Ezrin is a target for oncogenic Kit mutants in murine erythroleukemia

Spi-1/PU.1 transgenic mice develop an acute erythroleukemia evolving as a 2-step process.¹  The first step (HS1 stage) is characterized by a large hepatosplenomegaly associated to a severe anemia. Spleen and liver are infiltrated by preleukemic proerythroblasts arrested in maturation that remain strictly dependent upon erythropoietin (Epo) for proliferation and survival in vivo as well as in vitro. Spi-1/PU.1 is an ETS transcription factor that plays a critical role in regulating the commitment of multipotent hematopoietic progenitors and the development of the B lymphoid and monocytic lineages.^2-5  The pathology that occurs in Spi-1/PU.1 transgenic mice thus indicates that overexpression of Spi-1/PU.1 in erythroid precursor cells impedes their maturation. However, later on during the disease progression, proerythroblastic cells emerge that are characterized by Epo autonomous growth and tumorigenicity in vivo. These proerythroblastic cells (HS2 stage) harbor gain-of-function mutations in the Kit gene encoding the tyrosine kinase receptor for stem cell factor (SCF).⁶  These mutations target the Kit kinase domain and confer a ligand-independent tyrosine kinase activity to the Kit receptor. Kit mutants cause growth factor–independent proliferation of proerythroblasts, thereby resulting in neoplastic transformation. Thus, the Spi-1/PU.1 model of erythroleukemic transformation involves the collaboration of at least 2 main events, one blocking differentiation and one inducing proliferation

To further investigate the oncogenic events associated with this leukemic transformation process, we performed a proteomic profiling analysis to identify proteins differentially expressed in cells during the preleukemic (HS1) and the leukemic (HS2) stage. We found that ezrin was highly expressed in HS2 cells as compared with HS1 cells, and we focused our study on this protein. Ezrin is a member of the ezrin-radixin-moesin (ERM) protein family.⁷  Initially described as a cytoskeleton organizer in epithelial cells, ezrin participates to substrate adhesion, cell survival, cell motility, and formation of cell-cell junctions.^8,9  The NH2-terminal region anchors ezrin in the plasma membrane, whereas the COOH-terminal domain interacts with the actin cytoskeleton. In inactive state, the actin and membrane-binding domains are masked through intramolecular and/or intermolecular self-associations.⁹  The functional activation of ezrin disrupts the closed conformation through a sequential process involving phosphatidyl^4,5 -bisphosphate (PIP₂) interaction, plasma membrane localization, and tyrosine/threonine phosphorylations.^10-13  Notably, extracellular signals such as those mediated by the hepatocyte growth factor (HGF), the epidermal growth factor (EGF), or the platelet-derived growth factor (PDGF) induce phosphorylation of ezrin in epithelial cells through stimulation of their transmembrane receptors.^14-16  Alternatively, ezrin can be activated by intracellular signaling factors such as the Src kinase Lck in T lymphocytes¹⁷  or the Src¹⁸  and Rho pathways¹⁰  in epithelial cells. Moreover, ezrin interacts with p85, the regulatory subunit of PI3-kinase, and activates the PI3K/AKT pathway.¹⁶  Thus, ezrin participates in signaling programs leading to its pleiotropic functions.

Several studies provide evidence that ezrin has a metastasis-promoting function in cancer. Increased ezrin expression has been associated with high metastatic potential in a variety of human and rodent cancers, including pancreatic adenocarcinomas, osteosarcomas, rhabdomyosarcomas, and breast carcinomas.^19-25  In the present study, we report that a dysregulation of ezrin is related to the progression of preleukemic proerythroblasts toward malignancy. We show that disruption of ezrin function by the ezrin mutants Y145F and Y335F leads to reduced proliferation and increased death of leukemic cells. Notably, we show that ezrin is a downstream signaling effector of the oncogenic forms of Kit expressed in the leukemic proerythroblasts. Together, our data suggest that ezrin might be an important element in the oncogenic processes associated to activating mutations in tyrosine kinases.

HS1 and HS2 spi-1 transgenic cell lines have been previously described.^1,6  For growth kinetics, cells were plated at 2 × 10⁵ cells/mL. PP1 and PP2 (Calbiochem, La Jolla, CA), LY294002 (PI3 kinase inhibitor; Sigma- Aldrich, Lyon, France), and imatinib mesilate (Novartis-Pharma, Basel, Switzerland) were added at the indicated concentrations in culture medium. Living cells in culture were separated from dead cells through ficoll (Histopaque-1077; Sigma-Aldrich) gradient centrifugation (30 minutes at 400g).

