Signal Transduction and Hormone-dependent Internalization of the Thyrotropin-releasing Hormone Receptor in Cells Lacking Gq and G11 *

The thyrotropin-releasing hormone (TRH) receptor was expressed in embryonic fibroblasts from mice lacking the α subunits of Gq and G11 (Fq/11 cells) to determine whether G protein coupling is necessary for agonist-dependent receptor internalization. Neither TRH nor agonists acting on endogenous receptors increased intracellular calcium unless the cells were co-transfected with the α subunit of Gq. In contrast, temperature-dependent internalization of [3H]MeTRH in Fq/11 cells was the same whether Gqα was expressed or not. A rhodamine-labeled TRH analog and fluorescein-labeled transferrin co-localized in endocytic vesicles in Fq/11 cells, indicating that endocytosis took place via the normal clathrin pathway. Cotransfection with β-arrestin or V53D β-arrestin increased TRH-dependent receptor sequestration. Fq/11 cells were co-transfected with the TRH receptor and a green fluorescent protein (GFP)-β-arrestin conjugate. GFP-β-arrestin was uniformly distributed in the cytoplasm of untreated cells and quickly translocated to the periphery of the cells when TRH was added. A truncated TRH receptor that lacks potential phosphorylation sites in the cytoplasmic carboxyl terminus signaled but did not internalize or cause membrane localization of GFP-β-arrestin. These results prove that calcium signaling by the TRH receptor requires coupling to a G protein in the Gq family, but TRH-dependent binding of β-arrestin and sequestration do not.

Many G protein-coupled receptors undergo ligand-driven internalization through the clathrin pathway. Internalization or sequestration appears to modulate various aspects of signal transduction such as desensitization, resensitization, and activation of mitogen-activated protein kinases (1)(2)(3). A fundamental question regarding internalization of G protein-coupled receptors is how the activated receptor is targeted to clathrincoated pits in response to agonist binding. Recently, this question has been answered with the discovery that ␤-arrestin acts as an adaptor (4,5), binding both to receptors and to clathrin (6,7). Because ␤-arrestin binds most strongly to activated, phosphorylated receptors (5), and receptor phosphorylation is promoted by agonist binding, ␤-arrestin can discrimi-nate between inactive and active receptors and target agonistoccupied receptors to clathrin-coated pits and the endocytic pathway.
Most receptors coupled to G proteins in the G q family are sequestered following the binding of an agonist but not an antagonist (8). There are exceptions to this generalization, however, because some G q -coupled receptors do not internalize at all, others internalize only in some cell contexts (9), and a few internalize when occupied by an antagonist (10,11). It is uncertain if coupling to G q proteins is required for receptor phosphorylation, ␤-arrestin recognition of agonist-occupied receptor, or receptor internalization.
Internalization of the receptor for thyrotropin-releasing hormone (TRH), 1 which signals through G q and G 11 to activate phospholipase C, has been studied extensively (12)(13)(14)(15)(16)(17)(18). The TRH receptor internalizes rapidly (t1 ⁄2 ϳ 2.5 min) through the clathrin pathway upon agonist binding (13). When its cytoplasmic, carboxyl-terminal tail is deleted, the TRH receptor binds ligand with high affinity and stimulates phospholipase C but does not internalize, showing that coupling to G q is not sufficient to cause internalization (15,17). It is not clear whether coupling to G q is necessary for receptor sequestration. Several investigators have approached this question by characterizing the internalization of mutant receptors using antibodies rather than hormones to promote internalization and blocking signal pathways pharmacologically (15,17). They have reached contradictory conclusions about whether receptor-G protein coupling is necessary for TRH receptor internalization.
In this report, we have taken advantage of a fibroblast cell line that lacks G q and G 11 to examine directly the role of these G proteins in the internalization of TRH receptor (19,20). The Fq/11 cells were derived from the embryos of mice in which the ␣ subunits of both G q and G 11 had been knocked out by targeted gene disruption. Our results show that G q or G 11 is absolutely required for calcium signaling but not for coupling between the activated TRH receptor and ␤-arrestin or for receptor internalization.
