Critical Role of the Platelet-derived Growth Factor Receptor (PDGFR) β Transmembrane Domain in the TEL-PDGFRβ Cytosolic Oncoprotein*

The fusion of TEL with platelet-derived growth factor receptor (PDGFR) β (TPβ) is found in a subset of patients with atypical myeloid neoplasms associated with eosinophilia and is the archetype of a larger group of hybrid receptors that are produced by rearrangements of PDGFR genes. TPβ is activated by oligomerization mediated by the pointed domain of TEL/ETV6, leading to constitutive activation of the PDGFRβ kinase domain. The receptor transmembrane (TM) domain is retained in TPβ and in most of the described PDGFRβ hybrids. Deletion of the TM domain (ΔTM-TPβ) strongly impaired the ability of TPβ to sustain growth factor-independent cell proliferation. We confirmed that TPβ resides in the cytosol, indicating that the PDGFRβ TM domain does not act as a transmembrane domain in the context of the hybrid receptor but has a completely different function. The ΔTM-TPβ protein was expressed at a lower level because of increased degradation. It could form oligomers, was phosphorylated at a slightly higher level, co-immunoprecipitated with the p85 adaptor protein, but showed a much reduced capacity to activate STAT5 and ERK1/2 in Ba/F3 cells, compared with TPβ. In an in vitro kinase assay, ΔTM-TPβ was more active than TPβ and less sensitive to imatinib, a PDGFR inhibitor. In conclusion, we show that the TM domain is required for TPβ-mediated signaling and proliferation, suggesting that the activation of the PDGFRβ kinase domain is not enough for cell transformation.

PDGFR␤ 3 is a single-spanning transmembrane glycoprotein that binds to its dimeric ligand PDGF. It belongs to the type III receptor tyrosine kinase family, which also comprises PDGFR␣, c-KIT, Flt3, and c-Fms (1). The canonical mechanism of activation of receptor tyrosine kinases requires ligand-induced dimerization, which brings two kinase domains close to each other and allows the phosphorylation of critical regulatory tyrosine residues in the activation loop of the catalytic core with a subsequent boost of the receptor kinase activity (2). Recent findings demonstrated that dimerization is not sufficient by itself to activate these receptors. The extracellular Ig-like domain D4 must reorganize and establish contacts between two neighbors receptors to provide optimal tyrosine kinase activation (3,4). The intracellular juxtamembrane domain is devoted to the inhibition of the catalytic activity in the absence of the ligand by interacting with the kinase domain and inhibiting its activity (5)(6)(7). Mutations in the juxtamembrane region can alleviate this inhibition and activate the receptor in a ligand-independent manner (8,9). The C-terminal tail of PDGFR␤ is also blocking the receptor phosphorylation by allosteric inhibition (10). The phosphorylated receptor tyrosine residues bind SH2 domain-containing proteins, such as the p85 subunit of phosphatidylinositol 3-kinase, STAT5, or phospholipase C␥. The cascade of events initiated by ligand binding will ultimately affect gene expression and modulate cell proliferation, differentiation, and motility (11)(12)(13).
PDGFR genes are found rearranged in a certain subset of chronic myeloid malignancies (14 -16). The resulting hybrid genes encode constitutively active forms of receptors that contain the receptor intracellular kinase domain fused to the N-terminal part of a partner that can differ from one hybrid receptor to another. Among these, the hybrid between the ETV6 (ets variant gene 6)/TEL (translocation-ets-leukemia) transcription factor and PDGFR␤ (TP␤) is the most recurrent one and is encountered in patients with chronic myelomonocytic leukemia (17). TP␤ is activated in a ligand-independent manner by enforced dimerization mediated by the pointed (PNT, also called SAM or helix-loop-helix) domain of TEL (18). The activation of the transcription factor STAT5 has been demonstrated to be crucial for the transforming potential of the hybrid protein in cell lines and in mouse models (19,20). Furthermore, the levels of both STAT5 and TP␤ proteins are critical for cell transformation (21). In particular, TP␤ is not efficiently degraded in cells, and we showed that its increased stability promotes cell proliferation (22). Recently, activation of ERK signaling proteins was indicated as a mediator of TP␤-induced stem cell differentiation (23).
