The type II transforming growth factor‐β (TGF‐β) receptor Ser/Thr kinase (TβRII) is responsible for the initiation of multiple TGF‐β signaling pathways, and loss of its function is associated with many types of human cancer. Here we show that TβRII kinase is regulated intricately by autophosphorylation on at least three serine residues. Ser213, in the membrane‐proximal segment outside the kinase domain, undergoes intra‐molecular autophosphorylation which is essential for the activation of TβRII kinase activity, activation of TβRI and TGF‐β‐induced growth inhibition. In contrast, phosphorylation of Ser409 and Ser416, located in a segment corresponding to the substrate recognition T‐loop region in a three‐dimensional structural model of protein kinases, is enhanced by receptor dimerization and can occur via an inter‐molecular mechanism. Phosphorylation of Ser409 is essential for TβRII kinase signaling, while phosphorylation of Ser416 inhibits receptor function. Mutation of Ser416 to alanine results in a hyperactive receptor that is better able than wild‐type to induce TβRI activation and subsequent cell cycle arrest. Since on a single receptor either Ser409 or Ser416, but not both simultaneously, can become autophosphorylated, our results show that TβRII phosphorylation is regulated intricately and affects TGF‐β receptor signal transduction both positively and negatively.
Transforming growth factor‐β (TGF‐β) is a multi‐potent cytokine that regulates many biological functions, including arrest of the cell cycle in G1, production of extracellular matrix components and differentiation of many cell types (Roberts and Sporn, 1990). The signaling cell surface type I (TβRI) (Franzen et al., 1993; Bassing et al., 1994) and type II (TβRII) (Lin et al., 1992) receptors for TGF‐β, as well as receptors for activins (Mathews and Vale, 1991; Attisano et al., 1992, 1993; Mathews et al., 1992; ten Dijke et al., 1994a) and bone morphogenic proteins (Estevez et al., 1993; Koenig et al., 1994; ten Dijke et al., 1994b; Liu et al., 1995; Rosenzweig et al., 1995), form a family of proteins with a cysteine‐rich extracellular domain, a single transmembrane segment and a cytoplasmic segment with a serine/threonine kinase domain. TβRII exists as a homodimer on the cell surface even in the absence of ligand (Chen and Derynck, 1994; Henis et al., 1994). However, in the absence of TβRI, TβRII homodimers are insufficient for TGF‐β signal transduction (Luo and Lodish, 1996). The Ser/Thr kinase activity of TβRII is constitutively active and the receptor is autophosphorylated (Lin et al., 1992). Binding of TGF‐β1 to the extracellular domain of TβRII does not affect receptor kinase activity (Wrana et al., 1994; Chen and Weinberg, 1995). Rather, it causes the formation of a heteromeric complex containing TβRI and TβRII (Wrana et al., 1992) followed by transphosphorylation of TβRI by the TβRII kinase (Wrana et al., 1994). Phosphorylation of TβRI by TβRII is thought to activate its kinase activity and allow it to phosphorylate and activate downstream substrate(s). Both TβRI and TβRII kinase activities are required for signal transduction (Lin et al., 1992; Wrana et al., 1992; Franzen et al., 1993; Bassing et al., 1994). Mink lung cells expressing TβRII with a point mutation that disrupts its ability to transphosphorylate TβRI are insensitive to TGF‐β stimulation (Carcamo et al., 1995).
Mutations or deletions in TβRII can result in a loss of sensitivity to TGF‐β, a process that frequently accompanies the progression of malignancy, and may contribute to the development of many types of human cancer (Kadin et al., 1994; Park et al., 1994; Sun et al., 1994; Markowitz et al., 1995; Mazars et al., 1995; Knaus et al., 1996). For example, colon cancer cells with microsatellite instability (RER+) harbor mutations in the TβRII gene which cluster within small repeated sequences in the TβRII gene and which are accompanied by loss of TβRII surface expression. This inactivation of TβRII is likely to be a critical step in tumor progression, rather than a simple correlate of the RER+ phenotype (Markowitz et al., 1995). Truncations of TβRII or point mutations resulting in a dominant‐negative TβRII were also found in gastric cancer and lymphoma (Park et al., 1994; Knaus et al., 1996). Therefore, TβRII is a tumor suppressor.
Here we focus on the regulation of TβRII activity by autophosphorylation, since autophosphorylation has been recognized as an important regulatory mechanism for protein kinases. Autophosphorylation on tyrosine, serine or threonine residues often induces critical structural changes around the catalytic center of the kinase that result in the activation of kinase activity. This mechanism is employed in the activation of both the insulin receptor tyrosine kinase (Hubbard et al., 1994) and MAP kinase ERK2 (Zhang et al., 1994). These activating autophosphorylation sites usually are located in the T‐loop, within the kinase domain in close proximity to the catalytic loop and putative substrate‐binding site. Autophosphorylation of tyrosine residues in kinases such as the platelet‐derived growth factor (PDGF) or epidermal growth factor (EGF) receptors also generates docking sites for downstream signal transduction proteins containing either SH2 domains or PTB domains (Marengere and Pawson, 1994; van der Geer and Pawson, 1995). Finally, autophosphorylation of certain tyrosine residues is required for endocytosis and down‐regulation of many receptor protein tyrosine kinases (Cadena et al., 1994; Nesterov et al., 1995).
