JAK2 and MPL protein levels determine TPO-induced megakaryocyte proliferation vs differentiation

Megakaryopoiesis is the cellular process leading to platelet production from the differentiation of hematopoietic stem cells (HSC). It can be divided into a proliferative stage that generates megakaryocyte (MK) progenitors and precursors, and a maturation stage in which differentiating MKs proliferate no more. These 2 stages can be driven by thrombopoietin (TPO), which therefore exerts both proliferative and antiproliferative effects. TPO binding to its cognate receptor, myeloproliferative leukemia (MPL), activates multiple downstream signaling pathways.^1-4  MPL being devoid of kinase activity, the receptor associates with intracytoplasmic tyrosine kinases, in particular janus kinase 2 (JAK2), for signal transduction.^5-7  JAK2 is not only essential for TPO-induced signal transduction, but also for MPL stability and cell-surface expression.⁸ 

In essential thrombocytemia (ET) and primary myelofibrosis (PMF), 2 myeloproliferative neoplasms (MPNs), the abnormal accumulation of MKs suggests that these cells have escaped the proliferation arrest associated with terminal steps of differentiation.^9-11  Mutations in JAK2 (JAK2^V617F) and MPL are detected in ∼60% and ∼5% of PMF and ET, respectively.⁹  In addition, MPL is downregulated in MKs and platelets of PMF patients, and this decrease is a more controversial issue in ET.^12-16 

We have shown previously that TPO triggers MK proliferation arrest and promotes a differentiation-associated senescence through a strong mitogen-activated protein kinase (MAPK) signaling.¹⁷  In the present study, we show that TPO-induced proliferation vs differentiation depends on JAK2 and MPL protein levels. When 1 of these proteins is expressed at low level, TPO induces a weak signal that promotes cell proliferation. At higher JAK2 and MPL levels, TPO promotes cell-cycle arrest and MK differentiation. The modulation of MPL and JAK2 expression levels may regulate the 2 steps of normal megakaryopoiesis, and their downregulation may explain the abnormal proliferation of MKs in MPNs. This model could also explain the paradoxical increase in the platelet count induced by suboptimal doses of JAK2 chemical inhibitors.¹⁸ 

The human megakaryoblastic UT7 cells expressing MPL (UT7-11oc1 to oc7 clones) were grown in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen, Cergy Pontoise, France). This medium was supplemented with 10% fetal bovine serum, antibiotics (100 IU/mL penicillin and 50 mg/mL streptomycin), and GM-CSF (5 ng/mL) or recombinant human TPO (hTPO) (10 ng/mL).

Blood samples were obtained after informed consent in accordance with the Declaration of Helsinki. Approval was obtained from the Assistance Publique des Hôpitaux de Paris. CD34⁺ cells were isolated using immunomagnetic beads (Miltenyi Biotec, Paris, France) and grown in a serum-free medium, supplemented with hTPO (10 ng/mL; a generous gift from Kirin, Tokyo, Japan).

Oligonucleotide JAK2 short hairpins (shRNA JAK2) listed in supplemental Table 1 (available on the Blood Web site) were synthesized (Eurogentec, Angers, France) and inserted into a pBlue Script containing the human H1 promoter. The H1-shJAK2 or H1-SCR (scramble control sequence, supplemental Table 1) cassette was inserted into a lentiviral vector (pRRLsin-PGK [Phosphoglycerate kinase]-eGFP-WPRE; Genethon, Evry, France). The cDNA encoding the human MPL^WT was cloned into the bicistronic retroviral vector pMX-IRES-CD4. The cDNA encoding the human JAK2 (JAK2^WT or JAK^V617F) was cloned into a modified lentiviral vector pRRL, generated by replacing the promoter PGK with the promoter MND (governing JAK2 expression) followed by the promoter EF1a governing eGFP expression (TMJ-JAK2 vector, kindly provided by Dr Chloé James).

Vesicular stomatitis virus glycoprotein pseudotyped viral particles were produced into 293EBNA or 293T cells. Cells were infected with concentrated retrovirus or lentivirus supernatants for 2 hours at a multiplicity of infection of 10 and sorted by flow cytometry (FACS Vantage; BD Biosciences, Mountain View, CA) 48 hours later based on eGFP or CD4 expression.

The raw global gene expression values (from Affymetrix GeneChip) for 24-hour TPO-exposed UT7-11oc1-7 cells were processed with the robust multiarray analysis algorithm using BioConductor software, version 2.3.