In TP␤ and in most of the described PDGFR␤ hybrids, but not in hybrids derived from other receptor tyrosine kinases, the transmembrane sequence of the receptor is retained in the * This work was supported by grants from the Bekales Foundation, from Action de Recherche Concerté e (Communauté Française de Belgique), and from the Salus Sanguinis Foundation, Belgium. 1 Supported by a Télé vie fellowship. 2
Constructs and Mutagenesis-TP␤ in pMSCV-eGFP vector was described elsewhere (18). All of the TP␤ mutants were created by site-directed mutagenesis using QuikChange XL-II kit (Stratagene) according to the manufacturer's instructions. All of the constructs were checked by sequencing. FLAG and HA tag were cloned at the 5Ј site of TP␤ into a previously introduced Age1 site replacing the ATG codon of TP␤. Human HA-PDGFR␤ was cloned in pEF-BOS-puro as described (25). The HA tag was inserted after the sequence encoding the signal peptide of PDGFR␤ in the position predicted by the SignalP software (26).
Transient transfections of BOSC, PAE, and HEK-293T cells were performed using the calcium phosphate method as described elsewhere (22). Ba/F3 and 32D cell lines stably expressing receptors were created as previously reported (30). Briefly, retroviral supernatants were generated by co-transfection of the BOSC packaging cell line with plasmid DNA encoding the receptor and the ecotropic envelope protein using the calcium phosphate method (22). After 48 h, the supernatants were harvested and used in a spin infection protocol. One million cells were centrifuged for 2 h at 37°C and 300 ϫ g in presence of viral supernatant (1 ml) and Polybrene (20 g; Sigma). The cells were then resuspended in medium with 10% FBS and IL3. After 24 h, GFP-positive cells were isolated by fluorescence-activated cell sorting and maintained in DMEM with FBS and IL3. All of the cells used in the experiments showed an equal level of GFP expression.
Alternatively, 10 7 32D cells were electroporated with 60 g of DNA and diluted in 30 ml of DMEM with 10% FBS and IL3 as described earlier (28,29). After 48 h, the cells were selected with 3 g/ml puromycin for 14 days. Homogenous 32D cell lines expressing PDGFR␤ were described previously (29).
In [ 3 H]thymidine incorporation assays, Ba/F3 cells stably expressing receptors were washed extensively and seeded in triplicate in a 96-well plate in DMEM with 10% FBS in the presence of the indicated growth factor. After 20 h, [ 3 H]thymidine (0.5 Ci/well; GE Healthcare) was added for 4 h. The cells were then harvested, and the incorporation of [ 3 H]thymidine was quantified using a TopCount instrument (PerkinElmer Life Sciences).
Ba/F3 cells stability assays were performed as described previously (22). Briefly, 2 ϫ 10 6 cells expressing the indicated hybrid receptors were incubated with 50 g/ml cycloheximide for the indicated periods of time. The samples were collected at each time point and analyzed by Western blot.
Cross-linking Treatment, Immunoprecipitation, and Western Blot-Cross-linking assays were performed in the presence of bis(sulfosuccinimidyl)-suberate (BS3; Pierce). Briefly, 2 ϫ 10 5 cells were washed once with ice-cold PBS and then lysed in 200 l of 50 mM HEPES, pH 7.5, 150 mM NaCl, glycerol 10% (w/v), Triton 1% (w/v), EDTA 1 mM, 1 mM Pefabloc (Roche Applied Science), 1 g/ml aprotinin, and 1 mM Na 3 VO 4 . After clearing by centrifugation, the lysates were incubated with 0.25, 0.5, or 1 mM BS3 for 1.5 h at 4°C. The reactions were stopped by the addition of 50 mM Tris-HCl for 15 min at room temperature. The samples were analyzed by Western blot as described (22). Quantification of bands was performed after incubation with secondary fluorescent antibodies (IRDye) using the Odyssey system (Li-Cor) or using the ImageJ software (31) on scans of BioMax films (Kodak).