Virtually nothing is known about TβRII autophosphorylation; we do not know which residues are autophosphorylated nor the functional consequences of TβRII autophosphorylation. In particular, we do not know whether autophosphorylation regulates signal transduction by TβRII. Here we report the identification of three major autophosphorylation sites in TβRII and show that they play important roles in both positive and negative regulation of the kinase activity and signal transduction of TβRII. Two of these are located near the putative catalytic loop and putative substrate‐binding site, in a segment corresponding to the T‐loop in other kinases. Thus, TβRII activity is regulated intricately by autophosphorylation and possibly by phosphorylation by other cellular kinases.
The TβRII kinase undergoes autophosphorylation on multiple serine residues in the cytoplasmic domain
TβRII kinase undergoes autophosphorylation on similar sites when expressed in mammalian cells and in bacteria (Chen and Weinberg, 1995). We therefore used bacterial expressed TβRII to map the sites of autophosphorylation. When expressed in and purified from Escherichia coli as a GST fusion protein, the complete cytoplasmic domain of TβRII became phosphorylated in vitro in the presence of [γ‐32P]ATP (Figure 1A, lane 1). The kinase‐inactive K277R mutant of the TβRII cytoplasmic domain (unable to bind ATP) failed to become labeled with [γ‐32P]ATP (Figure 1A, lane 2), indicating that phosphorylation of TβRII resulted from autophosphorylation. Phosphoamino acid analysis of the autophosphorylated TβRII cytoplasmic domain indicated that autophosphorylation occurred predominantly on serine residues (Lin et al., 1992) (Figure 1B). Two‐dimensional tryptic peptide mapping revealed multiple spots (Figure 1C), suggesting that there may be more than one site of autophosphorylation.
Ser213 is a major site of autophosphorylation
To map the autophosphorylation sites in TβRII, the 32P‐labeled cytoplasmic domain of TβRII, removed from GST by thrombin digestion, was digested with trypsin and the resulting peptides separated by HPLC (Figure 2A). Only two adjacent HPLC fractions contained 32P‐labeled peptides (Figure 2B). Amino acid sequencing of the two fractions revealed that they both contained a mixture of two peptides that differed only by one amino acid at the amino‐terminus (Figure 2C). Since trypsin does not cleave clusters of basic amino acid residues efficiently, it is likely that the longer of the two peptides resulted from a partial tryptic digestion. These peptides contained only one serine residue: Ser213. To confirm that Ser213 was indeed phosphorylated, radio‐sequencing of the two HPLC fractions was carried out. This assay monitored the release of 32Pi from the peptide mixture after each cycle of Edman degradation. As expected, 32Pi was released after five cycles of Edman degradation, presumably from the shorter peptide, and reached the highest level after the sixth cycle due to the presence of the longer peptide with its N‐terminal arginine residue (Figure 2D).
To confirm biochemically that Ser213 is phosphorylated, Ser213 was mutated to Ala (S213A mutant) and Asp (S213D mutant), respectively. The S213A mutant mimics an unphosphorylated Ser213, and the S213D mutant, with its negatively charged side chain at position 213, resembles phosphorylated Ser213. The mutations were introduced into both the full‐length TβRII and a chimeric receptor (EpoR–TβRII) containing the extracellular domain of the erythropoietin receptor (EpoR) and the cytoplasmic domain of TβRII (Luo and Lodish, 1996). When expressed transiently in 293T cells, S213A displayed markedly weakened autophosphorylation, as evidenced by both an in vitro autophosphorylation assay (Figure 2E) and two‐dimensional tryptic mapping (Figure 2G). In contrast, the S213D mutants underwent autophosphorylation (Figure 2E). Two‐dimensional tryptic peptide mapping of the 32P‐labeled S213D mutant revealed efficient phosphorylation of spots b and c in a manner indistinguishable from the wild‐type receptor (Figure 2G). However, spots a and a′ were missing, indicating that they correspond to the two tryptic peptides containing phosphorylated Ser213. Mutation of Ser213 to alanine impaired the kinase activity, and phosphorylation on spots b and c was greatly reduced.
Therefore, Ser213 is a major site of TβRII autophosphorylation. The negative charge conferred by the phosphorylation of this site, which is mimicked by the S213D mutation, is required for kinase activity and autophosphorylation of other serine residues represented by spots b and c. Ser213 is located in the membrane‐proximal region of TβRII outside the kinase domain.
Phosphorylation of Ser213 in TβRII is required for transducing TGFβ growth inhibitory signals
To investigate the functions of autophosphorylation at Ser213 in TGF‐β receptor‐mediated growth inhibition, TβRIIs containing the S213A or S213D mutations were introduced into two different mammalian cell lines and analyzed for their ability to mediate TGF‐β‐induced growth inhibition.
