Molecular Mechanisms of RET Receptor–Mediated Oncogenesis in Multiple Endocrine Neoplasia 2B

The RET (rearranged in transfection) proto-oncogene encodes a receptor tyrosine kinase (RTK) with important roles in cell growth, differentiation, and survival. RET is expressed in cell lineages derived from the neural crest and in the tissues of the urogenital system (reviewed in ref. 1). Activation of the RET receptor is a multistep process, involving interaction with a soluble ligand of the glial cell line–derived neurotrophic factor (GDNF) family (GFL) and a nonsignaling cell surface–bound coreceptor of the GDNF family receptors α (GFRα). GFLs and GFRαs form complexes, which, in turn, bind to RET, triggering its dimerization and resulting tyrosine kinase activation (1).

Activating point mutations of RET cause multiple endocrine neoplasia type 2 (MEN 2), an inherited cancer syndrome characterized by medullary thyroid carcinoma (reviewed in ref. 2). The disease has three clinically distinct subtypes, ranging from the later onset, less severe, familial medullary thyroid carcinoma, characterized only by medullary thyroid carcinoma, to the more severe MEN 2A, characterized by medullary thyroid carcinoma, the adrenal tumor pheochromocytoma, and parathyroid hyperplasia. MEN 2B, the earliest onset and most aggressive form of MEN 2, is characterized by medullary thyroid carcinoma and pheochromocytoma, as well as by an array of developmental abnormalities, including marfanoid habitus, mucosal neuromas, ganglioneuromatosis of the intestinal tract, and myelinated corneal nerves (3). Morbidity and early mortality due to MEN 2B are very high. Management of both MEN 2B and sporadic medullary thyroid carcinoma is by surgical intervention. Treatment is complicated, as medullary thyroid carcinoma is prone to metastasis and is often refractory to both radiation and chemotherapy (4). Novel strategies, such as kinase inhibitors and small molecules, have as yet largely lacked specificity or efficacy at biologically tolerated dosages (1). Despite early genetic identification, “cure” is reported in <20% of MEN 2B cases (3).

Although all MEN 2 subtypes arise from mutations of RET, there are strong genotype/phenotype associations with specific mutations identified in each disease subtype. MEN 2A and familial medullary thyroid carcinoma mutations are primarily substitutions of one of several cysteine residues in the RET extracellular domain (2). More than 95% of MEN 2B cases are caused by a single germ-line mutation that results in substitution of a threonine for the normal methionine at residue 918 (M918T) in the RET kinase domain (5, 6). The same mutation occurs somatically in 50% to 70% of sporadic medullary thyroid carcinoma, where it can be associated with more aggressive disease and poor prognosis (7). The M918T mutation has been predicted either to induce a conformational change in the kinase catalytic core, leading to the activation of RET without ligand induced dimerization, or to alter the substrate specificity of RET, so that it preferentially binds substrates of cytoplasmic tyrosine kinases, such as SRC, or both (6, 8). Previous studies have predicted that the M918T mutation leads to a pattern of RET tyrosine phosphorylation, adaptor protein binding, and downstream signaling that differs in many respects from those associated with wild-type RET (WT-RET; refs. 6, 9). For example, in the absence of RET ligand, the M918T MEN 2B mutant (2B-RET) induces phosphorylation of proteins that interact with CRK and NCK, including the cytoskeletal protein paxillin, which seems to be more phosphorylated in the presence of 2B-RET than unactivated WT-RET (10). Together, these data have led to speculation that the M918T mutation may contribute to the increased activation of known RET signaling pathways or to activation of distinct, but as yet unidentified, pathways that may account for the earlier onset, broader phenotype, and increased severity of MEN 2B (10, 11).

Despite these predictions, the molecular consequences of the M918T mutation, and how this single amino acid substitution leads to the changes in RET activation or interactions that cause MEN 2B, have been difficult to confirm. Thus, elucidation of the mechanisms of 2B-RET function and dysfunction in MEN 2B require a better understanding of the molecular properties of both WT-RET and 2B-RET and of the activation and interactions of these kinases. Here, we have compared the biochemical, thermodynamic, cellular biological, and structural properties of WT-RET and 2B-RET to identify the underlying differences and similarities in these receptors. Our data show that the M918T mutation has multiple distinct and complementary effects on 2B-RET function including increasing intrinsic kinase activity, partially releasing kinase autoinhibition, and facilitating ligand-independent phosphorylation of 2B-RET receptors.

