Activated G Protein Gαs Samples Multiple Endomembrane Compartments*♦

Heterotrimeric G proteins are localized to the plasma membrane where they transduce extracellular signals to intracellular effectors. G proteins also act at intracellular locations, and can translocate between cellular compartments. For example, Gαs can leave the plasma membrane and move to the cell interior after activation. However, the mechanism of Gαs translocation and its intracellular destination are not known. Here we use bioluminescence resonance energy transfer (BRET) to show that after activation, Gαs rapidly associates with the endoplasmic reticulum, mitochondria, and endosomes, consistent with indiscriminate sampling of intracellular membranes from the cytosol rather than transport via a specific vesicular pathway. The primary source of Gαs for endosomal compartments is constitutive endocytosis rather than activity-dependent internalization. Recycling of Gαs to the plasma membrane is complete 25 min after stimulation is discontinued. We also show that an acylation-deacylation cycle is important for the steady-state localization of Gαs at the plasma membrane, but our results do not support a role for deacylation in activity-dependent Gαs internalization.

Heterotrimeric G proteins are localized to the plasma membrane where they transduce extracellular signals to intracellular effectors. G proteins also act at intracellular locations, and can translocate between cellular compartments. For example, G␣ s can leave the plasma membrane and move to the cell interior after activation. However, the mechanism of G␣ s translocation and its intracellular destination are not known. Here we use bioluminescence resonance energy transfer (BRET) to show that after activation, G␣ s rapidly associates with the endoplasmic reticulum, mitochondria, and endosomes, consistent with indiscriminate sampling of intracellular membranes from the cytosol rather than transport via a specific vesicular pathway. The primary source of G␣ s for endosomal compartments is constitutive endocytosis rather than activity-dependent internalization. Recycling of G␣ s to the plasma membrane is complete 25 min after stimulation is discontinued. We also show that an acylation-deacylation cycle is important for the steady-state localization of G␣ s at the plasma membrane, but our results do not support a role for deacylation in activity-dependent G␣ s internalization.
Heterotrimeric G proteins are best known for their role in transducing signals from the extracellular environment across the plasma membrane (1). However, it is becoming increasingly apparent that these signaling molecules have important functions in other cellular locations (2). Therefore, it is important to know how G proteins are distributed among various intracellular compartments, and how these proteins traffic between the plasma membrane, the cytosol, cytoskeletal elements, and intracellular organelles both in their resting state and when they are activated.
The most well studied example of activity-dependent G protein trafficking occurs in the retina, where intense illumination promotes the dissociation of transducin from rod outer segment disks and the subsequent translocation of both G␣ t and G␤␥ into the rod inner segment (3). Activity-dependent trans-location of G␣ and G␤␥ subunits occurs in other cells as well (4,5), most notably for G␣ s , the subunit responsible for stimulation of adenylate cyclase. Stimulation of a cognate receptor or inhibition of GTPase activity promotes activation and translocation of G␣ s from the plasma membrane to the cell interior (6 -8).
One important yet poorly understood aspect of G␣ s internalization is the intracellular destination of this G protein. Cell fractionation studies have shown that a large amount of constitutively active G␣ s is found in the soluble fraction, suggesting that active G␣ s may simply become soluble in the cytosol (8,9). However, only a small amount of G␣ s enters the soluble fraction after receptor-mediated activation, consistent with vesicle-mediated internalization and continued association with membranes. Imaging studies have shown that internalized G␣ s can appear to be diffuse in the cytosol (8,9), but association with intracellular vesicles has also been observed (10 -12). Therefore, the trafficking itinerary of active G␣ s has not been clearly defined. This question has taken on greater significance with the recent realization that receptors can continue to activate G␣ s after agonist-induced endocytosis (13)(14)(15)(16). It is known that active receptors and G␣ s internalize via distinct pathways (8,10,11,17), but it is unclear how receptors and G␣ s ultimately reach a common intracellular compartment. It is also thought that internalized G␣ s recycles to the plasma membrane (8), but very little is known about how this occurs. A related issue is the possible role of deacylation in activitydependent G␣ s internalization. G␣ s subunits are anchored to the plasma membrane in part by palmitoylation of a single N-terminal cysteine residue (18,19), and activation dramatically enhances the rate of palmitate turnover (20 -22). A cytosolic thioesterase, acyl protein thioesterase 1 (APT1), 2 has been shown to catalyze deacylation of G␣ s in vitro, and this activity is inhibited by association with G␤␥ (23). These findings suggest that activation may promote APT1-mediated net removal of palmitate and thereby release active G␣ s from the plasma membrane, but direct evidence supporting this model has been not been reported.