Hep3B hepatoma cells express TβRI and TβRII as well as the type III (betaglycan) TGF‐β receptor, and undergo growth inhibition in response to TGF‐β stimulation (Figure 3A, left panel: mock). A mutant Hep3B cell line, Hep3B.TR, lacks endogenous TβRII and therefore is unresponsive to TGF‐β (Figure 3A, right panel: mock) (Inagaki et al., 1993). Expression of wild‐type TβRII in Hep3B.TR cells restored growth inhibition to TGF‐β (Inagaki et al., 1993 and Figure 3A, right panel). In contrast, expression of TβRII(S213A) in Hep3B.TR cells failed to restore growth inhibition to TGF‐β (Figure 3A, right panel), even though the level of expression of TβRII(S213A) was comparable with, if not higher than, that of a similarly expressed exogenous wild‐type TβRII (data not shown). Interestingly, when introduced into the TGF‐β‐responsive Hep3B cells containing wild‐type endogenous TβRII, the mutant TβRII(S213A) receptor blocked TGF‐β‐induced growth inhibition (Figure 3A, left panel), suggesting that the S213A mutant is functioning in a dominant‐negative manner.
In contrast, the TβRII(S213D) mutant was similar to the wild‐type TβRII in its ability to restore growth inhibition to Hep3B.TR cells (Figure 3A, right panel). Expression of TβRII(S213D) in Hep3B cells had no effect on the growth response to TGF‐β (Figure 3A, left panel).
Previously we constructed and analyzed chimeric receptors containing the extracellular domain of the EpoR and the transmembrane and cytoplasmic domains of TβRII (E‐RII) or TβRI (E‐RI) (Luo and Lodish, 1996). When co‐expressed in TGF‐β‐responsive Ba/F3 cells, E‐RI and E‐RII mediate Epo‐induced growth inhibition in a manner similar to that mediated by the endogenous TGF‐β receptors in these cells in response to TGF‐β stimulation (Luo and Lodish, 1996) (Figure 3B). Consistent with the results in Hep3B/TR cells, co‐expression of E‐RII(S213A) with E‐RI did not result in Epo‐induced growth inhibition. In contrast, co‐expression of E‐RII(S213D) with E‐RI did result in Epo‐induced growth inhibition, though not to the same extent as with wild‐type E‐RII (Figure 3B).
Mutation of Ser213 in TβRII does not affect its ability to form a heteromeric complex with TβRI on the cell surface
The inability of the mutant TβRII(S213A) to mediate growth inhibitory responses may be due to a deficiency in cell surface expression, to an inability to form a heteromeric complex with TβRI or to a lack of kinase activity necessary for the activation of TβRI. To investigate these possibilities, 293T cells co‐expressing TβRI and S213A or S213D mutant TβRII receptors were incubated with [125I]TGF‐β1 followed by cross‐linking and immunoprecipitation with an antibody specific for the cytoplasmic domain of TβRII. Similarly to wild‐type TβRII (Figure 4, lane 2), both S213A and S213D mutant receptors were expressed on the cell surface and were capable of binding to ligand. Furthermore, all could form normal heteromeric complexes with TβRI, as evidenced by co‐immunoprecipitation of [125I]TGF‐β1 cross‐linked to TβRI by anti‐TβRII antibodies (Figure 4, lanes 3 and 4).
Mutation of Ser213 in TβRII decreases its ability to transphosphorylate the TβRI cytoplasmic domain
Since mutation of Ser213 to alanine in TβRII decreased its ability to autophosphorylate, we next asked whether the S213A mutation affected the ability of TβRII to transphosphorylate TβRI. Since Hep3B cells express too low a level of TGF‐β receptors to allow biochemical analyses, Ba/F3 cells expressing chimeric EpoR–TGF‐βRs were used in this experiment. Our previous experiments established that Epo induced transphosphorylation of E‐RI by E‐RII in a manner similar to TGF‐β‐induced transphosphorylation of TβRI by TβRII (Luo and Lodish, 1996) (Figure 5A, lanes 1 and 2). In cells expressing the E‐RII(S213A) mutant receptor, Epo‐induced transphosphorylation was reduced to a level <20% of that observed in cells expressing wild‐type E‐RII (Figure 5A, lanes 3 and 4), even though similar amounts of E‐RI were expressed in these cells (Figure 5A, lower panel). E‐RII(S213D) transphosphorylated E‐RI in response to Epo stimulation to an extent comparable with that of the wild‐type E‐RII (Figure 5A, lanes 5 and 6). These experiments suggest that the S213A mutant has a reduced ability to phosphorylate the TβRI cytoplasmic domain, a step crucial to the activation of and subsequent signal transduction by TβRI.
Homodimerization of the RII cytoplasmic domain enhances its autophosphorylation
Since TβRII exists as a homodimer on the cell surface, an important question is whether dimerization of TβRII has any effect on its autophosphorylation and whether autophosphorylation occurs by an inter‐ or intra‐molecular mechanism. Previously it has been impossible to study this question since dimerization of TβRII is constitutive even in the absence of ligand. With the EpoR–TβRII chimeric receptor, which exists as a monomer normally and requires Epo to dimerize, we could control the dimerization process and address these questions.
