Coated Pit-mediated Endocytosis of the Type I Transforming Growth Factor-β (TGF-β) Receptor Depends on a Di-leucine Family Signal and Is Not Required for Signaling*

Background: The role of TGF-β receptor endocytosis in signaling is controversial. Results: Two clathrin-mediated endocytosis signals induce internalization of the type I TGF-β receptor (TβRI). Endocytosis-defective TβRI mutants signal, whereas endocytosis-enhanced TβRI exhibits reduced Smad signaling. Conclusion: TβRI endocytosis is dispensable for Smad signaling and correlates with reduced response. Significance: Understanding the role of TGF-β receptor endocytosis is relevant to cell biology and cancer. The roles of transforming growth factor-β (TGF-β) receptor endocytosis in signaling have been investigated in numerous studies, mainly through the use of endocytosis inhibitory treatments, yielding conflicting results. Two potential sources for these discrepancies were the pleiotropic effects of a general blockade of specific internalization pathways and the scarce information on the regulation of the endocytosis of the signal-transducing type I TGF-β receptor (TβRI). Here, we employed extracellularly tagged myc-TβRI (wild type, truncation mutants, and a series of endocytosis-defective and endocytosis-enhanced mutants) to directly investigate the relationship between TβRI endocytosis and signaling. Our findings indicate that TβRI is targeted for constitutive clathrin-mediated endocytosis via a di-leucine (Leu180-Ile181) signal and an acidic cluster motif. Using Smad-dependent transcriptional activation assays and following Smad2/3 nuclear translocation in response to TGF-β stimulation, we show that TβRI endocytosis is dispensable for TGF-β signaling and may play a role in signal termination. Alanine replacement of Leu180-Ile181 led to partial constitutive activation of TβRI, resulting in part from its retention at the plasma membrane and in part from potential alterations of TβRI regulatory interactions in the vicinity of the mutated residues.

The roles of transforming growth factor-␤ (TGF-␤) receptor endocytosis in signaling have been investigated in numerous studies, mainly through the use of endocytosis inhibitory treatments, yielding conflicting results. Two potential sources for these discrepancies were the pleiotropic effects of a general blockade of specific internalization pathways and the scarce information on the regulation of the endocytosis of the signaltransducing type I TGF-␤ receptor (T␤RI). Here, we employed extracellularly tagged myc-T␤RI (wild type, truncation mutants, and a series of endocytosis-defective and endocytosisenhanced mutants) to directly investigate the relationship between T␤RI endocytosis and signaling. Our findings indicate that T␤RI is targeted for constitutive clathrin-mediated endocytosis via a di-leucine (Leu 180 -Ile 181 ) signal and an acidic cluster motif. Using Smad-dependent transcriptional activation assays and following Smad2/3 nuclear translocation in response to TGF-␤ stimulation, we show that T␤RI endocytosis is dispensable for TGF-␤ signaling and may play a role in signal termination. Alanine replacement of Leu 180 -Ile 181 led to partial constitutive activation of T␤RI, resulting in part from its retention at the plasma membrane and in part from potential alterations of T␤RI regulatory interactions in the vicinity of the mutated residues.
Cytokines of the transforming growth factor-␤ (TGF-␤) family regulate cellular homeostasis, affecting multiple biological processes such as development, differentiation, immune responses, cell growth arrest, and tissue/organ regeneration and maintenance (1)(2)(3)(4)(5)(6)(7). They are involved in a broad range of disorders, including cancer (where they function both as tumor suppressors and tumor promoters), vascular and skeletal diseases, primary pulmonary hypertension, and pre-disposition to angioproliferative disorders (8 -13). TGF-␤ signals by binding to two receptor serine/threonine kinases, type I and type II (T␤RI 3 and T␤RII), sometimes assisted by structurally diverse co-receptors ranging from the type III TGF-␤ receptor proteoglycan (betaglycan), and endoglin to glycosylphosphatidylinositol-anchored co-receptors (1, 3, 8, 11, 12, 14 -17). The type I and type II receptors are essential and sufficient for the canonical signaling via Smad proteins and for many non-Smad signaling responses. TGF-␤ binding triggers T␤RII-mediated phosphorylation and activation of T␤RI, which proceeds to phosphorylate the TGF-␤ receptor-activated Smads (Smad2/ 3). The phosphorylated Smad2/3 then associate with Smad4 and translocate into the nucleus, where together with other transcription factors they positively and negatively regulate the transcription of a large array of target genes in a cell contextdependent manner (7, 12, 14, 18 -21). In general, endocytosis regulates receptor cell surface levels of various receptors, potentially affecting signaling output (22)(23)(24)(25)(26)(27)(28)(29)(30). Studies on the endocytosis of active T␤RI and T␤RII were hampered by the inability to use iodinated TGF-␤ to follow their internalization, which is masked by the abundance of receptors and extracellular matrix-associated proteins that bind the ligand. Internalization of the TGF-␤ receptors was studied originally by chimeric GM-CSF/TGF-␤ receptors (25,27,(31)(32)(33), which may differ from the native receptors, as suggested by the failure of the latter to interact with the chimeric receptors (31). We and others studied the endocytosis of extracellular (EC) epitope-tagged T␤RII and T␤RI (24,26,34,35). A general feature that emerged from most studies on TGF-␤ receptor internalization is that clathrin-mediated endocytosis (CME) is the major route of their internalization (24 -27, 34 -36); a potential contribution of caveolar internalization is more controversial and was * This work was supported in part by Israel Science Foundation Grants 158/09 reported in some studies but not in others (24 -27, 34 -36). In addition, the type III TGF-␤ receptor has been demonstrated to modulate T␤RII and T␤RI endocytosis via interaction with ␤-arrestin2 (37,38). Importantly, the roles of TGF-␤ receptor endocytosis in signaling are still controversial. Thus, CME was linked to either enhanced (25,26) or reduced (27,39) Smad signaling in response to TGF-␤, although other reports suggested that Smad signaling is insensitive to inhibition of CME (40,41). Moreover, opposite effects of caveolar endocytosis on TGF-␤ non-Smad signaling were reported (41,42). However, most of these studies have relied on treatments that result in a general inhibition of CME or caveolar endocytosis. Under such conditions, the inhibitory effects are not restricted to the TGF-␤ receptors and can alter the trafficking and cellular distribution of multiple signaling proteins. Such effects could seriously limit data interpretation.