In immunoprecipitation experiments, the cells were lysed 24 h after transfection in 25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 10% glycerol, 1% Triton X-100, 1 mM Pefabloc, 1 g/ml aprotinin, and 1 mM Na 3 VO 4 . The cell lysates were centrifuged full speed for 20 min at ϩ4°C. In co-immunoprecipitation experiments with p85, the lysates were cleared with protein A/G for 1 h before being incubated overnight with the antibody. In double tag immunoprecipitations, the cell lysates were incubated overnight with 1 g of antibody at ϩ4°C after the centrifugation. Antibody complexes were collected by adding protein A/G Ultralink (Pierce) for 1 h at ϩ4°C, washed extensively, and then analyzed by Western blot.
Flow Cytometry and Immunofluorescence Staining-For cell surface staining, about 5 ϫ 10 5 cells were incubated for 1 h at 4°C with primary antibody diluted in Hanks' buffer complemented with 3% FBS and 1% NaN 3 (HAFA buffer). Conditions without antibody were included as a control. After one wash with HAFA, the cells were incubated with secondary antibody conjugated to phycoerythrin for 45 min at 4°C in the dark. After one more wash, the cells were analyzed by flow cytometry.
In intracellular staining experiments, the cells were washed extensively and incubated for 4 h in absence of IL3 and, in some experiments, with 500 nM imatinib. As a positive control, some cells were restimulated with IL3 for 5 min after starvation. The cells were fixed with 2% formaldehyde in PBS for 10 min at 37°C and then permeabilized with methanol on ice for 30 min. Following two washing steps with HAFA buffer, the cells were incubated with the antibody conjugated to Alexa-Fluor 647 (BD Transduction Laboratories) for 1 h at room temperature. The cells were washed and analyzed by flow cytometry. A condition without antibody was included as an additional control. The average of at least two independent experiments is shown with S.D., and a Student's t test was applied. In double staining experiments, the cells were at first incubated for 1 h with anti-PDGFR␤ antibody and, after two washing steps, with anti-phospho-STAT5 Alexa-Fluor 647 and anti-rabbit phycoerythrin. Staining with the two antibodies separately was also included as a control. One representative experiment is shown (n ϭ 4).
Immunofluorescence staining of transfected PAE cells was performed as follows: the cells grown on coverslips were fixed with 4% formaldehyde in PBS for 20 min and then washed three times with cold PBS. Permeabilization was performed with 5% FBS and 0.5% saponin in PBS for 1 h. The cells were then incubated with anti-PDGFR␤ antibody overnight at ϩ4°C. The cells were washed three times with TBST (50 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20, pH 7.6) and then incubated for 1 h with antirabbit antibodies conjugated to phycoerythrin. The coverslips were washed, mounted on slides, and observed with a fluorescence microscope (630ϫ magnification).
In Vitro Phosphorylation Assays-The receptors were immunoprecipitated from transiently transfected 293T cells as described above. To test receptor autophosphorylation, the cells were treated with 1 M imatinib for 4 h prior to lysis. For in vitro kinase reactions, the immunoprecipitated receptors were incubated with 50 M ATP (Fermentas) in 50 l of 50 mM HEPES, pH 7.4, and 10 mM MgCl 2 . To test kinase activity toward an exogenous substrate, 10 g of dephosphorylated myelin basic protein (MBP; Active Motif) were added to the reaction mixture. The reactions were incubated for 15 min at 30°C, then stopped by the addition of Laemmli buffer, and analyzed by Western blot. Phosphorylated receptors and MBP were separated on 8 and 18% gels, respectively.

RESULTS
The PDGFR␤ Transmembrane Domain Is Critical for TP␤mediated Proliferation of Hematopoietic Cells-The PDGFR␤ gene breakpoints that are associated with translocations and reported in hematological diseases are located either before or after the exon sequence encoding the receptor TM domain. Most of the translocation products reported so far retain the TM domain in the fusion protein (Fig. 1A, right panel). To understand the significance of this observation, we analyzed the role of the TM domain in TP␤, which is the most recurrent The position marked by an asterisk is usually in the intron preceding exon 11, which encodes the TM sequence, but breakpoints in introns before exon 9 or 10 were also reported. The sequence of the TM domain is shown in capital letters. In the right panel are listed the fusion partners of PDGFR␤ with the corresponding breakpoint positions in PDGFR␤. B, Ba/F3 or 32D cell lines were transduced with TP␤ or ⌬TM-TP␤ using a bicistronic retroviral vector encoding GFP and sorted according to GFP levels. The cells were incubated for 24 h in absence of IL3. Proliferation was measured by a [ 3 H]thymidine incorporation assay as described under "Experimental Procedures." All of the cell lines proliferated equally in the presence of IL3 (data not shown). S.D. were calculated from triplicate cultures in a representative experiment. Vector-transfected cells were used as a control. C, total cell lysates derived from the Ba/F3 cell lines used in B were analyzed by Western blot (WB) with anti-PDGFR and anti-␤-actin antibodies. The total cell lysates derived from the 32D cell lines were analyzed similarly (shown in Fig. 2A).