Figure 2F shows that homodimerization of the RII cytoplasmic domain by Epo enhances its ability to undergo autophosphorylation. In this study, Ba/F3 cells expressing E‐RII or E‐RII(S213D) were treated with Epo to induce dimerization, and autophosphorylation of RII was examined by an in vitro kinase assay. Stimulation with Epo induced a rapid increase in the autophosphorylation of both E‐RII and E‐RII(S213D), reaching the maximal level within 1 min after stimulation (Figure 2F). Since autophosphorylation of E‐RII(S213D) was also increased by Epo‐induced dimerization, the dimerization‐enhanced autophosphorylation occurs on serine residues other than Ser213. Two‐dimensional tryptic peptide mapping of 32P‐labeled E‐RII(S213D) before and after Epo stimulation did not reveal any new spots (data not shown), suggesting that the increased autophosphorylation occurs on the peptides represented by spots b and c.
Ser409 and Ser416 are additional sites of autophosphorylation
The identity of these additional autophosphorylation sites could not be determined using the HPLC and peptide sequencing approach described above for the identification of Ser213 because peptides containing these sites could not be recovered from the HPLC column (data not shown). We therefore generated point mutations in the cytoplasmic domain of TβRII, changing 24 out of a total of 29 serine residues individually to alanine (for details, see Materials and methods). The mutant receptors were transfected transiently into 293T cells and isolated by immunoprecipitation followed by in vitro autophosphorylation and two‐dimensional tryptic mapping to determine the identity of spots b and c. Surprisingly, none of the single serine to alanine substitutions exhibited any alteration in phosphorylation of spot b (data not shown). These data indicated that spot b was probably composed of peptides containing multiple serine residues which are capable of being phosphorylated. We therefore analyzed mutants containing multiple point mutations.
The experiment shown in Figure 6 suggested that spot b represents a peptide containing either phosphorylated Ser409 or phosphorylated Ser416. The maps shown in this and subsequent experiments focus on the area containing spots b and c and do not include spots a and a′. When both Ser409 and Ser416 were mutated simultaneously to alanine (S409,416A), spot b was not labeled with 32P after in vitro phosphorylation (Figure 6A). In contrast, mutation of either Ser409 or Ser416 individually to alanine did not affect the pattern. In particular, we did not observe a peptide with the mobility expected of one in which both Ser409 and Ser416 were phosphorylated, nor was any phosphopeptide missing in either the S409A or S416A mutants (data not shown). Thus, either Ser409 or Ser416 can become phosphorylated, but not both simultaneously on any one receptor. Importantly, mutations at Ser409 and Ser416 did not decrease the kinase activity of RII significantly (Figure 6B), and all receptors underwent autophosphorylation at spot c (Figure 6A) as well as at spots a and a′ (data not shown). Mutation of Ser416 to alanine even enhanced autophosphorylation of the receptor slightly (Figure 6B).
Phosphorylation of Ser409 is essential for TβRII signaling while phosphorylation of Ser416 is inhibitory
To examine the functions of phosphorylation of Ser409 and Ser416 in TβRII signal transduction, Ba/F3 cells stably expressing E‐RI together with a chimeric E‐RII receptor containing either an S409A, S409D, S416A, S416D mutation or the S409,416A double mutation were stimulated with Epo, and the growth of these cells was examined 4 days later. Mutation of Ser409 to alanine markedly reduced the ability of E‐RII to mediate growth inhibition in the presence of E‐RI (Figure 3C). In contrast to the S213D mutant, which exhibited a partial growth inhibitory response, the S409D mutant failed to mediate cell cycle arrest when co‐expressed with E‐RI (Figure 3C). This suggests that phosphorylated Ser409 affects the function of RII in a manner different from that of phosphorylated Ser213.
Importantly, mutation of Ser416 to alanine rendered E‐RII hypersensitive in that the growth of Ba/F3 cells containing this mutant E‐RII was inhibited to a greater extent than was that of the cells expressing wild‐type E‐RII (Figure 3C). These results suggest that phosphorylation at Ser409 enhances the ability of E‐RII to mediate growth inhibition, but that phosphorylation at Ser416 inhibits the ability of E‐RII to transduce a growth inhibitory signal. Therefore, Ser409 appears to be a positive regulatory site while Ser416 is an inhibitory site. Consistent with the inhibitory function of Ser416, mutation of Ser416 to aspartic acid, which mimicked fully phosphorylated Ser416, completely abolished the ability of the E‐RII receptors to mediate growth inhibition (Figure 3C). Interestingly, when both Ser409 and Ser416 were mutated to alanine, the resultant E‐RII(S409,416A) could transduce growth inhibitory signals, albeit at a level lower than normal E‐RII (Figure 3C). Thus, it appears that E‐RII(S409,416A) functions as an ‘average’ of the two single Ser→Ala mutations.
