Report

Activity-Dependent Internalization of Smoothened Mediated by ß-Arrestin 2 and GRK2

See allHide authors and affiliations

Science  24 Dec 2004:
Vol. 306, Issue 5705, pp. 2257-2260
DOI: 10.1126/science.1104135

Abstract

Binding of Sonic Hedgehog (Shh) to Patched (Ptc) relieves the latter's tonic inhibition of Smoothened (Smo), a receptor that spans the cell membrane seven times. This initiates signaling which, by unknown mechanisms, regulates vertebrate developmental processes. We find that two molecules interact with mammalian Smo in an activation-dependent manner: G protein–coupled receptor kinase 2 (GRK2) leads to phosphorylation of Smo, and β-arrestin 2 fused to green fluorescent protein interacts with Smo. These two processes promote endocytosis of Smo in clathrin-coated pits. Ptc inhibits association of β-arrestin 2 with Smo, and this inhibition is relieved in cells treated with Shh. A Smo agonist stimulated and a Smo antagonist (cyclopamine) inhibited both phosphorylation of Smo by GRK2 and interaction of β-arrestin 2 with Smo. β-Arrestin 2 and GRK2 are thus potential mediators of signaling by activated Smo.

Hedgehog (Hh) signaling is mediated by regulation of a protein called Smoothened (Smo) that spans the cell membrane seven times (7MS), activation of which sets in motion transcriptional events that control growth and patterning in vertebrate development (1, 2). Dysregulated Smo activity also leads to several forms of cancer (37). Hh binds to a receptor that spans the cell membrane 12 times, Patched (Ptc), and relieves inhibitory control of Smo by Ptc. However, almost nothing is known of the mechanisms operating just downstream of Smo to mediate and modulate its actions. β-Arrestins are cytosolic proteins that bind to most activated 7MS receptors after the receptors have been phosphorylated by GRKs, which promotes internalization of the receptors and some forms of signaling (8, 9). Elements that regulate receptor functions often show activity-dependent interaction with the receptor. Thus, we tested the hypothesis that β-arrestins and GRKs might interact with and regulate Smo.

β-Arrestin 2 tagged with green fluorescent protein (βarr2-GFP), when expressed alone, was diffusely distributed throughout the cytoplasm of human embryonic kidney 293 (HEK293) cells (Fig. 1A). Expression of βarr2-GFP together with a fusion protein of Smo with a portion of Myc (Myc-Smo), which is constitutively active when overexpressed in mammalian cells (10, 11), led to redistribution of βarr2-GFP from the cytosol to the plasma membrane, where it was found in a punctate pattern (Fig. 1B). Under these conditions, 32 ± 6% of cells expressing βarr2-GFP demonstrated such translocation (Fig. 1C). This pattern is similar to that observed after recruitment of βarr2-GFP to other activated 7MS receptors (12, 13). Furthermore, in such cells, more Smo was detected in the cytosol (fig. S1, B to D) than was seen when Smo was expressed alone (fig. S1A). β-Arrestin1-GFP, when expressed alone, was also evenly distributed in the cytosol (fig. S2A); however, no recruitment to Smo was observed (fig. S2B).

Fig. 1.

Localization of βarr2-GFP to the plasma membrane in cells overexpressing Smo. Confocal images of βarr2-GFP expressed alone (A) or with Myc-Smo (B) in HEK293 cells. (C) Effects of Ptc, ShhN, and GRK2 on recruitment of βarr2-GFP to the plasma membrane. βarr2-GFP was expressed with Myc-Smo (bar 1), Myc-Smo and FLAG-Ptc (bar 2), Myc-Smo and FLAG-Ptc (cocultured with HEK293 expressing ShhN, an active form of Shh) (bar 3), or Myc-Smo and GRK2 (bar 4) in HEK293 cells. Data are presented as the percentage of βarr2-GFP–expressing cells with recruitment of βarr2-GFP and are the means ± SEM of three independent experiments. *P < 0.05 (compared with bar 1) and **P < 0.005 (compared with bar 2) (unpaired t test). Scale bar, 10 μm.

