasm ratios were normalized by those obtained in cells overexpressing only GRK3ct-Venus.
Quantitative results of both 2PPM and translocation imaging experiments were expressed as mean Ϯ S.E. Statistical significance was evaluated using Student's t test. The normality of the data was tested and confirmed by D'Agostino and Pearson omnibus K 2 normality test.
Electrophysiology-Membrane current recordings were performed by conventional whole cell patch clamp technique (22) using an EPC10 USB amplifier (HEKA Elektronik Dr. Schulze GmbH, Germany). Patch pipettes were pulled from borosilicate glass capillaries (GC150T-10, Harvard Apparatus) using a vertical puller (PC-10, Narishige, Japan) and had a resistance of 2-6 M⍀ when filled with the pipette solution. Data were acquired and analyzed using Patchmaster software (HEKA Elektronik Dr. Schulze GmbH). I GIRK was measured as an inward current using a holding potential of Ϫ90 mV. The internal (pipette) solution composition was 100 mM potassium aspartate, 40 mM KCl, 5 mM MgATP, 10 mM HEPES, 5 mM NaCl, 2 mM EGTA, and 1 mM MgCl 2 (pH 7.3). The external solution composition was 120 mM NaCl, 20 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , and 10 mM HEPES (pH 7.3). The results were presented as mean Ϯ S.E. Statistical significance was evaluated using Student's t test. The normality of the data were tested and confirmed by D'Agostino and Pearson omnibus K 2 normality test.

RESULTS
We investigated, both by 2PPM and GRK3ct-FP translocation, fluorescently labeled G␣i/o subunits (G␣i/o-FP) of two published designs: GAP43-CFP-G␣ (23) and G␣-L91-YFP (11,12). In the GAP43-CFP-G␣ constructs, the N terminus of the native G␣, including a lipidation motif, is replaced by a GAP43 lipidation motif and a cyan fluorescent protein (Fig. 2, A and B, and supplemental Fig. 1). In the G␣-L91-YFP constructs, a yellow fluorescent protein is inserted into the ␣a-␣b loop after the Leu-91 residue (Fig. 2, C and D, and supplemental Fig. 1). Both groups of constructs carry a C351I or C352I mutation, rendering them insensitive to pertussis toxin and allowing their selective activation in the presence of endogenously expressed G␣ subunits. Both construct designs have been shown previously to be functional, judged by their ability to inhibit cAMP production and activate GIRK channels (11,23). Published data on FRET between FP-labeled G␣i/o-L91-YFP and CFP-labeled G␥2 or G␤1 subunits (11,12) have suggested that G␣i-L91-YFP containing heterotrimers did not dissociate during activation, whereas the G␣o-L91-YFP complex dissociated.

Two-photon Polarization Microscopy Allows Observation of the Interactions between Heterotrimeric G Protein Subunits-
To investigate whether 2PPM allows observation of the interactions between FP-tagged G␣i/o subunits and G␤1␥2 dimers, we carried out 2PPM observations of HEK293 cells transfected with G␣i/o-FP constructs both alone and in combination with G␤1 and G␥2 subunits. The results of our 2PPM observations are illustrated in Fig. 2 and summarized in Fig. 3. Briefly, none of the four investigated GAP43-CFP-G␣ constructs showed LD when expressed alone ( Fig. 2A). In contrast, when coexpressed with G␤1 and G␥2, all investigated GAP43-CFP-G␣ constructs exhibited pronounced LD (Figs. 2B and 3, A-D). The results of 2PPM observations of G␣ constructs of the G␣-L91-YFP design were less uniform and revealed unexpected structural differences between the investigated constructs despite their identical design and high DNA and amino acid sequence similarity (repeatedly verified by DNA sequencing). When expressed alone, the G␣i2-L91-YFP construct showed no LD, whereas the G␣i1-L91-YFP, G␣i3-L91-YFP, and G␣o-L91-YFP constructs of the same design exhibited significant LD (Fig. 2C). When coexpressed with G␤1 and G␥2, the G␣i2-L91-YFP construct again showed no LD, whereas the G␣i1-L91-YFP, G␣i3-L91-YFP, and G␣o-L91-YFP constructs exhibited more pronounced LD than when expressed alone (Figs. 2D and 3, E-H). Thus, with the exception of G␣i2-L91-YFP, 2PPM can be used The maximum dichroic ratio r max (and log 2 (r max )) is obtained by calculating the mean values of r (and log 2 (r)) from horizontal and 1/r (and log 2 (1/r)) from vertical parts of the cell outline.
to observe interactions between the studied FP-labeled G␣i/o subunits and the G␤1␥2 complex.