Mutations of Ser409 and Ser416 affect the transphosphorylation of the TβRI cytoplasmic domain
Mutation of Ser409 or Ser416 did not affect their ability to form a heteromeric complex with TβRI on the cell surface (Figure 4, lanes 5–7). Rather, they affect the ability of RII kinase to transphosphorylate the TβRI cytoplasmic domain. In cells co‐expressing E‐RI and E‐RII(S409A), we consistently observed an elevated level of E‐RI phosphorylation even in the absence of ligand (Figure 5B, lane 3). However, this phosphorylation of TβRI did not result in the transduction of growth inhibitory signals. Importantly, addition of Epo did not significantly increase the phosphorylation of E‐RI by E‐RII(S409A) (Figure 5B, lane 4). Similar results were observed in cells expressing E‐RII(S409D) (Figure 5B, lanes 9 and 10). These results suggest that phosphorylation of TβRI does not always lead to activation of its kinase activity and signal transduction. It is likely that this ligand‐independent phosphorylation of TβRI occurs on a site different from those phosphorylated in a ligand‐dependent manner in cells expressing wild‐type E‐RII.
In contrast, cells expressing the mutant E‐RII(S416A) receptor exhibited an increased Epo‐induced phosphorylation of E‐RI (about twice as much as in cells expressing E‐RII, as evidenced by the ratio of 32P incorporated relative to the amount of 35S‐labeled protein), and consequently the mobility of the phosphorylated E‐RI in cells expressing E‐RII(S416A) was slower than that in cells expressing wild‐type E‐RII (compare lanes 2 and 6 in Figure 5B). This enhancement in E‐RI phosphorylation is consistent with the increased ability of the E‐RII(S416A) receptor to mediate growth inhibition (Figure 3C). The double mutant E‐RII(S409,416A) phosphorylated E‐RI in a manner that appears to be an average of the single mutants S409A and S416A. In particular, both an elevated level of ligand‐independent phosphorylation of E‐RI (similar to that seen with the S409A mutant) (Figure 5B, lane 7), and an increase in 32P‐labeled E‐RI with a slower mobility (similar to that seen with the S416A mutant) (Figure 5B, lane 8) were observed. Since the E‐RII(S409,416A) double mutant was weaker than wild‐type E‐RII and ERII(S416A) in inducing cell cycle arrest, and since ERII(S409A) is defective in signal transduction, the ligand‐independent phosphorylation of E‐RI induced by E‐RII(S409A) is probably inhibitory.
Interestingly, mutation of Ser416 to Asp abolished the ability of E‐RII to autophosphorylate, and also to phosphorylate E‐RI (Figure 5B, lanes 11 and 12). Since the S416D mutant mimics E‐RII with a phosphorylated Ser416, this result further supports the conclusion that phosphorylation at Ser416 is inhibitory to the kinase activity of TβRII.
Autophosphorylation of TβRII occurs by both intra‐ and inter‐molecular mechanisms
Studies with other protein serine and tyrosine kinases demonstrate both intra‐ and inter‐molecular mechanisms of autophosphorylation. An intra‐molecular mechanism involves phosphorylation of the same peptide chain, while an inter‐molecular mechanism involves the kinase domain of one polypeptide chain phosphorylating amino acid residues on another of the same type. To examine which mechanism is involved in TβRII autophosphorylation, two mutant EpoR–TβRIIs were co‐transfected into 293T cells (Figure 7A). One is the kinase‐inactive E‐RII(K277R), whose ATP‐binding site was destroyed by the Lys→Arg mutation. This mutant still retains all serine phosphorylation sites. The second mutant, E‐RII(S409,416A), has lost the Ser409 and Ser416 autophosphorylation sites, but still retains kinase activity. If autophosphorylation occurs by an inter‐molecular mechanism, the kinase‐active E‐RII(S409,416A) will phosphorylate serine residues on E‐RII(K277R), which can be detected by two‐dimensional tryptic mapping. In contrast, if autophosphorylation only occurs intra‐molecularly, no phosphorylation of Ser409 or Ser416 will be observed because E‐RII(K277R) cannot phosphorylate itself and E‐RII(S409,416A) lacks these phosphorylatable serine residues.
Figure 7B shows that autophosphorylation at Ser409 or Ser416 can occur via an inter‐molecular mechanism. Expression of E‐RII(K277R) alone did not result in receptor autophosphorylation (data not shown). As shown in Figure 6A and also in Figure 7B, mutation of both Ser409 and Ser416 to alanine resulted in the loss of 32P‐labeled spot b (Figure 7B). However, when E‐RII(K277R) was co‐expressed with E‐RII(S409,416A), stimulation by Epo resulted in the phosphorylation of Ser409 and Ser416 on E‐RII(K277R), as evidenced by the radiolabeling of spot b (Figure 7B). This indicates that phosphorylation of Ser409 and/or Ser416 can occur via an inter‐molecular mechanism.
In a similar experiment, we co‐transfected 293T cells with the E‐RII(K277R) together with the E‐RII(S213D) receptor. Stimulation with Epo did not result in the radiolabeling of spots a and a′ (data not shown), suggesting that autophosphorylation at Ser213 occurred only by an intra‐molecular mechanism.
The finding that phosphorylation on Ser409 and/or Ser416 can occur inter‐molecularly is consistent with the enhancement of phosphorylation at Ser409 or Ser416 upon dimerization of the RII cytoplasmic domain (Figure 2F). In contrast, phosphorylation at Ser213 does not depend on receptor dimerization, and in this sense can be considered constitutive.