PDGFR␤ hybrid form found in chronic myeloid malignancies. We created a mutant of TP␤ devoid of TM sequence (⌬TM-TP␤) and examined its activity in the Ba/F3 and 32D hematopoietic cell lines. In line with previous reports, expression of TP␤ in these cell lines supported short term proliferation in the absence of IL3 as detected in thymidine incorporation assays. TP␤-expressing cells could also be maintained in culture without IL3 for longer periods of time, becoming cytokine-independent cells. Expression of the ⌬TM-TP␤ mutant instead resulted in a dramatic reduction of IL3-independent cell growth in the short term proliferation assay (Fig. 1B). ⌬TM-TP␤ cells could not reproducibly generate cytokine-independent cell lines. Our results suggest that the TM sequence has an important role in TP␤ transforming properties in hematopoietic cells.
TP␤ Is Not a Membrane-spanning Protein-In TP␤, the sequence encoding the signal peptide and the extracellular domain of PDGFR␤ is replaced by the sequence encoding the N-terminal part of TEL. Even though this makes the cell surface expression of the protein an unlikely event, the presence of the TM sequence could still allow the insertion into membranes.
To check the plasma membrane localization of the fusion protein, we created tagged versions of TP␤ and ⌬TM-TP␤ introducing an N-terminal HA tag and analyzed their expression by flow cytometry (Fig. 2A). HA-tagged wild-type PDGFR␤, which was used as positive control for the staining, showed a strong surface signal with the anti-HA antibody, whereas both forms of hybrid TP␤, with and without the TM domain, were negative for the anti-HA staining. Because the absence of cell surface expression of the hybrids did not exclude localization at the level of intracellular membranes, PAE cells were transfected with the receptors, stained with anti-PDGFR␤ antibody, and analyzed by fluorescence microscopy (Fig. 2B). Activated wildtype PDGFR␤ showed a typical punctuated pattern, which results from the internalized receptor complexes that are formed in the presence of the ligand (32). TP␤and ⌬TM-TP␤expressing cells appeared with a diffuse staining in the cytoplasm and did not show accumulation of the proteins in any particular subcellular compartment like the endoplasmic reticulum or the Golgi. This was confirmed by confocal microscopy (data not shown) and was in line with a previous report (33). Altogether these observations confirmed that TP␤ is not a membrane protein.
Deletion of the TM Domain Affects TP␤ Cross-linking-The pointed domain of TEL mediates oligomerization of TP␤, which is essential for the constitutive activation of the hybrid and transformation of hematopoietic cell lines (18). Recently, it has been shown that a purified peptide containing the TM domain of PDGFR␤ is able to form dimers in vitro (34). We sought to determine whether the TM sequence could contribute to the oligomerization in the context of the hybrid receptor. We initially tested the ability of ⌬TM-TP␤ to self-associate by using a co-immunoprecipitation approach with two different tags. FLAG-tagged and HA-tagged versions of the hybrid were co-transfected in 293T cells, alone or in combination, as illustrated in Fig. 3A. After immunoprecipitation with anti-HA antibodies, the interaction between two differently tagged proteins was examined by Western blot against the FLAG tag and vice versa. As expected, HA-and FLAG-TP␤ were co-immunoprecipitated (Fig. 3A, lane 5, two upper panels). Similar results were obtained with ⌬TM-TP␤, indicating that TP␤ selfassociation is retained in the ⌬TM-TP␤ mutant (lane 6, two upper panels). To further assess the oligomerization of ⌬TM-TP␤, we performed cross-linking experiments in the presence of BS3, a cross-linker that has been previously used to study the oligomerization of TP␤ (35). Cell lysates obtained from Ba/F3 cells expressing TP␤ or ⌬TM-TP␤ were incubated with increasing doses of BS3, and oligomer formation was visualized by Western blot (Fig. 3B). As a negative control we used the ⌬PNT-TP␤ mutant, which lacks the domain required for oligomerization. As previously reported, high molecular species of TP␤ protein were visible after treatment with the crosslinker, whereas only the monomeric form of ⌬PNT-TP␤ was visible, indicating that the TM domain alone does not induce aggregation in the absence of the PNT domain. In addition, ⌬TM-TP␤ could still form oligomers with a similar pattern of bands compared with TP␤ but with decreased signal intensity.