Here we map the subcellular distribution of inactive and active G␣ s using bioluminescence resonance energy transfer (BRET). We find that inactive G␣ s is constitutively located on the plasma membrane and on multiple endomembrane compartments, including endosomes. Upon activation, G␣ s reversibly translocates from the plasma membrane to several endomembrane compartments, including the endoplasmic reticulum, mitochondria, and endosomes. Neither inhibition of APT1 with a broad-spectrum serine hydrolase inhibitor nor mutation of the critical cysteine prevented G␣ s translocation. Our results are consistent with a model wherein active G␣ s dissociates from the plasma membrane, translocates via a nonvesicular mechanism, and indiscriminately samples intracellular membranes. Our results do not support or rule out a role for thioesterase-mediated net depalmitoylation in translocation of active G␣ s .

Results
The Subcellular Distribution of G␣ s -BRET can occur between non-associated membrane-bound donors and acceptors provided that they both are located in the same compartment and the acceptor is present at sufficient density. If an array of acceptors targeted to different membrane compartments is coexpressed (one at a time) with a membrane-bound donor, then a map of the donor's subcellular localization can be drawn (24). The BRET signal between a donor protein of interest and acceptor membrane markers depends on the propensity of the protein of interest to associate with different membranes, but also on the density of acceptors expressed on the surface of each compartment, the efficiency of each marker as a BRET acceptor, and the relative surface area of each compartment. Therefore, this method can only provide a qualitative map of subcellular distribution. Nevertheless, BRET produced at a given membrane compartment is directly related to the fraction of the donor associated with that compartment; therefore this method can be used to track changes in the subcellular distribution of membrane proteins over time in populations of living cells (24 -26).
To use this method to determine the subcellular distribution of G␣ s , we replaced residues 73-84 in the ␣ helical domain of this subunit with the BRET donor Rluc8 (27). Previous studies have shown that insertions as large as fluorescent proteins in this region do not interfere with G␣ s function (10,12). We coexpressed G␣ s -Rluc8 with unlabeled G␤ 1 and G␥ 2 and six standard membrane compartment markers that were fused to the BRET acceptor Venus. These markers targeted the cytosolic surface of the plasma membrane, Golgi apparatus, endoplasmic reticulum, and early, late, and recycling endosomes. Confocal imaging verified that each endomembrane marker labeled morphologically appropriate and distinct intracellular structures under conditions similar to those used for BRET experiments (Fig. 1A).
We observed robust BRET between G␣ s -Rluc8 and the plasma membrane acceptor (Fig. 1B), and much weaker BRET signals between G␣ s -Rluc8 and endomembrane acceptors. The smallest endomembrane signals were observed at the Golgi apparatus, whereas the largest signals were observed at recycling endosomes (Fig. 1B). These results are consistent with previous studies that have shown that G s is present at the plasma membrane, but also traffics constitutively through endosomal structures (4, 17, 28).