Here we show that TβRII kinase is regulated intricately by autophosphorylation on at least three serine residues (Table I): Ser213, in the membrane‐proximal segment outside the kinase domain, and Ser409 and Ser416, located in the substrate‐binding site in a segment corresponding to the T‐loop in a three‐dimensional structural model of protein kinases (Figure 8). Phosphorylation of Ser213 is required for the kinase activity of TβRII and its ability to phosphorylate TβRI. In contrast, phosphorylation of Ser409 does not significantly affect the kinase activity of TβRII since the S409A mutant can still undergo autophosphorylation, yet its ability to signal growth inhibitory signals is impaired dramatically. Therefore, the stimulatory effects of (phospho) Ser213 and Ser409 are achieved through distinct mechanisms. Phosphorylation of Ser416, on the other hand, inhibits the kinase activity of TβRII, since mutation of this serine to alanine enhances ligand‐induced phosphorylation of TβRI and results in a hyperactive receptor that is more able than wild‐type to induce TβRI activation and subsequent cell cycle arrest. Also, the S416D mutant, which mimics the fully phosphorylated Ser416, is deficient in autophosphorylation as well as phosphorylation of TβRI. Since on a single receptor either Ser409 or Ser416, but not both simultaneously, can become autophosphorylated, our results show that TβRII phosphorylation is regulated intricately and affects TGF‐β receptor signal transduction both positively and negatively.
Autophosphorylation at Ser213 also occurs via a different mechanism from that at Ser409 and Ser416. Phosphorylation at Ser213 occurs intra‐molecularly while Ser409 and Ser416 can be phosphorylated via an inter‐molecular mechanism. This accounts for the enhancement of autophosphorylation at Ser409 and Ser416 by dimerization. More importantly, it suggests an important function for a TβRII homodimer on the cell surface (Chen and Derynck, 1994; Henis et al., 1994). Despite the fact that TβRII homodimer alone is not sufficient for signal transduction, homodimerization of TβRII may play an important role in the regulation of TβRI activation. Dimerization of the RII cytoplasmic domain leads to increased phosphorylation of Ser409 or Ser416, which, in turn, modulates TβRI activation. Since phosphorylation of the two serine residues has opposite effects and since both are potential substrates for other cellular kinases, it is possible that, by regulating the ratio of phosphorylation at these two serines in response to an ever‐changing cellular environment, the extent of substrate phosphorylation can be fine‐tuned in an accurate and reversible manner. In particular, the ratio of phosphorylation at Ser409 and Ser416 could alter in response to TGF‐β stimulation and this may provide a mechanism for the up‐ or down‐regulation of receptor signal transduction.
Phosphorylation at Ser213
Ser213 is the primary site of autophosphorylation in TβRII and is required for the kinase activity of TβRII. Unlike many stimulatory autophosphorylation sites commonly seen in other protein kinases, which are located in the kinase domain, Ser213 is located outside the kinase domain, in a 56 amino acid segment between the membrane‐spanning and the kinase domain. This segment is poorly if at all conserved among receptors of the TGF‐β receptor family, in terms of both its length and amino acid sequence. Therefore, Ser213 could be a regulatory autophosphorylation site specific to TβRII. Phosphorylation at Ser213 could lead to conformational changes in the kinase domain of TβRII, resulting in the activation of its kinase activity. Since phosphorylation at Ser213 occurs via an intra‐molecular mechanism, it will be interesting to determine the three‐dimensional structure of the complete cytoplasmic domain of TβRII and learn how it folds such that Ser213 becomes accessible to the catalytic region of the kinase domain in the same molecule to be autophosphorylated. The three‐dimensional structure of the TβRII cytoplasmic domain may also tell us how phosphorylation at Ser213 activates the kinase activity of TβRII.
Phosphorylation at serines 409 and 416
Ser409 and Ser416 are only six amino acid residues apart, yet they exert opposite effects on the signal transduction of TβRII. Phosphorylation of Ser409 is stimulatory while phosphorylation of Ser416 is inhibitory to TβRII function. Since Ser409 and Ser416 are found in the same tryptic peptide, we are not able to determine the relative phosphorylation on these residues. However, we speculate that Ser409 normally is phosphorylated to a higher extent than Ser416, since E‐RII(S409,416A), lacking both serine residues, was much less able to mediate growth inhibitory responses than was the wild‐type E‐RII chimeric receptor.
By homology modeling with cAMP‐dependent protein kinase (PKA) (Knighton et al., 1991), Ser409 and Ser416 are located within the kinase domain in the so‐called T‐loop region that is involved directly in binding to the substrate peptide (Figure 8). Comparison of amino acid sequences of this region between TβRII and PKA reveals an ‘insert’ of seven amino acids that is specific for TβRII (amino acid residues 408–414), and this sequence could play an important role in the ability of TβRII to bind specific substrates. Deletion of this peptide insert markedly decreased the kinase activity of TβRII, and the resultant mutant receptor was defective in mediating growth inhibitory responses by TGF‐β (data not shown). Ser409 is located within this insert, while Ser416 is just outside it.