Altogether these experiments showed that the TM domain is not absolutely required for TP␤ oligomerization but favors its cross-linking.
⌬TM-TP␤ Signal Transduction Is Decreased-We next analyzed the phosphorylation of the ⌬TM-TP␤ mutant. Total cell lysates from Ba/F3 cell lines were immunoblotted with general anti-phosphotyrosine antibodies. Both TP␤ and ⌬TM-TP␤ were strongly phosphorylated, as shown in Fig. 4A. To compare the signaling properties of the two proteins, we performed intracellular staining of phospho-STAT5 and phospho-ERK1/2 (Fig. 4B), which are two important signaling mediators of TP␤ (19,20,23). As expected, TP␤ induced the phosphorylation of both proteins in Ba/F3 cells. This effect was inhibited by treatment with imatinib (500 nM), which blocks selectively the PDGFR␤ activity at that concentration (Fig. 4B). ⌬TM-TP␤ showed a much lower percentage of positive cells in comparison with TP␤ in the staining for activated STAT5 and ERK. These data indicated that ⌬TM-TP␤ was phosphorylated but unable to signal effectively, in line with its inability to sustain cell growth.
Tyrosine 581, which is located within the juxtamembrane domain of PDGFR␤, has been described, together with tyrosine 579 and 775, as a docking site for STAT5 (36). Among the three sites, the sequence surrounding tyrosine 581 fits better the suggested consensus sequence for STAT5 recruitment (25), and its mutation in wild-type PDGFR␤ has the strongest effect on STAT5 phosphorylation (36). Therefore, we assessed the phosphorylation of PDGFR␤ tyrosine 581 using a phospho-specific antibody in Western blot experiments (24). Fig. 4C shows that tyrosine 581 was phosphorylated to a similar extent in ⌬TM-TP␤ and TP␤. As a control for the specificity of the antibody, we used a kinase-inactive mutant, K654R TP␤ (Fig. 4C) (37). The observation that ⌬TM-TP␤ is strongly phosphorylated on tyrosines, including the most important STAT5 docking site, suggested that the inability of ⌬TM-TP␤ to signal was not due to a defect in phosphorylation.
Increased Degradation of ⌬TM-TP␤-Because we observed that the level of expression of ⌬TM-TP␤ in Ba/F3 cells was lower than TP␤ in some cell lines (Fig. 2, for instance), we analyzed the stability of the protein by inhibiting protein synthesis with cycloheximide. As shown in Fig. 5A, ⌬TM-TP␤ was degraded faster than TP␤, suggesting that the TM deletion destabilized the protein. To test whether the impaired oncogenic activity ⌬TM-TP␤ resulted from its lower level of expression, we transduced the ⌬TM-TP␤ cell line a second time and resorted the cells that had an increased GFP level. We obtained the ⌬TM-TP␤2X cell line, which presents a higher level of expression as shown by Western blot and flow cytometry in Fig.  5. The protein level of ⌬TM-TP␤2X was more comparable with the level of TP␤. Additional bands were observed on the ⌬TM-TP␤2X Western blot, possibly because of proteolysis, alternative translation start, or post-translational modifications (38). We then analyzed the different Ba/F3 cell lines in IL3-independent proliferation assays (Fig. 5C). Both ⌬TM-TP␤ and ⌬TM-TP␤2X had a much reduced level of proliferation in the absence of IL3 compared with TP␤. We next tested the signaling properties of these cells and performed intracellular staining for receptor and phospho-STAT5. We calculated the percentage of cells with activated STAT5 within the cell population positive for the receptor and GFP (Fig. 5D). The percentage of phospho-STAT5-positive cells was on average 3.6 Ϯ 1.1 times higher in TP␤ compared with ⌬TM-TP␤ 2X (n ϭ 4). Altogether, these findings indicated that ⌬TM-TP␤ was expressed at a lower level, as a result of increased degradation. However, this may not be the only factor explaining the defect in ⌬TM-TP␤ signaling and proliferation.