UT7-11oc1-7 and CD34⁺ cells were rinsed in PBS, stained for 30 minutes at 4°C with anti-CD41–APC, anti-CD41-FITC, anti-CD42–PE, or an irrelevant mouse immunoglobulin G1 (IgG1) (BD Biosciences). MPL cell-surface expression was explored using the mouse monoclonal anti-MPL IgG2b, clone 1.78¹⁹  (a generous gift from Amgen, Thousand Oaks, CA), and indirect immunofluorescence with a FITC-goat anti-mouse Ig. Cell samples were analyzed by flow cytometry using FACS Vantage.

Immunoblotting was performed as described previously.¹⁷  Antibodies used: Actin (Sigma-Aldrich), JAK2 (no. 3230), phosphoJAK2 (no. 3776), phosphoSTAT5 (no. 9351), phosphoSTAT3 (no. 9134), phosphoSTAT1 (no. 9171), phosphoERK (no. 4377), ERK (no. 9102), p21CIP1 (no. 2947), PARP (no. 9542), caspase 3 (no. 9662) (Cell Signaling, Beverly, MA), p53 (no. SC-6243), CD42b (no. SC59051) (Santa Cruz Biotechnology, Heidelberg, Germany), TPOR/c-Mpl (EMD Millipore, Billerica, MA), and HSC70 (Stressgen). Proteins were detected using enhanced chemiluminescence (ECL; Pierce Perbio, Brebières, France) and were quantified using ImageJ software (National Institutes of Health, Bethesda, MD).

Total RNA was extracted using a Trizol RNA isolation kit (Invitrogen). Transcription into cDNA was performed using random hexamers and SuperScript II reverse transcriptase (Invitrogen). All PCR reactions used Taqman PCR Master Mix (Applied Biosystems, Foster City, CA) to a final volume of 20 μL. Each cDNA sample was analyzed in triplicate in the ABI PRISM 7900 Sequence Detection System (Applied Biosystems).

Detection of SA-β-galactosidase activity was performed as described previously.¹⁷ 

AZD1480 was kindly provided by AstraZeneca Pharmaceuticals (Waltham, MA). Stock solution was diluted in dimethylsulfoxide (DMSO; Sigma-Aldrich) and was then diluted in culture medium for use. Novartis (Basel, Switzerland) kindly provided ruxolitinib. Stock solution was diluted in water with 0.5% methocellulose and 0.05% tween 80 for use.

C57BL/6 mice (Harlan France, Gannat, France) aged 8 to 9 weeks were treated with vehicle (0.5% methocellulose, 0.05% Tween 80) or ruxolitinib (INCB018424) twice daily by oral gavage (20 or 60 mg/kg body weight/day). Blood samples were collected after 5 days of treatment. Hematologic parameters were measured using an automated counter (MS9; Schloessing Melet, Cergy Pontoise, France). Animal experiments were performed in accordance with guidelines established by the Institutional Animal Committee.

Student t test was used in all experiments and was considered significant if P < .05.

The human megakaryoblastic cell line UT7, transduced to express stably the TPO receptor MPL, has been largely used to study megakaryocytic differentiation.^20,20-22  We have previously described the UT7-MPL clone named UT7-11oc1 in which TPO induces a proliferation arrest and a MK phenotype with senescence markers. These cells proliferate under GM-CSF exposure.¹⁷  From the UT7-11oc1 clone cultured in the presence of TPO, we derived the UT7-11oc2 subclone that escapes proliferation arrest (Figure 1A). This clone proliferates in the presence of TPO (Figure 1A) as well as in the presence of GM-CSF (Figure 1B). Morphologic changes (Figure 1C), SA-β-galactosidase activity (supplemental Figure 1A), expression of CD41 (Figure 1D), and p21^CIP1 protein expression (Figure 1E and supplemental Figure 1B) observed in TPO-treated UT7-11oc1 cells are decreased or are no more observed in TPO-treated UT7-11oc2 cells. In addition, TPO-induced activation of the MAPK pathway is decreased in UT7-11oc2 cells (Figure 1E and supplemental Figure 1B). These results led us to use UT7-11oc2 cells to explore how cells can escape TPO-induced proliferation arrest.

