FIP1L1-PDGFRα Imposes Eosinophil Lineage Commitment on Hematopoietic Stem/Progenitor Cells*

Although leukemogenic tyrosine kinases (LTKs) activate a common set of downstream molecules, the phenotypes of leukemia caused by LTKs are rather distinct. Here we report the molecular mechanism underlying the development of hypereosinophilic syndrome/chronic eosinophilic leukemia by FIP1L1-PDGFRα. When introduced into c-KithighSca-1+Lineage- cells, FIP1L1-PDGFRα conferred cytokine-independent growth on these cells and enhanced their self-renewal, whereas it did not immortalize common myeloid progenitors in in vitro replating assays and transplantation assays. Importantly, FIP1L1-PDGFRα but not TEL-PDGFRβ enhanced the development of Gr-1+IL-5Rα+ eosinophil progenitors from c-KithighSca-1+Lineage- cells. FIP1L1-PDGFRα also promoted eosinophil development from common myeloid progenitors. Furthermore, when expressed in megakaryocyte/erythrocyte progenitors and common lymphoid progenitors, FIP1L1-PDGFRα not only inhibited differentiation toward erythroid cells, megakaryocytes, and B-lymphocytes but aberrantly developed eosinophil progenitors from megakaryocyte/erythrocyte progenitors and common lymphoid progenitors. As for the mechanism of FIP1L1-PDGFRα-induced eosinophil development, FIP1L1-PDGFRα was found to more intensely activate MEK1/2 and p38MAPK than TEL-PDGFRβ. In addition, a MEK1/2 inhibitor and a p38MAPK inhibitor suppressed FIP1L1-PDGFRα-promoted eosinophil development. Also, reverse transcription-PCR analysis revealed that FIP1L1-PDGFRα augmented the expression of C/EBPα, GATA-1, and GATA-2, whereas it hardly affected PU.1 expression. In addition, short hairpin RNAs against C/EBPα and GATA-2 and GATA-3KRR, which can act as a dominant-negative form over all GATA members, inhibited FIP1L1-PDGFRα-induced eosinophil development. Furthermore, FIP1L1-PDGFRα and its downstream Ras inhibited PU.1 activity in luciferase assays. Together, these results indicate that FIP1L1-PDGFRα enhances eosinophil development by modifying the expression and activity of lineage-specific transcription factors through Ras/MEK and p38MAPK cascades.

During the last decade, it has become clear that hematopoietic growth factors regulate only growth and survival of hematopoietic cells, whereas lineage-specific transcription factors, such as GATA-1, GATA-3, PU.1, Pax-5, C/EBP␣, and C/EBP⑀, crucially control the lineage commitment and lineage-specific differentiation. For example, granulocyte colony-stimulating factor signaling induced megakaryopoiesis in granulocyte colony-stimulating factor receptor-transgenic mice (1). Also, erythropoietin (EPO) 2 was found to promote terminal granulocytic differentiation in EPO receptor-transgenic mice. From these data, we speculated that signal transduction molecules activated by hematopoietic growth factors would not influence the lineage commitment of hematopoietic stem cells/progenitor cells (HSCs/HPCs) or subsequent lineage-specific differentiation (2). However, it has very recently been shown that the MEK/ERK pathway is involved in myeloid lineage commitment (3). Also, PKB (c-Akt) was shown to be involved in lineage decision during myelopoiesis (4). In addition, FLT3-activating mutations were proved to inhibit C/EBP␣ activity through ERK1/2-mediated phosphorylation (3,5). These results suggest that signal transduction molecules activated by hematopoietic growth factors or their genetic mutations would not only promote growth and survival but also influence lineage commitment and subsequent differentiation of hematopoietic cells.