An alternative approach to assess the role of TGF-␤ receptor endocytosis in signaling, which circumvents the above problems, is the use of TGF-␤ receptor mutants with defective or enhanced endocytosis. Here, we applied this approach to T␤RI, which directly phosphorylates Smad2/3. We find that T␤RI undergoes constitutive internalization via CME, mediated by an internalization signal from the di-leucine family (Leu 180 -Ile 181 ), which functions in conjunction with an acidic cluster motif. Using T␤RI mutants lacking these signals (CME-defective) or containing an additional strong CME targeting motif (YRIL; CME-enhanced), we show that T␤RI endocytosis is not required for Smad signaling; rather, it correlates with reduced signaling in response to TGF-␤, in line with a role in signal termination. An additional effect of mutating Leu 180 -Ile 181 to alanine is partial constitutive activation of T␤RI, which apart of retaining the receptor at the plasma membrane, may be most likely due to the proximity of this sequence (located in the ␣GS1 region) to the binding sites of FKBP12 (43), Smads (44,45), and additional regulatory proteins (46).
Plasmids-Expression vectors encoding human T␤RI (in pcDNA3) or T␤RII (in pcDNA1) with an EC myc epitope tag were described by us earlier (34,48,51,52). Constitutively active myc-T␤RI-T204D was generated by inserting the T204D point mutation into myc-T␤RI by the QuikChange mutagenesis kit (Stratagene). The TGF-␤-responsive luciferase reporter constructs p3TP-Luc(ϩ) in pGL3 (53) and (CAGA) 12 -Luc in pGL3ti (54) were a gift from P. Knaus (Free University of Berlin, Germany). The latter construct is considered to be highly specific for TGF-␤-mediated Smad transcriptional activation, due to the specific binding of Smad3 and Smad4 to the CAGA boxes in the promoter (54).
Mutagenesis-The truncation, alanine substitution, and endocytosis-enhanced mutants of human myc-T␤RI employed in this study are depicted in Fig. 1. The C-terminal truncation mutants (T␤RI-V158, -P164, -D179, and -D183) were generated by PCR using QuikChange, with the myc-T␤RI plasmid serving as a template, as described by us earlier for T␤RII mutagenesis (34). A forward primer recognizing the T7 promoter region of the plasmid was employed, and one of the following primers (containing a stop codon and a NotI restriction site) served as the reverse primer (the consensus stop codon is depicted in boldface letters, and the NotI restriction site is underlined): 1) T␤RI-V158, 5Ј-TTTTCCTTTTGCGGCCGC-CTACACTCGATGGTGAATGACAGTGC-3Ј; 2) T␤RI-P164, 5Ј-GGAGCGGCCGCTTAGTCCTCTTCATTTGG-3Ј; 3) T␤RI-D179, 5Ј-TTTTCCTTTTGCGGCCGCCTAGTCTTT-CGTAGTACCCTCTG-3Ј; and 4) T␤RI-D183, 5Ј-TTTTCCT-TTTGCGGCCGCCTAATCATAAATTAAGTCTTTCAACG-3Ј. The PCR products were digested with HindIII and NotI and inserted into pcDNA3 digested with the same enzymes. Alanine substitution mutants of myc-T␤RI were generated using the same mutagenesis kit; the forward primer for each mutant is shown (with Ala codon in boldface letters), and the complementary sequence served as the reverse primer as follows: 1) T␤RI-2A, here, myc-T␤RI in pcDNA3 served as a template, with the forward mutagenesis primer 5Ј-TTGAAAGACGCT-GCTTATGATATGACAACGTCAGGTTC-3Ј; 2) T␤RI-3A, the template was myc-T␤RI in pcDNA3, with the forward mutagenesis primer 5Ј-AGTGCCAAATGCAGCGGCCCCT-TCATTAGATCGCCCTTTTATTTC-3Ј; and 3) T␤RI-3A2A, here, T␤RI-3A served as a template, employing the same mutagenesis primers used to generate T␤RI-2A. For making the endocytosis-enhanced T␤RI-YRIL mutant, a primer targeting the T7 promoter was used as the forward primer. A reverse primer (as follows) was designed such that the DNA sequence encoding the YRIL peptide (italics) was inserted between the original C-terminal residue of myc-T␤RI (Met 503 ) and the stop codon (boldface), immediately followed by a HindIII restriction site (underlined): 5Ј-CAACAGGAAGGCATCAAATGTATC-GGATTTTATAAAAGCTTGGG-3Ј. The PCR product was digested with HindIII and inserted into pcDNA3 digested with the same enzyme. The T␤RI-2A-YRIL was generated by using T␤RI-YRIL as template with the same mutagenesis primers used to generate T␤RI-2A. All mutants were verified by sequencing.
Internalization Measurements-COS7 or R1B-L17 cells growing on glass coverslips were transfected with plasmids encoding one of the myc-T␤RI-based constructs. After 24 h, the internalization of the myc-tagged receptors was quantified by the point confocal method employing the fluorescence recovery after photobleaching setup under nonbleaching illumination conditions as described by us earlier (34). Briefly, the cell surface receptors were labeled at 4°C as described for immunofluorescent labeling of myc-tagged receptors. Labeled cells were either fixed immediately with 4% paraformaldehyde or warmed to 37°C for the indicated periods to allow endocytosis; they were then transferred back to 4°C, fixed, and mounted for immunofluorescence as above. Endocytosis was quantified by measuring the reduction in the fluorescence intensity levels at the plasma membrane, focusing the laser beam through the ϫ63 objective at defined spots (1.86 m 2 ) in the focal plane of the plasma membrane away from vesicular staining, passing the fluorescence through a pinhole in the image plane to make it a true FIGURE 1. Schematic representation of T␤RI and its truncation, alanine substitution, and endocytosis-enhanced mutants. The construction of the mutants, starting from myc-T␤RI in pcDNA3, is described under "Experimental Procedures." The numbers to the right of the truncation mutants (not drawn to scale) indicate the last amino acid residue after which the stop codon was inserted. Human T␤RI consists of 503 amino acids (for sequence see Ref. 73). The last residue of the EC domain is at position 126, and the putative transmembrane (TM) domain ends at residue 146, followed by the cytoplasmic (CY) domain. The GS region consists of residues 175-205. The diagram at the top depicts the sequence of the membrane-proximal region present in whole or in part in the T␤RI truncation mutants along with the amino acid residues most relevant for various mutants. Underlined is the ␣GS1 region. The numbers above or below specific amino acid residues represent their location in the T␤RI sequence. Substitutions by alanine or the addition of the YRIL peptide sequence are indicated by boldface letters.
