PDGFRA Alterations in Cancer: Characterization of a Gain-Of-Function V536E Transmembrane Mutant As Well As Loss-Of-Function and Passenger Mutations

Home , PDGFRA, Receptor tyrosine kinase

ORIGINAL ARTICLE PDGFRA alterations in cancer: characterization of a gain-of-function V536E transmembrane mutant as well as loss-of-function and passenger mutations

AI Velghe1, S Van Cauwenberghe1, AA Polyansky2,3, D Chand4, CP Montano-Almendras1, S Charni1, B Hallberg4, A Essaghir1 and J-B Demoulin1

Activating mutations in the platelet-derived growth factor (PDGF) receptor alpha (PDGFRA) have been described in patients with gastrointestinal stromal tumors or myeloid malignancies associated with hypereosinophilia. These patients respond well to imatinib mesylate, raising the question as to whether patients with a PDGF receptor mutation in other tumor types should receive a tyrosine kinase inhibitor treatment. We characterized 10 novel somatic point mutations in PDGFRA that have been reported in isolated cases of glioblastoma, melanoma, acute myeloid leukemia, peripheral nerve sheath tumors and neuroendocrine carcinoma. The PDGFRA transmembrane domain mutation V536E stimulated Ba/F3 cell growth and signaling via ERK and STAT5 in the absence of ligand. This mutant, identified in glioblastoma, was strongly inhibited by imatinib. Modeling suggested that the mutation modulates the packing of the transmembrane domain helices in the receptor dimer. By contrast, two mutations in highly conserved residues affected the receptor traffic to the cell surface or kinase activity, thereby preventing the response to PDGF. The other mutations had no significant impact on the receptor activity. This functional analysis matched the predictions of SIFT and PolyPhen for only five mutations and these algorithms do not discriminate gain from loss of function. Finally, an E996K variant that had been identified in a melanoma cell line was not expressed in these cells. Altogether, several newly identified PDGFRA mutations do not activate the receptor and may therefore represent passenger mutations. Our results also underline the importance of characterizing novel kinase alterations in cancer patients.

Oncogene (2014) 33, 2568–2576; doi:10.1038/onc.2013.218; published online 10 June 2013 Keywords: PDGF receptor; tyrosine kinase inhibitors; glioblastoma; leukemia; GIST

INTRODUCTION (GIST, 5–10%).3,4 Most mutations are located in the juxtamembrane Platelet-derived growth factor (PDGF) receptors are receptor domain or in the activation loop and are likely to disrupt the tyrosine kinases that are required for the embryo development.1 inhibited conformation of the kinase domain while stabilizing Two genes, PDGFRA and PDGFRB, encode the receptor a and b the active one. Activated fusions genes, such as FIP1L1-PDGFRA, isoforms, which are highly homologous and share a common were described in myeloid malignancies associated with architecture. They consist of five N-terminal immunoglobulin (Ig)-like hypereosinophilia (MMAH).5–7 In most cases, the fusion disrupts domains, which are binding to various isoforms of PDGF, a single the juxtamembrane domain, which is enough to activate the transmembrane domain and a split protein tyrosine kinase domain. receptor. Alternatively, the fusion partner includes an Although the exact structure of these receptors has not been solved oligomerization domain, which activates the kinase activity, such yet, the mechanism of activation could be modeled based on as in most PDGFRB fusions. Recently, somatic PDGFRA point mutation analysis and the partial structure of related receptors such mutations were also found in hypereosinophilia.8 Interestingly, FLT3 and KIT. In the absence of ligand, the juxtamembrane domain some of the reported mutations were not transforming when and the activation loop adopt a conformation that keeps the kinase tested in cell lines. Gain-of-function PDGFRA mutations were also domain inactive. PDGF binding induces the receptor dimerization reported in inflammatory fibroid polyps.9 In addition, PDGFRA is and both inhibitory domains undergo large conformation changes activated by overexpression in glioblastoma of the proneural and become phosphorylated on tyrosines, stabilizing the active subtype.10 An in-frame deletion of 81 amino acids in the Ig-like receptor state. In PDGF receptor b (PDGFRB), the C-terminal tail was domains 4 and 5 was found to activate PDGFRA in glioblastoma also suggested to have an inhibitory role.2 This was not investigated patients with a receptor gene amplification.11,12 in PDGFRA; its C-terminus being the most divergent part compared The PDGF receptor kinase activity is effectively inhibited by with PDGFRB. imatinib mesylate (Glivec), a tyrosine kinase inhibitor developed Activating point mutations in PDGFRA have been described in against BCR-ABL, which turned out to be at least 10 times more a small subset of patients with gastrointestinal stromal tumors active on PDGF receptors.7 This drug induces long-term remissions