We tested whether Ptc, an inhibitor of Smo (14), might itself bind β-arrestin–GFP. Ptc covalently tagged with a FLAG epitope, when expressed alone, was distributed primarily at the plasma membrane (fig. S1E). Localization of β-arrestin1–GFP (15) or βarr2-GFP (Fig. 2A) was not altered in cells also expressing Ptc, and expression of βarr2-GFP did not alter the distribution of Ptc (fig. S1, F to H). However, expression of Ptc together with Myc-tagged Smo and βarr2-GFP markedly inhibited localization of βarr2-GFP at the plasma membrane (Fig. 1B vs. Fig. 2B). Under these conditions, only 4 ± 2% of cells expressing βarr2-GFP demonstrated translocation (Fig. 1C). These data suggest that βarr2-GFP binds to the active form of Smo. Secreted Sonic hedgehog ligand (Shh) binds to Ptc and blocks its inhibitory effect on Smo activity (10, 16). Thus, Shh treatment might be expected to restore association of βarr2-GFP with Smo in cells expressing Ptc, Smo, and βarr2-GFP. Therefore, we mixed cells expressing the secreted active form of Shh (ShhN) with cells expressing Ptc, Smo, and βarr2-GFP and cultured them for 12 hours. Of the βarr2-GFP–expressing cells, 22 ± 2% showed translocation of βarr2 under these conditions, compared with only 4 ± 2% when the added cells had been transfected with empty vector rather than ShhN (Fig. 1C).

Fig. 2.

Ptc and cyclopamine inhibit membrane recruitment of βarr2-GFP, and SAG relieves Ptc inhibition of recruitment of βarr2-GFP to the plasma membrane. Confocal images of βarr2-GFP expressed with FLAG-Ptc (A), FLAG-Ptc and Myc-Smo (B), and Myc-Smo (C and D) in HEK293 cells. Cells were left untreated (A to C) or treated with 8 μM cyclopamine (D) at 37°C for 5 min. Recruitment of βarr2-GFP to Smo was ablated by treatment with cyclopamine (C versus D), but not with dimethylsulfoxide (DMSO), a vehicle for cyclopamine. (E) Effect of cyclopamine on interaction of βarr2-GFP with Smo. HEK293 cells stably expressing Myc-Smo and βarr2-GFP were left untreated (lane 1) or treated with cyclopamine (6 μM, lane 2) at 37°C for 1 hour. Cell extracts were immunoprecipitated with anti-Myc affinity gel. Immunoprecipitates were immunoblotted with antibodies against βarr2 (A2CT) (top) or antibodies against Myc (middle). Whole-cell lysates were immunoblotted with A2CT antibodies (bottom). (F and G) Confocal images of cells stimulated with 0.3 μM SAG at 37°C for 0 min (F) and 30 min (G). HEK293 cells were transfected with βarr2-GFP, FLAG-Ptc, Myc-Smo, and GRK2. Scale bar, 10 μm. Representative images or a blot of three independent experiments are shown.

To further explore whether βarr2-GFP binds to the active form of Smo, we used cyclopamine, a compound that acts as a direct antagonist of Smo (10). Recruitment of βarr2-GFP to the plasma membrane was abolished when cells were treated with 8 μM cyclopamine for 5 min at 37°C (Fig. 2, C and D). Further evidence for interaction of βarr2 and Smo within a protein complex and for the inhibitory effect of cyclopamine on such an interaction was obtained from cellular coimmunoprecipitation studies in which cyclopamine virtually eliminated the constitutive association of these two proteins (Fig. 2E).

In Drosophila, stimulation by Shh leads to activation and phosphorylation of Smo (17). However, the kinase responsible for Smo phosphorylation has not been identified. Because the interaction of β-arrestin with 7MS receptors is generally initiated by GRK-mediated phosphorylation of the receptor (8, 9), we tested whether Smo phosphorylation might be mediated by GRK2, a ubiquitously expressed member of this kinase family. Expression of Myc-Smo in [32P]orthophosphate-labeled HEK293 cells revealed phosphorylation of Smo by endogenous kinases (Fig. 3). Transfection of the cells with small interfering RNA (siRNA) directed against GRK2, which decreased GRK2 expression by ∼95% (Fig. 3), led to a decrease in Smo phosphorylation relative to that in cells treated with control siRNA (Fig. 3; fig. S3). These data suggest that GRK2 contributes to phosphorylation of Smo in vivo.

Fig. 3.