Clonogenic assays were performed in 1% methylcellulose medium (MethoCult M3134; Teku-bio, Le Perray en Yvelines, France) supplemented with 1% fetal bovine serum (FBS) and Epo (1 U/mL). A total of 400 growing cells were seeded into 1.5 mL medium in 30-mm dishes and colonies (50 cells at least) were counted on day 6 of incubation.

Cells were observed using a Nikon Eclipse TE300 microscope (Nikon, Champigny-sur-Marne, France) with a 40× objective. Images were acquired with a Nikon Coolpix 950 and processed using Adobe Photoshop (Adobe Systems, San Jose, CA).

The human vesicular stomatitis virus glycoprotein (VSVG)–tagged ezrin constructs (wild-type [WT], N-terminal (Nter), Y145F, and Y335F) were cloned into the pEF-neo vector by standard cloning procedures. pEF-BOS-Kit^D814Y and pEF-BOS-Kit^WT were described previously.⁶  pMKIT-Flt3^D835Y was supplied by Dr P. Dubreuil (Inserm, Marseille, France),²⁶  and pcDNA3-myc-TEL-Jak2 was described previously.²⁷  A total of 2 μg of plasmids were transfected in COS-7 cells with Lipofectamine-PLUS reagent according to manufacturer instructions (Invitrogen, Cergy, France). Spi-1 transgenic proerythroblasts were nucleofected with 5 μg of plasmids using an Amaxa nucleofector and G16 programs (Amaxa Biosystems, Köln, Germany). Transfectants were selected in growth medium containing 800 μg/mL G418 (Invitrogen, Cergy Pontoise, France).

Cells were fixed with phosphate-buffered saline (PBS) containing 4% paraformaldehyde for 10 minutes at room temperature, washed in PBS, and stained for 10 minutes with 50 mg/mL Hoechst 33342 (Molecular Probes, Eugene, OR) in PBS. Apoptotic cells were examined under an inverted fluorescence microscope.

The following primary antibodies were used: the monoclonal anti-VSVG (VSVG P5D4 clone; Roche Diagnostics, Mississauga, ON), a polyclonal antibody directed against the carboxy-terminal domain of ezrin,²⁸  a rabbit anti-TEL immune serum,²⁷  a rabbit immune anti-Kit serum,²⁶  the antiphosphotyrosine 4G10 clone (Upstate Biotechnology, Lake Placid, NY), and antibodies raised against the phosphorylated forms of Akt (S473) and Erk1/Erk2, against the constitutive Akt and Erk1/Erk2, and against the activated form of caspase 3 (D175; Cell Signaling, Beverly, MA). An anti–β-actin antibody was purchased from Sigma, anti-Flt3 (N20), anti-moesin, anti-radixin, and anti-merlin antibodies were purchased from Teku bio) and the anti-BrdU polyclonal antibody was purchased from Dako (Trappes, France).

Cells (2 × 10⁷ cells/mL) were lysed in ice-cold TNE buffer (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM EDTA, 20 mM NaF, 25 mM β-glycerophosphate, 1 mM Na-pyrophosphate, 0.1 mM Na₃VO₄, 1% NP40, and protease inhibitors [Amersham, Orsay, France]). Lysates were centrifuged at 18 000g for 15 minutes at 4°C. Supernatants were incubated with 1 μg of ezrin antibody and 20 μL of protein G beads (Ultra-link Immobilized protein G plus, Invitrogen) for 2 hours at 4°C, then washed extensively with TNE buffer containing 0.1% NP40. Whole-cell extracts or immunoprecipitates were fractionated by SDS-PAGE, blotted, and visualized as previously described.²⁹ 

Exponentially growing cells were fixed in 70% cold ethanol and treated with RNAse prior DNA staining with propidium iodure (PI). To estimate cell-cycle time, cells were treated with BrdU, fixed immediately in cold ethanol or kept at 37°C for 3 hours after BrdU removal prior to fixing. Immunodetection of incorporated BrdU, estimation of S and G₂/M phases, and potential doubling time (Tpot) calculations were achieved as described.³⁰ 

Flow cytometry analyses were performed using a FACScalibur (Becton Dickinson, Meylan, France). Data were analyzed with FlowJo (Treestar, San Carlos, CA) software for Akt labeling and BrdU and ModFitLT (Verity, Topsham, ME) software for G₁/S/G₂ determination.