Homology modeling. Three-dimensional models of the RET tyrosine kinase domain (residues 709-988) were constructed using the crystal structure of the autoinhibited and active insulin receptor tyrosine kinase (IRK) as structural templates. RET and IRK sequences were aligned using CLUSTAL W (12) and manually adjusted for optimal structure prediction. The homology modeling program MODELLER (13) was used to read the alignment files of RET to autoinhibited IRK (Protein Data Bank code: IRK) and active IRK (Protein Data Bank code: 1IR3). Up to 100 hundred models were generated for each template, and the “best fit” model was selected as the model with the lowest value of the MODELLER objective function. The “best” active and autoinhibited models of RET were further refined by a series of energy minimization steps with the Discover module of the InsightII software (Accelrys Software, San Diego, CA). Models were energy minimized using 100 cycles of steepest-descent algorithm and evaluated for stereochemical quality by PROCHECK.

Expression constructs. Tetracycline-inducible RET expression constructs, generated by fusing cDNAs encoding a myristylation signal, two dimerization domains, and the intracellular portion (amino acid 658 to COOH terminus) of RET (icRET), have been described (14). Full-length human RET9 cDNA was cloned into pcDNA3.1 (Invitrogen, Burlington, Ontario, Canada). Constructs were validated for binding of known RET substrates (data not shown). Site-specific RET mutants were generated in WT-RET constructs by overlapping PCR, as described (15). Glutathione S-transferase (GST) fusion constructs encoding residues 664 to 1,072 of WT-RET and 2B-RET (with the M918T mutation) were generated in a modified pGEX-4T-3 vector. Constructs were verified by direct sequencing (Cortec, Kingston, Ontario, Canada). Expression and GST fusion constructs for GFRα1 (15), SHC (16), NCK (17), GRB10 (18), SRC (Y527F; ref. 19), and signal transducers and activators of transcription 3 (STAT3; ref. 20) have been described.

Protein purification and biophysical analyses. Purification of GST-RET proteins and biophysical analyses are described in Supplementary Methods.

Cell culture and transfection. HEK293 cells expressing the reverse-tetracycline transcriptional activator (Tet-on)(BD Biosciences, Mississauga, Ontario, Canada) were grown in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (Sigma, Oakville, Ontario, Canada) and 1 μg/mL doxycycline. SYF cells (19) were grown in the same medium without doxycycline. Constructs were transiently transfected into cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Full-length RET constructs were cotransfected with a GFR1 expression construct (15) and treated with 100 ng/mL GDNF (Promega, Madison, WI) for 15 minutes before harvesting. AP20187 dimerizer (1 μmol/L; ARIAD, Cambridge, MA) was added to induce icRET dimerization 30 minutes before harvesting.

Immunoprecipitations and Western blotting. Total protein lysates were harvested 48 hours after transfection and suspended in 20 mmol/L Tris-HCl (pH 7.8), 150 mmol/L NaCl, 1 mmol/L sodium orthovanadate, 1% Igepal, 2 mmol/L EDTA, 1 mmol/L phenylmethylsulfonyl fluoride, 10 μg/mL aprotonin, and 10 μg/mL leupeptin (15). Protein assays were carried out using the bicinchoninic acid protein assay kit (Pierce, Rockford, IL).

RET expression was detected using the C-19 antibody (Santa Cruz Biotechnology, Santa Cruz CA) and RET tyrosine phosphorylation was detected using an anti-phospho-RET antibody (Cell Signaling, Beverly, MA) that specifically recognizes phosphorylation of primary tyrosine residue Y905 (21) and an anti-phoshotyrosine antibody, pY99 (Sigma), which was also used to detect phospho-paxillin. For immunoprecipitations, lysates were incubated with a 1:50 dilution of the appropriate primary antibody with agitation for 2 hours at 4°C, mixed with Protein AG (Santa Cruz Biotechnology), and incubated on ice for 2 hours with shaking. Immunoprecipitates were pelleted, washed, and resuspended in SDS-PAGE sample buffer. Samples were denatured, separated on 10% SDS-PAGE gels, and transferred to nitocellulose membranes (Bio-Rad, Mississauga, Ontario, Canada; ref. 15). For our low-stringency nonreducing conditions, lysates were prepared in sample buffer in the absence of β-mercaptoethanol and separated on 6% SDS-PAGE gels, without any denaturation step.