1de Duve Institute, Universite´ catholique de Louvain, Brussels, Belgium; 2Department of Structural and Computational Biology, Max F Perutz Laboratories, University of Vienna, Vienna, Austria; 3MM Shemyakin and Yu A Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Moscow, Russia and 4Department of Molecular Biology, Umeå University, Umeå, Sweden. Correspondence: Professor J-B Demoulin, de Duve Institute, Universite´ catholique de Louvain, MEXP B1.74.05, Avenue Hippocrate 75, BE-1200 Brussels, Belgium. E-mail: [email protected] Received 4 October 2012; revised 21 March 2013; accepted 4 April 2013; published online 10 June 2013 Characterization of novel PDGFRA mutations AI Velghe et al 2569 in myeloid malignancies associated with hypereosinophilia in the mutant is not properly trafficking to the cell surface, in agreement presence of a PDGF receptor fusion gene.13,14 GISTs are also with the flow cytometry data (Figure 1b). sensitive to imatinib, although resistant mutations such as D842V have been described.4 New molecules may overcome resistance. V536E constitutively activates the PDGF receptor By contrast, imatinib has failed to improve survival of glioblastoma patients.15,16 We next tested the proliferation of transfected Ba/F3 cells using a The success of imatinib in cancers carrying an activating sensitive tritiated thymidine incorporation test. The V536E mutation PDGFRA variant, that is, GIST and MMAH, raises the question as to in the transmembrane domain was the only one supporting Ba/F3 whether all cancer patients with a PDGF receptor mutation should proliferation in the absence of cytokine, like the D842V positive receive a tyrosine kinase inhibitor treatment. Even though these control (Figure 2a). In the presence of PDGF, the expression of WT patients are rare, next-generation sequencing technologies will or mutant receptors allowed Ba/F3 cell growth to a similar extent, allow the systematic detection of many different mutations in most except W349C, as expected from the lack of surface protein cancers. These new technologies have raised great expectations for expression, and F808L, which did not respond to stimulation. the development of a personalized treatment of each cancer. Similar results were obtained by counting cells in liquid cultures in Ten novel somatic point mutations in PDGFRA have been the absence or presence of PDGF (data not shown). All cell lines reported in a variety of cancers, namely isolated cases of grew similarly in the presence of interleukin-3 (IL3). glioblastoma, melanoma, acute myeloid leukemia (AML), peri- Constitutive signaling by PDGFRA-V536E was confirmed using a pheral nerve sheath tumors and neuroendocrine carcinoma luciferase reporter assay driven by serum response elements in (Table 1).10,17–20 The mutations were found in single cases, MCF7 cells (Figure 2b). In line with the proliferation results, most except W349C and V536E, which were reported in two mutants were comparable to the WT receptor and only weakly patients.10,21 In most cases, a polymorphism was carefully stimulated the reporter activity, except W349C and F808L, which excluded. However, they were not functionally characterized and did not produce a significant reporter signal. Similar results were it is not clear whether these patients could have benefited from obtained using a promoter driven by signal transducer and imatinib therapy. The mutations are located in various domains of activator of transcription (STAT) in PAE cells (Supplementary the receptor, including regions where activating mutations have Figure 1). The ability of the mutant receptors to respond to PDGF not been reported yet, such as the transmembrane domain and was also tested by co-transfecting a PDGF-B cDNA, which the C-terminal tail. The goal of the present study was to confirmed the inability of the two mutants mentioned above to characterize them. respond to PDGF. To further demonstrate that V536E is a transforming mutation, we tested its ability to mediate the formation of transformed fibroblasts foci. Low-passage NIH3T3 cells were transfected with RESULTS the V536E mutant compared with activated D842V, WT and W349C prevents PDGFRA glycosylation and traffic to the cell W349C receptors. In line with the Ba/F3 experiment, PDGFRA- surface V536E induced the formation of foci, by contrast to WT and Mutants were stably transfected in Ba/F3 cells, which proliferate in W349C receptors (Figure 3a). the absence of cytokine when they express a constitutively We further studied the V536E mutant signaling in Ba/F3 cells by activated PDGF receptor.4,22 As a positive control, we also analyzing STAT5, extracellular signal regulated kinase (ERK)-1/2 expressed PDGFRA-D842V, a constitutively activated mutant and AKT phosphorylation by western blot. The expression of found in GISTs.4 The mutant Ba/F3 lines were sorted to remove V536E or D842V induced the phosphorylation of ERK and STAT5 the fraction of cells that did not express PDGFRA. Their cell surface but not AKT (Figure 3b). By contrast, cell stimulation with PDGF expression after sorting is shown in Figure 1a. W349C was not induced the phosphorylation of ERK and AKT but not STAT5, detectable at the cell surface and could not be sorted. Next, the suggesting that signal transduction by V536E differed from the WT receptor expression level was analyzed by western blot receptor. The phosphorylation of the receptor itself was then (Figure 1b). All mutants were expressed at a level similar or analyzed in transfected HEK293T cells (Figure 3c). V536E higher than D842V. Two receptor bands are detected by western phosphorylation on tyrosines was increased compared with the blots in transfected Ba/F3 cell extracts and represent different WT receptor, confirming that this mutation constitutively activates N-glycosylation levels, as described.23 The heavier form corresponds the receptor. to the fully glycosylated receptor, which is matured in the Golgi PDGFRA-D842V is resistant to imatinib, whereas juxtamembrane and expressed at the cell surface. Unlike wild-type (WT) PDGFRA mutants are highly sensitive to this drug. We have also shown and the other mutants, W349C was detectable as a single band previously that a transmembrane deletion shifts the sensitivity of corresponding to immature glycosylation, indicating that this the ETV6-PDGFRB fusion to imatinib.22 The ability of imatinib to

Table 1. PDGFRA mutations characterized in the present study

Mutation Domain Cancer type Impact Reference

P130S Ig-like 2 Malignant peripheral nerve sheath tumor — 17 W349C Ig-like 4 Glioblastoma Intracellular 10 V469A Ig-like 5 Malignant peripheral nerve sheath tumor — 17 V536E Transmembrane Glioblastoma Gain of function 10 F808L Kinase Acute myeloid leukemia M1 Kinase dead 18 D842V Activation loop Gastrointestinal stromal tumor Gain of function N870S Kinase Acute myeloid leukemia M1 — 18 G829R Kinase Glioblastoma — 19 E996K C-terminal Metastasized melanoma cell line Not expressed in the cell line 19 D1071N C-terminal Neuro-endocrine carcinoma of the lung — 19 S1049CH* C-terminal Glioblastoma — 20 Abbreviations: Ig, immunoglobulin; *, stop codon.

& 2014 Macmillan Publishers Limited Oncogene (2014) 2568 – 2576 Characterization of novel PDGFRA mutations AI Velghe et al 2570

Figure 1. W349C perturbs PDGFRA traffic to the cell surface. Ba/F3 cells were stably transfected with PDGFRA mutants and sorted. Receptor expression was tested by flow cytometry on intact cells (a) or by western blot (b). The average fluorescence intensity compared with the wild-type (WT) receptor (set at 100%) is shown (from a minimum of two experiments).