Cells (10⁷ in 0.5 mL α-MEM containing 2% FBS) were injected subcutaneously into 8-week-old female nude mice. Tumor nodules were taken off after 3 weeks and weighed.

To assess changes in protein expression that could play a role during the leukemic progression of spi-1 transgenic proerythroblasts, we performed a comparative proteomic profiling between preleukemic HS1 and leukemic HS2 cells. Protein extracts were resolved by bidirectional gel electrophoresis (Supplementary materials, available on the Blood website; see the Supplemental Materials link at the top of the online article). The differentially detected protein spots were subjected to trypsin digestion, and the resulting peptides were analyzed by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry. We focused our study on ezrin, whose expression was higher in HS2 than in HS1 extracts. This observation was further confirmed by Western blot analysis of extracts prepared from independent HS1 and HS2 cell lines. Indeed, ezrin expression was 3- to 5-fold increased in extracts from HS2 cells compared with extracts from HS1 cells, except for the 622HS2 cell line (Figure 1A). Notably, among the HS2 cell lines, cell line 622 was previously characterized as autocrine for Epo.¹ 

Ezrin belongs to the ERM family of proteins that includes radixin, moesin, and merlin, the product of the tumor-suppressor gene NF2. When the expression level of these different proteins was analyzed by Western blotting with appropriate antibodies, no significant variation was seen between preleukemic and leukemic cell extracts, except for ezrin (Figure 1B).

Thus, an overexpression of ezrin appears as a characteristic feature of the malignant status of proerythroblasts in spi-1 transgenic mice.

To approach whether the overexpression of ezrin played a role in HS2 cells, we ectopically overexpressed ezrin in HS1 cells and studied the changes in proliferation. To this end, constructs encoding the VSVG-tagged wild-type ezrin (VSVG-ezrin^WT) and/or the neomycine resistance gene (NeoR) were stably transfected in 663 HS1 cells and G418-selected clones were amplified. Among those, 2 clones from each transfection were studied in more details (Figure 1; Table S2). Since similar results were obtained, only data with one clone 663HS1-Ez^WT and one clone 663HS1-neo are shown herein. Ezrin expression was controlled by Western blotting using first an anti-VSVG antibody and then an anti-ezrin antibody (Figure 1C). The exogenous VSVG-ezrin^WT and the endogenous ezrin migrate similarly as an 80-kDa protein. Ezrin quantification indicated an approximately 3-fold increase in 663HS1-Ez^WT compared with 663HS1-neo cells. Given the increased expression of ezrin detected in HS2 cells (Figure 1A), the expression level of enforced ezrin appeared physiologically relevant. As illustrated in Figure 1D, transduced 663HS1-Ez^WT cells proliferated in the presence of Epo quite similarly to control 663HS1-neo cells. Epo starvation resulted in a massive cell death, demonstrating that the survival and proliferation of 663HS1-Ez^WT cells remained strictly dependent on Epo. This indicated that ezrin overexpression was not sufficient to abolish the Epo dependency of HS1 cells for growth and survival. We also determined that ezrin overexpression did not induce Epo hypersensitivity in HS1 cells when cultured in the presence of limiting concentrations (0.01 and 0.05 U/mL) of Epo (data not shown).

To gain insights into a role for ezrin in leukemic HS2 proerythroblasts, we used a dominant-negative form of ezrin, corresponding to the N-terminal domain (amino acids 1-309) of the protein.¹⁵  An expression vector for the VSVG-tagged ezrin mutant (VSVG-ezrin^Nter) was stably-transfected into 606HS2 cells (clone 606HS2-Ez^Nter). Controls were 606HS2 cells overexpressing VSVG-ezrin^WT (clone 606HS2-Ez^WT) and/or NeoR (clone 606HS2-neo). A total of 2 cell clones from each transfection were studied in Figure 2 and Table S2. A 40-kDa protein corresponding to VSVG-ezrin^Nter, and the 80-kDa protein corresponding to VSVG-ezrin^WT were detected in transfected cells by Western blotting using an anti-VSVG antibody (Figure 2A). We then analyzed the proliferation rate of HS2 cells in which ezrin^Nter expression was enforced. A marked reduction of approximately 3-fold over 48 hours was seen in the number of living 606HS2-Ez^Nter cells compared with control 606HS2-Ez^WT and 606HS2-neo cells (Figure 2B), indicating that the overexpression of a dominant-negative form of ezrin strongly affects the proliferation of leukemic cells. Next, we investigated the consequences of ezrin^Nter expression on the viability of HS2 cells. Figure 2C shows that about 35% of 606HS2-Ez^Nter cells were dead after 48 hours of culture, whereas no mortality was observed in 606HS2-Ez^WT and 606HS2-neo cells. 606HS2-Ez^Nter cells died by apoptosis as shown by the presence of fragmented nuclei detected by Hoechst staining (Figure 2D). Accordingly, the mature active form (17 kDa) of caspase-3 processed during apoptosis³¹  was detected in extracts from 606HS2-Ez^Nter cells on Western blotting (Figure 2E).