GST pull-down assays. GST fusion proteins expressed in E. coli were eluted with a polyprep column (Bio-Rad) in 100 mmol/L glutathione elution buffer. For GST pull-down assays, 5 μg of GST fusion protein and GST sepharose beads (Amersham) were incubated with whole-cell lysates at 4°C for 3 hours with agitation. Bound proteins were resolved by SDS-PAGE as described above.

SRC kinase assay. Transfected SYF cells were harvested after 24 hours and immunoprecipitated with anti-RET or anti-SRC (Cell Signaling) antibodies. SRC kinase assay was done using a SRC kinase kit according to the manufacturer's instructions (Cell Signaling; ref. 22). The SRC kinase specific activity was calculated from the specific counts (total counts minus nonspecific counts). Nonspecific counts were determined by doing parallel assays in the absence of immunoprecipitates.

Soft agar colony formation assay. Soft agar colony formation assays were done as described (23). Briefly, HEK293 cells were transiently transfected with each of the icRET constructs or a control plasmid (pTRE2). Approximately 5 × 10⁴ cells were resuspended in 0.2% top agar in medium and plated on 0.4% bottom agar in medium. AP20187 (100 nmol/L) was added in both the lower and upper layers and culture medium supplemented with 100 nmol/L AP20187 was added to the top layer every 2 to 3 days. Colonies were counted after 2 weeks. Statistical significance was confirmed by one-way ANOVA.

Comparison of WT-RET and 2B-RET structure and activity. To identify the molecular mechanisms underlying MEN 2B, we first investigated whether the WT-RET and the oncogenic 2B-RET mutant differed in overall structure. We used circular dichroism (CD), a form of spectroscopy used to determine the secondary structure of molecules, to compare WT-RET and 2B-RET. Common secondary structure motifs exhibit distinctive CD spectra in the far-UV range whereas near-UV provides a fingerprint of protein tertiary structure. Using purified recombinant WT-RET and 2B-RET proteins, we compared the relative fraction of each molecule that was in α-helix, β-sheet (antiparallel/parallel), β-turn, or other (random) conformations. Far-UV and near-UV CD spectra of WT-RET and 2B-RET had profiles corresponding to globular proteins with defined secondary structure, comparable with those of typical kinases (Supplementary Fig. S1). Both far-UV and near-UV CD spectra of 2B-RET were superimposable on those of WT-RET, suggesting that there were no significant global changes in the secondary or tertiary structure of RET due to the M918T mutation.

We next asked whether WT-RET and 2B-RET differed in intrinsic properties such as ATP binding, thermostability, and enzyme kinetics. We used isothermal titration calorimetry, a thermodynamic technique for the evaluation of interactions between different molecules, to compare the ATP binding affinity of WT-RET and 2B-RET using purified recombinant proteins (Supplementary Fig. S2A). In isothermal titration calorimetry, heat released on interaction of two molecules can be used to determine dissociation constant (K_d), reaction stoichiometry (N), and thermodynamic variables including binding enthalpy (ΔH; Table 1). We found that the ATP equilibrium dissociation constant (K_d) for 2B-RET (15.3 μmol/L) was significantly lower than that of WT-RET (192 μmol/L), indicating that 2B-RET has >10-fold greater affinity for ATP (Table 1). Consistent with this, the heat of enthalpy was also more favorable for ATP binding to 2B-RET than WT-RET. As anticipated for RTKs, stoichiometry calculated by isothermal titration calorimetry suggested a single ATP binding site for WT-RET receptor monomers (N = 1.11). Interestingly, however, 2B-RET seemed to be associated with two ATP molecules (N = 2), suggesting that these molecules may associate in solution to form dimers that associate with two ATP molecules (Table 1). We further confirmed this using analytic gel filtration, which allows size separation of dimers and monomers in solution. Our data confirmed the presence of significantly more dimers for 2B-RET than WT-RET (Supplementary Fig. S3).

We compared the kinetic properties of purified WT-RET and 2B-RET in in vitro kinase assays (Supplementary Fig. S4). When kinase activity was compared over a range of ATP concentrations, we found that WT-RET required much higher ATP concentrations (K_M 188.2 ± 17 μmol/L) to achieve the same level of autophosphorylation as 2B-RET (K_M 105 ± 11 μmol/L; Table 1). The M918T mutation seemed to increase the stability of the enzyme-ATP complex and relatively strengthened ATP-binding. The result is a significant overall increase in the kinase activity of 2B-RET as compared with WT-RET.