Deletion of the TM Domain Increases the in Vitro Kinase Activity of TP␤-The experiments described above showed that TM deletion strongly impaired the ability of TP␤ to activate signaling and transform hematopoietic cells but did not abolish receptor phosphorylation. We noticed that the general tyrosine phosphorylation of ⌬TM-TP␤ was slightly higher than TP␤. This observation prompted us to compare the activity of the kinase domain of the ⌬TM-TP␤ mutant versus TP␤ in vitro. In a first set of experiments, we compared the ability of immunopurified receptors to autophosphorylate in the presence of ATP. Receptor-transfected cells were initially treated with imatinib, which prevented phosphorylation of tyrosine residues. The receptors were then immunoprecipitated, washed, and subjected to in vitro kinase assays in the absence or presence of ATP. As an additional negative control, we used the kinase-dead mutant K654R TP␤. The receptor phosphorylation was then quantified by immunoblotting with anti-phosphotyrosine antibodies. As shown in Fig. 6A, the autophosphorylation of the ⌬TM-TP␤ protein was increased compared with TP␤. Surprisingly, the treatment with imatinib did not completely dephosphorylate ⌬TM-TP␤. This might be due to a difference in the sensitivity of ⌬TM-TP␤ to the drug in comparison with TP␤. In agreement with this hypothesis, when we exposed transduced Ba/F3 cells to increasing doses of imatinib, a higher concentration was required to inhibit the proliferation of cells expressing ⌬TM-TP␤ (Fig. 6C). This difference in imatinib IC 50 might reflect a change in the protein conformation, which could bind imatinib less efficiently (39).
We next assessed the activity of hybrid receptors toward the exogenous substrate MBP in an in vitro kinase assay (Fig. 6B). In this experiment, as in the autophosphorylation assay, the ⌬TM-TP␤ mutant performed better than TP␤. This was particularly clear in the quantitative analysis of the results. In conclusion, the ⌬TM-TP␤ mutant harbored an increased in vitro kinase activity toward itself and toward exogenous substrates, in comparison with TP␤.
⌬TM-TP␤ Can Recruit Signaling Proteins-The strong in vitro catalytic activity of ⌬TM-TP␤ was surprising because the mutant was unable to signal efficiently in cells. This discrepancy could be explained by a defective recruitment of signaling proteins by ⌬TM-TP␤. We could not detect the association with STAT5 with TP␤ by co-immunoprecipitation (data not shown), most likely because phosphorylated STAT5 quickly dissociates from the receptor and migrates to the nucleus. Then we tested the association with another signaling protein, the phosphatidylinositol 3-kinase regulatory subunit p85, in Ba/F3 cell lines. As shown in Fig. 7, endogenous p85 was co-immunoprecipitated with TP␤ as well as with ⌬TM-TP␤. As an additional control, we used cells expressing the FIP1L1-PDGFR␣ hybrid protein, which also interacted with p85. This experiment shows that the ⌬TM-TP␤ mutant has the ability to recruit signaling proteins such as p85 TM Sequence Requirements for Transformation by TP␤-To determine which part of the TM sequence was required for TP␤ activation, we performed sequential deletions and tested the mutants in Ba/F3 cell proliferation assays. As shown in Fig. 8A, deletion of 14 amino acid residues had no impact on cell growth, whereas deletion of 21 amino acids recapitulated the Vector-transduced cells were used as a control. All of the samples were analyzed on the same Western blot, and some lanes were cut out of the final image for clarity. C, the same cell lines as in B were used in a proliferation assay as described for Fig. 1B. All of the cell lines proliferated equally in the presence of IL3 (data not shown). The cells transduced once or twice with TP␤ proliferated at a similar level (data not shown). D, TP␤ and ⌬TM-TP␤2X cells were stained for phospho-STAT5 and PDGFR␤. The dot plots show the percentages of cells expressing GFP and PDGFR␤. The histogram shows the phospho-STAT5 staining in the population positive for GFP and for the PDGFR␤ indicated by rectangles. TP␤ is shown with a bold line, ⌬TM-TP␤ is shown with a thin line, and control is shown with a dashed line. One representative experiment is shown (n ϭ 4). Vector-expressing cells were used as a control. effect of the ⌬TM mutation. These results suggested that a short hydrophobic sequence is enough to preserve TP␤ activity. Deletion of 4 and 7 amino acids consistently resulted in much higher levels of protein expression. Thus, we could not make a conclusion regarding the effect of these deletions on cell proliferation.