Phosphorylation of Smo mediated by GRK2. Decreased phosphorylation of Smo in cells lacking GRK2. HEK293 cells were transfected with control siRNA and DNA empty vector (lane 1), or Myc-Smo (lanes 2 and 3) along with control siRNA (lane 2) or siRNA directed against GRK2 (lane 3). Cells were incubated with [32P]orthophosphate at 37°C for 1 hour. Proteins from cell extracts were either immunoblotted with antibodies against GRK2 (top) or immunoprecipitated with anti-Myc affinity gel. Immunoprecipitates were either immunoblotted with antibodies to Myc (middle) or processed for autoradiography (bottom). A representative blot of three independent experiments is shown.

Because GRK2 influences phosphorylation of Smo, we tested its effect on the recruitment of βarr2-GFP to Smo. In the absence of coexpressed GRK2, 32 ± 6% of cells expressing βarr2-GFP demonstrated βarr2-GFP translocation to the plasma membrane (Fig. 1C). Expression of GRK2 with Smo increased the percentage of cells with recruitment of βarr2-GFP to the plasma membrane to 78 ± 4% (Fig. 1C). Amounts of Smo expressed were similar in the presence or the absence of overexpressed GRK2 (15). GRK2 alone had no effect on βarr2-GFP translocation in the absence of overexpressed Smo (15).

An agonist for Smo, benzo[b]thiophene-2-carboxamide, 3-chloro-N-[4-(methylamino)cyclohexyl]-N-{[3-(4-pyridinyl)phenyl]methyl}-(9CI) (SAG), has been identified that relieves the inhibitory effect of Ptc on Smo (18). Incubation of cells with SAG (0.03 nM to 3 μM) at 37°C for 1 hour did not further enhance recruitment of βarr2-GFP to the plasma membrane in HEK293 cells expressing Smo and βarr2-GFP, probably because overexpressed Smo is already constitutively active (11, 15). However, exposure of cells to SAG (0.3 μM) at 37°C for 30 min or 1 hour did relieve Ptc inhibition of translocation of βarr2-GFP to the plasma membrane by Ptc in HEK293 cells expressing Ptc, Smo, GRK2, and βarr2-GFP (Fig. 2, F and G; fig. S4). The increased recruitment of βarr2-GFP to the plasma membrane was dependent on the concentration of SAG used, with 50% maximal effect at ∼30 nM (fig. S5).

These experiments indicate that recruitment of βarr2-GFP to Smo faithfully monitors the activation state of Smo. Because endogenous GRK2 appears to phosphorylate Smo (Fig. 3; fig. S3), we next tested whether Smo phosphorylation in the presence of endogenous or transfected GRK2 also reflects the activation state of Smo. Expression of GRK2 with Smo led to more phosphorylation of Smo (Fig. 4). In the absence of GRK2 coexpression, stimulation of cells with SAG for 15 min resulted in no change of Smo phosphorylation; however, in the presence of GRK2 expression, SAG stimulation led to a 1.3-fold increase of Smo phosphorylation (Fig. 4). Exposure of cells to the antagonist cyclopamine for 15 min led to a 30% decrease of Smo phosphorylation in the absence of coexpressed GRK2 and a 55% decrease in the presence of coexpressed GRK2 (Fig. 4). Expression of Ptc with Smo in the presence or absence of coexpressed GRK2 resulted in a 35 or 30% decrease in Smo phosphorylation, respectively (Fig. 4). These data indicate that phosphorylation of Smo also reflects its activation state.

Fig. 4.

Effects of GRK2, Ptc, SAG, and cyclopamine on Smo phosphorylation. HEK293 cells were transfected with vector (bar 1), or Myc-Smo (bars 2 to 9), and FLAG-Ptc or GRK2 (bars 5 to 9) as indicated. Cells were labeled with [32P]orthophosphate at 37°C for 1 hour and then left untreated or treated with 0.3 μM SAG (bars 3 and 7) or 2 μM cylopamine as indicated at 37°C for 15 min. Proteins from cell extracts were immunoprecipitated with anti-Myc affinity gel. Immunoprecipitated Smo was processed for autoradiography. Data are presented as fold changes of Smo phosphorylation over that in cells transfected with control vector. *P < 0.005 (compared with bar 1); **P < 0.02 and ***P < 0.0001 (compared with bar 2); ****P < 0.005 (compared with bar 6) (unpaired t test). The results shown are the means ± SEM of three independent experiments.