Figure 2. PDGFRA V536E is constitutively activated, whereas W349C and F808L do not respond to PDGF. (a) Transfected Ba/F3 cells were extensively washed and cultured in the absence of cytokine or in the presence of PDGF for 24 h. Thymidine incorporation was quantiﬁed to evaluate cell proliferation. One representative experiment out of three is shown. All cell lines proliferated similarly in the presence of interleukin-3 (data not shown). (b) MCF7 cells were transiently transfected with the indicated receptor and a luciferase construct driven by a serum-responsive element. The average of three experiments is shown with s.e.m. V536E signiﬁcantly induced the luciferase activity in the absence of cytokine compared with wild type (WT; Po0.05, Student’s t-test).

block proliferation and signaling induced by V536E was analyzed Impact of V536E on PDGFRA dimerization and structure of the in Ba/F3 and MCF7 cells. Figure 4 shows that imatinib completely transmembrane domain inhibited Ba/F3-V536E cell growth with an IC50 similar to the one Mutations similar to V536E in the transmembrane helix of other reported for FIP1L1-PDGFRA and much lower than the concentra- receptors were suggested to promote dimerization.24 Therefore, tion required to block BCR-ABL.7 The activation of serum- we tested the impact of the V536E mutation on PDGFRA responsive elements by V536E in luciferase assays was also dimerization. Receptor dimers were chemically crosslinked dampened by imatinib. These experiments demonstrate that before cell lysis and western blot analysis (Figure 5a). This V536E is highly sensitive to imatinib. experiment suggested that a small fraction of the receptor could

Oncogene (2014) 2568 – 2576 & 2014 Macmillan Publishers Limited Characterization of novel PDGFRA mutations AI Velghe et al 2571

Figure 3. Cell transformation and constitutive signaling by PDGFRA V536E. (a) Low-passage NIH3T3 cells were transfected with the indicated PDGFRA receptor or empty vector. After 3 weeks, transformed ﬁbroblasts foci were stained with crystal violet, photographed (upper panel) and scored (lower panel, average of three independant experiments performed in triplicate cultures, arbitrary units). (b)Ba/F3 cells transfected PDGFRA wild type (WT), V536E or D842V were stimulated with PDGF or IL3 for 15 min or left untreated. Cells were lysed and protein extracts were analyzed by western blot (WB) with antibodies recognizing active phosphorylated STAT5, AKT or ERK. Antibodies against total STAT5, AKT and ERK2 were used as controls. (c) HEK293T cells transiently transfected with the indicated receptor were lysed and analyzed by western blot with anti-phospho-tyrosine or with anti-PDGFRA antibodies (left panels). Alternatively, PDGFRA was immunoprecipitated (IP) from lysates before WB (right panels).

be crosslinked in the absence of ligand. Treatment with PDGF then increased the dimer formation, as expected from published results.25 Importantly, no major difference was observed between WT and mutant receptors. To understand better the structural origin of the receptor constitutive activation induced by the V536E mutation, we performed molecular modeling of possible conformations of PDGFRA transmembrane helix dimer using PREDDIMER approach.26 Transmembrane helices of PDGFRA were predicted to form several distinctive top-scoring dimer conformations similar to those of PDGFRB (Figure 5b and Supplementary Figure 2). We found that the substitution of respective valine for glutamic acid is likely to change the packing efficiency of the predicted dimer conformations and may induce a repopulation of the dimer conformational ensemble. Taking into account an observed gap between packing efficiencies of the transmembrane dimer conformations (calculated according to the scoring function) in mutant V536E receptor, the left-handed conformation is likely to be predominant, whereas the right-handed one is predicted to be less populated (Figure 5c). In contrast, we observed the reverse situation for WT receptor. In addition, the packing efficiency of both WT conformations are relatively more similar to each other. Finally, the mutation has no effect on the packing efficiency of the parallel transmembrane dimer conformation (Figure 5c). Previously, nuclear magnetic resonance-derived left-handed conformation of PDGFRB transmembrane dimer (Figure 5b) was proposed to correspond to activated state of the receptor.27 Figure 4. V536E is sensitive to inhibition by imatinib. (a) Ba/F3 cells Taking together this assumption with results of our modeling, we expressing the indicated receptor were treated with imatinib for speculate that the V536E mutation might favor PDGFRA 24 h. Thymidine incorporation was quantified to evaluate cell proliferation. (b) MCF7 cells were transiently transfected with the transmembrane dimer in the activated state and obstruct the indicated receptor, a luciferase reporter construct controlled by conformational transition of the receptor to the inactive one. serum-responsive elements and a control b-galactosidase reporter. Similar situation has been recently observed by modeling for Cells were treated with imatinib for 24 h. The luciferase activity was ErbB2-Neu* V664Q transmembrane mutant,26 where the measured and divided by the b-galactosidase activity. analogues mutation increased the dimer packing efficiency

& 2014 Macmillan Publishers Limited Oncogene (2014) 2568 – 2576 Characterization of novel PDGFRA mutations AI Velghe et al 2572

Figure 5. V536E mutation does not affect receptor dimerization but modulates packing of transmembrane domains. (a) Ba/F3 cells expressing the indicated receptor were incubated with PDGF-BB for 1 h on ice. Cells were washed and treated with bis(sulfosuccinimidyl)-suberate (BS3) to crosslink membrane protein complexes. PDGF receptors were isolated from cell protein extracts using wheat germ agglutinin-coupled beads. Western blot was performed with anti-PDGFRA antibodies. (b) Top-scoring predicted conformation of the mutant PDGFRA transmembrane dimer (magenta) superimposed with left-handed nuclear magnetic resonance (NMR) conformation (green) of PDGFRB dimer. Crossing angles of dimer conformations and backbone root-mean-square deviations (RMSDs) between them are given. Peptides are shown in C cartoon-and-stick representation. (c) Normalized packing score (FSCOR) of the wild type (WT) and the mutant PDGFRA transmembrane dimers plotted as functions of the dimer crossing angle (w). For each dimer, three top-scoring conformations (numbered ‘1’, ‘2’ and ‘3’) are shown. See Supplementary Figure 2 for conformation details. Left- and right-handed dimer conformations are shown in cartoon-and-stick representation.