Using the same method, we derived 5 other clones (UT7-11oc3-7) that escape the antiproliferative action of TPO (supplemental Figure 1C). All these clones remain dependent on cytokines for their growth, as evidenced by cellular apoptosis induced by a 24-hour cytokine starvation (supplemental Figure 1D). We compared gene expression after a 24-hour exposure to TPO in UT7-11oc1 and the 6 UT7-MPL subclones UT7-11oc2-7. Nonsupervised analysis separated 2 groups of proliferative clones, UT7-11oc2 and UT7-11oc3 on one hand, and the other clones on the other, possibly indicating distinct evasion mechanisms (Figure 1F). One of the downregulated genes was JAK2. Because MPL and JAK2 are key actors in TPO-induced biological response, we studied the level of MPL and JAK2 proteins in these different clones. We observed a decrease in MPL expression in UT7-11oc2 cells and a decrease in JAK2 protein level in UT7-11oc3-7 cells, compared with UT7-11oc1 (Figure 1G). These decreases correlated with a decrease in the corresponding mRNA expression (supplemental Figure 1E). MPL cell-surface expression, evaluated by flow cytometry, was decreased in all studied UT7-11oc2-5 clones as compared with UT7-11oc1 cells (Figure 1H and supplemental Figure 1F). This was confirmed by considering the upper band on MPL immunoblots, which corresponds to the mature, glycosylated form of the receptor (Figure 1G). The decreased membrane expression of MPL in UT7-11oc3-5 clones was in accordance with the role of JAK2 in MPL stability and trafficking.⁸ 

To demonstrate directly that MPL cell-surface expression was the major determinant in the cellular response to TPO, we sorted UT7-11oc1 and UT7-11oc2 cells depending on their MPL membrane expression (supplemental Figure 2A for UT7-11oc1). However, fluorescence intensity for UT7-11oc2 MPL^high cells peaked below that of UT7-1oc1 MPL^low cells (622 vs 750, not shown). With this strategy, we could not modulate the proliferative phenotype under TPO for both clones (supplemental Figure 2B). Thus we used another strategy, which objective was to increase MPL or JAK2 expression, by retrovirus or lentivirus transduction, in UT7-11oc2 to 11oc5 clones. Of note, whereas this approach led to a main increase in MPL protein expression (Figure 2A), MPL cell-surface expression remained in the same order of magnitude for infected cells (supplemental Figure 2C). The increased expression of MPL did not promote p21^CIP1 protein expression in response to GM-CSF in any clone (Figure 2A). In UT7-11oc2 and UT7-11oc3 but not in UT7-11oc4 and 5, MPL-increased expression restored the cell-cycle arrest in response to TPO (Figure 2B). This effect was associated with an increase in p21^CIP1 expression (Figure 2C), morphologic changes (not shown), and SA-β-galactosidase activity (Figure 2D). MPL overexpression also increased ERK phosphorylation after TPO treatment (see clone UT7-11oc2 on Figure 2E). Similar to MPL, the increase in JAK2 expression did not promote p21^CIP1 protein expression in response to GM-CSF in any clone (Figure 2F). After JAK2 overexpression, TPO induced a decrease in cell proliferation (Figure 2G), an increase in p21^CIP1 protein expression (Figure 2H), morphologic modifications (not shown), and SA-β-galactosidase activity (Figure 2I) in all clones except UT7-11oc2. JAK2 overexpression also increased ERK phosphorylation after TPO treatment (see UT7-11oc4 clone, Figure 2J). Altogether, MPL and JAK2 expression levels determined the MAPK cellular response to TPO and the antiproliferative effect of the cytokine.

To explore further this hypothesis, that is, the proliferation arrest requests a strong signaling, we studied the effects of various TPO concentrations on UT7-11oc1 cells. At 0.1 ng/mL, TPO promoted UT7-11oc1 cell proliferation, whereas higher concentrations did not (supplemental Figure 3A). A proliferative effect of highest concentrations of TPO was restored by combination with the JAK2 chemical inhibitor AZD1480 at 1 μM (supplemental Figure 3B), a concentration that inhibits cell signaling (supplemental Figure 3C). Of note, higher concentrations of AZD1480 (>5 μM) inhibited cell proliferation, as expected (supplemental Figure 3D). Restoration of TPO-induced cell proliferation was associated with the inhibition of morphologic changes (supplemental Figure 3E). We also knocked down JAK2 in UT7-11oc1 cells using 4 lentiviral shRNAs whose efficacy was checked at the mRNA (Figure 3A) and protein (Figure 3B) levels. As expected, given the role of JAK2 in MPL cell trafficking, the decreased expression of JAK2 (Figure 3B) correlated with a decreased membrane expression of MPL (Figure 3C). A slight decrease in JAK2 expression with shJAK2-1 and shJAK2-2 promoted cell proliferation in response to TPO (Figure 3D) while inhibiting TPO-induced p21^CIP1 expression (Figure 3E), morphologic changes (Figure 3F), and CD41 expression (Figure 3G) without altering growth under GM-CSF (Figure 3D). The important decrease in JAK2 expression by shJAK2-3 and −4 markedly inhibited the growth under GM-CSF but still increased cell growth under TPO in comparison with the control cell line (Figure 3D). Altogether, the response of UT7-MPL cells to TPO depends on the MP/JAK2 expression level, a higher expression being requested for TPO to arrest cell proliferation and to promote cell differentiation (see model on Figure 4A). According to this demonstration, a decrease in JAK2 expression can be sufficient to escape the antiproliferative effect of TPO.