It has previously been shown that delivery of G␣ s to the plasma membrane during biosynthesis depends on association with G␤␥ (29). Consistent with this, we found that BRET produced by G␣ s -Rluc8 at the plasma membrane decreased significantly when this subunit was expressed without exogenous G␤ 1 and G␥ 2 subunits (Fig. 1B). The BRET signal originating from recycling endosomes was also significantly smaller without exogenous G␤␥, whereas BRET originating from the endoplasmic reticulum was slightly but significantly larger. Trafficking of G␣ s -Rluc8 to the plasma membrane in the absence of exogenous G␤␥ could have been due to association with endogenous G␤␥ dimers. To test this idea, we constructed a subunit (G␣ s -Rluc8 IEKϩ) bearing mutations at the N-terminal G␤␥ interface, as this mutant was shown previously to be highly defective with respect to plasma membrane association (29). Similarly, we found that BRET produced by G␣ s -Rluc8 IEKϩ at the plasma membrane was reduced even further than that produced by wild-type G␣ s -Rluc8 in the absence of exogenous G␤␥, and BRET produced at recycling endosomes was similarly impaired (Fig. 1B). These results are consistent with a role for G␤␥ in membrane targeting of nascent G s heterotrimers (29), and demonstrate that BRET can be used to map the subcellular distribution of G␣ s -Rluc8.
Activity-dependent Redistribution of G␣ s -Several studies have shown that a pool of active G␣ s leaves the plasma membrane and either enters the cytosol or associates with intracellular vesicles (6 -12). Because the intracellular destination of G␣ s is uncertain, we compared the subcellular distribution of G␣ s before and after activation. Stimulation of coexpressed ␤ 2 adrenergic receptors (␤ 2 ARs) with the agonist isoproterenol for 10 min decreased BRET produced by G␣ s -Rluc8 at the plasma membrane by ϳ40% ( Fig. 2A). Isoproterenol also increased BRET at several endomembrane compartments. The largest increase (ϳ100%) was observed at the endoplasmic reticulum, whereas small but significant increases (ϳ10%) were observed at late and recycling (but not early) endosomes ( Fig. 2A). The relatively large BRET increase at the endoplasmic reticulum could reflect a specific trafficking itinerary for active G␣ s -Rluc8. Alternatively, this difference could simply reflect the large surface area of the endoplasmic reticulum, which would allow this compartment to capture a large fraction of G␣ s -Rluc8 if this protein is released into the cytosol to randomly sample endomembranes. In this case, isoproterenol should also induce large BRET increases at other intracellular organelles that have a large surface area. After the endoplasmic reticulum, the mitochondria have the largest collective surface area in HEK 293 cells (Fig. 1A), and indeed isoproterenol induced a 39 Ϯ 6% increase in BRET between G␣ s -Rluc8 and an acceptor located on the surface of mitochondria (p Ͻ 0.001, n ϭ 7). Taken together these results are consistent with the notion that some active G␣ s -Rluc8 is released into the cytosol and randomly samples endomembranes. We then examined the subcellular distribution of G␣ s -Rluc8 after more prolonged activation of ␤ 2 ARs. Stimulation with isoproterenol for 4 h decreased BRET at the plasma membrane and increased BRET at the endoplasmic reticulum (Fig. 2, C and  D), similar to what was observed after stimulation for 10 min. However, prolonged activation significantly decreased BRET at all endosomal compartments (Fig. 2, C and D), i.e. the opposite of what was observed after stimulation for 10 min. Treatment with cholera toxin for 2-4 h produced a similar redistribution of G␣ s -Rluc8 away from the plasma membrane and endosomes and toward the endoplasmic reticulum (Fig. 3). Cholera toxin also occluded BRET changes induced by acute stimulation with isoproterenol. These results indicated a biphasic effect of activation on G␣ s -Rluc8 abundance in endosomes.