relative to the WT and favored the left-handed over the PDGFRA protein or transcript, compared with positive controls right-handed dimer conformations. (Figure 7). In conclusion, the E996K mutation, which did not confer any growth advantage in our assays, was not expressed in The F808L mutant is devoid of kinase activity LB373-MEL cells. Proliferation and luciferase assays suggested that the F808L mutant was unable to respond to PDGF stimulation, even though DISCUSSION it was expressed at the cell surface. To test the kinase activity of Among 10 uncharacterized mutations, V536E was the only this mutant, we first overexpressed it in HEK293T cells. Figure 6 gain-of-function variant that we identified. To our knowledge, shows that F808L was not significantly phosphorylated on this is the first cancer mutation described in the PDGF receptors tyrosines by contrast to WT and D842V receptors. Then, we transmembrane domain. A similar V664E mutation in ErbB2/Neu performed an in vitro kinase assay after receptor immunopreci- was shown to induce the receptor constitutive tyrosine kinase pitation. The phosphorylation of the WT and D842V receptors was activity.24 Activating point mutations in the transmembrane enhanced in the presence of ATP, but F808L remains unpho- segment have also been found in other receptor tyrosine sphorylated, indicating it is a kinase-dead mutant (Figure 6). 3 kinases, such as c-KIT and FGFR3. The insertion of two correctly positioned glutamic residues in a simplified valine-based The E996K mutant is not expressed in the cell line in which it was transmembrane domain is enough to activate PDGFRB and identified ErbB2/Neu.28 In addition, the importance of the transmembrane The E996K mutant was identified in the LB373-MEL cell line, domain for PDGFRB activation was underlined by studies showing derived from a melanoma metastasis.19 The mutation was absent that this domain is essential for transformation by the ETV6- in immortalized lymphocytes from the same patient confirming PDGFRB fusion and by the bovine papillomavirus E5 protein.22 It that it was an acquired cancer variant. We confirmed the presence was recently shown that the PDGFRB transmembrane domain is of a heterozygous somatic mutation in the genomic DNA of able to form a stable left-handed helical dimer in artificial LB373-MEL (data not shown), but we were unable to detect the membranes.27 In this structure, V543, which is homologous to

Oncogene (2014) 2568 – 2576 & 2014 Macmillan Publishers Limited Characterization of novel PDGFRA mutations AI Velghe et al 2573 Further experiments are required to decipher the exact mechanism of constitutive PDGF receptor activation by this mutation. Based on initial reports, PDGFRA mutation screenings in GIST usually focus on exons 12 and 18, which encode the juxtamembrane domain and the activation loop.4 We suggest that sequencing exon 11, encoding the transmembrane domain, may reveal additional clinically relevant PDGFRA mutations in GIST. Two mutants, F808L and W349C, did not show any sign of activation and did not respond to PDGF stimulation for different reasons. Interestingly, these two residues were perfectly conserved in vertebrate PDGFRA as well as in the related human receptors PDGFRB, c-KIT, FLT3, CSF1R and VEGFRs (Supplementary Table I). In the c-KIT structure, W349 mediates key interactions with distant amino acids.31 This mutation may therefore interfere with PDGFRA folding. The W349C protein appeared as a single immature band and failed to traffic to the plasma membrane. Although a cysteine at position 349 may induce the formation of a Figure 6. F808L is a kinase dead mutant. (a) HEK293T cells were new aberrant cysteine bridge, we failed to detect covalent W349C transiently transfected with the indicated receptor. Four hours dimeric complex. W349 is also located close to two predicted before lysis, cells were treated with imatinib 0.5 mM. PDGFRA was N-glycosylation sites (N353LT and N359LT). The mutation may immunoprecipitated (IP) from lysates and analyzed by western blot (WB) with anti-phospho-tyrosine or anti-PDGFRA antibodies. thus block the glycosylation of these sites or prevent the traffic to (b) Alternatively, immunoprecipitates were incubated with ATP the Golgi where the receptor undergoes its final glycosylation (50 mM) in kinase buffer for 10 min. Receptor phosphorylation was steps. In agreement with its absence at the cell surface, W349C did analyzed by WB and quantified using the Odyssey technology. The not respond to the addition of exogenous PDGF. In addition, ratio between the receptor phosphorylation and the total receptor co-transfection of a PDGF-B cDNA did not activate the mutant expression was calculated. It is normalized to the wild-type (WT) receptor either. This may result from the disruption of the Ig-like levels (average of three experiments). domain 4, which is essential for PDGFRA signal transduction.32 The second defective mutation, F808L, is located nearby the activation loop and disrupted the kinase activity. A similar kinase dead mutation has been described in the receptor tyrosine kinase ALK in neuroblastoma.33 Even if F808L was supporting leukemia development by an unknown mechanism independent of the receptor kinase activity, leukemia patients carrying this mutation are unlikely to benefit from imatinib therapy. Altogether, these two variants likely represent passenger mutations. Nevertheless, we cannot rule out the possibility that they provide an advantage to tumor cells in a particular molecular environment. Seven mutations produced no particular phenotype and may also be classified as passenger mutations. These residues are conserved in mammalian PDGFRA except V469, which is replaced by an isoleucine in some species (Supplementary Table I). Interestingly, this valine is substituted by an alanine in Tetraodon nigroviridis, mimicking the V469A mutation found in a malignant Figure 7. E996K is not expressed in the cell line in which it was peripheral nerve sheath tumor. These seven amino acids are identified. (a) LB373-MEL cells were stained with anti-PDGFRA poorly conserved in receptors related to PDGFRA. Remarkably, antibodies and phycoerythrin-linked anti-mouse-IgG. Cells were analyzed by flow cytometry. 32D cells transfected with PDGFRA N870 is replaced by a serine in PDGFRB, supporting our conclusion were used as a positive control. (b) RNA was extracted from LB373- that N870S does not affect PDGF receptors significantly. The MEL cells and analyzed by reverse transcription–PCR using primers original report did not formally exclude that this allele corre- specific for PDGFRA or RPLP0 (as a control). HepG2 cells, which do sponds to a rare polymorphism rather than a somatic mutation.18 not express PDGFRA, and AG01518 human fibroblasts were used as In the case of E996K, we had access to the original cell line in negative and positive controls, respectively. which this somatic mutation had been identified,19 but which did not express any detectable PDGFRA protein or transcript. This observation supports the hypothesis that E996K is a passenger PDGFRA V536, is located at the edge of the interface (Figure 5a). mutation, although we cannot rule out the possibility that the The PDGFRA-V539E mutant is likely to populate this dimeric receptor was expressed in the original tumor cells and was lost conformation by increasing the packing efficiency of the during the adaptation to cell culture. transmembrane helix dimer, as suggested by modeling. A similar Three mutations are located in the C-terminal tail, which is situation was predicted for the rat ErbB2-Neu* V664Q transmem- unique to PDGFRA, except the last five amino acids, which are also brane mutant.26 Recently, active and inactive dimer conformations present in PDGFRB.34 This C-terminal motif, which is lost in the of the EGF receptor transmembrane domain have been reported, S1049CH* frame shift mutant,20 was suggested to recruit PTEN via supporting the hypothesis that the conformation of this domain NHERF adaptor proteins, thereby limiting the activation of the has a critical role in receptor tyrosine kinase activation.29,30 Our PI3K pathway.35 However, we did not observe any significant model also predicts that the orientation and the angle between increase in PKB phosphorylation downstream this mutant (data the two helices in the left-handed dimer is affected by the V536E not shown). The role of the C-terminal domain of PDGFRA remains mutation, which may optimally positions the two kinase domains to be established. facing each other in the receptor dimer. In conclusion, we suggest The occurrence of passenger mutations in cancer genome was that the V536E substitution affects the conformation of the highlighted by high throughput sequencing projects.19,36 transmembrane domain dimer rather than dimerization itself. Mutation recurrence in multiple patients has been widely used