Consistent with this increased ATP binding, we found that 2B-RET seemed to be more highly autophosphorylated both in the presence and absence of ligand stimulation as compared with activated WT-RET (Fig. 1). Further, when we investigated known RET substrates (e.g., paxillin, SHC, STAT3, NCK1, and GRB10), we found that although each of these bound both WT-RET and 2B-RET, all bound more 2B-RET, as we would expect due to the higher relative levels of 2B-RET autophosphorylation. However, when relative differences in autophosphorylation levels were taken into account, we found no differences in the binding of any tested substrate to phospho-WT-RET or phospho-2B-RET (Fig. 1).

To assess more localized changes in interactions between residues and domains due to the M918T mutation, we next used differential scanning calorimetry to compare the overall thermal stability and structural flexibility of WT-RET and 2B-RET conformations over a range of temperatures (20-120°C) and in the presence or absence of ATP (Supplementary Fig. S2B). In the absence of ATP, we found that WT-RET was more conformationally stable or rigid than 2B-RET, as indicated by a higher melting temperature [midpoint of denaturation transition (T_m), 65.10°C; Table 1]. The relatively higher energy requirement for denaturation is consistent with the tightly folded, more stable conformation of an autoinhibited (inactive) kinase structure (24, 25). 2B-RET had a lower melting temperature (T_m, 62.65°C), suggesting a relatively more flexible protein conformation with a less rigid overall structure. Interestingly, in the presence of ATP, we found that WT-RET and 2B-RET had similar melting temperatures (Table 1), indicating that the conformations of ATP-bound (active) WT-RET and 2B-RET enzymes were more similar than those of their autoinhibited forms. The lower melting temperature of the active forms was consistent with the thermodynamic effects of the release of autoinhibition in other kinases (24, 25). Together, these data suggested that, in addition to altered kinase activity, changes in intramolecular interactions and a potential reduction in conformational rigidity might play significant roles in the functional effects of the M918T mutation in 2B-RET.

Homology modeling of WT-RET and 2B-RET. To provide a framework for our structural and functional comparisons of WT-RET and 2B-RET, we generated three-dimensional homology models for RET. A BLAST search, using the amino acid sequence of the RET tyrosine kinase domain (residues 709-988), identified multiple similar RTKs with significant homology (Fig. 2A). The RET tyrosine kinase domain shares 53%, 52%, and 39% homology, respectively, with that of fibroblast growth factor (FGF) receptor 1, FGF receptor 2, and IRK. Of these, IRK was the only kinase for which the crystal structure is available for both active and autoinhibited forms, providing templates for homology modeling. Thus, we generated three-dimensional homology models of the kinase domain for each of WT-RET and 2B-RET in both their active and autoinhibited forms based on the corresponding IRK structures. Ribbon diagrams of the tyrosine kinase domain of RET show a bilobed structure that is common to all protein kinases (Fig. 2B).

Methionine 918 lies within the substrate-binding pocket of RET, in the P+l loop, adjacent to the activation loop (Fig. 2), and substitution of this residue with thronine seems to alter the shape of the substrate-binding pocket in our models (Supplementary Fig. S5). This would most likely reflect changes in amino acid interactions caused by replacement of a large methionine residue with a smaller, polar, threonine residue. Our models show that, in the autoinhibited form of WT-RET, M918 interacts with residues in the activation (I913) and P+1 (P914) loops and with neighboring residues further downstream (Y928, S922, and L923), which in turn interact with P+1 and catalytic loop residues, forming a tightly closed, autoinhibited structure (Fig. 3A). On activation of WT-RET, interactions between M918 and some of these residues (S922 and Y928) are reduced in strength or lost (predicted distance between residues exceeds 3.5 Å) while new interactions (V915) form, resulting in less tightly constrained, more open conformation. Interestingly, in the autoinhibited 2B-RET, T918 does not interact with I913, P914, or Y928, although it does interact with V915, and retains the interactions with S922 and L923 that are seen in autoinhibited WT-RET (Fig. 3A). Thus, T918 in autoinhibited 2B-RET retains some of the normal interactions of M918 in autoinhibited WT-RET but also acquires some of the properties and interactions associated with the active WT-RET.