The PDGFR␤ TM sequence contains two serine and one threonine residue within the hydrophobic stretch (boxed residues in Fig. 8B). In particular, threonine 545 has been shown to be important for a productive interaction between PDGFR␤ and the E5 protein of bovine papilloma virus in membranes (40), and serine 536 is one of the amino acids lost in ⌬TM21 compared with ⌬TM14. To analyze the role of these three amino acids in the activation of TP␤, we mutated the serine residues into alanine and the threonine into a valine and tested transduced Ba/F3 cell lines for IL3-independent proliferation. As shown in Fig. 8B, the cells expressing the mutants proliferated similarly compared with cells expressing TP␤, suggesting that these polar residues are dispensable for the activation of TP␤.
Interruption of the Juxtamembrane Domain Can Rescue the Activity of ⌬TM-TP␤-Breakpoints falling before the TM domain are generally located in the large intron between exons 10 and 11, which encodes the TM domain, as illustrated for TP␤ in Fig. 9A. This particular position could either reflect the requirement of a TM domain for proper activation of the hybrids or a certain weakness in this chromosomal region. Breakpoints falling within the exon 12, which encodes the juxtamembrane region, have been reported in a few cases such as the PRKG2-PDGFR␤ fusion, in which approximately one-third of the juxtamembrane (JM) domain is deleted (41).
We generated a deletion in TP␤ removing the TM domain and a portion of the juxtamembrane domain (exon12-TP␤; Fig.  9A) and tested transduced Ba/F3 cells for cytokine-indepen-  dent proliferation. We observed that the exon12-TP␤ mutant sustained cell growth to an extent similar to TP␤ (Fig. 9B), although it was expressed at a lower level. The phosphorylation of exon12-TP␤ was stronger than TP␤, similar to what we observed for ⌬TM-TP␤ (Fig. 9B). Altogether our observations suggested that the presence of a TM domain in PDGFR␤ hybrids is not an absolute requirement, provided that the JM domain is also deleted.

DISCUSSION
Here, we show that the transmembrane domain of PDGFR␤ has a crucial role in the transformation of hematopoietic cells by TP␤. This was surprising because such a hydrophobic stretch can destabilize cytosolic proteins. Nevertheless, evidence from a number of receptor studies have shown that TM sequences are important for orientation and stabilization of active dimeric membrane receptors (34,(42)(43)(44).
Deletion of the TP␤ TM sequence did not seem to change the subcellular localization of the protein, because cell surface and intracellular staining indicated that TP␤ and ⌬TM-TP␤ reside in the cytosol, in line with previous reports (33). In addition, the ⌬TM14-TP␤ mutant, which retained only 10 residues of the TM sequence, was still active, although its TM domain is most likely too short to span a lipid bilayer.
The ⌬TM-TP␤ protein was expressed at a lower level, because of an increased degradation, which likely contributes to its lack of transforming activity, in line with published results (22). However, other mechanisms must be involved, because increasing the expression of ⌬TM-TP␤ did not augment cell signaling and proliferation.