Megakaryopoiesis is a dynamic process with a proliferation step followed by a maturation step, both induced by TPO (Figure 5A). We cultured CD34⁺ cells in a serum-free medium in the presence of TPO to trigger their MK differentiation, and we observed a differentiation-associated increase in MPL and JAK2 mRNA expression levels (Figure 5B). JAK2 protein expression increased with differentiation, whereas that of MPL remained stable or decreased at the end of maturation (Figure 5C-D). Analysis of cell signaling along megakaryocytic differentiation identified an activation of STAT5 that peaked at day 5, a sustained phosphorylation of STAT3, and a phosphorylation of ERK1/2 that peaked between day 5 and day 7, preceding p21^CIP1 expression (Figure 5C). These observations—compared with the results obtained in the UT7-MPL clones—suggest that the differentiation-associated increase in MPL and JAK2 expression promotes a switch from proliferation to the differentiation step of TPO-driven megakaryopoiesis.

According to the model in which high JAK2 and MPL levels induce an antiproliferative signal, it is expected that the expression level of these 2 proteins is decreased in PMF and ET. Accordingly, immunoblot analyses of these proteins in platelets show a decrease in their expression in the context of JAK2^V617F and MPL^W515 mutations in PMF and ET samples (Figure 6A-C). Quantification of the JAK2 protein expression confirms a significant decrease in platelets of PMF JAK2^V617F (P < .05), ET JAK2^V617F (P < .01), and ET MPL^W515 (P < .05), compared with platelets from healthy donors (Figure 6D). Similarly, quantification of MPL protein expression indicates a significant decrease in platelets of PMF JAK2^V617F (P < .01), ET JAK2^V617F (P < .01), and ET MPL^W515 (P < .05) compared with platelets from healthy donors (Figure 6E). Surprisingly, JAK2 and MPL expression was not decreased in patients without JAK2^V617F and MPL^W515L/K, and an increase in MPL expression was even observed in platelets of ET patients without any mutation in JAK2 and MPL (P < .05) (Figure 6E). The decreased expression of JAK2 protein in PMF and ET (Figure 6F) was independent of any change at the mRNA level, suggesting a posttranslational mechanism (Figure 6G). The mechanism of MPL-decreased expression (Figure 6H) remained unclear as it was associated to a decreased mRNA level in PMF (P < .05) but not in ET (nonsignificant) (Figure 6I). To explore the mechanism underlying the decrease in JAK2 protein expression in the context of a JAK2^V617F mutation, we expressed the mutant protein in UT7-MPL cells. We had noticed that, in UT7-11oc1 cells, TPO specifically induced a decrease in JAK2 protein expression (supplemental Figure 4A), following its phosphorylation (supplemental Figure 4B), without any decrease in JAK2 mRNA (supplemental Figure 4C). This effect was not observed with other cytokines, including GM-CSF and erythropoietin (supplemental Figure 4A-B,E). Cell pretreatment with the proteasome inhibitor lactacystin before TPO treatment partially restored JAK2 protein expression, arguing for a proteasomal degradation of JAK2 on TPO exposure (supplemental Figure 4D). Finally, whereas overexpression of wild-type JAK2 in UT7-11oc1 cells induced the expected increase in JAK2 protein expression, that of JAK2^V617F protein induced a paradoxical decrease in JAK2 protein level (supplemental Figure 4F), suggesting that JAK2^V617F could promote the degradation of the wild-type protein.