Activation of G Proteins Leads to a Variety of LD Responses in G␣i/o-FP Constructs-
To determine whether activated G␣i/ o-FP subunits interact with G␤1␥2 dimers, we carried out 2PPM imaging of G protein activation in cells transfected with a G␣i/o-FP, G␤1, G␥2, and an ␣2aAR-FP. The presence of ␣2aAR-FP did not affect the LD of G␣i/o-FP constructs, either in presence or absence of G␤1 and G␥2. In all constructs of the GAP43-CFP-G␣i/o design, application of NE caused a robust decrease but not a complete disappearance of LD in the CFP moiety of the G␣i/o constructs (Fig. 3, A-D). In all four GAP43-CFP-G␣i/o constructs, the LD remaining after receptor activa-tion was considerably higher than in the respective subunits expressed alone. Thus, upon activation, the GAP43-CFP-G␣i/o subunits appear to enter a state distinct both from the nonactivated heterotrimer and from the fully dissociated G␣ subunit. Our 2PPM results cannot reveal the identity of this activated state, which may consist of rearranged heterotrimeric G proteins, partially dissociated heterotrimers (the 2PPM data being consistent with ϳ70% dissociation), or G␣ monomers in a conformation distinct from that of non-activated single G␣ subunits.
Unlike the GAP43-CFP-G␣i/o constructs, the constructs of the G␣i/o-L91-YFP design exhibited a variety of responses upon receptor activation (Fig. 3, E-H). Modest decreases of LD Fluorescence elicited by horizontal and vertical excitation beam polarizations is colored red and green, respectively (double-headed arrows), with the color bar indicating the dichroic ratio r. The absence of red or green color in the images indicates the absence of LD. B, same as A, but for GAP43-CFP-G␣i/o constructs coexpressed with G␤1 and G␥2 subunits. LD (a red/green pattern) is apparent in the outlines of cells expressing all four GAP43-CFP-G␣i/o constructs. C, same as A, but for G␣i/o-L91-YFP constructs. LD is present in cells expressing the G␣i1-L91-YFP and G␣i3-L91-YFP constructs but not the G␣i2-L91-YFP or G␣o-L91-YFP constructs. D, same as C, but the G␣i/o-L91-YFP constructs were coexpressed with G␤1 and G␥2 subunits. LD is present in cells expressing all but the G␣i2-L91-YFP construct. Interestingly, the distribution of fluorescence intensities (localization of red/green parts of the cell outline) indicates that although the YFP fluorophore in the G␣i1-L91-YFP and G␣i3-L91-YFP constructs is oriented close to perpendicular to the cell membrane, in the G␣o-L91-YFP construct it is close to being parallel to the cell membrane. In G␣i2-L91-YFP, the fluorophore is either in a disordered orientation or close to the magic angle, 52.0 degrees, to the cell membrane.
upon activation could be discerned in G␣i1-L91-YFP and G␣i3-L91-YFP. The G␣i2-L91-YFP construct did not show a statistically significant change in LD upon activation. Activation of the G␣o-L91-YFP-containing complex caused a decrease in LD to levels indistinguishable from those observed in the G␣o-L91-YFP subunit expressed alone. These LD measurements confirm a structural heterogeneity among the studied G␣i/o-L91-YFP constructs that is not caused by sequence irregularities. Structural heterogeneity among the G␣i/o-L91-YFP constructs is also reflected in the extent of basal FRET between the G␣i/o-L91-YFP and G␤1-CFP subunits (Fig. 4). The observed structural heterogeneity within the G␣i/o-L91-YFP group of constructs is suggestive of functional heterogeneity.
Our results indicate that there are important differences among the studied constructs, both between and within the two groups of constructs. The differences may involve not only structure but also function of the constructs. In the absence of a clear pattern, no conclusion regarding subunit rearrangement or dissociation upon activation can be drawn from these experiments alone.