Activating mutations of the tyrosine kinases (TKs), such as c-Kit, platelet-derived growth factor receptor (PDGFR), FLT3, and c-ABL, are provoked by several mechanisms, including chromosomal translocations and various mutations involving their self-regulatory regions. These mutations are often involved in the pathogenesis of various types of hematologic malignancies. BCR-ABL is known to cause chronic myelogenous leukemia and acute lymphoblastic leukemia. Most patients with PDGFR␤ rearrangement reveal common clinical features resembling chronic myelogenous leukemia or choronic myelomonocytic leukemia. In contrast, FLT3 mutations (ITD and point mutations in the TK domain) are primarily detectable in acute myeloid leukemia (AML) or myelodysplastic syndrome (6 -8). Also, c-KIT mutations in the TK domain (Asp 816 3 Val, Tyr, Phe, or His) are found in patients with aggressive mastocytosis, myelodysplastic syndrome, and AML (9 -15). Although these leukemogenic TKs (LTKs) activate a common set of downstream signaling molecules, such as Ras/ MAPK, PI3-K/Akt/mTOR, and STATs, the mechanisms by which LTKs cause different disease phenotypes remain to be clarified.
The concept of "cancer stem cell" has widely been recognized and validated in various types of cancers, including breast cancer, brain tumors, colon cancer, lung cancer, and malignant melanoma. This concept was originally established in AML as a "leukemic stem cell (LSC)" (23,24). In this concept, LSCs are defined as specific leukemic cells that can cause leukemia when transplanted into NOD/SCID mice. In AML, although leukemic blasts often display relatively homogenous features, they are organized in a hierarchy. Among them, LSCs reveal the most immature CD34 ϩ CD38 Ϫ phenotype similar to normal HSCs, whereas several antigen expressions are different. LSCs, which account for only 0.2-1.0% of AML cells in the bone marrow (BM), have both abilities to self-renew and to produce restrictedly differentiated leukemia cells, thereby maintaining themselves and yielding leukemia cells composing the majority (23,25,26). It is still unclear whether LSCs originate solely from HSCs or are generated from nonstem immature cells that have acquired de novo self-renewal ability. It has been shown that, although common myeloid progenitors (CMPs) and granulocyte/monocyte progenitors (GMPs) have very limited life spans, several leukemogenic oncogenes, such as MLL-ENL, MOZ-TIF2, and MLL-AF9, have an ability to immortalize these cells, thereby enabling them to act as LSCs (27,28). On the other hand, although LSCs in a chronic phase of chronic myelogenous leukemia are at an HSC level, chronic myelogenous leukemia cells at a CMP/GMP level can act as LSCs in an accelerated phase, suggesting that additional gene mutations can change the main LSC population during disease progression. From these findings, it is now speculated that the leukemia phenotype is determined by the biologic property of the mutated gene and/or the lineage and the differentiation state of LSCs.
In an attempt to analyze the molecular mechanisms by which each LTK causes leukemia with the specific phenotype, we introduced FIP1L1-PDGFR␣, which plays a causal role in HES/ CEL, into murine HSCs and various types of HPCs. As a result, we found that FIP1L1-PDGFR␣ specifically enhanced eosinophil development from HSCs/HPCs and imposed the lineage conversion to eosinophil lineage on megakaryocyte/erythroid progenitors (MEPs) and common lymphoid progenitors (CLPs) through Ras/MEK and p38 MAPK cascades by modifying the expression and activity of lineage-specific transcription factors.
Retrovirus Transduction-The conditioned media containing high titer retrovirus particles were prepared as described previously (33). Briefly, an ecotropic packaging cell line, 293gp, kindly provided by Dr. H. Miyoshi (RIKEN BioResource Center, Tsukuba, Japan), was transfected with each retrovirus vector by the calcium phosphate coprecipitation method. After 12 h, the cells were washed and cultured for 48 h. To produce lentivirus, 293T cells were transfected with each shRNA expression vector together with a packaging vector (pCAG-HIVgp) and a lentivirus envelope and Rev construct (pCMV-VSV-G-RSV-Rev), both of which were provided by Dr. Miyoshi. Then the supernatant containing virus particles was collected, centrifuged, and concentrated 50-fold in volume. The precultured murine BM cells were infected with each retrovirus in the RPMI1641 medium supplemented with the same medium containing protamine sulfate for 48 h in 6-well dishes coated with Ret-roNectin (Takara Bio Inc., Shiga, Japan).
Colony Assays-Cells were seeded into methylcellulose medium (MethoCult GF M3434; Stem Cell Technologies, Vancouver, Canada) at a density 2.5 ϫ 10 2 cells/35-mm dish and were cultured with 5% CO 2 at 37°C. All cultures were performed in triplicate, and the numbers of colonies were counted after 10 days.