Down-regulation of Cell-surface myc-T␤RI-Down-regulation of T␤RI was assayed by measuring the time dependence of the reduction in the level of biotinylated cell-surface receptors (57). Mv1Lu cells in 10-cm plates were transfected by Lipofectamine TM 2000 with 4 g of DNA of an expression plasmid for myc-T␤RI or empty vector. After 24 h, cells were rinsed three times with ice-cold HBSS/HEPES and incubated for 45 min with 0.5 mg/ml membrane-impermeable sulfo-NHS-LC-Biotin (Pierce) in borate buffer (10 mM boric acid, 150 mM NaCl, pH 8). Cells were rinsed twice with cold HBSS/HEPES containing 15 mM glycine, followed by incubation at 37°C in complete growth medium for the indicated times (see figure legends). Cells were lysed on ice with RIPA buffer as described (38). After removal of cell debris, 6% (v/v) of each lysate was taken for SDS-PAGE and Western blotting to probe total myc-T␤RI (0.6 g/ml ␣myc) and ␤-actin (1:20,000) followed by peroxidase-G␣M (1:5000) as above. The remainder of each lysate was subjected to immunoprecipitation by 2 g of ␣myc IgG and protein G-Sepharose (30 l) as described (58). Immunoprecipitates were rinsed and subjected to SDS-PAGE (10% gel) and blotting as described under "Smad Phosphorylation Assay," using peroxidase-streptavidin (Invitrogen; 1:2500) and ECL to detect biotinylated myc-T␤RI. Densitometry was as described under "Smad Phosphorylation Assay." Transcriptional Activation Assays-Luciferase reporter transcriptional activation assays were performed on Mv1Lu-de-rived cell lines using two protocols, adapted for unperturbed and endocytosis-inhibited cells. Unperturbed cells in 6-well plates were co transfected with the following: (i) 1 g of luciferase reporter construct, p3TP-Luc(ϩ) or (CAGA) 12 -Luc; (ii) 0.3 g of pRL-TK (Renilla luciferase, Promega); and (iii) 1 g of empty vector or a vector encoding a myc-T␤RI construct or myc-T␤RII. At 12 h post-transfection, the cells were split in parallel onto glass coverslips in 6-well plates (to assess the level of the myc-TGF-␤ receptor at the cell surface) and onto 96-well plates (for the luciferase assays). After another 24 h, the cells in the 96-well plates were serum-starved (2 h), stimulated (or not) with 300 pM TGF-␤1 (18 h in starvation medium), lysed, and analyzed by the DLR assay system. The results were normalized for transfection efficiency using the Renilla luminescence. To calibrate for the expression levels of the myc-TGF-␤ receptors at the cell surface, the latter were determined on the coverslipplated cells after fluorescent labeling of the cell-surface receptors, using the point-confocal fluorescence intensity measurement as described under "Internalization Measurements." Transcriptional activation assays on cells subjected to treatments that inhibit endocytosis were conducted similarly, except that the stimulation with TGF-␤1 (300 pM) was for a shorter period (6 h), during which the cells were subjected (or not) to the inhibition treatment (for treatment description, see "Treatments Affecting Internalization" below).
Treatments Affecting Internalization-Experiments employing internalization-inhibitory treatments were conducted on COS7 or R1B-L17 cells using two protocols, one for testing the effects of a given treatment on endocytosis and the other for studying the effects on transcriptional activation. All endocytosis assays were conducted in HBSS/HEPES/BSA, and all treatments were initiated by a 15-min preincubation (37°C) with the inhibitory drug/medium. The cells were retained under the inhibitory condition throughout the fluorescent labeling and endocytosis measurement. For the luciferase assays, which were conducted in starvation medium, the endocytosis-disrupting treatment was started upon addition of TGF-␤1, and the cells were kept under the same conditions until the lysis step (6 h). The treatments employed were as follows: (i) hypertonic treatment to disrupt the structure of clathrin-coated pits (59) was conducted in HBSS/HEPES/BSA (for endocytosis studies) or starvation medium (for transcriptional activation assays) supplemented with 0.45 M sucrose (34); (ii) treatment with CPZ, which redistributes AP2 from the plasma membrane to endosomes (60), was carried out using 100 M (COS7 cells) or 50 M (R1B-L17 cells) CPZ; (iii) DYN treatment, which inhibits dynamin-dependent internalization (61), was done using 80 M DYN; and (iv) nystatin treatment to inhibit caveolar endocytosis (26, 27, 62) employed 25 g/ml of the drug.