& 2014 Macmillan Publishers Limited Oncogene (2014) 2568 – 2576 Characterization of novel PDGFRA mutations AI Velghe et al 2574 as a criterion to select driver cancer mutations. In this respect, presence of G418 (3 mg/ml). Transfected cell populations were sorted by after the initial submission of the present manuscript, the V536E flow cytometry after staining with a monoclonal anti-PDGFRA antibody mutation was described in a second glioblastoma patient.21 (R&D systems, Minneapolis, MN, USA, MAB1264) and anti-mouse However, W349C was also reported in two patients.10 The immunoglobulin labeled with phycoerythrin (Jackson Laboratories, Bar recurrence criterion is also too restrictive because unique cancer Harbor, ME, USA). A few days later, homogeneous PDGFRA expression was confirmed using the same antibodies by flow cytometry. For proliferation variants may be clinically relevant. For instance, several patients assays, cells were washed three times to remove IL3 and seeded in 96-well with non-recurrent PDGFR fusions genes were successfully treated 4 3 plates (10 cells/well in 200 ml of medium with 10% serum) without with imatinib. To predict the impact of single-nucleotide variant cytokine, with PDGF-BB (20 ng/ml, Peprotec, London, UK) or IL3 (200 U/ml). on protein function, several algorithms based on sequence After 20 h, [3H]-thymidine (0.5 mCi, GE Healthcare, Diegem, Belgium) was homologies and structural data have been developed, such as added to each well for 4 h. Cells were harvested and the radioactivity SIFT and PolyPhen.37 For five out of eight mutations present in the incorporated into DNA was counted using a TopCount instrument (Perkin database, PolyPhen matched our results (Supplementary Table I). Elmer, Waltham, MA, USA). In this set of mutations, the conservation with receptors related to LB373-MEL cells were kindly provided by Francis Brasseur (Ludwig PDGFRA (FLT3, KIT, CSF1R and VEGFRs) was best correlated with Institute for Cancer Research, Brussels, Belgium) and were cultured in Iscove-Dulbecco medium with fetal calf serum (10%) and essential amino the observed functional impact. These tests do not predict acids. Cells were analyzed by flow cytometry as described above. mRNA whether a given variant leads to a gain of function or a loss of was extracted using a Qiagen kit and PDGFRA expression was measured by function, a piece of information that is essential in oncology. quantitative reverse transcription–PCR as described41 using AG01518 Therefore, a thorough experimental analysis of cancer mutations human fibroblasts and HEPG2 hepatocarcinoma cells as positive and remains essential. negative controls, respectively.42 The expression of the house keeping After the initial submission of this paper, three novel PDGFRA gene RPLP0 was used as control. The sequence of the 996 codon was mutations—among many others—were reported in diffuse large confirmed by sequencing of multiple clones obtained after PCR 38 amplification of genomic DNA using the oligonucleotides 50-CGTGT B-cell lymphoma. Based on the criteria mentioned above, they 0 0 0 may also represent passenger mutations (Supplementary Table I). GGACTCAGACAATGCA-3 and 5 -GGCAGAGGAATGATGTAGCCA-3 in PDGFRA exon 22. Our results confirm the occurrence of activating mutations in glioblastoma, in which two activated PDGFRA variants have been previously described.11,12 Whether these patients could be more sensitive to kinase inhibitors remains to be studied. By contrast, Transformation assay our observations do not support the existence of driver PDGFRA Low-passage NIH3T3 cells (American Type Culture Collection, Manassas, 33,43 4 point mutations in acute myeloid leukemia or malignant VA, USA) were transfected as described. Briefly, 4.5 Â 10 cells were peripheral nerve sheath tumors. In the latter tumors, other seeded the day before into collagen-coated 12-well plates and transfected alterations, namely PDGFRA gene amplification or autocrine for 6 h with 0.5 mg of DNA (pEF-MYC-CYTO vector containing PDGFRA WT, D842V, V536E or W349C) and 1.4 ml of Lipofectamine 2000 in 0.3 ml of Opti- loops, may activate WT and mutant receptors.17 Such alterations 10 MEM (both from Invitrogen). Twenty-four hours after transfection, three- have also been described in glioblastomas. The presence of both fifth of the cells from each well were transferred to wells in collagen-coated activating and passenger alterations in glioblastoma, which was six-well plates and kept in Dulbecco’s modified Eagle’s medium (10% fetal also recently reported in idiopathic hypereosinophilia, may reflect calf serum and 0.5 mg/ml G418) until the cells reached confluence. a particularly high PDGFRA mutation rate in these cancers. In line Thereafter, cells were kept in Dulbecco’s modified Eagle’s medium with this hypothesis, the proneural subtype of glioblastoma, in (5% fetal calf serum and 0.25 mg/ml G418) for approximately 10 days. which PDGFRA is expressed, is associated with a higher frequency Three weeks after transfection, the cells were fixed using methanol and of the hypermutator phenotype.10 plates were stained with crystal violet. Altogether, we identified a number of key residues for PDGFRA function and characterized a novel type of imatinib-sensitive PDGFRA-activating mutation in the transmembrane segment. Our Western blot, immunoprecipitation and crosslinking work underlines the importance of characterizing PDGF receptor These experiments were performed essentially as described.22,44 Briefly, mutations in a systematic manner. This is a major issue in the Ba/F3 cells expressing PDGFRA were washed and incubated in the absence development of personalized targeted therapies. of IL3 for 4 h. In signaling studies, cells were restimulated for 15 min with PDGF-BB (25 ng/ml) or IL3 (200 U/ml) at 37 1C, washed and lysed in SDS–polyacrylamide gel electrophoresis sample buffer. For crosslinking studies, 2 Â 106 cells were stimulated with PDGF-BB (100 ng/ml) for 1 h on MATERIALS AND METHODS ice to prevent endocytosis. Cells were washed with phosphate-buffered Reagents and plasmids saline and treated with bis(sulfosuccinimidyl)-suberate (Pierce/Thermo- The human PDGFRA cDNA was subcloned in pEF-MYC-CYTO (Invitrogen/ scientific, Rockford, IL, USA, 1 mM) for 1 h on ice in 200 ml phosphate- Life Technologies, Gent, Belgium), which allowed a high expression level in buffered saline. Glycine (3.5 mM, final concentration) was then added to Ba/F3, compensating the fast turn-over of the receptor in these cells.39 A stop the reaction before protein extraction in lysis buffer. PDGF receptors Flag tag was inserted after the signal peptide. Mutations were introduced were isolated with wheat germ agglutinin-coupled agarose beads (Sigma- by Quickchange (Stratagene/Agilent technologies, Diegem, Belgium) and Aldrich, St Louis, MO, USA). Beads were extensively washed and proteins confirmed by sequencing (Macrogen, Seoul, Korea). Western blots were were eluted with sample buffer for western blot analysis with anti-PDGFRA performed as described34 with anti-PDGFRA (1:200, Santa Cruz, Santa Cruz, antibodies. CA, USA, #951), PY99 (1:200, Santa Cruz), anti-phospho-Y694-STAT5 (1:1000, Signalway Antibodies, College Park, MD, USA), anti-STAT5, anti- phospho-AKT, anti-AKT, anti-phospho-ERK-1/2 (1:1000, Cell Signaling Technologies, Danvers, MA, USA) and anti-ERK2 (1:1000, EET rabbit Luciferase assays antiserum34). Imatinib mesylate was purchased from LC Laboratories MCF7 and PAE cells were plated in six-well plates, respectively, at 300 000 (Woburn, MA, USA). and 150 000 cells/ well. After one day, cells were transfected with a luciferase construct controlled by serum response elements (0.5 mg) or STAT-response elements (GRR5, 0.5 mg),44 pDRIVEchEF1-RU5 Cell culture and transfections (b-galactosidase reporter, Invitrogen, 0.5 mg), pEF-MYC-CYTO-PDGFRA HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium with (0.25 mg) and PDGF-B (0.25 mg)2 in the presence of Fugene HD (for MCF7, fetal calf serum (10%) and antibiotics (Invitrogen). Cells were transfected Promega) or Lipofectamine 2000 (for PAE, Invitrogen). After 24 h, cells were using the calcium phosphate method as described.39 Ba/F3 cells were lysed and luciferase activity was measured recommended by the cultured in the same medium supplemented with IL3 (200 U/ml). Cells manufacturer (Promega). b-Galactosidase levels were used to normalize were electroporated as described elsewhere40 and selected in the the results as described.42