The VSVG-ezrin^Nter expression construct was also stably-transfected into 663HS1 cells. The Epo-dependent proliferation rates, and the viability of 663HS1-Ez^Nter and 663HS1-Ez^WT cells were compared. Figure 2F shows that the overexpression of VSVG-ezrin^Nter in HS1 cells induced a profound impairment of proliferation and survival.

Altogether, these findings indicate that the expression of ezrin^Nter acts in a dominant-negative way on HS2 as well as HS1 cells, suggesting that endogenous ezrin participitates in controlling growth and survival of proerythroblasts.

Ezrin is known to be a substrate for tyrosine kinases. We recently showed that mutations in the kit gene occurred in 86% of HS2 tumors. These mutations target either amino acid residue 814 (D814Y) or 818 (D818Y), causing the autophosphorylation of the SCF receptor and consequently the activation of downstream mitogen-activated protein (MAP) kinase and PI3 kinase pathways.⁶ 

As shown in Figure 1, ezrin was found overexpressed in HS2 cells harboring Kit mutations, but not in the 622 HS2 cells that were autocrine for Epo and not mutated for Kit.⁶ 

This prompted us to investigate whether Kit mutants were involved in the phosphorylation of ezrin. At first, the tyrosine phosphorylation status of ezrin was investigated in 2 HS2 leukemic cell lines, one harboring the Kit mutation D814Y (606HS2 cells) and one harboring the D818Y mutation (931HS2 cells). Ezrin was immunoprecipitated and analyzed by Western blotting using antiphosphotyrosine antibodies. Tyrosine phosphorylation of ezrin was detected in both cell lines treated with pervanadate (Figure 3A). Moreover, this phosphorylation was abolished when the 931HS2 cells were treated with PP1 or imanitib mesylate (IM), 2 compounds that we previously used as inhibitors of the Kit^D818Y tyrosine kinase activity.⁶  In the 606 HS2 cells, ezrin phosphorylation was abolished by PP1, but not altered by IM as expected from the resistance of Kit^D814Y to IM.⁶  Thus, ezrin is phosphorylated on tyrosine in HS2 cells, and this phosphorylation is associated with the Kit mutant activities.

To investigate a potential interaction between ezrin and Kit^D814Y, we used COS-7 cells that do not express germinal Kit. Cells were transiently transfected with expression vectors coding for VSVG-ezrin^WT and/or Kit^D814Y, and coimmunoprecipitation experiments with an anti-VSVG antibody were performed. Tyrosine-phosphorylated proteins associated with ezrin were assessed by immunoblotting using an antiphosphotyrosine, an anti-Kit, and an anti-VSVG antibody, successively. Then, the expression of transfected proteins was controlled by immunoblotting of whole-cell extracts (WCEs) with anti-Kit and anti-VSVG antibodies. As shown in Figure 3B, a 140-kDa tyrosine-phosphorylated protein corresponding to Kit was coimmunoprecipitated with tyrosine-phosphorylated VSVG-ezrin in Kit-transfected cells. This phosphorylation of ezrin was dependent on Kit kinase activities because a treatment of COS-7 cells coexpressing ezrin and Kit^D814Y by PP1 or PP2 abolished the tyrosine phosphorylation of ezrin (Figure 3C), whereas an IM treatment did not.

Altogether, these data argue for ezrin as a direct substrate of Kit tyrosine kinase activity in vivo.

To test whether ezrin phosphorylation was limited to the Kit mutants, 2 constitutively activated tyrosine kinases were assayed: the mutated Flt3 receptor (Flt3^D835Y) frequently identified in human acute myeloid leukemias (AML),³²  and the TEL-Jak2 fusion protein expressed in chronic myeloproliferative disease³³  or acute lymphoblastic leukemias.³⁴  The expression vectors were transfected into COS-7 cells along with the VSVG-ezrin expression vector, and the expression of transfected proteins was controlled by immunoblotting of WCEs with appropriate antibodies (Figure 3D). The immunoprecipitated ezrin was tyrosine-phosphorylated when expressed with Flt3^D835Y but not with TEL-Jak2 (Figure 3D; top subpanel). Altogether, the data indicate that ezrin is phosphorylated on tyrosine by either Kit^D814Y or Flt3^D835Y, these proteins being constitutively activated mutants of the class III tyrosine kinase receptor family.