T␤RI Is Endocytosed Mainly through Clathrin-coated Pits
Irrespective of Its Activation-The major aim of this study was to explore the interdependence between the endocytosis of T␤RI, which is the receptor responsible for Smad activation, and its signaling output. First, we set to define the relative contribution of CME and caveolar endocytosis to T␤RI internalization. T␤RI carrying an EC myc epitope tag (myc-T␤RI) was expressed in COS7 cells; the cell surface receptors were labeled at 4°C as described under "Experimental Procedures," followed by shifting the cells to 37°C for defined periods, and measuring T␤RI endocytosis by the point-confocal assay (34). The initial surface distribution of myc-T␤RI prior to warming to 37°C was homogeneous, typical of cell-surface labeling in the absence of endocytosis ( Fig. 2A). Upon incubation at 37°C, the fluorescence shifted in a time-dependent manner to a vesicular pattern characteristic of endocytic vesicles (Fig. 2, B-D). Quantitative measurement by the point-confocal method (34) of fluorescent-labeled myc-T␤RI remaining at the cell surface as a function of the incubation time at 37°C yielded an internalization half-time (t1 ⁄ 2 ) of ϳ13 min (Fig. 2G), close to that found by us earlier for T␤RII (34). Notably, myc-T␤RI internalization was fully blocked by various treatments that inhibit CME, including incubation with sucrose-containing hypertonic medium, CPZ, or DYN (which inhibits both CME and caveolar endocytosis) (Fig. 2, E and G). Conversely, treatment with nystatin (an inhibitor of caveolar endocytosis) had no effect (Fig. 2, F and G). We next extended the experiments on myc-T␤RI endocytosis to R1B-L17 cells, which are a T␤RI-deficient cell line derived from the TGF-␤-responsive Mv1Lu epithelial cells (63), providing a highly suitable system for TGF-␤ signaling studies (50). The sensitivity of R1B-L17 cells to the various endocytosis-inhibiting treatments were similar to those of COS7 cells (supplemental Fig. 1A), with a somewhat faster endocytosis rate in untreated R1B-L17 cells (t1 ⁄ 2 ϳ9 min). To explore whether T␤RI endocytosis depends on its activation, the experiments were repeated with constitutively active myc-T␤RI-T204D; the endocytosis of this mutant and its sensitivity to various inhibitors were identical to myc-T␤RI (Fig. 2H), demonstrating that T␤RI endocytosis does not depend on its activation state. In line with these results, the internalization rate of T␤RI was not affected by ligand (200 pM TGF-␤1) in either COS7 or R1B-L17 cells (data not shown). Although it is still possible that a subpopulation of T␤RI can associate with lipid raft domains and/or caveolae (26,64), these results indicate that the major pathway for T␤RI endocytosis is via clathrin-coated pits, in accord with several former reports on the endocytosis of TGF-␤ receptors (24,25,27,34,35,39,40).
To determine the effects of the treatments that inhibit T␤RI endocytosis on TGF-␤-mediated Smad signaling, we employed the DLR assay using TGF-␤-responsive luciferase reporter constructs. The experiments were conducted in R1B-L17 cells, where the mock-transfected cells lacking T␤RI provide a direct control for potential T␤RI-independent activation by a given treatment. The various CME-inhibiting treatments employed (CPZ, DYN, or hypertonic treatment with sucrose) did not inhibit Smad-responsive transcriptional activation. Rather, they induced variable degrees of constitutive activation (supplemental Fig. 1, B and C), in line with reports on continued or enhanced TGF-␤ signaling in cells subjected to CME-blocking treatments (39 -41). The hypertonic treatment (sucrose) induced some transcriptional activation even in the absence of T␤RI, most likely due to effects involving reduction of the cellular and nuclear volumes by partial dehydration. These variable effects suggest that although endocytosis-blocking treatments are useful for establishing the relative contribution of FIGURE 2. T␤RI is internalized mainly through clathrin-coated pits. COS7 cells were transfected with myc-T␤RI; after 24 h, they were either left untreated, or subjected to an internalization-inhibiting treatment (hypertonic sucrose-supplemented medium, CPZ, DYN, or nystatin; see "Experimental Procedures"). The surface receptors on live intact cells were then labeled at 4°C (time 0) with ␣myc followed by Alexa 546-G␣M FabЈ and incubated for defined intervals at 37°C before being returned to 4°C and fixed (see "Experimental Procedures"). A-F, typical images of myc-T␤RI during endocytosis. Bar, 20 m. The time of incubation at 37°C is designated for each panel. E and F depict cells treated to inhibit CME (sucrose-containing hypertonic medium) or caveolar endocytosis (nystatin), respectively. G and H, quantitative measurements of the endocytosis of myc-T␤RI (G) and constitutively active myc-T␤RI-T204D (H). The fluorescence intensity remaining at the cell surface was measured by the point confocal method ("Experimental Procedures"), focusing the laser beam on defined spots in the plasma membrane focal plane, away from vesicular staining. Results are mean Ϯ S.E. of 150 -200 cells for each time point, taking for each sample the intensity at time 0 as 100%. Because all treatments that inhibit CME (hypertonic treatment, CPZ or DYN) abrogated T␤RI internalization to the same extent, only the results of one such treatment (ϩ sucrose) are shown. specific endocytosis pathways to the internalization of a given receptor, the interpretation of their effects on signaling is confounded by the alteration of the trafficking and cellular distribution of multiple signaling proteins by the same treatment. To bypass this problem, we turned to an alternative approach, based on measuring the signaling responses of endocytosis-defective or endocytosis-enhanced T␤RI mutants.
Identification of the Motifs Required for T␤RI Endocytosis-To generate endocytosis-defective T␤RI mutants, it was first necessary to identify the endocytosis motif(s) responsible for its internalization. Because CME-targeting motifs are cytoplasmic, and T␤RI internalization proceeds mainly via CME (Fig.  2), we initially generated a series of myc-T␤RI mutants with progressive truncations of the cytoplasmic domain (generated by introducing stop codons at the desired positions). These mutants are depicted in Fig. 1 (first four mutants). The first truncation mutant, T␤RI-D183, was chosen to retain the Leu 180 -Ile 181 sequence, based on our previous studies that identified an LI di-leucine family motif on T␤RII as its major CMEtargeting signal (34). T␤RI-D183 was internalized in COS7 cells as fast as the wild type T␤RI (Fig. 3), suggesting that no internalization signals are located beyond Asp 183 . Notably, further truncation at Asp 179 (eliminating Leu 180 -Ile 181 ) reduced the endocytosis rate ϳ2-fold (Fig. 3), indicating that Leu 180 -Ile 181 may play a role in T␤RI internalization. To validate this assumption, we have replaced Leu 180 -Ile 181 by alanines in the context of the full-length T␤RI to generate the T␤RI-2A alanine-replacement mutant (depicted in Fig. 1). The internalization rate of T␤RI-2A was compromised to the same extent as T␤RI-D179 (Fig. 3).