The cross-linking of ⌬TM-TP␤ induced a pattern of bands that were similar in size to those observed with TP␤ but was consistently less efficient. This might reflect a decreased oligomerization of ⌬TM-TP␤. However, the co-immunoprecipitation experiment was unaffected by the TM deletion. In addition, ⌬TM-TP␤ phosphorylation and kinase activity were not reduced, as one would expect if oligomerization was impaired. For these reasons, we speculate that decreased cross-linking may reflect an altered disposition of the polypeptides, which exposes fewer residues to the cross-linker, decreasing the efficiency of the cross-linking reaction. This is further supported by recent data showing that the purified TM domain of PDGFR␤ can be cross-linked (34). Thus, it is possible that the TM domains are clustered in the TP␤ oligomer, even though they are not required for the oligomerization process itself, driven by the PNT domain.
The in vitro kinase assays showed that the ⌬TM-TP␤ kinase activity was enhanced, whereas the sensitivity to imatinib was reduced in comparison with TP␤. It is known that imatinib binds to the kinase ATP-binding pocket in its inactive state. The reduced sensitivity of ⌬TM-TP␤ to imatinib suggested that the conformation of its ATP-binding pocket is modified in a way that makes it unable to fit imatinib as efficiently as in TP␤. Noticeably, the fact that the removal of the hydrophobic sequence from TP␤ reduced protein stability could also indicate that the ⌬TM mutation alters TP␤ folding. These data suggest a model in which the TM domain contributes to the activation of TP␤ by imposing to the PDGFR␤ kinase domain a conformation that is optimal for signaling. Further work on the PDGFR structure may indicate whether this is a valid hypothesis.
We observed a reduced activation of STAT5 and ERK1/2 in ⌬TM-TP␤ cells, which contrasted with its enhanced kinase activity and phosphorylation. ⌬TM-TP␤ protein was also able to associate with signaling proteins like p85, which is expected if the hybrid receptor is properly phosphorylated. It was also able to phosphorylate an exogenous substrate such as MBP, at least in vitro, but failed to induce the phosphorylation of STAT5 in cells. One possible explanation for this discrepancy could be that the hyperactivation of the ⌬TM-TP␤ kinase results in the recruitment of a negative regulator, such as a tyrosine phosphatase.
The TM sequence of PDGFR␤ is predicted to adopt an ␣-helical conformation and to determine the relative orientation of the catalytic subunits in the dimeric receptor (34,45,46). The way in which two adjacent kinase domains face each other was shown to be critical for receptor activation in experiments where a dimerization motif was shifted in the PDGFR␤ TM domain and caused periodic activation of the receptor (47). Deletion of two amino acids produces a rotation that abolishes the activity of PDGFR␤ and other receptors such as Neu or the erythropoietin receptor (47,48). In TP␤, the deletion of two amino acids was expected to cause a rotation of two adjacent kinase domains of half a turn, but we did not observe any inactivation of TP␤, which argues against a role for the TM domain in providing proper orientation to the kinase subunits in this case. This might be related to the fact that the pointed domain induces the polymerization of TP␤, whereas the wild-type receptor undergoes dimerization (34,35,49).
The constitutive activation of the kinase domain of receptor tyrosine kinases is generally believed to be sufficient for malignant transformation of cells. However, in the present report we showed that ⌬TM-TP␤ presented a high kinase activity in vitro and is phosphorylated in cells but is unable to activate STAT5 and ERK1/2 and to support cell proliferation. This indicates that the constitutive activation of the kinase domain is not enough to transform cells.
Our results with the exon12-TP␤ mutant showed that a breakpoint in exon 12, with the consequent deletion of part of the inhibitory JM domain, is able to overcome the lack of the TM domain. This goes in line with previous reports describing the disruption of the JM domain as an alternative mechanism of activation for PDGFR␤, independently from ligand-induced dimerization and from fusion with oligomerization domains (50). Thus, the TM domain seemed to be required only in PDGFR␤ hybrids that have an intact JM domain, which represent the majority of the cases described so far.
Most receptor tyrosine kinase fusion products include a dimerization domain in addition to the kinase domain. The present work suggests that the linker sequence between these two domains may also play an important role in the efficient activation of the oncoprotein. This is in line with recent reports pointing to key conformational changes in the region between the ligand-binding domain and the kinase domain of PDGFR␤ and c-KIT upon receptor dimerization (3,4). In conclusion, our work revealed a new role for the PDGFR␤ TM domain in the context of the cytosolic TP␤ protein and possibly of other hybrid oncoproteins derived from PDGFR␤.