The T‐loop region is also the location where conserved autophosphorylation sites are found in many protein serine and tyrosine kinases, including PKA (Knighton et al., 1991), ERK2 (Zhang et al., 1994) and the insulin receptor (Hubbard et al., 1994). Autophosphorylation of one or more of these sites is required for activation of kinase activity. However, mutation of Ser409 to alanine did not significantly affect TβRII kinase activity, since the receptor still underwent autophosphorylation. Furthermore, although mutation of Ser409 to aspartate mimics phosphorylated serine, S409D could not activate TβRI for downstream signal transduction. This indicates that while phosphorylated Ser409 is essential for receptor signal transduction, it is not only the negative charge of the phosphate that is important. Phosphorylation at Ser409 could change the conformation of this region of the kinase, thereby affecting the binding of its substrates, but this change is not (solely) due to the negative charge of the phosphate group. Alternatively, phosphorylated Ser409 could be involved directly in the recruiting and binding of the substrates in a manner similar to that of phosphotyrosine binding to SH2 or PTB domains (Marengere and Pawson, 1994; van der Geer and Pawson, 1995). This latter hypothesis becomes especially intriguing following the recent discoveries that phosphoserine can mediate protein–protein interactions between the Raf1 and 14‐3‐3 proteins (Muslin et al., 1996). Phosphorylated Ser409 could be involved directly in the binding and activation of TβRI, or it could bind and titrate cellular inhibitors away from TβRI. It may also mediate binding and consequently phosphorylation of other proteins that transduce signals from the TGF‐β receptors.
In contrast, phosphorylation of Ser416 could lead to an unfavorable conformation of the catalytic region for autophosphorylation and/or substrate phosphorylation. This negative effect could be mimicked fully by mutating Ser416 to aspartate. Elimination of this effect by changing Ser416 to alanine increased both TβRII autophosphorylation and transphosphorylation of TβRI. Neither on receptors isolated from transfected cells nor on GST–kinase fusion proteins were we able to detect a phosphopeptide corresponding to one in which both Ser409 and Ser416 were phosphorylated; spot b on our two‐dimensional maps corresponds to a peptide in which either Ser409 or Ser416 are phosphorylated. A key question, therefore, is how the relative phosphorylation of these two serine residues is controlled. Since these could be phosphorylated by an inter‐molecular mechanism, the precise geometry by which the cytosolic domains of two TβRII receptors are juxtaposed in a homodimer could be important.
In cells expressing the mutant S409A or S409D receptors, we consistently observed an elevated level of TβRI phosphorylation in the absence of ligand stimulation. However, this phosphorylation of TβRI did not result in any significant inhibition of growth, since the cells grew normally. In contrast, in cells expressing the S416A mutant, ligand stimulation resulted in enhanced phosphorylation of TβRI and enhanced growth inhibition. These results suggest that phosphorylation of E‐RI occurs on different residues when induced by wild‐type, S416A or S409A E‐RIIs and with different consequences. TβRI phosphorylation may not always lead to activation of receptor signal transduction.
Phosphorylation at multiple serine residues
Besides Ser213, Ser409 and Ser416, there are additional autophosphorylation sites in TβRII, in particular, those corresponding to spot ‘c’ on the two‐dimensional tryptic maps. Therefore, the regulation of TβRII is not as simple as had been thought before. Multiple autophosphorylation events have been observed in receptor tyrosine kinases including PDGF receptor (Kashishian et al., 1992) and colony‐stimulating factor‐1 receptor (van der Geer and Hunter, 1993). These phosphorylated tyrosine residues mediate interactions with different downstream signaling molecules and thus provide a mechanism that enables one receptor to initiate multiple signaling pathways. Since TGF‐β receptor can mediate complex biological responses, autophosphorylation on multiple serine residues may serve as a possible mechanism for the activation of different signal transduction pathways leading to these responses. Although identified as autophosphorylation sites, these serine residues may also serve as targets for other cellular kinases, and thus the activity of TβRII could be up‐ or down‐modulated in response to an array of intracellular signals. The identification of these important serine residues serves as a starting point to dissect further the regulatory mechanisms and signal transduction pathways of the TGF‐β receptors.
Materials and methods
Ba/F3, a pro‐B cell line, was grown in RPMI medium supplemented with 10% fetal bovine serum (FBS) and 10% WEHI‐3 cell‐conditioned medium as a source of interleukin‐3. 293T, a human kidney cell line, and Bosc and Bing, two 293T cell‐based retroviral packaging cell lines (Pear et al., 1993) were maintained in Dulbecco–Vogt modified Eagle‘s medium (DMEM) supplemented with 10% FBS. Hep3B, a human hepatoma cell line, was grown in minimum Eagle's medium (MEM) supplemented with 10% FBS. The Hep3B.TR cell line (Inagaki et al., 1993) was derived from Hep3B cells and lacks endogenous TβRII. It was maintained in the same medium as Hep3B.