We then studied the effect of JAK2 chemical inhibitors on megakaryocytopoiesis. A low concentration of the JAK2 inhibitor AZD1480 (0.2 µM) induced an increase in MK production from cord blood CD34⁺ cells exposed to TPO (Figure 7A). At this concentration, AZD1480 inhibits TPO-induced cell signaling (Figure 7B) and stimulates the MK production from adult CD34⁺ cells exposed to TPO (Figure 7C), without affecting differentiation marker expression (Figure 7D), or the percentage of platelets-generating MKs (not shown). Higher concentrations of AZD1480 (eg, 1 µM) completely inhibited cell proliferation and differentiation (Figure 7A). Finally, we treated C57BL/6 mice twice daily by oral gavage with vehicule (0.5% methocellulose, 0.05% tween 80) or the JAK2 inhibitor ruxolitinib at 20 or 60 mg per kg of body weight (mpk) per day (Figure 7E). As expected, given the essential role of JAK2 in TPO-induced signal transduction, ruxolitinib at 60 mpk induced a decrease in the platelet number (798 ± 226.109 vs 1018 ± 112.109) (Figure 7F, P < .05). However, a lower dose of ruxolitinib (20 mpk) induced a paradoxical and significant increase in the platelet count (1298 ± 241.109) (Figure 7F, P < .01).

Cellular response to cytokines depends on the nature, duration, strength, and cellular localization of the intracellular signals that are generated. It is generally assumed that physiologically, a low-intensity signal promotes only cell survival, a weak or moderate signal induces cell proliferation, whereas a strong signal induces cell-cycle arrest by different ways including differentiation, senescence, and apoptosis.^23,23-27  Physiological MK differentiation has been shown to request a strong activation of the MAPK pathway.^21,22,28  Here, we propose that TPO, the central cytokine in this process, induces a weak signal in MK progenitors, which promotes their proliferation, while inducing a strong signal in more mature cells, which induces their terminal differentiation. Alteration in this dual effect of TPO, as a result of a deregulated JAK2 and MPL expression, may account for the abnormal MK proliferation that characterizes some MPNs.

The proof of concept of this dual regulation was obtained in the UT7-MPL cell line in which the ability to proliferate in response to TPO was acquired in subclones wherein the expression of MPL or JAK2 protein was decreased. MPL/JAK2 expression determined whether TPO triggered cell-cycle arrest with acquisition of an MK phenotype (higher MPL/JAK2 expression), or whether it promoted cell proliferation without differentiation (lower MPL/JAK2 expression). TPO-induced cell-cycle arrest was associated with a strong activation of the MAPK pathway in accordance with previous studies,^21,22  whereas the MAPK signal was weaker when TPO was inducing proliferation. Altogether, the ability of some UT7-MPL clones exposed to TPO to proliferate depended on JAK2 and MPL protein expression, which determines downstream cell signaling intensity. The dose-dependent effect of TPO on cell proliferation and the ability of the chemical JAK2 inhibitor AZD1480 at suboptimal doses to promote proliferation of cells that might otherwise undergo differentiation further argued for the proposed hypothesis. Although flow cytometry analysis of MPL membrane expression identified a threshold in TPO-induced cellular response, we cannot rule out the selection of another genetic event that accounts for the differential response of the studied subclones to TPO. Nevertheless, we demonstrated that shRNA-mediated decrease in JAK2 protein expression, which correlates with a decrease in MPL cell-surface expression, can be sufficient to escape the antiproliferative action of TPO.

MPL and JAK2 protein expression was found to increase regularly along normal MK differentiation. Based on the observations made in the UT7-MPL model, we propose that the MPL/JAK2 expression could reach a threshold at which the signal induced by TPO switches from proliferation to cell-cycle arrest and terminal maturation. According to this model, every event leading to a decrease in MPL or JAK2 expression or in signaling intensity may extend the proliferative stage and might induce an amplification of the progenitor cell compartment in MPNs. Accordingly, a thrombocytosis is observed in transgenic mice expressing reduced levels of MPL,^29,30  and a decrease in MPL expression was observed in platelets and MKs of patients with MPNs, with either PMF or ET.^13-15  This decrease in MPL expression was obvious in patients harboring a JAK2^V617F or a MPL^W515 mutation, whereas MPL expression was increased in ET patients without any of these mutations. As we were not able to evaluate reliably the cell-surface expression of MPL in these patients, we cannot exclude a decreased expression in their MK progenitors related to a trafficking defect. In any event, these results may explain the controversial reports of MPL expression in ET.^12-16  The mechanisms leading to an MPL expression decrease could depend on the disease as MPL mRNA expression was decreased in PMF whereas it was normal in ET, suggesting that a transcriptional mechanism could operate in PMF. The frequent mutations in epigenetic regulators identified in this disorder could participate in MPL deregulation.^31,32 