Functional analyses of mutant RET proteins. Our models suggested that the M918T mutation may loosen the tight interactions among residues in the activation, catalytic, and P+1 loops thought to maintain WT-RET in an autoinhibited conformation. If relaxation of autoinhibition is a significant contributor to aberrant 2B-RET function, we would predict that mutations of other residues involved in these interactions in WT-RET would mimic the effects of M918T and lead to active receptors with similar properties to 2B-RET. Based on predictions from our models, we selected four critical residues (I913, P914, S922, and Y928; Fig. 3A) that interact with M918, and generated alanine substitution mutants at each position on a WT-RET background. As Y928 has been shown to act as a phosphospecific binding site in RET-mediated STAT signaling (26), we also generated a Y928F mutant, which would not bind these substrates but which would not alter the overall shape of the substrate-binding pocket. We used previously described RET expression constructs (14) that contain the intracellular domain of RET linked to dimerization domains and a membrane localization signal (icRET) that can be activated by treatment with a bivalent dimerizing agent to induce receptor dimerization (27). We found that mutant RET proteins were expressed at slightly lower levels than either WT-RET or 2B-RET (Fig. 3B), as has previously been noted for RET site-specific mutants (e.g., refs. 21, 28, 29). However, although WT-RET, 2B-RET, and the I913A and S922A mutant proteins were phosphorylated at relatively similar levels on induction of RET dimerization when expression was taken into account, there was some reduction in the relative phosphorylation levels of Y928F and Y928A RET mutants and almost complete loss of phosphorylation of the P914A mutant (Fig. 3B). Binding efficiency of WT-RET and RET mutants for a GST-tagged SHC-SH2 domain, a known RET substrate, correlated with autophosphorylation levels, with reduced or absent substrate binding for I914A and Y928A mutants (Fig. 3B).

We next tested the effects of mutating the predicted RET critical residues on anchorage-independent growth in soft agar colony formation assays. As predicted from our kinase activity data, we found that cells expressing 2B-RET formed significantly more colonies than those expressing a ligand/dimerizer activated WT-RET, or any of the critical residue mutants (P < 0.01; Fig. 3C). Interestingly, the I913A mutant, as well as the P914A and Y928A RET mutants, had significantly reduced colony-forming ability as compared with either 2B-RET (P < 0.01) or WT-RET (P < 0.05), and not appreciably different from an empty vector control (pTRE). However, despite expressing relatively less RET protein, the Y928F mutant had a comparable colony forming efficiency to activated WT-RET, whereas the S922A mutant had an increased colony forming efficiency relative to WT-RET (P < 0.05) and intermediate to that of activated WT-RET and 2B-RET (Fig. 3C). Together, these data show that substitution of I913 or P914 and the Y928A mutation do not mimic the M918T mutation, but in fact significantly impair one or all RET functions, suggesting that these residues have important structural or functional roles in addition to participating in the tight interactions that maintain the autoinhibited conformation of RET. Conversely, the Y928F and S922A mutations, which are predicted to reduce the sum of interactions between residues in the substrate binding pocket, catalytic loop, and activation loop (Fig. 3A), seem to retain significant autophosphorylation ability, substrate binding, and transforming potential, consistent with the type of functional effects associated with the M918T mutation in 2B-RET. These data suggest that mutations that reduce the tight interactions seen in the autoinhibited form of WT-RET can contribute to the activation or transforming potential of RET.