Finally, we determined that ezrin could also be a target for the wild-type Kit stimulated by its ligand SCF. In COS-7 cells cotransfected with expression vectors coding for ezrin and Kit^WT, ezrin is tyrosine-phosphorylated in the presence of SCF, though with a poor efficiency (Figure 3E).

Ezrin has 2 major tyrosine residues (Y145 and Y353) that can be phosphorylated in response to stimulation by some growth factors, such as EGF, and are important for ezrin signaling in epithelial cells.¹²  To determine whether residues Y145 and Y353 are involved in ezrin function in HS2 cells, we generated stable transfectants of 606HS2 cells and 931HS2 cells with expression constructs coding for VSVG-ezrin^Y145F and VSVG-ezrin^Y353F. The expression level of transfected proteins were analyzed by Western blotting using an anti-VSVG and an anti-ezrin antibody, successively. A total of 2 independent clones of 606HS2-Ez^Y145F and 606HS2-Ez^Y353F and 2 independent clones of 931HS2-Ez^Y145F and 931HS2-Ez^Y353F cells were analyzed in Figure 2 and Table S2. Whatever the constructs (Ezrin^WT, Ezrin^Y145F, or Erzin^Y353F), the expression level of ezrin was less than 3-fold over the control (Figure 4A).

The growth kinetics of 606HS2-Ez^Y145F and 606HS2-Ez^Y353F cells were compared with those of 606HS2-Ez^WT and 606HS2-neo cells. A significant reduction in cell proliferation was seen at 48 and 72 hours (Figure 4B). Then, cell viability was analyzed by trypan blue exclusion (Figure 4C). Although ezrin^Y145F reduced the expansion of 606HS2 cells, no alteration in cell viability was seen even after 72 hours. In contrast, 28% plus or minus 5% of cells expressing ezrin^Y353F were dead after 72 hours of culture. These cells died by apoptosis as deduced from the detection of cleaved caspase-3 in cell extracts (Figure 4D). These observations were confirmed in clonogenic assays. The cloning efficiency and the size of colonies derived from 606HS2-Ez^Y145F and 606HS2-Ez^Y353F cells were significantly reduced compared with 606HS2-Ez^WT and 606HS2-neo control cells (Figure 4E). In addition, morphologic examination of cells forming the colonies showed striking differences. Whereas 606HS2-Ez^Y145F cells had no evident cell morphology alterations, 606HS2-Ez^Y353F cells exhibited morphologic characteristics of cell death. The ability of ezrin^Y353F to induce cell death impeded further biological and biochemical investigations with HS2 cells transfected with Ezrin^Y353F.

The relevance of the reduced 606HS2-Ez^Y145F cell proliferation in vitro to tumor development in vivo was tested. 606HS2-Ez^Y145F, 606HS2-Ez^WT, or 606HS2-neo cells were injected subcutaneously into nude mice, and the tumor growth was measured 3 weeks after graft. The tumors formed by 606HS2-Ez^Y145F cells were significantly smaller (around 3-fold) than tumors formed by 606HS2-Ez^WT cells or 606HS2-neo cells (Figure 4F), indicating that expression of ezrin^Y145F mutant in HS2 cells reduced their tumorigenic potential in vivo.

Altogether, these data suggest that Y145 and Y353 phosphorylations are required for the function of ezrin in HS2 leukemic cells, and that these tyrosine phosphorylations participate in different pathways to control either the growth or the survival of leukemic proerythroblasts.

We investigated further whether the proliferative disadvantage of 606HS2-Ez^Y145F cells was related to a modification in the cell cycle. A flow cytometric analysis of the DNA content after PI incorporation at 24 hours or at 48 hours (similar results not shown) shows that cells transfected with ezrin^Y145F or ezrin^WT or neo vector exhibited a very similar repartition into the phases of cell cycle (Figure 5A).