In Vitro Immortalization Assays for HPCs-Immortalization assays of HPCs in vitro were performed as previously described (34). In brief, 10 4 cells were plated in 1.1 ml of methylcellulose medium (Methocult M3434). After the 1 week of culture, colony numbers were counted, and single-cell suspensions of colonies (10 4 cells) were subsequently replated under identical conditions. Replating was repeated every week in the same way.
Semiquantitative RT-PCR Analysis-Total RNA was isolated from 5 ϫ 10 4 FACS-sorted GFP-positive cells using TRIzol reagent (Invitrogen). RT-PCR was performed using SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. The cDNA product (1 l) was resuspended in 20 l of the PCR buffer containing 0.5 units of TaqGold  The PCR products were electrophoresed in agarose gels containing ethidium bromide, and their amounts were analyzed with a Fluor Imager595 and ImageQuant software (Amersham Biosciences).
Measurement of Phosphorylation of Intracellular Signaling Molecules-Phosphorylation of intracellular molecules was assessed using Phosflow technology (BD Biosciences) according to the manufacturer's recommendation. Briefly, cells were fixed with Phosflow Fix Buffer and incubated at 37°C for 10 -15 min. After permeabilization at room temperature for 10 min, cells were washed twice with Phosflow Perm/Wash Buffer and incubated at room temperature for 10 min. After the binding reaction to each antibody, cells were washed once with Phosflow Perm/Wash buffer, resuspended in 500 l of BD Pharmingen stain buffer (BD Bioscience), and then subjected to flow cytometric analysis. All experiments were repeated independently at least three times, and reproducibility was confirmed.
Statistical Analyses-Statistical analyses were performed using Student's t test.

Effects of FIP1L1-PDGFR␣ on the Growth and Survival of
Murine KSL Cells-To investigate the effects of LTKs on the growth, differentiation, and survival of HSCs/HPCs, we constructed bicistronic retrovirus vectors for FIP1L1-PDGFR␣ and TEL-PDGFR␤, which express these cDNAs together with EGFP through the internal ribosome entry site in the infected cells. At first, we introduced these retrovirus vectors into KSL cells. After a 48-h infection, 55-65% of KSLs were found to be GFP-positive in all of transfectants by flow cytometric analysis (data not shown). Next, we isolated retrovirus-infected cells as GFP-positive cells and cultured them in the medium with or without SCF, TPO, FLT3L, and IL-6. As shown in Fig. 1A (left), neither FIP1L1-PDGFR␣ nor TEL-PDGFR␤ further augmented cytokine-dependent growth of KSLs. However, these LTKs enabled KSLs to survive and proliferate under cytokine-deprived conditions at least for 96 h, whereas mock (an empty retrovirus)-infected KSLs rapidly led to apoptosis in this condition (Fig. 1A, right).
Next, we performed colony assays using these retrovirus-infected KSLs. After 2-day retrovirus infection, GFP-positive cells were sorted and plated into methylcellulose medium containing the cytokine mixture (EPO, TPO, SCF, granulocyte colony-stimulating factor, and IL-3), and numbers of colonies were counted after 10 days. As shown in Fig. 1B, the total number of colonies that developed from FIP1L1-PDGFR␣or TEL-PDGFR␤-infected KSLs was increased by 40 -50% as compared with that from mock-infected KSLs. Also, these colonies were larger than those yielded from mock-infected KSLs (data not shown). However, the proportion of CFU-GEMM, CFU-GM, CFU-G, CFU-M, and BFU-E was roughly the same among three transfectants, indicating that these LTKs scarcely influence the lineage commitment and differentiation of KSLs in colony assays performed in this cytokine combination (Fig. 1B).