Some Constitutively Active Mutants of the G␣i/o-FP Constructs Interact with G␤1␥2-To ascertain the ability of activated G␣i/o subunits to interact with G␤␥ dimers, we created mutants of all tested G␣i/o-FP constructs bearing a Gln-to-Leu mutation in the third switch region (Q204L in G␣i1, G␣i3; Q205L in G␣i2, G␣o). This mutation has been shown (24) to abolish the GTPase activity of G␣ subunits, rendering the mutants constitutively active. We observed the mutated G␣ subunits by 2PPM both in the presence and absence of overexpressed G␤1 and G␥2 subunits. The results are summarized in Fig. 3.
Thus, our 2PPM observations indicate that two of the CA mutants of the GAP43-CFP-G␣ design (G␣i2 and G␣i3) can interact (albeit likely only weakly) with G␤1␥2. Two of the CA mutants of the G␣i-L91-YFP design (G␣i1 and G␣i3) showed  signs of strong interaction with G␤1␥2. The other four CA mutants did not show signs of this interaction. Similar to the diverse results obtained with the non-mutated G␣i/o-FP constructs, the results with the CA mutants reveal structural and functional differences among the investigated constructs, even within a particular design.
Prior to activation of the ␣2aAR receptor, GRK3ct-FP showed cytoplasmic localization in all experiments. The LD of each G␣i/o-FP construct was indistinguishable from its LD when coexpressed with only G␤1 and G␥2. Upon receptor activation by NE, GRK3ct-Venus translocated to the cell membrane in experiments involving each of the four G␣i/o-FP constructs of the GAP43-CFP-G␣i/o design (Fig. 5, B and D-G). This translocation was accompanied by the appearance of a small but discernible LD in GRK3ct-Venus (Fig. 5C) and also by changes of LD in the GAP43-CFP-G␣i/o constructs, consistent with activation of the G protein complex. In contrast, no GRK3ct-Cerulean translocation could be observed in experiments involving G␣i-L91-YFP constructs (Fig. 5, D-G), although, at least in G␣i1-L91-YFP, activation of the G protein construct was evidenced by changes in its LD. Activation of the G␣o-L91-YFP complex led to GRK3ct-Cerulean translocation, albeit weaker than in the GAP43-CFP-G␣i/o constructs. These results suggest that constructs of the GAP43-CFP-G␣i/o design dissociate upon activation. Among the G␣i/o-L91-YFP constructs, only G␣o-L91-YFP dissociates upon activation.
To assess the effects of the GRK3ct-FP presence on G protein dissociation, we investigated whether coexpression of a GRK3ct-FP probe affects the LD of G␣i1-FP both in an inactive and activated state. If GRK3ct significantly shifted the equilibrium between dissociated and non-dissociated G protein subunits, the effect should be reflected in the LD of the G␣-FP construct. However, we have not been able to observe such an effect (Fig. 6), which suggests that results obtained by using a GRK3ct-FP probe accurately reflect the extent of G protein dissociation that is present even in the absence of this probe.
Complexes of Non-modified Gi/o Proteins Dissociate upon Activation-To ascertain whether complexes of non-modified Gi/o proteins dissociate upon activation, we observed intracellular localization of GRK3ct-Venus in cells transfected with G␣i/o, G␤1, G␥2, ␣2aAR, and GRK3ct-Venus. Prior to activa-  (20)). The sites of interactions of G␤1␥2 with G␣i1 only are colored yellow (34), with GRK2ct only shown blue, and with both G␣i1 and GRK2ct shown in green (20). A major overlap between the G␤1␥2 surfaces interacting with G␣i1 and GRK2ct is apparent. B, two-photon microscopy images of yellow fluorescence of a typical cell transfected with the GAP43-CFP-G␣o, G␤1, G␥2, ␣2AR, and GRK3ct-Venus constructs imaged before and after application of 1 M NE and a ratio image of the two images color-coded to express the fluorescence intensity ratio by hue. Translocation of GRK3ct-Venus, evidenced by a fluorescence intensity decrease in the cytoplasm accompanied by an increase in the cell membrane regions, is apparent. Scale bars ϭ 5 m. C, 2PPM images of GRK3ct-Venus before and after application of 1 M NE (data acquired simultaneously with the images in B). Coloring is as in Fig. 1 tion, GRK3ct-Venus appeared cytoplasmic in experiments involving each of the four investigated non-labeled G␣i/o proteins. Upon treatment with NE, GRK3ct-Venus uniformly translocated to the cell membrane to an extent comparable with that observed in experiments on constructs of the GAP43-CFP-G␣i/o design (Fig. 5, D-G). These results indicate that, in living cells, upon activation, all four investigated non-modified G␣i/o subunits dissociate from G␤1␥2.