JAK2 protein expression has not been often studied in MPNs. We detected a decrease in JAK2 protein expression in platelets of MPN patients, correlated with the presence of a JAK2^V617F or a MPL^W515 mutation, and without any change in mRNA expression, suggesting a posttranscriptional mechanism that may be related to the constitutive activation and phosphorylation of JAK2. Given the role of JAK2 in MPL stability and traffic,⁸  JAK2-decreased expression could participate in the decreased cell-surface expression of the TPO receptor. In the context of heterozygosity, observed in about half of ET, JAK2^V617F may also induce the degradation of the wild-type JAK2, consecutively decreasing MPL membrane expression. This mechanism could occur in addition to the recently reported proteasomal degradation following MPL ubiquitination by JAK2^V617F.³³  Here, we provide some evidence that JAK2^V617F can induce a decrease in total JAK2 expression in UT7-MPL cell line. The reduced cell signaling intensity during the first stage of megakaryopoiesis (resulting from a decreased MPL/JAK2 expression) would then extend the proliferative stage, allowing the progenitor cells to perform additional mitosis, leading to increased MK production.

The discrepancies between MPNs and congenital amegakaryocytic thrombocytopenia could be explained by the role of MPL in the HSC compartment. Gain-of-function MPL mutations seen in MPNs may confer an advantage to the HSC (survival, self-renewal, exit from quiescence) by activating signaling, whereas loss-of-function MPL mutations described in congenital amegakaryocytic thrombocytopenia, which impede TPO signaling, may result in a defect in HSC and hematopoietic cell production, resulting in amegakaryocytosis.

The proposed model suggests that weak signaling may be involved in the abnormal MK proliferation. In favor of this hypothesis, low concentrations of the JAK2 inhibitor AZD1480 induced an increase in the MK number generated in vitro in the presence of TPO, without any change in MK markers and proplatelet formation, which is a result in apparent contradiction with the thrombocytopenia described in clinical trials.³⁴  As expected, higher concentrations of the JAK2 inhibitor induced cell death. Thus, a suboptimal inhibition of JAK2 could possibly increase platelet production. Accordingly, whereas ruxolinitib, a JAK2 inhibitor used in clinics, induced thrombocytopenia in mice treated with high doses, the drug paradoxically increased the platelet count when used at suboptimal doses. An increased platelet count was also reported in mice treated with MRLB-11055, a highly potent inhibitor of JAK2,³⁵  whereas an increased platelet count was observed in some PMF patients treated with lower doses of ruxolitinib.¹⁸ 

In conclusion, we propose that the expression level of JAK2 and MPL regulates the switch from proliferation to differentiation in the megakaryocytic lineage (Figure 4B). According to this model, the downregulation of JAK2 and MPL combined with the autonomous signaling resulting from mutations may extend the proliferative stage of megakaryopoiesis and enhance MK production. Similarly, an incomplete inhibition of JAK2, as obtained with suboptimal doses of currently developed inhibitors, may induce a paradoxical thrombocytosis.

The authors thank Dr C. James for viral vector construction, Philippe Dessen, Justine Guégan, and Thomas Robert for gene-array analysis, INSERM, the Foundation Aging Research Centre, and the Association Laurette Fugain for support (R.B.), and the Ligue Nationale Contre le Cancer for support (E.S. and W.V. teams). The authors thank Kerin Brewery (Tokyo, Japan) for the generous gift of recombinant TPO.

Contribution: R.B. designed research, performed research, analyzed data, and wrote the paper; D.R.-W. designed research, performed research, and analyzed data; C.T., H.A., C.L., F.P., C.W., P.R., Y.L., and J.-B.M. performed research and analyzed data; S.G., E.S., and W.V. designed research, analyzed data, and wrote the paper; and S.N.C. analyzed data and wrote the paper.

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

Correspondence: Stéphane Giraudier, INSERM UMR 1009, Institut Gustave Roussy 114, Rue Edouard Vaillant, 94805 Villejuif, France; e-mail: stephane.giraudier@hmn.aphp.fr.