Enforced Eosinophil Development by FIP1L1-PDGFR␣
still kept colony-forming activities even after the third and fourth plating (FIP1L1-PDGFR␣ versus mock at the third plating, p Ͻ 0.01; at the fourth plating, p Ͻ 0.01), although these activities were rather reduced (Fig. 1C, left). On the other hand, even if FIP1L1-PDGFR␣ or TEL-PDGFR␤ was introduced, CMPs could not form any colony at the third plating, as was the case with mock-infected CMPs (Fig. 1C, right). To evaluate leukemogenic potential of FIP1L1-PDGFR␣-transduced KSLs in vivo, we transplanted these cells into lethally irradiated mice in combination with freshly prepared competitor KSLs. As a result, although none of the mice transplanted with mocktransduced KSLs developed leukemia or MPD, FIP1L1-PDGFR␣-transduced KSLs developed MPD in three mice and acute leukemia in one mouse of five recipient mice within 15 weeks after transplantation (Table 1). However, in agreement with the previous report (16), none of the five recipient mice developed eosinophilic disorders. In addition, none of the 10 mice transplanted with FIP1L1-PDGFR␣-transduced CMPs developed MPD or leukemia (data not shown). Together, these results indicate that FIP1L1-PDGFR␣ can confer the ability of cytokine-independent growth/survival on KSLs and enhance their self-renewal, whereas it cannot immortalize CMPs in vitro or in vivo.
To examine whether Gr-1 ϩ CD125 ϩ cells that developed from FIP1L1-PDGFR␣-transduced KSLs are actually eosinophil precursors, we further cultured these KSLs with a cytokine mixture containing IL-5 for an additional 5 days. As a result, most of FIP1L1-PDGFR␣-transduced but not mock-or TEL-PDGFR␤-transduced KSLs came to possess large granule characteristics of mature eosinophil in the MG staining, which were positive for the eosinostain (Fig. 2B). Furthermore, after 10-day cultures, we examined the mRNA expression of eosinophil-related genes, GATA-1, IL-5R␣, and C/EBP⑀, by RT-PCR analysis using sorted GFP-positive cells. As shown in Fig. 2C, IL-5R␣ and C/EBP⑀ mRNAs were detected only in FIP1L1-PDGFR␣transduced KSLs. Also, GATA-1 mRNA was more intensively expressed in FIP1L1-PDGFR␣-transduced KSLs than in mockor TEL-PDGFR␤-transduced KSLs. These data indicate that Gr-1 ϩ CD125 ϩ cells that developed from FIP1L1-PDGFR␣transduced KSLs can indeed differentiate into mature eosinophils.
Effects of FIP1L1-PDGFR␣ on Differentiation of CMPs, MEPs, and CLPs-It was previously shown that eosinophil precursors stochastically develop from HSCs through MMP, CMP, and GMP (40,41). Therefore, at first, we examined whether FIP1L1-PDGFR␣ can enhance the development of eosinophils from CMPs. For this purpose, we isolated CMPs from murine BM mononuclear cells by FACS using several markers (Fig. 3A). Then we introduced FIP1L1-PDGFR␣ into these cells and cultured them with SCF, IL-6, FLT3L, and TPO for 6 days. As was the case with KSLs, FIP1L1-PDGFR␣ remarkably enhanced the development of Gr-1 ϩ CD125 ϩ cells from CMPs compared with mock cultures (57% versus 6%, p Ͻ 0.01; Fig. 3B).

JOURNAL OF BIOLOGICAL CHEMISTRY 7727
for kinase activation and for transforming properties of FIP1L1-PDGFR␣ (42). To determine the role of FIP1L1 in FIP1L1-PDGFR␣-enhanced eosinophil development, we gen-erated two artificial chimeric constructs, FIP1L1-PDGFR␤ and TEL-PDGFR␣, in which FIP1L1 in FIP1L1-PDGFR␣ and TEL in TEL-PDGFR␤ were completely replaced (Fig. 4A). In addition, we generated retrovirus vectors for constitutively active PDGFR␣ (PDGFR␣V561D and PDGFR␣D842V), which are considered to be causative mutations of gastrointestinal stromal tumors (43) (Fig. 4A). When expressed in a murine IL-3-dependent cell line, Ba/F3, all of the four PDGFR mutants conferred IL-3-independent growth on these cells (data not shown). Also, Western blot analysis demonstrated that these PDGFR mutants phosphorylated various cellular proteins, including themselves (data not shown), indicating that these proteins act as constitutively active tyrosine kinases.