Fq/11 cells are fibroblasts derived from G q ␣/G 11 ␣ knockout mice and were generously provided by Dr. Melvin Simon (California Institute of Technology, Pasadena, CA) (19,20). Fq/11 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and were passaged by trypsinization. Transfection of Fq/11 cells was achieved with LipofectAMINE PLUS according to the manufacturer's suggestions. In cotransfection assays, 1 g of pCDM8mTRHR and 2 g of pCMVGq or pRSVCAT were used. Primary mouse embryonic fibroblasts were a gift from Dr. Dianqing Wu (University of Rochester, Rochester, NY) and were cultured and transfected similarly.
Methods-Ca 2ϩ imaging was carried out as described (22,23). To measure radioligand binding, cells in 35-mm dishes were incubated in Hank's buffered saline solution containing 15 mM HEPES, pH 7.4 containing 5 nM [ 3 H]MeTRH with or without a 1000-fold molar excess of unlabeled hormone. Dishes were then washed and incubated in ice-cold 0.5 M NaCl, 0.2 M acetic acid (pH 2.8) for 1 min, and radioactivity in the acid wash (surface) and the cells (internalized) was quantified (13). Binding data shown are typical of results in 2 to 4 independent experiments.
Staining with Rhod-TRH and fluorescein-labeled transferrin (FITCtransferrin) was carried out as described (12). In colocalization experiments, cells were incubated in Hepes-buffered saline solution with 170 nM Rhod-TRH and 5 g/ml fluorescein-labeled transferrin (FITC-transferrin) at 37°C for 1 h. Control experiments, in which cells were stained with either Rhod-TRH or FITC-transferrin alone, showed that bleedthrough was negligible.
To examine translocation of the ␤-arrestin-green fluorescent protein (GFP) conjugate (␤arr2GFP), Fq/11 cells were transfected with 2 g of pCDM8mTRHR or pAd/CMVmTRHRC335STOP and 0.4 g of p␤arr2GFP/dish. About 24 h after transfection, cells were washed and incubated with Hepes-buffered saline solution at 37°C. Cells were stimulated with 1 M TRH for up to 20 min. Localization of ␤arr2GFP in live cells was followed microscopically with a 40ϫ objective and a fluorescein filter. Microscopic results are typical of data obtained in 3 to 10 independent experiments.

RESULTS
We transfected Fq/11 cells with cDNA encoding the TRH receptor and a 2-fold higher concentration of either a control plasmid or plasmid encoding G q ␣ and then measured the ability of various agonists to increase the concentration of cytoplasmic free Ca 2ϩ , [Ca 2ϩ ] i , in individual cells loaded with the Ca 2ϩ reporter Fura-2. Following transfection with the TRH receptor alone, Fq/11 cells failed to respond to TRH or to a mixture of agonists acting on endogenous receptors (thrombin, endothelin, bombesin, and bradykinin) (Figs. 1 and 2 and Table I). When transfected with the TRH receptor and the ␣ subunit of G q , a significant number of cells responded to TRH (15%) and to the mixture (11%) (Figs. 1 and 2 and Table I). Subsequent experiments showed that bradykinin was responsible for most of the activity in the agonist mixture (Table I). TRH caused a consistently larger increase in [Ca 2ϩ ] i than the mixture. Based on [ 3 H]MeTRH binding, the number of TRH receptors expressed was within 50% of the same in cultures transfected with a control plasmid or plasmid encoding G q ␣ (data not shown). To determine whether intracellular Ca 2ϩ stores were adequate in the nonresponsive cells, we added the Ca 2ϩ chelator BAPTA to reduce extracellular Ca 2ϩ to ϳ4 nM and then added a low concentration of ionomycin to release Ca 2ϩ from intracellular stores. All of the Fq/11 cells tested responded to ionomycin with a large increase in Ca 2ϩ except those cells that had previously responded to TRH, which had predictably smaller responses because their Ca 2ϩ stores had been partially depleted.