SAG 1.1 (19) and cyclopamine regulated Smo trafficking as well (fig. S6). When a fusion of Smo with yellow fluorescent protein (Smo-YFP) was expressed in HEK293 cells, it was expressed at the plasma membrane and in the cytosol, possibly because of the presence of partially processed or constitutively internalized Smo (fig. S6A). Treatment of cells with SAG 1.1 for 1 hour led to internalization of Smo-YFP (fig. S6B). The internalized Smo-YFP was recycled back to the plasma membrane after treatment of the cells with cyclopamine for 4 hours (fig. S6C). SAG 1.1 did not affect the rate of internalization of the transferrin receptor (15).

β-Arrestins target 7MS receptors such as β2-adrenergic receptor (β2AR) to clathrin-coated pits to mediate their internalization (8). To test whether Smo undergoes similar regulation, we expressed both Smo-YFP and β2AR-RFP (red fluorescent protein) in the same HEK293 cells and stimulated them for 1 hour with SAG 1.1 and isoproterenol. Both receptors were internalized and colocalized (fig. S7). Expression of a dominant-negative mutant of dynamin (in which Lys44 was replaced by Ala) also blocked internalization of Smo-YFP stimulated by SAG 1.1 (fig. S8). Moreover, in cells coexpressing Myc-Smo and βarr2-GFP, endogenous clathrin and βarr2-GFP colocalized in a punctate pattern at the plasma membrane (Fig. 5, A to C). These data indicate that expression of βarr2 causes association of Smo with clathrin-coated pits for internalization.

Fig. 5.

Internalization of Smo-YFP via clathrin-coated pits. (A to C) βarr2 targets Smo to clathrin-coated pits. Confocal images of βarr2-GFP (A) and endogenous clathrin (B) in the same HEK293 cell overexpressing Myc-Smo and βarr2-GFP. Arrows point to where both βarr2 and clathrin were localized (C). (D to I) HEK293 cells were transfected with control siRNA (D and E), siRNA directed against βarr2 (F and G), and siRNA directed against GRK2 (H and I) and subsequently transfected with Smo-YFP, which was detected by confocal microscopy (D to I). Cells were treated with 0.3 μM SAG 1.1 for 0 min (D, F, H) and 30 min (E, G, I) at 37°C. Scale bar, 10 μm. Representative images of three independent experiments are shown.

We used siRNA directed against βarr2 or GRK2 to reduce the amounts of endogenous βarr2 or GRK2 expressed in HEK293 cells [to ∼10% of that in control cells (fig. S9, A and B)]. Smo-YFP expressed in cells treated with control siRNA was distributed at the plasma membrane and in the cytosol (Fig. 5D). Stimulation of cells with SAG 1.1 for 30 min resulted in internalization of Smo-YFP (Fig. 5, D and E). However, SAG 1.1-induced internalization of Smo-YFP was abolished in cells transfected with siRNA directed against βarr2 or GRK2 (Fig. 5, F to I). The percentage of cells showing SAG 1.1-induced internalization of Smo was increased in cells overexpressing βarr2 or GRK2 (fig. S10).

Here, we have demonstrated that βarr2 and GRK2 mediate clathrin-dependent internalization of Smo. However, it is possible that they may also modulate or mediate aspects of Smo signaling as is the case for other 7MS receptors (8, 20, 21). Indeed, β-arrestin 2 knockdown in zebrafish embryos by morpholino antisense leads to a lethal developmental phenotype (22) that is remarkably similar to that seen after genetic knockouts of either Smo or Gli2 (2325).

Although Smo is reported to activate Gαi directly or indirectly in frog melanophores (26), no genetic evidence to support coupling of Smo to G proteins has been reported. Several cytosolic components downstream of Smo such as Costal2 (Cos2), Fused (Fu), Suppressor of Fused, and Cubitus interruptus (Ci) have been identified in Drosophila, and the protein complex containing Cos2, Fu, and Ci has been recently reported to associate with Smo via Cos2 (2730). However, βarr2 and GRK2 interact with mammalian Smo in an activation-dependent manner and, thus, may provide a platform for development of screening assays to discover ligands that directly regulate the activity of this important oncogenic receptor that might be useful as therapeutic agents.

Supporting Online Material

www.sciencemag.org/cgi/content/full/306/5705/2257/DC1

Materials and Methods

Figs. S1 to S10

References and Notes

References and Notes

Stay Connected to Science

Navigate This Article