Oncogene (2014) 2568 – 2576 & 2014 Macmillan Publishers Limited Characterization of novel PDGFRA mutations AI Velghe et al 2575 Modeling methods therapy in patients with progressive pretreated glioblastoma resistant to standard Spatial structures of the dimers formed by transmembrane helical dose temozolomide. J Neurooncol 2010; 96: 393–402. fragments of the WT PDGFRA and its V536E mutant were built using 17 Holtkamp N, Okuducu AF, Mucha J, Afanasieva A, Hartmann C, Atallah I et al. PREDDIMER algorithm starting from their sequences (see Figure 5b) as Mutation and expression of PDGFRA and KIT in malignant peripheral nerve described elsewhere.26 For calculations of backbone root-mean-square sheath tumors, and its implications for imatinib sensitivity. Carcinogenesis 2006; deviations between the predicted models and PDGFRB transmembrane 27: 664–671. dimer, its nuclear magnetic resonance structure was used (2L6W27). 18 Hiwatari M, Taki T, Tsuchida M, Hanada R, Hongo T, Sako M et al. Novel missense Packing efficiencies of the predicted dimers were estimated according to mutations in the tyrosine kinase domain of the platelet-derived growth factor C values of PREDDIMER scoring function (FSCOR). Normalized values of FSCOR receptor alpha(PDGFRA) gene in childhood acute myeloid leukemia with for a combined set of three top ranked of PDGFRA-WT and PDGFRA-V536E t(8;21)(q22;q22) or inv(16)(p13q22). Leukemia 2005; 19: 476–477. C dimer conformations were calculated as: FSCOR ¼ (FSCOR À /FSCORS)/ 19 Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G et al. std.FSCOR, where /FSCORS and std.FSCOR are the mean and standard Patterns of somatic mutation in human cancer genomes. Nature 2007; 446: deviation within the set. 153–158. 20 Rand V, Huang J, Stockwell T, Ferriera S, Buzko O, Levy S et al. Sequence survey of receptor tyrosine kinases reveals mutations in glioblastomas. Proc Natl Acad Sci CONFLICT OF INTEREST USA 2005; 102: 14344–14349. The authors declare no conflict of interest. 21 Alentorn A, Marie Y, Carpentier C, Boisselier B, Giry M, Labussiere M et al. Prevalence, clinico-pathological value, and co-occurrence of PDGFRA abnormalities in diffuse gliomas. Neuro Oncol 2012; 14: 1393–1403. ACKNOWLEDGEMENTS 22 Toffalini F, Hellberg C, Demoulin JB. Critical role of the platelet-derived growth factor receptor (PDGFR)-beta transmembrane domain in the TEL-PDGFRbeta We thank Francis Brasseur (Ludwig Institute for Cancer Research, Brussels, Belgium) cytosolic oncoprotein. J Biol Chem 2010; 285: 12268–12278. for the LB373-MEL cell line and Stefan Constantinescu (de Duve Institute) for advices. 23 Keating MT, Harryman CC, Williams LT. Platelet-derived growth factor receptor This work was supported by the Salus Sanguinis Foundation and a grant from ‘Action inducibility is acquired immediately after translation and does not require de Recherches Concerteés’ (Communaute´ Française de Belgique, Belgium). glycosylation. J Biol Chem 1989; 264: 9129–9132. 24 Burke CL, Lemmon MA, Coren BA, Engelman DM, Stern DF. Dimerization of the p185neu transmembrane domain is necessary but not sufficient for transforma- REFERENCES tion. Oncogene 1997; 14: 687–696. 1 Andrae J, Gallini R, Betsholtz C. Role of platelet-derived growth factors in 25 Yang Y, Yuzawa S, Schlessinger J. Contacts between membrane proximal regions physiology and medicine. Genes Dev 2008; 22: 1276–1312. of the PDGF receptor ectodomain are required for receptor activation but not for 2 Chiara F, Bishayee S, Heldin CH, Demoulin JB. Autoinhibition of the platelet- receptor dimerization. Proc Natl Acad Sci USA 2008; 105: 7681–7686. derived growth factor beta-receptor tyrosine kinase by its C-terminal tail. J Biol 26 Polyansky AA, Volynsky PE, Efremov RG. Multistate organization of transmem- Chem 2004; 279: 19732–19738. brane helical protein dimers governed by the host membrane. JAmChemSoc 3 Toffalini F, Demoulin JB. New insights into the mechanisms of hematopoietic cell 2012; 134: 14390–14400. transformation by activated receptor tyrosine kinases. Blood 2010; 116: 27 Muhle-Goll C, Hoffmann S, Afonin S, Grage SL, Polyansky AA, Windisch D et al. 2429–2437. Hydrophobic matching controls the tilt and stability of the dimeric platelet- 4 Corless CL, Schroeder A, Griffith D, Town A, McGreevey L, Harrell P et al. PDGFRA derived growth factor receptor (PDGFR) beta transmembrane segment. J Biol mutations in gastrointestinal stromal tumors: frequency, spectrum and in vitro Chem 2012; 287: 26178–26186. sensitivity to imatinib. J Clin Oncol 2005; 23: 5357–5364. 