To determine whether internalization of the TRH receptor also requires G protein coupling, we measured sequestration of [ 3 H]MeTRH to an acid/salt-resistant compartment in Fq/11 cells with and without transfection of G q ␣. When Fq/11 cells were transfected with only the TRH receptor and then incu-bated with [ 3 H]MeTRH at 37°C, 32% of specifically bound hormone was acid/salt resistant, in comparison to only 12% when binding was performed at 0°C (Fig. 3). When cells were transfected with the receptor and G q ␣, the extent of internalization was the same (Fig. 3). Internalization of [ 3 H]MeTRH was unaffected by treatment with pertussis toxin, indicating that G i and G o are not involved. We also transfected embryonic fibroblast cultures obtained from wild type mice of the same strain as the knockout mice (MEF cells) with the TRH receptor cDNA and followed [ 3 H]MeTRH internalization. The wild type fibroblasts expressed TRH receptors at the same level as the Fq/11 cells, and they also internalized [ 3 H]MeTRH by a pertussis toxin-insensitive pathway.
Because the TRH receptor normally internalizes in a clathrindependent manner, we compared internalization of FITCtransferrin, a classical marker for the coated pit pathway, and a bioactive, rhodamine-labeled TRH analog, Rhod-TRH, in Fq/11 cells. Rhod-TRH fluorescence in Fq/11 cells was dim, consistent with the low level of [ 3 H]MeTRH binding. When Fq/11 cells were transfected with only the TRH receptor and incubated with Rhod-TRH at 0°C, fluorescence was on the cell surface (Fig. 4). Incubation with excess TRH blocked the membrane fluorescence, indicating that it was because of receptorbound Rhod-TRH. When these cells were incubated with Rhod-TRH at 37°C, Rhod-TRH was concentrated inside the cell, indicating that internalization of the liganded receptor had taken place. The bright specks that can be seen in all panels were associated with transfection reagents and were also seen in mock-transfected cells (data not shown). Internalization of FITC-transferrin was rapid and extensive, with nearly all the label in an endocytic compartment by 5 min (data not shown). Cells were also incubated simultaneously with Rhod-TRH (red) and FITC-transferrin (green) at 37°C. There was extensive colocalization of the two labels, shown in yellow and orange (Fig. 5). These findings all indicate that TRH drives internalization of its receptor via the normal coated pit pathway without G protein interaction.
To establish whether overexpression of ␤-arrestin would increase sequestration, we transfected Fq/11 cells with the TRH receptor with or without ␤-arrestin. Cells transfected with the receptor and ␤-arrestin internalized more [ 3 H]MeTRH (56%) than cells transfected with receptor alone (32%) (Fig. 6), suggesting that ␤-arrestin was functioning as an adaptor in the absence of G protein coupling. A mutant ␤-arrestin, V53D, has been shown to act in a dominant negative manner in some settings but not in others. V53D ␤-arrestin increased internalization of [ 3 H]MeTRH as effectively as wild type arrestin (Fig. 6).
To determine whether G protein coupling is needed for ␤-arrestin binding, we transfected Fq/11 cells with the TRH receptor and a GFP-␤-arrestin conjugate and followed the localization of GFP-␤-arrestin before and after activating the receptor. GFP-␤-arrestin was uniformly distributed in the cytoplasm of untreated cells (Fig. 7). Following the addition of TRH, GFP-␤-arrestin quickly translocated to the periphery of the cells, suggesting that it bound to the agonist-occupied receptor in the absence of G protein interaction. Once translocated to the plasma membrane, much of the GFP-␤-arrestin remained there for up to 1 h in the continued presence of TRH (data not shown). TRH did not cause translocation of GFP-␤-arrestin in cells expressing a carboxyl-terminal-truncated TRH receptor (Fig. 7, also see below). Even at a concentration of 10 nM, TRH rapidly and effectively translocated GFP-␤-arrestin (Fig. 8). The cells shown in Figs. 7 and 8 were fairly dim. In cells with bright GFP fluorescence, translocation was not as evident, presumably because the bright cells expressed more GFP-␤-arrestin, and the proportion translocated in response to TRH was small. We never observed translocation of GFP-␤-arrestin to the membrane following activation of the endogenous bradykinin receptor (Fig. 8). We have found that agonist mixtures activating endogenous receptors in transfected 293 cells or GH3 cells do not cause translocation of GFP-␤-arrestin, although they cause

FIG. 3. Internalization of [ 3 H]MeTRH in Fq/11 cells and in normal primary mouse fibroblasts.