Previous studies have shown that ␤ 2 ARs and G␣ s localize to different intracellular structures after receptor activation (8,10). To demonstrate this with BRET, we expressed ␤ 2 AR-Rluc8 and our standard panel of membrane-localized acceptors. Prior to stimulation, ␤ 2 AR-Rluc8 produced robust BRET at the plasma membrane and Golgi apparatus, and smaller signals at other endomembrane compartments (Fig. 4A). As was the case with G␣ s -Rluc8, stimulation with isoproterenol for 30 min decreased BRET from ␤ 2 AR-Rluc8 at the plasma membrane (25). Isoproterenol also increased BRET from ␤ 2 AR-Rluc8 at endosomes, but these increases were much larger (ϳ50 -100%)   than those observed with G␣ s -Rluc8, and were evident at all three endosomal compartments. Unlike what we observed with G␣ s -Rluc8, isoproterenol did not increase BRET from ␤ 2 AR-Rluc8 at the endoplasmic reticulum (Fig. 4A). We next directly compared ␤ 2 AR-Rluc8 and G␣ s -Rluc8 trafficking at both short (30-min) and long (4-h) time points. The abundance of both proteins at the plasma membrane decreased progressively over the course of 4 h (Fig. 4B). In contrast, although both proteins were more abundant in recycling endosomes 30 min after agonist stimulation, only ␤ 2 AR-Rluc8 continued to accumulate in endosomes at 4 h, whereas G␣ s -Rluc8 was depleted from this compartment. These results underscore the conclusion that ␤ 2 ARs and G␣ s are found in common endomembrane compartments, but traffic there via different mechanisms after receptor activation. We next examined the kinetics of activity-dependent redistribution of G␣ s -Rluc8 in more detail, focusing on the reciprocal changes in BRET at the plasma membrane and endoplasmic reticulum. At room temperature (ϳ25°C), stimulation with isoproterenol decreased BRET at the plasma membrane with a half-life (t1 ⁄ 2 ) of 95 s, and increased BRET at the endoplasmic reticulum with a similar half-life of 113 s (Fig. 5A). This time course is similar to the rate at which G␣ s -fluorescent protein chimeras translocate from the plasma membrane after receptor activation (10,12), but much slower than conformational changes associated with activation (30). The addition of the ␤ 2 AR antagonist propranolol allowed recovery of BRET signals in both of these compartments (Fig. 5A). Full recovery required ϳ25 min for both the plasma membrane (t1 ⁄ 2 ϭ 320 s) and the endoplasmic reticulum (t1 ⁄ 2 ϭ 256 s; Fig. 5B).
Mechanisms of Reversible G␣ s Translocation-The mechanisms responsible for the translocation of G␣ s away from and back to the plasma membrane after a period of activation are poorly understood. In particular, the roles of vesicular transport and palmitate cycling have not been firmly established. Therefore, we took advantage of the BRET assay to test these mechanisms. Vesicular transport is arrested by low temperatures (31); therefore we studied the loss and recovery of BRET from G␣ s -Rluc8 at the plasma membrane in cells incubated at 37°C, room temperature (ϳ25°C), and 15°C. Lowering the temperature significantly increased the baseline BRET signal from G␣ s -Rluc8 at the plasma membrane, but did not affect the BRET decrease produced by isoproterenol (p Ͼ 0.05, n ϭ 9; Fig.  6A). In contrast, lowering the temperature partially inhibited recovery of BRET at the plasma membrane, which after 15 min reached 90 Ϯ 1% of the control value at 37°C, but only 69 Ϯ 1%  of the control value at 15°C (p Ͻ 0.001). The latter result suggested that vesicular transport may play a role in the return of internalized G␣ s to the plasma membrane. However, disruption of microtubules with nocodazole had no effect on G␣ s -Rluc8 internalization or recovery, and disruption of the Golgi apparatus with brefeldin A had only a small (but significant) effect on recovery of BRET at the plasma membrane (81 Ϯ 0% of control versus 85 Ϯ 0% of control for DMSO; p Ͻ 0.001; Fig.  6B), suggesting that G␣ s recycles to the plasma membrane via a transport mechanism other than the classical secretory or recycling vesicular pathways.