28 Bell CA, Tynan JA, Hart KC, Meyer AN, Robertson SC, Donoghue DJ. Rotational 5 Demoulin JB, Montano-Almendras CP. Platelet-derived growth factors and their coupling of the transmembrane and kinase domains of the Neu receptor tyrosine receptors in normal and malignant hematopoiesis. Am J Blood Res 2012; 2: 44–56. kinase. Mol Biol Cell 2000; 11: 3589–3599. 6 Medves S, Demoulin JB. Tyrosine kinase gene fusions in cancer: translating 29 Arkhipov A, Shan Y, Das R, Endres NF, Eastwood MP, Wemmer DE et al. mechanisms into targeted therapies. J Cell Mol Med 2012; 16: 237–248. Architecture and membrane interactions of the EGF receptor. Cell 2013; 152: 7 Cools J, DeAngelo DJ, Gotlib J, Stover EH, Legare RD, Cortes J et al. A tyrosine 557–569. kinase created by fusion of the PDGFRA and FIP1L1 genes as a therapeutic target 30 Endres NF, Das R, Smith AW, Arkhipov A, Kovacs E, Huang Y et al. Conformational of imatinib in idiopathic hypereosinophilic syndrome. N Engl J Med 2003; 348: coupling across the plasma membrane in activation of the EGF receptor. Cell 1201–1214. 2013; 152: 543–556. 8 Elling C, Erben P, Walz C, Frickenhaus M, Schemionek M, Stehling M et al. Novel 31 Yuzawa S, Opatowsky Y, Zhang Z, Mandiyan V, Lax I, Schlessinger J. Structural imatinib-sensitive PDGFRA-activating point mutations in hypereosinophilic basis for activation of the receptor tyrosine kinase KIT by stem cell factor. Cell syndrome induce growth factor independence and leukemia-like disease. Blood 2007; 130: 323–334. 2011; 117: 2935–2943. 32 Omura T, Heldin CH, Ostman A. Immunoglobulin-like domain 4-mediated 9 Huss S, Wardelmann E, Goltz D, Binot E, Hartmann W, Merkelbach-Bruse S et al. receptor-receptor interactions contribute to platelet-derived growth factor- Activating PDGFRA mutations in inflammatory fibroid polyps occur in exons 12, induced receptor dimerization. J Biol Chem 1997; 272: 12676–12682. 14 and 18 and are associated with tumour localization. Histopathology 2012; 61: 33 Schonherr C, Ruuth K, Eriksson T, Yamazaki Y, Ottmann C, Combaret V et al. The 59–68. neuroblastoma ALK(I1250T) mutation is a kinase-dead RTK in vitro and in vivo. 10 Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD et al. Integrated Transl Oncol 2011; 4: 258–265. genomic analysis identifies clinically relevant subtypes of glioblastoma 34 Demoulin JB, Seo JK, Ekman S, Grapengiesser E, Hellman U, Ronnstrand L et al. characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 2010; Ligand-induced recruitment of Na þ /H þ -exchanger regulatory factor to the 17: 98–110. PDGF (platelet-derived growth factor) receptor regulates actin cytoskeleton 11 Clarke ID, Dirks PB. A human brain tumor-derived PDGFR-alpha deletion mutant is reorganization by PDGF. Biochem J 2003; 376: 505–510. transforming. Oncogene 2003; 22: 722–733. 35 Takahashi Y, Morales FC, Kreimann EL, Georgescu MM. PTEN tumor suppressor 12 Ozawa T, Brennan CW, Wang L, Squatrito M, Sasayama T, Nakada M et al. PDGFRA associates with NHERF proteins to attenuate PDGF receptor signaling. EMBO gene rearrangements are frequent genetic events in PDGFRA-amplified J 2006; 25: 910–920. glioblastomas. Genes Dev 2010; 24: 2205–2218. 36 Frohling S, Scholl C, Levine RL, Loriaux M, Boggon TJ, Bernard OA et al. Identifi- 13 David M, Cross NC, Burgstaller S, Chase A, Curtis C, Dang R et al. Durable cation of driver and passenger mutations of FLT3 by high-throughput DNA responses to imatinib in patients with PDGFRB fusion gene-positive and BCR-ABL- sequence analysis and functional assessment of candidate alleles. Cancer Cell negative chronic myeloproliferative disorders. Blood 2007; 109: 61–64. 2007; 12: 501–513. 14 Metzgeroth G, Walz C, Erben P, Popp H, Schmitt-Graeff A, Haferlach C et al. Safety 37 Ramensky V, Bork P, Sunyaev S. Human non-synonymous SNPs: server and survey. and efficacy of imatinib in chronic eosinophilic leukaemia and hypereosinophilic Nucleic Acids Res 2002; 30: 3894–3900. syndrome: a phase-II study. Br J Haematol 2008; 143: 707–715. 38 Zhang J, Grubor V, Love CL, Banerjee A, Richards KL, Mieczkowski PA et al. Genetic 15 Paulsson J, Lindh MB, Jarvius M, Puputti M, Nister M, Nupponen NN et al. Prog- heterogeneity of diffuse large B-cell lymphoma. Proc Natl Acad Sci USA 2013; 110: nostic but not predictive role of platelet-derived growth factor receptors in 1398–1403. patients with recurrent glioblastoma. Int J Cancer 2011; 128: 1981–1988. 39 Toffalini F, Kallin A, Vandenberghe P, Pierre P, Michaux L, Cools J et al. The fusion 16 Dresemann G, Weller M, Rosenthal MA, Wedding U, Wagner W, Engel E et al. proteins TEL-PDGFRbeta and FIP1L1-PDGFRalpha escape ubiquitination and Imatinib in combination with hydroxyurea versus hydroxyurea alone as oral degradation. Haematologica 2009; 94: 1085–1093.