Left, Fq/11 cells were transfected with plasmids encoding TRH receptor (1 g/dish) and control plasmid (2 g/dish) (Con) or TRH receptor (1 g/dish) and G q ␣ (2 g/dish) (G q ␣). Right, mouse embryonic fibroblasts (MEF) from wild type animals and Fq/11 cells were transfected with TRH receptor alone. Cells were incubated for 60 min at 0 or 37°C with 5 nM [ 3 H]MeTRH, and the percent of specifically bound hormone internalized was determined. [Ca 2ϩ ] i responses comparable with those evoked by TRH. 2 These results suggest that the TRH receptor is more effective than most in translocating GFP-␤-arrestin.
We also evaluated the effect of a TRH antagonist, chlordiazepoxide, on GFP-␤-arrestin localization. Chlordiazepoxide (10 M) did not cause translocation of GFP-␤-arrestin by itself but blocked GFP-␤-arrestin translocation initiated by a low dose of TRH, 10 nM (Fig. 9). When added at 10 M, TRH overcame the effect of the antagonist and caused GFP-␤-arrestin to move to the membrane. The K d values of TRH and chlordiazepoxide for the TRH receptor are ϳ10 nM and 3 M, respectively. ␤-Arrestin is believed to bind more tightly to phosphorylated receptors (6,7), and although the phosphorylation sites on the TRH receptor have not been mapped, most potential phosphorylation sites are found in the intracellular carboxyl-terminal tail after residue 341. We therefore transfected cells with GFP-␤arrestin and a truncated form of the TRH receptor (C335 STOP) that lacks these potential phosphorylation sites. The truncated receptor was expressed and capable of transducing a signal, because TRH increased [Ca 2ϩ ] i if G q ␣ was cotransfected with the C335STOP receptor (Table I). Activation of the truncated TRH receptor with TRH did not cause GFP-␤-arrestin to move to the membrane, however (Fig. 7).
Because it has been suggested that a receptor-␤-arrestin complex can signal directly to the mitogen-activated protein kinase pathway (3), we asked whether TRH can activate mitogen-activated protein kinase in Fq/11 cells expressing the TRH receptor with and without G q ␣. We did not detect significant TRH effects on the abundance of phosphorylated mitogen-activated protein kinases on Western blots using cells that had been serum-starved for 24 h and treated with 1 M TRH for up to 20 min, regardless of expression of G q ␣. Because background mitogen-activated protein kinase activity was high and TRH receptor levels were low, these results do not rule out the possibility that TRH activates the mitogen-activated protein kinase pathway weakly in transfected Fq/11 cells, but any response must be very small.

DISCUSSION
Previous studies have led to contradictory conclusions concerning the requirement for G proteins in the sequestration of G protein-coupled receptors coupled to the G q family (15,17,24,25). By using cells lacking the ␣ subunits of G q and G 11 , we have obtained definitive evidence that the TRH receptor undergoes agonist-dependent sequestration in the absence of its normal signal transducers, G q and G 11 . It is improbable that the receptor internalized as a result of coupling to some other G protein. Although the TRH receptor has been reported to exert effects on ion channels via G i and G o (26 -28), pertussis toxin did not alter agonist-dependent internalization in Fq/11 cells. The TRH receptor has been reported to couple to phospholipase C via G s in oocytes (29), but neither TRH nor bradykinin evoked any calcium response in Fq/11 cells unless the cells were transfected with G q ␣. The TRH receptor was not overexpressed in mouse fibroblasts. Assuming that 15% of Fq/11 cells expressed receptors, because 15% of the cells exhibited a calcium response to TRH when G q ␣ was co-transfected, we calculate that there were no more than 30,000 TRH binding sites/cell. This is below the level in pituitary cells (at least 50,000) and well below the levels in other transfected cell models (HEK, COS, HeLa, Chinese hamster ovary, C6) (16,30,31).