APT1 removes the palmitoyl anchor from G␣ s in vitro (23), and activation of G␣ s increases the rate of palmitate turnover in cells (20 -22). Therefore, it has been suggested that APT1-dependent depalmitoylation may be important for activity-dependent translocation of G␣ s . To test this idea, we incubated cells with the broad-spectrum lipase inhibitor hexadecylfluorophosphonate (HDFP), which has previously been shown to inhibit more than 20 lipases, including APT1, and to attenuate basal G␣ s palmitate turnover (32). Treatment with HDFP for 4 h decreased BRET from G␣ s -Rluc8 at the plasma membrane, and increased BRET from G␣ s -Rluc8 at the endoplasmic reticulum, Golgi apparatus, and late endosomes (Fig. 7A). This result is consistent with a role for palmitate cycling in the steady-state localization of G␣ s at the plasma membrane, as suggested previously for other palmitoylated membrane proteins (33). However, HDFP did not prevent isoproterenol-induced translocation of G␣ s -Rluc8 from the plasma membrane to the endoplasmic reticulum (Fig. 7A), suggesting that translocation did not require APT1 or another HDFP-sensitive hydrolase.
To further examine the possible role of palmitate cycling, we studied G␣ s -Rluc8 subunits that were mutated to carry alternative N-terminal lipid anchors. We constructed G␣ s -Rluc8 variants that were either myristoylated in addition to the usual palmitoylation (G␣ s -Rluc8 ϩmyr), or myristoylated but not palmitoylated (G␣ s -Rluc8 ϩmyr-palm) (9). A previous study has shown that G␣ s ϩmyr subunits are more strongly localized to the plasma membrane than wild-type G␣ s and are resistant to activity-dependent internalization, whereas G␣ s ϩmyr-palm subunits are weakly localized to the plasma membrane, but are still susceptible to activity-dependent internalization (9). In agreement with these results, we found that G␣ s -Rluc8 ϩmyr produced significantly more BRET at the plasma membrane than wild-type G␣ s -Rluc8 (Fig. 7B), and this signal was relatively resistant to isoproterenol (8 Ϯ 1% decrease; Fig. 7C). In contrast, G␣ s -Rluc8 ϩmyr-palm produced significantly less BRET at the plasma membrane than wild-type G␣ s -Rluc8, and significantly more BRET at the endoplasmic reticulum (Fig.  7B). Despite this weak localization at the plasma membrane, BRET from G␣ s -Rluc8 ϩmyr-palm at the plasma membrane still decreased by 36 Ϯ 3% after stimulation with isoproterenol, and this coincided with a significant increase in BRET at the endoplasmic reticulum and recycling endosomes (Fig. 7D).

Discussion
The results of the present study extend previous studies of G␣ s subcellular localization and trafficking in several important ways. Our results show that G␣ s is present on the cytosolic surface of several endomembrane compartments both before and after activation. Previous studies have indicated that G proteins associate with intracellular membranes during biosynthesis (4) and that G␣ s undergoes constitutive clathrin-independent endocytosis (17). Our results confirm the presence of this subunit at the endoplasmic reticulum and at least three distinct endosomal compartments. Our results also show that receptormediated activation of G␣ s leads to only small increases in the availability of this subunit at endosomes, and this only after relatively short-term receptor activation. Activation for 4 h actually decreased the availability of G␣ s at endosomes, presumably because the source of endosomal membranes (the plasma membrane) was depleted of G␣ s . Therefore, we conclude that constitutive endocytosis is the primary supplier of G␣ s to endosomes. It is now recognized that G protein-coupled receptors (GPCRs) can continue to activate G s in endosomes (13)(14)(15)(16), and our results suggest that internalized receptors largely activate heterotrimers that are already present in these compartments. Although it has been known for 30 years that a fraction of G␣ s leaves the plasma membrane after activation (6 -8), the intracellular destination of this subunit has not been clearly established. Imaging studies have suggested that G␣ s internalizes into vesicles (10 -12), and one study identified Rab11-positive (recycling) endosomes as a target compartment (10). However, other imaging studies have suggested a more diffuse cytosolic (i.e. soluble) distribution of active G␣ s (8,9). Our finding that internalization of activated G␣ s persists at 15°C suggests that vesicular trafficking is not involved, and is consistent with previous findings with more specific inhibitors of endocytic trafficking (8,10). Moreover, the magnitudes of agonistinduced BRET increases at endosomes are too small to be consistent with robust endocytosis into these compartments. This interpretation is supported by the observation that G␣ s -Rluc8 appears at the endoplasmic reticulum and mitochondria without an appreciable delay after leaving the plasma membrane. Our results are thus consistent with a model wherein the affinity of activated G␣ s for membranes is decreased (but not abolished), and subunits diffuse through the cytosol to randomly sample membrane compartments. Because the endoplasmic reticulum has the largest surface area of any endomembrane compartment, this organelle captures the largest fraction of the newly available G␣ s . Indiscriminate sampling of intracellular membranes by activated G␣ s is consistent with previous imaging studies (8 -12), which have not identified a consensus intracellular target for this subunit.