To obtain information on the extent of Gi/o complex dissociation upon activation, we compared the intracellular localization of GRK3ct-Venus in cells overexpressing GRK3ct-Venus, G␤1, G␥2, but not G␣i/o or a receptor, with cells overexpressing GRK3ct-Venus, G␤1, G␥2, a G␣i/o, and ␣2aAR, activated by addition of norepinephrine ( Fig. 7 and supplemental Fig. 1). In the former cells, virtually all overexpressed G␤1␥2 should be available for binding by GRK3ct-Venus. Thus, this experiment provides information on GRK3ct-Venus localization when virtually all present G␤1␥2 are G␣-free and allows calibration of our GRK3ct results. In the latter cells, upon NE application, the extent of GRK3ct-Venus membrane localization reached 70 -75% of that observed in the calibration experiment. Therefore, we estimate that, upon activation, 70 -75% of non-modified, overexpressed Gi/o complexes dissociate.
Because the process of GTP hydrolysis is considerably slower than Gi/o dissociation (25), we can consider the Gi/o heterotrimers and the dissociated subunits to be in a simple two-state equilibrium described by an equilibrium dissociation constant. This approximation then allows extrapolation of our results obtained with overexpressed Gi/o proteins to Gi/o proteins present at endogenous concentrations (about four times lower than concentrations of overexpressed Gi/o subunits (supplemental Fig. 2)). Our calculations (see supplemental data) indicate that, when present at endogenous concentrations, 85-90% of Gi/o molecules dissociate upon activation.
Dissociation of G Protein Heterotrimers Is Not Required for Activation of Downstream Effectors-To determine how dissociation of G protein heterotrimers relates to their functional activity, we investigated the ability of constitutively active G␣-FP constructs to regulate the activity of GIRK channels (Fig. 8). We transfected HEK293 cells with either the GAP43-CFP-G␣i1(Q204L) or G␣i1(Q204L)-L91-YFP construct, along with constructs GIRK1-C-CFP and GIRK4 (together encoding a functional GIRK channel), either with or without the G␤1 (or G␤1-YFP) and G␥2 constructs. Our results show that the presence of neither the constitutively active G␣i1 subunit (GAP43-CFP-G␣i1(Q204L) or G␣i1(Q204L)-L91-YFP) affected the GIRK channel activity induced by the presence of endogenous free G␤␥ dimers or overexpressed G␤1, G␥2 subunits. In contrast, the presence of non-activated G␣i1 subunits (GAP43-CFP-G␣i1 or G␣i1-L91-YFP) greatly reduced GIRK1/4 channel activity, presumably by sequestering free G␤␥ dimers. Activation by 1 M NE led to a pronounced increase of GIRK activity in both types of G␣i1-FP constructs. The observed GIRK channel activation by G␣i1-L91-YFP constructs is not likely to be due to a small fraction of dissociated G␣i1 complexes because GIRK activation is a stoichiometric process (26), and a small fraction of dissociated G␣i complexes would only activate a small fraction of GIRK channels. Because our 2PPM and GRK3ct translocation data show that the G␣i1(Q204L)-L91-YFP and activated G␣i1-L91-YFP constructs interact with G␤␥ dimers, we postulate the existence, at least in the G␣i-L91-YFP line of constructs, of G␣-G␤␥ interactions that do not affect downstream signaling events. Only limited G␣-G␤␥ interactions are, however, present in other constitutively active FPlabeled G␣i/o constructs as well as in agonist-activated, nonmodified G␣i/o constructs. Therefore, we conclude that, although heterotrimer dissociation is not strictly required for downstream signaling, dissociation is the likely physiological mechanism of G␣i/o activation.

DISCUSSION
The question of whether Gi/o proteins dissociate into a free G␣ subunit and a G␤␥ dimer upon activation has been widely disputed (2, 5-16, 19, 27-29). Results of in vitro studies have largely supported dissociation of activated Gi/o protein heterotrimers (8,9,28,29), whereas results of cell culture experiments have been mixed, with several studies (11)(12)(13) yielding evidence for Gi/o complex rearrangement rather than dissociation. We tried to investigate the process of Gi/o protein activation in conditions as close to natural as possible by observing G protein heterotrimers labeled with only a single FP and with no label at all.