A polyclonal antibody directed against the extracellular domain of the EpoR (anti‐EpoR) was raised in rabbits against a GST fusion protein containing the EpoR extracellular domain (Klingmuller et al., 1995). A polyclonal rabbit antiserum against a peptide corresponding to a C‐terminal sequence of the human TβRII (anti‐RII) and a polyclonal antiserum specific for a cytoplasmic peptide in human ALK5 (anti‐RI) were described previously (Moustakas et al., 1995).
The following serine residues in the cytoplasmic domain of TβRII were mutated to alanine: Ser198, 199, 213, 225, 228, 229, 268, 286, 295, 343, 352, 353, 363, 382, 383, 401, 409, 416, 432, 441, 449, 502, 548 and 553. Point mutations of serine residues of TβRII were introduced by PCR with Pfu polymerase. All constructs were confirmed by sequencing.
Transfection and infection
The pLXSN and pBABEpuro DNA constructs were transfected into 293T, Bosc or Bing cells using the Ca2+ phosphate precipitation method described previously (Pear et al., 1993). Infection of Ba/F3 cells was performed using procedures described previously (Luo and Lodish, 1996). Infected cells were selected in a 24‐well cluster plate in growth medium containing either 600 μg/ml G418 (Gibco) for the chimeric E‐RII cDNAs or 1.5 μg/ml puromycin (Sigma) for the chimeric E‐RI cDNAs. Infected Hep3B and TR cells were selected in medium containing 600 μg/ml G418.
Growth inhibition assay
Thirty thousand Ba/F3 cells were grown for 4 days in 2 ml of medium containing varying concentrations of growth factors in one well of a 24‐well culture plate. Viable cells were then counted in a Coulter Counter. The extent of growth inhibition by the various factors was calculated from the ratio of the number of viable cells compared with cultures without any added factor. Similarly, 5×104 Hep3B cells were seeded in each well of a 6‐well cluster plate. Then, 24 h later, various concentrations of TGF‐β1 were added and the growth of the cells was evaluated 5 days later by cell counting.
Radioiodination of TGF‐β and binding–cross‐linking assay
Iodination of TGF‐β1 and the binding and cross‐linking assay were carried out as described previously (Moustakas et al., 1995).
Ba/F3 cells expressing the chimeric receptors were washed, resuspended at a concentration of 5×107 cells/ml and stimulated with 100 U of Epo (Amgen) at room temperature for 1–10 min. Cells were then lysed in NP‐40 lysis buffer [1% NP‐40, 50 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 50 mM NaF, 10 μg/ml aprotinin and 2 mM phenylmethylsulfonyl fluoride (PMSF)] at a concentration of 1×107 cells/ml for 20 min at 4°C. Immunoprecipitation and deglycosylation with PNGase F were carried out as described previously (Luo and Lodish, 1996).
In vitro kinase assay
In vitro kinase reactions using immunoprecipitates of the chimeric receptors bound to protein A–Sepharose were carried out in 20 μl of kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl2) in the presence of 10 μCi [γ‐32P]ATP (3000 Ci/mmol) for 10 min at 30°C. The reaction was terminated by the addition of 5 mM EDTA, and the immune complexes were eluted from the beads by boiling in SDS gel sample buffer. The labeled chimeric receptors were then re‐immunoprecipitated with antibodies to the extracellular domain of the EpoR.
Phosphoamino acid analysis and two‐dimensional tryptic peptide mapping
TβRII or E‐RII were labeled in vitro with [γ‐32P]ATP, separated by SDS–PAGE and transferred to a PVDF or a nitrocellulose membrane. Phosphoamino acid analysis was performed on proteins immobilized on PVDF membranes as described (Kamps, 1991). Tryptic digestion and two‐dimensional phosphopeptide mapping were carried out with protein bound to nitrocellulose as described previously (Luo et al., 1991).
Mapping of autophosphorylation sites
Bacterially expressed GST‐RII, a fusion protein of GST and the TβRII cytoplasmic domain (Luo et al., 1995), was digested with trypsin and the resulting peptides separated by reverse phase HPLC using a C18 column. Peptide fractions were collected based on absorbance at 210 nm and counted for the presence of 32P. The 32P‐labeled peptides were then subjected to amino acid sequencing and radio‐sequencing. To monitor the release of 32Pi from the peptide during radio‐sequencing, fractions collected after each cycle of Edman degradation were spotted on a TLC plate, and the TLC plate was exposed to an X‐ray film.
We are grateful to Richard Cooke at the MIT Biopolymer center for HPLC, peptide sequencing and radio‐sequencing analyses. We also thank Drs S.Brockman, U.Klingmüller, X.Liu, A.Sirotkin, H.Wu and Q.Zhou for critical reading of the manuscript. We thank Dr U.Klingmüller for the generous gift of affinity‐purified anti‐EpoR antibodies, Celtrix Laboratories, Inc. and R&D Systems for kindly providing TGF‐β1, and Amgen and Kirin Corp. for providing Epo. This work was supported in part by NIH grant R01 CA‐63260 to H.F.L. K.L. was supported by a postdoctoral fellowship from the National Institutes of Health and is currently a fellow of the Mary Ingraham Bunting Institute of Radcliffe College.
- Copyright © 1997 European Molecular Biology Organization