Our results may also help to explain why, in cell fractionation studies, the appearance of G␣ s in the soluble fraction is more evident after chronic activation (by mutation or ADP-ribosylation) than after receptor-mediated activation (8,9). Immediately after receptor activation, we observe a significant loss of G␣ s at the plasma membrane, but no loss from other membrane compartments. After 4 h, G␣ s is further depleted from the plasma membrane, and is also less abundant at endosomes. One possible explanation for this is that over time, endosomes either recycle to the plasma membrane or fuse with endosomes bearing active receptors, thus exposing their complement of G␣ s to active receptors. In other words, the G␣ s that is constitutively located on intracellular membranes is transiently protected from receptor-mediated activation, and would not immediately appear in the soluble fraction.
Recycling of activated G␣ s to the plasma membrane has been observed in only one prior study (8), and very little is known about the time course of recycling or the mechanisms involved. Here we show that recycling occurs over the course of 25 min, persists at low temperatures, and is resistant to disruption of microtubules and the Golgi apparatus. These results rule out a role for the classical secretory pathway in the return of activated G␣ s to the plasma membrane, and argue against a role for vesicular trafficking in general. It is possible that free G␣ s subunits sample all endomembrane compartments and passively redistribute back to the plasma membrane upon deactivation, as is thought to be the case for translocating G␤␥ dimers (34). However, the rate at which G␣ s recycles to the plasma membrane is 3-fold slower than the rate at which it leaves this compartment, suggesting that different mechanisms are involved in recycling and the initial translocation. It is also possible that the route of G␣ s recycling to the plasma membrane is similar to the (still poorly defined) route taken by nascent G␣ s during biosynthesis, as the initial delivery of this G protein to the plasma membrane also persists after disruption of the Golgi apparatus (35). Although there is precedent for non-vesicular delivery of palmitoylated proteins to the plasma membrane (36,37), additional studies will be required to fully determine how G␣ s traffics to and from the plasma membrane. Finally, our results fail to support the hypothesis that internalization of activated G␣ s depends on depalmitoylation. Activation of G␣ s greatly increases the rate of palmitate turnover on this subunit, suggesting that net depalmitoylation may lead to detachment of G␣ s from the plasma membrane (8,20,21). Although a great deal of circumstantial evidence supports this hypothesis, it has not been directly tested. We show here that HDFP, a lipid hydrolase inhibitor that decreases constitutive palmitate turnover on G␣ s in cells (32), fails to prevent agonist-induced internalization of this subunit. That HDFP induced a redistribution of G␣ s in unstimulated cells suggests that the inhibitor was effective and that a constitutive acylation-deacylation cycle is important for steady-state localization of G␣ s at the plasma membrane, as is the case for other palmitoylated proteins (33). It is possible that an as yet unidentified HDFP-resistant enzyme is responsible for activity-dependent depalmitoylation and internalization of G␣ s . However, it is also possible that the increase in palmitate turnover upon activation is not the primary mechanism of G␣ s translocation. This possibility is consistent with a report indicating that activation increases the rate of palmitate turnover on G␣ s but does not change the fraction of palmitoylated G␣ s (38). This alternative is also supported by our observation that a G␣ s mutant that lacks the cysteine necessary for palmitoylation undergoes a similar activity-dependent redistribution from the plasma membrane to intracellular compartments (9), although it is possible that this mutant and wild-type G␣ s translocate via different mechanisms. Although not conclusive, our results suggest that a factor other than palmitoylation may dictate the affinity of activated G␣ s for the plasma membrane.