The partial inhibition of the endocytosis of T␤RI-2A and T␤RI-D179 suggests that T␤RI may contain additional endocytosis signals. One possibility is that T␤RI may also undergo caveola-like endocytosis. Yet the failure of nystatin, a known blocker of caveolar endocytosis, to inhibit the internalization of T␤RI(WT) (Fig. 2, F and G) argues against this possibility. This notion is strongly supported by the demonstration that the residual endocytosis of T␤RI-2A is insensitive to nystatin but is fully blocked by CME-blocking treatments (see Fig. 3G). This suggests that in addition to Leu 180 -Ile 181 , there is another CMEtargeting signal in the membrane-proximal region of T␤RI. To identify this signal, we further truncated T␤RI to generate T␤RI-P164 and T␤RI-V158 (Fig. 1), and we measured their endocytosis in COS7 cells. Although the internalization of T␤RI-P164 was similar to that of T␤RI-2A, the endocytosis of T␤RI-V158 was nearly eliminated (Fig. 3). This suggests that FIGURE 3. Endocytosis of T␤RI truncation and alanine replacement mutants. COS7 cells were transfected with the indicated myc-T␤RI mutants (for structure, see Fig. 1) followed by labeling the cell-surface myc-tagged receptor population at 4°C as in Fig. 2. They were then shifted to 37°C for the indicated times to allow endocytosis. A-E, typical images of cells expressing specific myc-T␤RI mutants after internalization for 20 min. A partial inhibition of the internalization (indicated by a reduction in the vesicular staining) relative to T␤RI(WT) was seen for T␤RI-2A and T␤RI-3A; similar results were obtained for the truncation mutants truncation mutants D179 and P164 (not shown). A nearly complete blockade (indicated by the smooth surface labeling after 20 min at 37°C) was observed for the T␤RI-3A2A and T␤RI-V158 mutants. Bar, 20 m. F, quantification of the endocytosis of the various myc-T␤RI mutants by point-confocal microscopy (see "Experimental Procedures" and Fig. 2). Results are mean Ϯ S.E. of 150 -200 cells in each time point. Intensity at time 0 for each sample was taken as 100%. G, endocytosis of T␤RI mutants is sensitive to CME inhibitors but not to an inhibitor of caveolar endocytosis. At 24 h post-transfection, cells were left untreated or treated with either nystatin, CPZ, or sucrose (hypertonic medium) as described under "Experimental Procedures." The surface receptors were then labeled at 4°C as above, followed by a 20-min incubation at 37 or 4°C (time 0) in media containing inhibitors where indicated (see under "Experimental Procedures"). Endocytosis of the myc-T␤RI mutants was quantified by the point confocal method. For each mutant and treatment, the fluorescence intensity of the same sample at time 0 was taken as 100%; the percentage of the fluorescence intensity remaining at the cell surface after 20 min at 37°C was subtracted to obtain the % internalization. By comparing treated with untreated cells for each mutant, nystatin had no significant effects in all cases (p Ͼ 0.05, Student's t test). Treatments that inhibit CME induced a highly significant reduction in the endocytosis of all T␤RI proteins (WT or mutants; ***, p Ͻ 10 Ϫ4 ), except the endocytosis-defective myc-T␤RI-3A2A mutant, whose internalization was marginal already in untreated cells. AUGUST 3, 2012 • VOLUME 287 • NUMBER 32 the additional endocytosis signal resides in the short segment between Val 158 and Pro 164 . Notably, this sequence contains an acidic amino acid cluster (EED, at positions 161-163); such a cluster was reported to act as a CME-targeting signal (65)(66)(67). Mutation of Glu 161 -Glu 162 -Asp 163 to alanines generated the T␤RI-3A mutant (Fig. 1), whose endocytosis rate was reduced to the same extent as T␤RI-2A (Fig. 3). Moreover, similar to T␤RI-2A, the residual endocytosis of T␤RI-3A was blocked by CME-inhibiting treatments but not by nystatin (Fig. 3G), in line with the CME-targeting character of both signals. Importantly, the endocytosis of the double T␤RI mutant containing both the 3A and 2A alanine replacement mutations (T␤RI-3A2A; for schematics, see Fig. 1) was drastically inhibited (Fig. 3). We conclude that the internalization of T␤RI in COS7 cells is mediated by a di-leucine and an acidic cluster CME signals, reminiscent of the situation encountered for insulin-regulated amino peptidase (67).

Endocytosis and Signaling of TGF-␤ Type I Receptor
The endocytosis-defective T␤RI mutants enable studies on the impact of specific inhibition of T␤RI endocytosis on signaling. Such studies can be complemented by the use of a T␤RI mutant that undergoes enhanced CME, revealing the effects of augmented T␤RI internalization on the signaling outcome. To this end, we grafted the strong CME signal YRIL, which belongs to the YXXZ (X indicates any amino acid; Z indicates hydrophobic amino acid) family of CME signals (68,69), onto the C terminus of T␤RI (T␤RI-YRIL; Fig. 1). In line with the CME targeting capability of the YRIL peptide, the endocytosis rate of T␤RI-YRIL was markedly enhanced relative to T␤RI(WT) (Fig.  4, A-D). The enhanced endocytosis is via the CME pathway, as shown by the ability of CME inhibitory treatments (but not nystatin) to block it (Fig. 4E).
Early studies reported some differences between the endocytosis of chimeric TGF-␤/GM-CSF receptors in different cell types (32,33), suggesting that TGF-␤ receptor endocytosis may depend on the cellular context. Thus, after the initial characterization of T␤RI endocytosis signals in the COS7 cell system, we turned to validate the results in Mv1Lu-derived cells, which are better suited for signaling studies. To this end, we measured the internalization of the full-length endocytosis-defective and endocytosis-enhanced T␤RI mutants in R1B-L17 cells. As in the case of COS7 cells, the Leu 180 -Ile 181 motif was required for T␤RI endocytosis, and addition of the strong CME-targeting YRIL signal to T␤RI markedly enhanced its endocytosis (Fig. 5). However, the role of the Leu 180 -Ile 181 signal in T␤RI internalization in the R1B-L17 cells is much more prominent than in COS7 cells, as shown by the near-complete loss of endocytosis in the T␤RI-2A mutant (Fig. 5). The notion that the Leu 180 -Ile 181 signal is sufficient for T␤RI endocytosis in R1B-L17 cells is supported by the very minor effect of the 3A mutation (eliminating the acidic cluster motif) on T␤RI endocytosis in these cells and by the similar and very low internalization rates of T␤RI-2A and T␤RI-3A2A (Fig. 5). The strong YRIL signal is predominant in T␤RI-YRIL endocytosis, as demonstrated by the similar endocytosis rates of T␤RI-YRIL and the T␤RI-2A-YRIL double mutant lacking the Leu 180 -Ile 181 signal (Fig. 5); the endocytosis rates of these two mutants were also identical in COS7 cells (data not shown). The partial difference between COS7 cells and Mv1Lu-derived cells in the usage of specific endocytosis signals may reflect cellular context differences in the repertoire of clathrin adaptor proteins.