Effects of FIP1L1-PDGFR␣ on the Expression and Activity of Lineage-specific Transcription Factors in KSLs-To further clarify the mechanism through which FIP1L1-PDGFR␣ enhanced eosinophil development, we analyzed the effects of FIP1L1-PDGFR␣ on the expression of GATA-1, GATA-2, C/EBP␣, and PU.1, all of which have been reported to be key transcription factors for eosinophil development (45)(46)(47). To detect the changes in the expression of these factors that precede the phenotypic change, we isolated mRNA from sorted GFP-positive KSLs after 48-h retrovirus infection and performed semiquantitative RT-PCR analysis, since an apparent phenotypic change was not observed until 4 days ( Fig. 2A, top). As shown in Fig. 6A, although the expression of PU.1 was not so different among three transfectants, FIP1L1-PDGFR␣ augmented the expression of C/EBP␣ (p Ͻ 0.01) and GATA-1 (p Ͻ

DISCUSSION
In this study, we found that TEL-PDGFR␣, but not FIP1L1-PDGFR␤, PDGFR␣D562V, or PDGFR␣D842V, promoted eosinophil development from KSLs as efficiently as FIP1L1-PDGFR␣. This result indicates that constitutive TK activity transmitted from chimeric structure of PDGFR␣ is necessary to augment eosinophil development. In agreement with our finding, novel mutations identified in CEL were restricted to the chimeric form of PDGFR␣ (i.e. KIF5B-PDGFR␣ formed by t(4; 10)(q12;p11), STRN-PDGFR␣ by t(2;4)(p24;q12), and ETV6-PDGFR␣ by t(4;12)(q2?3;p1?2)). As for the roles for downstream signaling molecules, the current results indicate that Ras/MEK and p38 MAPK play essential roles in FIP1L1-PDGFR␣-induced eosinophil development. However, this finding seems to be inconsistent with the fact that Ras/MEK is activated by various LTKs and normal hematopoietic growth factors. As for this reason, because FIP1L1-PDGFR␣ more intensely activated MEK/ERK and p38 MAPK than TEL-PDGFR␤, we speculated that leukemogenic signals transmitted from chimeric PDGFR␣ would be quantitatively and qualitatively different from those from wild type TKs or other LTKs, thereby specifically promoting eosinophil development. In addition to the regulation of neoplastic cell proliferation, ERK has also been implicated in the control of signaling cascades associated with eosinophilia in asthma. Duan et al. (48) reported that an MEK inhibitor dramatically inhibited OVAinduced lung tissue eosinophilia and airway hyperresponsiveness. Also, p38 MAPK is important for the induction of eosinophilia and function of terminal differentiated eosinophils in allergic airway inflammation (49,50). In addition, our data suggest that p38 MAPK would regulate eosinophil development at the early stage of hematopoiesis. Further studies to elucidate the crucial signal transduction mechanisms that control eosinophil development will provide a better rationale for the design of drug therapy not only for FIP1L1-PDGFR␣-associated HES/ CEL but also for allergic inflammation.
Our in vitro studies showed that FIP1L1-PDGFR␣ confers cytokine independence on KSLs and enhances their self-renewal activity, whereas it did not immortalize CMPs. In addition, although FIP1L1-PDGFR␣-transduced KSLs caused MPD in recipient mice, FIP1L1-PDGFR␣-transduced CMPs did not. These results indicate that FIP1L1-PDGFR␣ cannot confer selfrenewal activity on CMPs and that the genetic alternation of FIP1L1-PDGFR␣ that causes CEL/HES occurs at an HSC level but not at a CMP level. In addition, we confirmed that mature eosinophils were generated from FIP1L1-PDGFR␣-transduced KSLs in the presence of IL-5, indicating that FIP1L1-PDGFR␣ does not impair terminal differentiation of eosinophils. Also, when expressed in MEPs or CLPs, FIP1L1-PDGFR␣ brought about lineage conversion to eosinophil lineage. Together, these results suggest that, although LSCs harboring FIP1L1-PDGFR␣ derived from HSCs would continuously produce an excess number of mature eosinophils, a part of the eosinophils might be derived from FIP1L1-PDGFR␣-harboring MEPs or CLPs.