& 2014 Macmillan Publishers Limited Oncogene (2014) 2568 – 2576 Characterization of novel PDGFRA mutations AI Velghe et al 2576 40 Demoulin JB, Uyttenhove C, Van Roost E, de Lestre B, Donckers D, Van Snick J phosphatidylinositol 3-kinase and sterol regulatory element-binding proteins. et al. A single tyrosine of the interleukin-9 (IL-9) receptor is required for STAT J Biol Chem 2004; 279: 35392–35402. activation, antiapoptotic activity, and growth regulation by IL-9. Mol Cell Biol 1996; 43 Schonherr C, Ruuth K, Kamaraj S, Wang CL, Yang HL, Combaret V et al. Anaplastic 16: 4710–4716. Lymphoma Kinase (ALK) regulates initiation of transcription of MYCN in neuro- 41 Montano-Almendras CP, Essaghir A, Schoemans H, Varis I, Noel LA, Velghe AI et al. blastoma cells. Oncogene 2012; 31: 5193–5200. ETV6-PDGFRB and FIP1L1-PDGFRA stimulate human hematopoietic progenitor 44 Medves S, Noel LA, Montano-Almendras CP, Albu RI, Schoemans H, proliferation and differentiation into eosinophils: role of NF-kappaB. Haematolo- Constantinescu SN et al. Multiple oligomerization domains of KANK1- gica 2012; 97: 1064–1072. PDGFRB are required for JAK2-independent hematopoietic cell proliferation 42 Demoulin JB, Ericsson J, Kallin A, Rorsman C, Ronnstrand L, Heldin CH. Platelet- and signaling via STAT5 and ERK. Haematologica 2011; 96: derived growth factor stimulates membrane lipid synthesis through activation of 1406–1414.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2014) 2568 – 2576 & 2014 Macmillan Publishers Limited