Smad Signaling by Endocytosis-defective and Endocytosis-enhanced T␤RI Mutants-The endocytosis-defective and endocytosis-enhanced T␤RI mutants described above enable studies that directly address the role of T␤RI endocytosis in signaling. To this end, we evaluated TGF-␤-mediated Smad signaling by transcriptional activation studies using the TGF-␤-responsive luciferase reporter constructs p3TP-Luc(ϩ) (53) and (CAGA) 12 -Luc (54). These studies employed R1B-L17 cells, which express T␤RII but lack functional T␤RI, and therefore do not respond to TGF-␤ unless they are transfected with T␤RI (63). The cells were co-transfected by one of the myc-T␤RI constructs (or empty vector; control), together with a luciferase reporter construct and pRL-TK (Renilla luciferase, serving as a transfection calibration control). Because the response to TGF-␤ is initiated by binding to the cell-surface receptors and the T␤RI endocytosis mutants could display different levels of expression at the cell surface, we measured in parallel the surface levels of the myc-tagged receptors by the point confocal method (see "Experimental Procedures"). Although these levels were rather similar (the differences were less than 15%), we have also calibrated the luciferase results accordingly (see "Experimental Procedures"). As shown in Fig.  6, the CME-defective mutations in T␤RI had a pronounced effect on Smad signaling in the absence of ligand; transfection with CME-defective mutants (T␤RI-2A and T␤RI-3A2A) Addition of the YRIL peptide markedly enhanced the internalization rate of T␤RI. E, effects of endocytosis-inhibitory treatments on myc-T␤RI-YRIL internalization. The measurements followed the protocol detailed in Fig. 3G. Data are presented as % internalization during 20 min at 37°C, calculated as in Fig.  3G. Asterisks indicate significant differences between treated and untreated (control) cells (***, p Ͻ 10 Ϫ4 ; Student's t test).
induced marked transcriptional activation already without ligand, an effect not shown by either T␤RI(WT) or the endocytosis-enhanced T␤RI-YRIL mutant. The constitutive nature of this activity is further validated by the lack of basal signaling in Mv1Lu cells in complete growth medium, as shown by the similarly low signaling levels in the absence and presence of the T␤RI kinase inhibitor SB431542 (Fig. 6, C and D). This notion is reinforced by the constitutive signaling of T␤RI-2A and -3A2A in DR26 cells (supplemental Fig. 2), which lack functional T␤RII and therefore do not respond to ligand. The ability of the endocytosis-defective mutants to activate the luciferase reporters without TGF-␤ stimulation was even somewhat higher than that of the constitutively active T␤RI-T204D mutant. However, in contrast to T␤RI-T204D, whose activity was insensitive to TGF-␤, the endocytosis-defective mutants retained a degree of TGF-␤ responsiveness (Fig. 6). Examination of the fold-increase in the luciferase activity following stimulation with TGF-␤ shows that this parameter is significantly lower for the endocytosis-defective mutants (ϳ2-fold increase; compare the black and white bar pairs for each receptor in Fig. 6). This effect is not due to low activation in the presence of ligand, but rather it reflects a smaller incremental increase in the presence of ligand due to high basal activity. This suggests that the CMEdefective mutants possess some degree of constitutive activity. Notably, the endocytosis-defective T␤RI-2A mutant expressed in R1B-L17 cells induced a moderate phosphorylation of Smad2/3 already in the absence of ligand, increasing significantly to at least the same level mediated by T␤RI(WT) upon stimulation with TGF-␤1 (Fig. 6E). To explore whether this constitutive activation of the 2A-containing T␤RI mutants stems from their defective endocytosis or from an independent effect induced by the 2A mutation, we employed the myc-T␤RI-2A-YRIL compound mutant, which bears the 2A mutation but undergoes efficient endocytosis by virtue of the YRIL signal. This mutant displayed constitutive activation as well as ligand-mediated stimulation, albeit at lower levels then the endocytosis-defective 2A mutants (Fig. 6). This suggests that the high constitutive activity levels of the latter mutants involve contributions from both mechanisms (reduced endocytosis and activation by the 2A mutation). A negative control is supplied by the T␤RI-V158 truncation mutant, which is inactive (lacks the kinase domain) but is also endocytosis-defective; indeed, the kinase-deficient T␤RI-V158 failed to activate the luciferase reporter constructs either in the absence or presence of TGF-␤.
To validate that the above effects are not specific to the R1B-L17 mutant cell line, we conducted the same experiments in the parental Mv1Lu cells. It is more difficult to detect the effects of the specific T␤RI mutants in these cells, because they respond to TGF-␤ via their endogenous receptors (see the vector-only transfected cells in Fig. 6). For this reason, the increase in signaling upon transfection with T␤RI(WT) in Mv1Lu cells was below the detection limit. However, this background issue does not extend to signaling in the absence of ligand, which is very low for the endogenous receptors. Fig. 6 demonstrates that expression of the endocytosis-defective T␤RI-2A in Mv1Lu cells also results in constitutive transcriptional activation of the Smad pathway already in the absence of ligand. A similar phenomenon is observed with the constitutively active T␤RI-T204D mutant (Fig. 6). Notably, although the T␤RI-2A-YRIL mutant exhibited partial constitutive activation, it was significantly lower (p Ͻ 0.05) than that of T␤RI-2A, in line with the results obtained in R1B-L17 cells.