Then, the cell-cycle time of 606HS2-Ez^Y145F and 606HS2-Ez^WT cells and 606HS2-neo cells was measured after pulse labeling with bromodeoxyuridine. The DNA-BrdU flow cytometry analysis permitted to estimate the potential doubling time and the duration of cell-cycle phases as described by Terry and White.³⁰  Calculation in Table 1 indicates a S phase of 15 hours and a doubling time of around 24 hours for 606HS2-Ez^Y145F cells, whereas the S phase was around 10 hours and the doubling time was less than 14 hours for 606HS2-Ez^WT and 606HS2-neo cells. This result is consistent with a delayed progression through each phase of the cell cycle in cells expressing ezrin^Y145F.

We previously showed the importance of PI3K/Akt and Erk1/Erk2 signaling pathways in the proliferation and survival of HS2 cells.²⁹  We thus evaluated whether the expression of ezrin^Y145F would influence the phosphorylation of Akt and Erk1/Erk2. As shown by fluorescence-activated cell sorter (FACS) analysis (Figure 5B), the expression of ezrin^Y145F did not change the levels of Akt phosphorylation, agreeing with the absence of apoptotic cells in Figure 4. The levels of phosphorylated Erk1/Erk2 were compared in total extracts from 606HS2-Ez^Y145F, 606HS2-Ez^WT, and 606HS2-neo cells by Western blotting using appropriate antibodies. Figure 5C shows that Erk1/2 phosphorylation was strongly reduced in 606HS2-Ez^Y145F cells compared with 606HS2-Ez^WT cells and 606HS2-neo cells. Thus, in agreement with the proliferation delay induced by ezrin^Y145F, the phosphorylation of ezrin on residue Y145 appears associated to activation of Erk1/2.

The erythroleukemia that develops in spi-1 transgenic mice evolves as a multistep process. The early step is characterized by an accumulation of erythroblasts arrested at a proerythroblastic stage that remain strictly dependent from Epo for survival and proliferation. These preleukemic cells (HS1) result from the ectopic overexpression of the transcription factor Spi-1/PU.1 in the proerythroblast. Later on, leukemic proerythroblasts (HS2 cells) characterized by autonomous growth and tumorigenicity emerge as a consequence of acquired somatic mutations in the tyrosine kinase receptor Kit. In the present work, we show that ezrin expression is enhanced in leukemic HS2 proerythroblasts compared with preleukemic HS1 cells, and we provide evidence that the oncogenic forms of Kit are responsible for ezrin phosphorylation. By gene transfer of dominant-negative ezrin variants, we demonstrate that inhibition of ezrin function promotes a delay in cell proliferation and alters the apoptosis-resistant phenotype of the HS2 leukemic cells. Together, these findings point to a new role for ezrin as a signaling player in the development of hematopoietic malignancies.

Functions of ezrin are well-defined in epithelial cells. Ezrin is a membrane cytoskeletal cross-linker that participates in several growth factor receptors signaling leading to cell survival, differen-tiation, motility, invasion, and cell adhesion. However, the role played by ezrin during normal erythroid differentiation remains poorly documented. No abnormality during erythroid development has been reported in ezrin knockout (vil2-null) mice, which die with an intestinal epithelium disorganization.³⁵  Nevertheless, the protein is detected in the marginal band of chicken erythrocytes, suggesting a role in the cytoskeletal organization.³⁶  Ezrin is expressed in murine erythrocytes (our personal unpublished data, May 2005), and the data reported herein show that ezrin, like other members of the ERM family, is expressed in spi-1 transgenic proerythroblasts. Strikingly, an enforced expression of the N-terminus domain of ezrin in both preleukemic and leukemic cells resulted in a reduction in cell number and the appearance of apoptotic cells, indicating that ezrin may play a crucial role in proerythroblasts by participating to the viability and proliferation. Because of unanticipated effects of this dominant-negative mutant, future experiments are needed to determine the precise function of ezrin in erythoid cells.

A recurrent feature found in HS2 leukemic cells was that ezrin was overexpressed in leukemic cells, suggesting a distinct role for this protein in the progression toward malignancy. This finding is consistent with several studies that have linked an increase in ezrin expression to a high metastatis potential in a variety of sarcomas and carcinomas.^19-25  In these solid tumors, elevated levels of ezrin have been attributed to a transcriptional up-regulation of the ezrin (vil2) gene. Notably, we observed that the transcription pattern of the vil2 gene was similar in HS1 and HS2 cells (our unpublished data from Northern blot and RQ-PCR analysis, March 2004). Moreover, nucleotidic sequencing did not reveal modifications in the primary structure of ezrin in HS2 cells, excluding the idea that mutations induce changes in the protein stability. The mechanism(s) by which ezrin become up-regulated during leukemogenesis remains unknown.