Characterization of 2B-RET preferred substrates. Previous studies have suggested that the M918T mutation might also alter 2B-RET substrate recognition such that it preferentially binds substrates of cytoplasmic kinases, such as SRC (6). To determine whether 2B-RET was more able to act on SRC substrates than was WT-RET, we examined their relative ability to phosphorylate a normal SRC substrate peptide (KVEKIGEGTYGVVYK, derived from p34^cdc2; ref. 22). As SRC itself is also a known RET substrate (30), we assessed RET-mediated phosphorylation of SRC substrates in SYF cells, which do not express the three SRC family kinases SRC, YES, or FYN (19), to avoid contamination of our assays by SRC kinase activity. Using a series of full-length RET expression constructs, we first confirmed that, on RET activation with GDNF and GFRα1, both WT-RET and 2B-RET were autophosphorylated (Supplementary Fig. S6) and, further, that 2B-RET was more highly autophosphorylated than WT-RET in SYF cells. Our data indicate that SRC family kinases are not required for RET phosphorylation and, further, that phosphorylation by SRC is not a major cause of the higher 2B-RET phosphorylation. Interestingly, following receptor activation, both WT-RET and 2B-RET phosphorylated the p34^cdc2 SRC substrate peptide (Supplementary Fig. S6B). Moreover, as seen for other RET substrates, the level of substrate phosphorylation correlated well with the relative levels of autophosphorylation of WT-RET and 2B-RET and was comparable to that of an activated SRC control. Similar results were observed using both a specific anti-phospho-RET antibody (pY905) and a pan-phosphotyrosine (pY99) antibody (not shown). When we controlled for the different autophosphorylation levels of WT-RET and 2B-RET, we saw no significant difference in relative SRC substrate phosphorylation (Supplementary Fig. S6B), suggesting that intrinsic differences in recognition of this specific SRC substrate by WT-RET and 2B-RET are minimal.

Monomeric and dimeric RET proteins. Previous studies have shown that mutant forms of RET containing an MEN 2A type cysteine substitution mutation in the extracellular domain (2A-RET) form dimers or higher molecular weight protein complexes constitutively without ligand stimulation, whereas WT-RET does not in the absence of ligand (31, 32). The active form of 2B-RET has been predicted to be either a constitutive dimer or cis-phosphorylating monomer (33). We compared the abilities of full-length WT-RET, 2A-RET, and 2B-RET to form higher molecular weight protein complexes under very low-stringency nonreducing conditions (Fig. 4). In the presence of GDNF, we found both WT-RET and 2B-RET primarily in high molecular weight complexes, presumed to include dimerized receptors and interacting proteins (Fig. 4A). However, in the absence of ligand, and without any form of protein denaturation (e.g., boiling), WT-RET is found primarily as monomers with minimal complex formation (Fig. 4B), whereas a significant fraction of 2B-RET and 2A-RET are present in higher molecular weight protein complexes. When we compared the phosphorylation of WT-RET, 2B-RET, and 2A-RET, we found that, in the absence of ligand, 2B-RET and 2A-RET are more phosphorylated than WT-RET (Fig. 4B), and that the higher molecular weight protein complexes formed by both 2B-RET and 2A-RET are phosphorylated and of similar size. Further, whereas phospho-RET is found in the high molecular weight complexes for both 2A-RET and 2B-RET, the phosphodimer/phosphomonomer ratio is higher for 2A-RET than 2B-RET, suggesting that a greater proportion of phospho-2B-RET occurs as phosphomonomers under nonreducing conditions. Under reducing conditions, the higher molecular weight protein complexes were not seen (Fig. 4B) whereas a modest denaturation step, such as boiling samples before SDS-PAGE, eliminated all but 2A-RET dimers (not shown). Thus, our data suggest that 2B-RET exists both as activated monomers and as dimers but that these dimers are less stable than those formed by 2A-RET through formation of covalent intermolecular bonds.

MEN 2B is the most severe form of multiple endocrine neoplasia type 2, characterized by early onset endocrine tumors and a broad range of developmental abnormalities (3). More than 95% of MEN 2B is caused by a methionine to threonine substitution of residue 918 in the kinase domain of RET (5) and the identical mutation occurs in >50% of sporadic medullary thyroid carcinoma, where it has been associated with a poorer prognosis (7). Despite predictions about the mechanisms of this mutation, experimental confirmation of its functional effects on RET has been lacking. In this study, we have characterized the thermodynamic and kinetic properties of the RET kinase and investigated the proposed molecular mechanisms underlying the M918T mutation. Our data indicate that this mutation has multiple related but distinct effects on the RET receptor.

Our data suggest that the primary effects of the M918T mutation may be to enhance the intrinsic kinase activity of 2B-RET. We found that the affinity of 2B-RET for ATP was >10-fold greater than that of WT-RET (K_d 15.3 versus 192 μmol/L, respectively). Further, WT-RET required a much higher concentration of ATP to achieve the same levels of autophosphorylation seen with 2B-RET (K_M 188 versus 105 μmol/L), suggesting that the 2B-RET-ATP complex was significantly more stable. Together, these effects would significantly increase the relative kinase activity of 2B-RET. As we would predict from these biophysical studies, we showed that 2B-RET is more highly autophosphorylated than activated WT-RET, and binds and phosphorylates correspondingly more of its substrates in multiple cell types (Figs. 1 and 3; Supplementary Fig. S6). Our data also confirmed that this is an intrinsic property of 2B-RET and was not dependent on the presence of other kinases, such as SRC family kinases, to activate RET because in vitro analyses using purified recombinant proteins and in vivo confirmation in SYF cells clearly showed that 2B-RET achieves the same high level of autophosphorylation, irrespective of the presence of SRC family kinases (Supplementary Fig. S6).