The translocation of Smad2/3 to the nucleus is a key early step in TGF-␤-mediated Smad signaling, which could be utilized to validate the constitutively active character of the endocytosis-defective T␤RI mutants. To this end, we expressed myc-T␤RI (WT or endocytosis mutants) in Mv1Lu cells, and we investigated the effects on Smad2/3 nuclear translocation. The transfected cells were stimulated for 30 min with TGF-␤1 or left untreated, and fluorescence microscopy was employed to score cytoplasmic versus nuclear localization of Smad2/3 in the transfected cells (identified by immunofluorescence using ␣myc to label myc-T␤RI). Typical results are shown in Fig.  7A, and data derived from several independent experiments integrating data from 300 cells are depicted in Fig. 7B. These studies demonstrate that although Smad2/3 is mainly cytoplasmic in unstimulated control (mock-transfected) cells, the vast majority of these cells (which express endogenous TGF-␤ receptors) display TGF-␤1-induced nuclear accumu- For each mutant, the intensity at time 0 was taken as 100%; the percentage of fluorescence intensity remaining at the cell surface after internalization was subtracted to obtain the % internalization. Asterisks indicate significant difference between the endocytosis of T␤RI(WT) and a given mutant (**, p Ͻ 0.01; Student's t test). lation of Smad2/3. In sharp contrast, the CME-defective mutants (T␤RI-2A and T␤RI-3A2A), but not T␤RI(WT) or endocytosis-enhanced T␤RI-YRIL, induced Smad2/3 nuclear translocation even prior to stimulation with TGF-␤ (Fig. 7). The T␤RI-2A-YRIL double mutant also exhibited constitutive Smad nuclear translocation, but to a lower extent than the endocytosis-defective 2A mutants (Fig. 7), in accord with the results in the luciferase transcriptional activation assays (Fig. 6). As expected, constitutively active T␤RI-T204D mutant also induced Smad2/3 nuclear translocation in the absence of ligand (Fig. 7).
These results demonstrate that T␤RI endocytosis is dispensable for signaling (Figs. 6 and 7 and supplemental Fig. 1). The endocytosis half-time measured for T␤RI(WT) in Mv1Lu cells is 9 min. Therefore, faster signaling output kinetics would support the notion that it is independent of T␤RI endocytosis. To that end, we measured the kinetics of TGF-␤-induced Smad2/3 nuclear translocation (an early TGF-␤ signaling event that can be measured with high sensitivity) in Mv1Lu cells. The results (Fig. 8, A and B) yield a rate at least 2-fold faster (t1 ⁄ 2 ϳ4 min). In addition, TGF-␤ receptor endocytosis was also linked earlier to receptor degradation (26,27); we therefore measured the degradation of cell-surface T␤RI in Mv1Lu cells, using a cell-surface biotinylation protocol (see "Experimental Procedures"). The results (Fig. 8, C and D) yield a degradation half-time of ϳ2 h both in the absence or presence of TGF-␤1, in accord with former reports (26,27). Thus, the degradation rate of T␤RI is much slower than its endocytosis rate in the same cells, in line with a role for T␤RI endocytosis in signal termination.

DISCUSSION
Numerous studies on the endocytosis of T␤RII and T␤RI have suggested CME as the major internalization pathway, although a potential contribution by caveolar endocytosis has been contentious (24 -27, 34 -36). Attempts to define the roles of TGF-␤ receptor endocytosis (especially T␤RI, which is the receptor that directly mediates downstream signaling to the Smad pathway) have led to conflicting results (25-27, 39 -41), largely due to reliance on treatments that inhibit altogether CME and/or caveolar endocytosis. In this study, we identified the previously unknown CME-targeting signals on T␤RI, generated endocytosis-defective and endocytosis-enhanced T␤RI mutants, and we employed them to directly investigate the relationship between T␤RI endocytosis and signaling. Our findings indicate that T␤RI endocytosis is dispensable for TGF-␤ signaling and may even play a negative role (enhanced endocytosis may reduce signaling). Notably, mutating the Leu 180 -Ile 181 endocytosis signal to alanines perturbed T␤RI internalization and endowed it with partial constitutive activity. Although retention of T␤RI at the plasma membrane contributes to this constitutive activity, an additional contributing factor may be the proximity of the mutated residues to T␤RI regions involved in regulatory interactions.
Former studies have measured the endocytosis rate of GM-CSF/T␤RI chimeric receptors dimerized by GM-CSF (32), but a direct quantitative measurement of T␤RI endocytosis was not reported. Using the point-confocal endocytosis assay (34) to measure the internalization of EC-tagged myc-T␤RI, we found that it undergoes constitutive endocytosis independent of its activation state ( Fig. 2 and supplemental Fig. S1A), in line with former reports that TGF-␤ receptor internalization does not depend on activation (26,27). The initial endocytosis rate (the fraction internalized per min, derived from the linear part of the internalization curve) (70) was 0.025 min Ϫ1 in COS7 cells, and 0.05 min Ϫ1 in R1B-L17 cells, in the same range reported for GM-CSF/T␤RI homodimers (32). These rates are also close to those of T␤RII internalization (32,34).
To facilitate the search for T␤RI endocytosis signals, it was desired to determine initially the pathways that contribute to its internalization. In both COS7 and Mv1Lu-derived cells, CME emerged as the major T␤RI internalization pathway, as evidenced by the blockade of T␤RI endocytosis in cells subjected to several CME inhibitory treatments and by the insignificant effect of nystatin, a caveolar endocytosis inhibitor (Figs. 2 and 3; supplemental Fig. S1). The identification of CME as the prominent T␤RI internalization pathway is in line with the large majority of earlier reports on the endocytosis of TGF-␤ receptors or their chimeric constructs (24 -27, 34, 35, 40), and it is further supported by the strong inhibitory effect of RNAi-mediated clathrin knockdown on T␤RI internalization (27). It differs from a previous publication that suggested that TGF-␤ receptors also undergo caveolar endocytosis (26); however, the latter report did not measure directly T␤RI endocytosis and employed a higher nystatin concentration (50 g/ml), which may also inhibit CME (27).