Strikingly, an ectopic overexpression of ezrin in preleukemic proerythroblasts is not sufficient to change the behavior of the cells in terms of Epo dependency for growth and survival. Altogether, these observations were consistent with the idea that the contribution of ezrin overexpression to oncogenesis required additional stimuli specific to leukemic cells.

Besides the observation that ezrin was overexpressed in HS2 leukemic proerythroblasts, we found that ezrin was tyrosine-phosphorylated. Activating mutations in the SCF receptor Kit is a hallmark of HS2 cells.⁶  Because ezrin overexpression was associated to Kit mutation in HS2 cells (Figure 1A), mutant Kit appeared as a potential player for the tyrosine phosphorylation of ezrin. As a first corroboration, we observed that the tyrosine phosphorylation of ezrin was abolished during treatments of HS2 cells by drugs (PP1 and IM) known to abolish the kinase activities of Kit.⁶  Consistent with a functional interaction, ezrin and Kit mutants were coimmunoprecipitated in extracts prepared from COS-7 cells cotransfected with both constructs and the phosphorylation of ezrin was sensitive to inhibitors of the kinase activity of Kit, confirming that the Kit kinase activity was important for the tyrosine phosphorylation of ezrin.

In epithelial cells, ezrin is phosphorylated by some tyrosine kinases. Specifically, the tyrosines 145 (Y145) and 353 (Y353) are phosphorylated in response to the stimulation by HGF.¹⁵  Phosphorylation of Y353 induces the activation of the PI3 kinase and its downstream effector Akt, providing a mechanism for the function of ezrin in cell survival.¹⁶  Notably, Y353 is not present in radixin and moesin. Therefore, the antiapoptotic function associated to Y353 phosphorylation appears specific for ezrin. Phosphorylation of Y145 participates in signaling involved in the control of cell spreading and proliferation.¹⁸  Y145 is a substrate for various kinases of the Src family, including Lck in T cells in response to T-cell receptor (TCR) activation¹⁷  and Src in epithelial cells in response to EGF stimulation.¹⁸  To address the importance of Y145 and Y353 in ezrin function in HS2 leukemic cells, we expressed ezrin mutants point-substituted (Y145F and Y353F) that mimic dephosphorylated ezrin. The massive apoptosis resulting from ezrin^Y353F expression shows that phosphorylation on residue Y353 induces a survival signal in leukemic cells. In contrast, expression of ezrin^Y145F in leukemic cells caused a major defect in proliferation both in vitro and in vivo due to an increase in the potential doubling time of the cells, which was associated with an extinction in MAP kinases Erk1/2 activation. Thus, Erk may be an effector of ezrin in cell proliferation, as reported by others in breast carcinoma cells.²⁵  The phosphorylation of both Y145 and Y353 residues provide distinct regulatory signals in controlling cell proliferation and survival, and these findings are in line with previously reported behavior of ezrin in epithelial cells. How Kit phosphorylates ezrin is unknown. Clearly, ezrin phosphorylation is not a usurped pathway for mutated Kit, since ezrin can also be phosphorylated downstream of SCF-stimulated Kit. Kit and ezrin coimmunoprecipitation in COS-7 cells makes ezrin a possible substrate for Kit, but phosphorylation of ezrin can also be carried out by a second kinase activated by Kit. Formally, it will be important to determine whether the tyrosine residues 145 and 353 of ezrin are substrates for Kit kinase in leukemic cells.

This study is the first that emphasizes a role for oncogenic kinases in the phosphorylation of ezrin in tumors. How might ezrin phosphorylation contribute to erythroleukemia associated to Kit mutations? Given the increasing evidences that ezrin is a signal transducer in functions as diverse as cell morphogenesis, adhesion, motility, proliferation, and survival depending on the cell stimuli and cell type,^9,37  it is likely that ezrin also behaves as a signal transducer in leukemic cells. The persistent signaling mediated by Kit mutants could block ezrin in a constitutively active form. Phosphorylated tyrosines of ezrin may serve as anchoring residues, enabling the building of constitutive multicomponent transmembrane complexes involved in the activation of signaling pathways including the MAP kinase pathway.