In constructs of the GAP43-CFP-G␣i/o design, both 2PPM experiments and GRK3ct-YFP translocation observations suggest that the Gi/o heterotrimers, to a large extent, dissociate upon activation. 2PPM experiments show a marked decrease, but not disappearance, of LD upon activation and reveal some (likely limited) interactions between CA mutants of the G␣i/o subunits and the G␤1␥2 dimers. Investigations of GRK3ct-YFP intracellular localization showed pronounced translocation upon GAP43-CFP-G␣i/o activation. The simplest explanation for these observations involves a substantial (ϳ70%) dissociation of the overexpressed GAP43-CFP-G␣i/o heterotrimers upon activation.
Experiments with constructs of the G␣i/o-L91-YFP design revealed heterogeneity in structure and molecular interactions. In G␣i1-L91-YFP and G␣i3-L91-YFP, little decrease of LD occurs upon activation, and a strong interaction exists between their CA mutants and the G␤1␥2 dimers (Fig. 3, E and G). Furthermore, no GRK3ct-CFP translocation can be observed upon activation of G␣i1-L91-YFP and G␣i3-L91-YFP. These results indicate no or only minor (Ͻ 10%) heterotrimer dissociation upon activation. This conclusion is also valid for the G␣i2-L91-YFP construct, in which, however, 2PPM shows structural characteristics substantially different from both G␣i1-L91-YFP and G␣i3-L91-YFP. Surprisingly, the results of both 2PPM and GRK3ct-CFP translocation experiments obtained with the G␣o-L91-YFP construct are consistent with extensive (albeit lower than in the GAP43-CFP-G␣o construct) heterotrimer dissociation upon activation. Our observations of structural heterogeneity among constructs of the G␣i/o-L91-YFP design add to the more subtle heterogeneity of results of published (12) and our own FRET experiments (Fig. 4). Published experiments with G␣i/o-L91-YFP constructs (7,11,12) have largely supported rearrangement rather than dissociation upon G protein activation. Our results cast doubts over the relevance of results obtained with the G␣i/o-L91-YFP constructs to non-modified G␣i/o proteins.
Experiments on non-labeled Gi/o proteins (using GRK3ct-FP translocation as a proxy for Gi/o heterotrimer dissociation) show that all tested overexpressed Gi/o proteins dissociate upon activation to an extent of ϳ70 -75%. Extrapolation to Gi/o proteins present at endogenous concentrations suggests that 85-90% of Gi/o complexes dissociate upon activation. No statistically significant shift of the equilibrium between G protein heterotrimers and dissociated subunits in the active state because of coexpression of GRK3ct-FP construct could be detected and is, therefore, likely to be negligible. The extent of Gi/o heterotrimer dissociation is consistent with the dissociated subunits being the activated Gi/o form that mediates signal transduction to downstream effectors (Fig. 9).
What is the functional significance of Gi/o protein dissociation upon activation? To find out, we tested the ability of the GAP43-CFP-G␣i1 and G␣i1-L91-YFP constructs and their constitutively active mutants to regulate activity of GIRK channels. Surprisingly, neither the constitutively active GAP43-CFP-G␣i1(Q204L) nor the G␣i1(Q204L)-L91-YFP construct reduce the activity of the GIRK1/4 channel (Fig. 8), although the G␣i1(Q204L)-L91-YFP construct does bind G␤␥ dimers (Fig. 3). This finding indicates that efficient downstream signaling is possible even by a G protein heterotrimer, albeit by one modified by an FP insertion. However, because our results indicate that non-modified G␣i/o subunits largely dissociate upon activation and support published results showing that free G␤␥ dimers activate GIRK channels, we postulate G␣i/o heterotrimer dissociation to be the major physiological mechanism of G␣i/o signal transduction to downstream effectors. It is, however, possible, perhaps even likely, that, to some extent, GIRK activation involves interactions between an activated G␣i/o  heterotrimer and the GIRK channel, possibly affecting the signaling kinetics (30,31).
In conclusion, our results explain published contradictory results obtained with constructs of different designs, bring in line in vitro and in vivo evidence, and appear to settle the issue of G protein dissociation within the Gi/o family. Our findings also open up the possibility that other heterotrimeric G proteins thought to undergo rearrangement, such as Gs (13,16) and Gq (32), dissociate upon activation. Apart from providing new information on the mechanisms of Gi/o signaling, our results also illustrate the uses and capabilities of 2PPM, a novel microscopy technique that allows observations of protein-protein interactions using a single fluorescent label.