Petrou et al. (15) suggested that TRH receptor internalization involves a G q -independent pathway requiring the CNC sequence in the carboxyl-terminal tail and perhaps an additional, G q -dependent pathway, based on the study of mutant receptors and the finding that antibody to an extracellular, amino-terminal epitope tag induced internalization but not signaling. They also suggested that G q ␣ and the TRH receptor internalize in the same vesicles. Milligan and co-workers (32) find that TRH down-regulates G q ␣ and causes an increased accumulation of G q ␣ in endocytic vesicles, but they do not think that the receptor and G q cointernalize because of large discrep-ancies in the half-times for receptor and G q ␣ internalization. Our results prove that G q/11 is not needed for internalization but do not establish whether the receptor co-internalizes with G q/11 when the appropriate G protein is available. Gershengorn and co-workers (18) suggest that G protein coupling and phospholipase C activation are both necessary for TRH receptor sequestration, because the phospholipase C inhibitor U73122 inhibits sequestration and several mutant receptors that bind TRH with high affinity but do not activate phospholipase C fail to internalize (18). An alternative interpretation is that mutants unable to activate phospholipase C do not internalize because they do not assume the activated conformation necessary for interaction with receptor kinases or ␤-arrestin as well as G protein. U73122 exerts multiple effects on calcium homeostasis and other processes (33)(34)(35) and may have impaired TRH-driven internalization by a nonspecific action. Uncertainty regarding the requirement for G q coupling is not limited to the TRH receptor, because U73122 inhibits the internalization of muscarinic receptors (25), and sequestration of bombesin receptors has been reported to require G protein interaction (24).
It was not previously known whether receptors needed to couple to G q either to promote an activated receptor conformation or to promote phosphorylation by receptor kinases to bind ␤-arrestin. Our results, which show that GFP-␤-arrestin moves to the membrane in response to TRH in Fq/11 cells, confirm the role of ␤-arrestin in TRH receptor sequestration and prove that the process does not require G protein interaction. By demonstrating that GFP-␤-arrestin does not bind to the carboxylterminal-truncated C335STOP receptor, we have established the molecular basis for its failure to internalize. These results suggest that phosphorylation sites between Thr-342 and Ser-391 are important for ␤-arrestin binding and receptor sequestration. Although ␤-arrestin appears to be involved in the internalization of most G protein-coupled receptors, there are exceptions, including the G q -coupled M1 and M3 muscarinic receptors (36).
The maximal extent of TRH receptor sequestration was relatively low in the Fq/11 cells and in wild type mouse fibroblasts, below 40%, whereas the extent of internalization is above 80% when the receptor is expressed in HEK293, HeLa, or C6 cells (16,18). Furthermore, TRH receptor sequestration in Fq/11 cells was increased by overexpression of ␤-arrestin. Because the extent of internalization correlates with the level of expression of ␤-arrestin and receptor kinases in a number of cell lines (37), it seems likely that ␤-arrestin was expressed at low levels in the Fq/11 cells. The failure of the V53D mutant ␤-arrestin to act as a dominant negative is not completely surprising, because the dominant negative effect is only partial for the ␤2-adrenergic receptor (5,7).
Previous work had shown that internalization of G s -and G i/ G o -coupled receptors can occur without G protein involvement. The ␤2-adrenergic receptor can internalize in cyc Ϫ and unc S49 cells (with absent or nonfunctional G s ␣, respectively) (38,39), and at least some G i/ G o -coupled receptors can internalize normally in pertussis toxin-treated cells (40). We have now shown that ␤-arrestin binding and internalization of G qcoupled receptors can also occur in the absence of G protein coupling. ␤-Arrestin apparently discriminates between the agonist-activated and the inactive TRH receptor either by recognizing the active conformation of the receptor or by recognizing a protein that interacts with the activated receptor. Further work will be required to establish the molecular basis of this process.