The increase in 2B-RET-ATP binding is intriguing, as the M918T mutation is located in the substrate binding region of the receptor, distant from the sequences of the ATP binding cleft (Fig. 2). Mutations that specifically increase ATP binding (as opposed to other activation mechanisms) and alter downstream signaling have been identified in oncogenic forms of other RTKs, such as epidermal growth factor receptor (EGFR), but these mutations are primarily localized within or close to the ATP binding cleft, where they directly affect the interactions between the receptor and ATP (34, 35). The location of M918, distant from this region, suggested a novel effect on RTK conformation that might contribute to altered ATP binding. Our initial analyses using CD did not identify overall global differences in the secondary or tertiary structures of WT-RET and 2B-RET (Table 1; Supplementary Fig. S1), suggesting that the M918T mutation might cause more local changes in interactions, possibly affecting RET autoinhibition, which indirectly affected ATP binding. In the absence of activation, RTK monomers adopt a closed “autoinhibited” conformation in which the activation loop blocks access to the substrate-binding pocket. This conformation is very energetically stable (high T_m) due to tight intramolecular interactions that result in a rigid conformation. On activation, the kinase becomes much more flexible as autoinhibiting interactions are released, and it takes on a more open conformation (lower T_m). Consistent with these predictions, when we compared thermal denaturation profiles generated by differential scanning calorimetry, we found that WT-RET had a higher T_m, suggesting a more rigid conformation, whereas 2B-RET had a lower T_m, suggesting a more flexible, partially open conformation. In the presence of ATP, T_m was much lower for both receptors, consistent with the more flexible, open conformation of an activated kinase. Together, these data suggested that 2B-RET has a more flexible conformation that may reflect partial loss of autoinhibition.

The more flexible structure predicted for 2B-RET may be due, in part, to a reduction in the sum of tight molecular interactions between residues in the activation loop, catalytic loop, and substrate binding P+1 loop, which are important in maintaining autoinhibition (Fig. 3). Our homology model and functional data on critical residues suggest some interactions that may contribute to this process. We found that mutation S922A significantly potentiates colony formation over an activated WT-RET, which is predicted to act in a similar fashion to 2A-RET receptors (ref. 36; P < 0.05, one way ANOVA), mimicking, at least in part, the effect of the M918T mutation in 2B-RET (Fig. 3C). Consistent with this, mutations of serine 922, to either phenylalanine or proline, have been detected in sporadic medullary thyroid carcinoma (37, 38), indicating that mutations of this residue may also be transforming in vivo. Conversely, whereas I913A and Y928A mutants retained some ability to autophosphorylate and bind substrate, they did not promote growth in soft agar assays, even on ligand stimulation (Fig. 3), suggesting a partial loss of some RET functions. These data are consistent with the M918T mutation having multiple distinct effects on RET structure and function, only some of which are mimicked by these critical residue mutations.