Phosphorylation on Y145 depends on the conformational activation of ezrin through binding to polyphosphoinositodes and threonine phosphorylation downstream of the Rho or protein kinase C (PKC) signaling pathways.^10,38  Although threonine 567 is phosphorylated in HS2 cells (data not shown), the mechanism of its phosphorylation is unknown. We previously showed that the constitutive activation of MAP kinases Erk1/2 in HS2 cells is determined by a PKC-dependent mechanism.²⁹  It is tempting to speculate that PKC participates in the process of ezrin activation during the signaling cascade initiated from Kit mutants and leading to Erk1/2 phosphorylation.

Besides ezrin phosphorylation by Kit, we found that the oncogenic form of Flt3 receptor (Flt3^D835Y) can also induce the phosphorylation of ezrin on tyrosine. Given that ezrin is also phosphorylated downstream to the PDGF receptor,¹⁴  it is possible to postulate that ezrin is a substrate for the class III tyrosine kinase receptors; 2 of them, Kit and Flt3, play a crucial role in hematopoiesis.³⁹  In this regard, the role of phosphorylated ezrin in proliferation and survival might be seen as a general process when tyrosine kinases are activated in cancer cells. The mutant forms of Kit in the HS2 tumors are also found in human mastocytosis and acute myeloid leukemia (AML). Gastrointestinal stromal tumors (GISTs) in humans are associated with Kit-activating mutations that affect the juxtamembrane domain. In these tumors, an up-regulation of ezrin has been correlated to the expression of Kit mutants through a transcriptional process.^40,41  As the ezrin phosphorylation level has not been determined in those tumor samples, it is possible that ezrin phosphorylation downstream of Kit mutants occurs similarly to what is observed in murine erythroleukemia. Concerning malignant hemopathies, gain-of-function mutations targeting Flt3^32,42  and Kit^43,44  are detected respectively in approximately 40% and 5% of AML, and are responsive of the acute proliferation of blasts. It will be of interest to make an extensive analysis of the level of expression and phosphorylation of ezrin in human leukemic samples.

Kit may play a role in hematopoietic cells through mechanisms similar to those involving Met in invasiveness and metastatic migration in solid tumors. Met signaling toward Erk is dependent on the formation of a complex with ezrin and adhesion molecules such as the hyaluronic acid receptor CD44.⁴⁵  CD44 has been implicated in metastasis in a variety of sudies.⁴⁶  CD44 is expressed in hematopoietic cells and has a role in homing and adhesion.⁴⁷  Its overexpression in malignant hematopoietic cells is frequent⁴⁸  and indicative of a bad prognostic. The constitutive anchorage of ezrin in the membrane may promote its interaction with signaling partners such as CD44. This hypothesis remains to be validated in future studies.

These past years, several works have underlined the importance of ezrin overexpression during cell transformation, invasion, and metastasis. Our findings show that tyrosine phosphorylation rather than overexpression is essential in a murine erythroleukemic process. They identify ezrin as an effector of oncogenic tyrosine kinases during a leukemic process and open new perspectives regarding the disruption of ezrin function as an approach for targeting leukemia cell growth and survival. More studies are required to elucidate which downstream pathways are affected by ezrin activation and which genes are regulated in response to this type of signaling in cancer.

The authors thank Z. Maciorowski (Cytometry Department at Institut Curie) for helpful assistance with flow cytometry and cell-cycle analysis and Janssen-Cilag (Issy-les-Moulineaux, France) for the gift of recombinant human Epo. We are grateful to C. Guillouf, P. Rimmelé, O. Kosmider, and D. Buet for many helpful discussions and F. Wendling for her comments on the manuscript.

This work was supported by the Institut National de la Santé et de la Recherche Médicale (Inserm) and Institut Curie, and by the Association pour la Recherche sur le Cancer (grant no. 3822), Fondation de France (grant no. 4778), Ligue contre le Cancer (équipe labellisée 2006 to F.M.-G.), Institut National du Cancer (INCa), and Association Christelle Bouillot. R.M. was a postdoctoral researcher supported by the Fondation de France.

Contribution: R.M. performed proteomic profiling, phenotypic characterization of the cell lines, molecular cloning, FACS, and cell-cycle analysis, and COS-7 transfection experiments and tumorigenic assays. L.H. performed proteomic profiling analysis. A.N. contibuted purification of anti-ezrin antibodies and co-immunoprecipitation assays. I.G. performed COS-7 transfection experiments. M.A. provided expertise in ezrin function. P.M. provided expertise in proteomic profiling. F.M.-G. designed research and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Françoise Moreau-Gachelin, Inserm U830–Institut Curie, 26 rue d'ulm, 75005 Paris, France; e-mail: framoreau@curie.fr.