Following the identification of CME as the main route for T␤RI internalization, we conducted a mutational analysis of myc-T␤RI in search of its CME-targeting signal(s). To this end, we first screened a series of cytoplasmic truncation mutants and then focused on alanine substitution mutations in regions whose omission inhibited T␤RI endocytosis (for schematics of the mutants, see Fig. 1). The results (Figs. 3 and 5) implicate two signals in T␤RI endocytosis as follows: (i) Leu 180 -Ile 181 , a dileucine family signal from the same family as the Ile 218 -Ile 219 -Leu 220 sequence identified earlier as the sole CME targeting signal in T␤RII (34); (ii) a three-residue Glu 161 -Glu 162 -Asp 163 negatively charged cluster in the membrane-proximal region preceding the GS domain, reminiscent of acidic cluster CMEtargeting motifs (65)(66)(67). Interestingly, a combination of a dileucine sequence and acidic cluster was shown to be required for the dynamic retention of insulin-regulated aminopeptidase in endosomes (67). Notably, the contribution of the two signals to T␤RI endocytosis differs between the COS7 and Mv1Luderived cells, with an equal weight for each signal in COS7 cells (Fig. 3), but a near-complete dependence on the Leu 180 -Ile 181 signal in R1B-L17 cells (Fig. 5). The impact of the two signals is in accord with the reported cell type differences in the internalization of chimeric TGF-␤ receptors (32,33), and it may reflect distinct repertoires of clathrin adaptor proteins in different cell lines.
The identification of the CME-targeting signals on T␤RI and the preparation of endocytosis-defective T␤RI mutants lacking these signals enabled direct investigation of the roles of T␤RI endocytosis in signaling, without resorting to treatments that induce a general block of endocytic pathways, thus avoiding potential side effects due to the altered cellular distribution of multiple signaling proteins. To this end, we expressed endocytosis-defective (T␤RI-2A, T␤RI-2A3A), endocytosis-enhanced (T␤RI-YRIL, T␤RI-2A-YRIL), or T␤RI(WT) in R1B-L17 cells (devoid of T␤RI), and we measured their effects on Smad-dependent transcriptional activation of TGF-␤ luciferase reporter constructs. The results (Fig. 6) lead to two important conclusions. First, T␤RI endocytosis is not required for signaling; in fact, the activation of the endocytosis-defective mutants is higher than that of T␤RI(WT), although the endocytosis-enhanced T␤RI-YRIL induces a lower level of transcriptional activation. This suggests that T␤RI internalization via coated pits plays a role in signal termination. Second, the endocytosis-defective mutants display partial constitutive activity; their expression in R1B-L17 cells results in high activation levels in the absence of ligand, whereas TGF-␤ can still induce further stimulation, unlike the constitutively active T␤RI-T204D mutant, whose activity is ligand-independent (Fig. 6). The constitutively active character of the endocytosis-defective Mv1Lu cells in 10-cm plates were transfected with myc-T␤RI or pcDNA3 (control) and subjected to surface biotinylation followed by incubation at 37°C for the indicated times as detailed under "Experimental Procedures." Immunoprecipitates of myc-T␤RI and cell lysates (6% v/v of the lysate) were resolved by SDS-PAGE and Western blotting using the indicated probes. C, Western blots of a representative experiment (one out of four). D, quantification (means Ϯ S.E., n ϭ 4) of the intensities of the biotinylated myc-T␤RI immunoprecipitated at the different time points, after normalization to the myc-T␤RI levels in the lysates. The normalized level in myc-T␤RI transfected cells at time 0 (2nd lane) in each experiment was taken as 100%. Identical results were obtained in the presence or absence of ligand; thus, only the latter are shown. mutants is observed also in the parental Mv1Lu cells (Fig. 6) and in DR26 cells (lacking ligand response due to the absence of T␤RII; supplemental Fig. 2), and it is validated by the constitutive nuclear translocation of Smad2/3 in these cells (Fig. 7). The notion that the retention of T␤RI endocytosis-defective mutants at the cell surface plays a role in their constitutive activation gains further support from the finding that addition of a strong endocytosis signal (YRIL) to the T␤RI-2A mutant attenuates its constitutive activation level (Figs. 6 and 7). In line with these findings, control experiments using a battery of CME-inhibitory treatments (including DYN, which inhibits CME and caveolar endocytosis), were also capable of inducing constitutive activation of the luciferase reporter constructs (supplemental Fig. S1). The specific activation levels and the extents of the residual responsiveness to TGF-␤ were highly variable, potentially reflecting different inhibition mechanisms by distinct treatments, leading to diverse effects on intracellular traffic. This is most likely one of the reasons for conflicting reports on the effects of endocytosis-inhibitory treatments on TGF-␤ signaling. Thus, several studies reported continued or enhanced TGF-␤ signaling in cells subjected to CME-inhibiting treatments (39 -41) or suggested that endocytosis is dispensable at least for the initial steps of TGF-␤ signaling, up to Smad/ SARA association (25) or Smad2/3 phosphorylation (40,71). However, there are also reports that dominant-negative dynamin or K ϩ depletion inhibit TGF-␤-mediated transcriptional activation (24 -26). These inconsistencies emphasize the advantage of using endocytosis-defective receptor mutants in place of general inhibitory treatments.
The partial constitutive nature of the endocytosis-defective T␤RI mutants may be a direct consequence of their retention at the cell surface; alternatively, the mutation of Leu 180 -Ile 181 to alanines in these mutants may have an additional independent activating effect on T␤RI. The results of this study suggest contribution by both mechanisms, because the T␤RI-2A-YRIL compound mutant, which carries the Leu 180 -Ile 181 to Ala 180 -Ala 181 mutation but undergoes efficient endocytosis via the YRIL signal, displays an intermediate level of activation (Figs. 6 and 7). The Leu 180 -Ile 181 peptide is localized in the ␣GS1 region of T␤RI, which was proposed to be involved in the activating conformational switch of T␤RI GS region following its phosphorylation by T␤RII (72). Thus, mutations in this region, which is also proximal to the binding sites of FKBP12 (43,72), Smads (44,45), and additional signal regulators (46), have the potential to partially emulate T␤RI activation.
In summary, this study demonstrates that T␤RI is endocytosed via the CME pathway, identifies the specific internalization motifs involved, and reveals that T␤RI endocytosis is not required for signaling. Based on the inverse correlation between T␤RI endocytosis and signaling, we propose that T␤RI internalization plays a role in signal termination.