Together, our models and our functional and thermodynamic analyses suggest that the M918T mutation causes a local conformational change in the 2B-RET kinase that partially releases autoinhibition, increasing ATP-binding. However, structural release of substrate blockade alone cannot account for potentiated 2B-RET function. Our data further show that, on ligand induced activation, both WT-RET and 2B-RET form high molecular weight complexes, but that 2B-RET and 2A-RET also form these dimers in the absence of ligand under low-stringency nonreducing conditions (Fig. 4). Further, whereas the ratios of RET dimers and monomers are similar for 2A-RET and 2B-RET, the proportion of phospho-RET in dimers is greater for 2A-RET than for 2B-RET. All RTKs are thought to exist in equilibrium between monomeric and dimeric pools, even in the absence of ligand (39, 40). In general, the dimeric forms are nonproductive interactions as these RTKs are locked in the autoinhibited state, and their formation is transient and energetically unfavorable (Fig. 5). For wild-type receptors, extracellular ligand binding stabilizes the formation of active dimers and leads to kinase stimulation. In 2A-RET, intermolecular disulfide bonds link extracellular domain residues and mimic this effect, leading to constitutively active RET dimers (8, 41). Interestingly, 2A-RET dimers seem to be stronger than those between 2B-RET monomers, which were detected in the absence of denaturation but were abolished by even a brief denaturation step (not shown). As described above, our data suggest that, in the presence of the M918T mutation, 2B-RET adopts an intermediate, partially open conformation, although it remains conformationally unlikely that these monomers can self-phosphorylate. However, the stoichiometry of 2B-RET-ATP binding and our analytic gel filtration data indicating that purified 2B-RET forms dimers, as well as the detection of phospho-2B-RET in high molecular weight protein complexes in the absence of ligand stimulation, suggest that the monomer-dimer equilibrium is shifted for 2B-RET, increasing the pool of transient dimers (Figs. 4 and 5). These complexes enable transphosphorylation of RET receptors, followed by phosphorylation of downstream targets; however, in the absence of conventional ligand and coreceptor, complexes are not as stable as dimers formed by 2A-RET or activated WT-RET. As a result, complex formation is transient and leads to a pool of phosphorylated monomeric 2B-RET, dissociated from the higher molecular weight complexes that may also be able to initiate downstream signals (Fig. 5). We would thus predict that the sum of both active dimers and monomers may also contribute, in part, to the increased activity of 2B-RET that we observe.

Finally, previous studies have suggested that 2B-RET may also recognize novel substrates or have a shift in substrate preference, allowing it to activate additional, or alternative, signaling pathways distinct from the targets of WT-RET (6, 10). However, these studies generally compared 2B-RET with unactivated WT-RET before the recognition of GDNF as the RET ligand. In the presence of ligand stimulation, we did not detect differences in the nature of WT-RET and 2B-RET substrates, and the relative differences in binding or phosphorylation of substrates correlated completely with the relative levels of WT-RET and 2B-RET autophosphorylation (Figs. 1 and 3; Supplementary Fig. S6). Substrates such as paxillin (Fig. 1), previously thought to be unique to 2B-RET (10), also seem to bind ligand-stimulated WT-RET, although at lower levels. We also did not detect a preference for binding of the SRC substrate p34^cdc2 by 2B-RET, although we cannot exclude differences in binding of WT-RET and 2B-RET to other SRC substrates. Together, these data are consistent with studies of RET in Drosophila, which failed to detect novel 2B-RET specific substrates (42), and, with the lack of novel substrates identified to date in numerous studies, suggest that activation of additional signaling pathways through strong interactions with uniquely 2B-RET substrates may not be a major contributing component of 2B-RET onco-activity.

In summary, we have shown that a combination of mechanisms, including increased kinase activity, partial release of autoinhibition, and a relative increase in ligand-independent formation of activated monomers and dimers, contributes to 2B-RET activity. Our data suggest that the effect of these distinct mechanisms on kinase activity may be at least partly additive, which may contribute to the relative severity and broader phenotype of MEN 2B as compared with that of other MEN 2 forms. Understanding the molecular mechanisms of 2B-RET has important implications for development of therapeutics with some specificity for mutant RET forms. Pharmacologic small-molecule inhibitors that target ATP binding sites have well-proven clinical usefulness but also have potential for broad side effects related to normal functions of the targeted kinase. As RET is known to be an important neuronal survival receptor (30), therapeutic targeting of mutant receptors, while sparing the wild-type molecule, would be advantageous. Mutations of EGFR that increase ATP binding have been shown to increase sensitivity to the inhibitor gefitinib in non–small-cell lung cancer (35), whereas MET molecules with a M1268T mutation that corresponds to M918T have proved more sensitive to ATP blocking inhibitors (43). Thus, specific targeting of ATP binding in RET may be a valuable tool to aid in the development of anticancer therapies for both MEN 2B and for sporadic tumors that harbor this mutation.

Grant support: Canadian Institutes of Health Research and the Canadian Cancer Society (L.M. Mulligan), and Canadian Institutes of Health Research traineeships in Transdisciplinary Cancer Research and Protein Function Discovery (T.S. Gujral).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Rob Campbell and Kim Munro for assistance, and J. McGlade, L. LaRose, J. Duyster, B. Elliott, and J. Darnell for providing us with expression constructs.