In a previous report, FIP1L1-PDGFR␣-transduced HSCs/ HPCs caused myeloproliferative disorder in the recipient mice like BCR-ABL-or TEL-PDGFR␤-transduced KSLs (16,51,52), which was rather different from simple eosinophilia observed in human HES/CEL. Also in our transplantation experiment, none of the five mice transplanted with FIP1L1-PDGFR␣-expressing KSLs developed eosinophilic disorders. However, we also observed that, whereas FIP1L1-PDGFR␣-introduced KSLs differentiated up to IL-5R␣ ϩ eosinophil precursors under the cultures without IL-5, supplement of IL-5 let these IL-5R␣ ϩ cells undergo eosinophilic terminal differentiation. In accord with this hypothesis, Yamada et al. (52) reported that transplantation of FIP1L1-PDGFR␣-transduced HSCs/HPCs obtained from IL-5 transgenic mice resulted in marked eosinophilia resembling HES/CEL in the recipient mice. Since p210BCR-ABL-transduced HSCs/HPCs did not cause eosinophilia even in the presence of IL-5 overexpression in the recipient mice, the induction of eosinophilia was attributable to FIP1L1-PDGFR␣, Together with our results, these lines of evidence suggest that, although FIP1L1-PDGFR␣ is a major etiologic factor causing Enforced Eosinophil Development by FIP1L1-PDGFR␣ eosinophilia, it is not sufficient to induce HES/CEL but requires additional events, such as IL-5 overexpression. In fact, some patients with FIP1L1-PDGFR␣-associated HES were complicated with T-cell lymphoma (53)(54)(55). The frequency of FIP1L1-PDGFR␣-induced HES/CEL was not as high (about 10%) as initially reported. However, similar LTK is supposed to be involved in the pathogenesis of HES/CEL, because imatinib is effective in some patients who do not have a FIP1L1-PDGFR␣ mutation (56). Also, a significant proportion of patients with HES/CEL have abnormal T-lymphocyte populations, such as CD3 ϩ CD4 Ϫ CD8 Ϫ and CD3 Ϫ CD8 ϩ T cells, which secret high levels of IL-5 (57). Currently, HES is categorized into two groups, "myeloproliferative variant" and "T-cell-mediated HES," and these groups are thought to be independent of each other (58,59). However, because T-cell differentiation might be perturbed by FIP1L1-PDGFR␣, it may be meaningful for the better understanding of the pathogenesis of HES/CEL to clarify the relationship between these two groups.
Iwasaki et al. (60) isolated eosinophil progenitors from murine BM, and they concluded that eosinophil developmental pathway would diverge from neutrophils and monocytes at the GMP stage. The lineage commitment of HSCs/HPCs and subsequent lineage-specific differentiation are crucially regulated by lineage-specific transcription factors, such as GATA-1, GATA-3, PU.1, C/EBP␣, and C/EBP⑀. Among them, GATA-1 and PU.1 are known to antagonize each other and induce differentiation to erythroid/megakalyocyte or myeloid lineage, respectively (61)(62)(63). The CEBP family (CEBP␣ and CEBP⑀) is essential for the differentiation to myeloid lineage (64 -66). FOG (Friend of GATA) and C/EBP␤ regulate the eosinophil lineage induction antagonistically (67). Furthermore, enforced expression of C/EBP␣ converts MEPs to eosinophils (68), and expression of PU.1 converts them to GMPs (61,67). Also, forced expression of GATA-1 in myeloid cells induces the formation of either MEPs or eosinophils, depending on the concentration of the factor (69). In addition, it was recently reported that C/EBP␣ expression followed by GATA-2 expression in GMPs is critical for eosinophil lineage specification (46). However, it is plausible that the mechanism of lineage commitment in leukemic cells is somewhat different from that in normal hematopoietic cells. In this study, we found that FIP1L1-PDGFR␣ enhanced the expression of GATA-1, GATA-2, and C/EBP␣ and suppressed PU.1 expression. Also, FIP1L1-PDGFR␣ suppressed transcription activities of PU.1. These results suggest that LTKs can influence the lineage commitment of HSCs/HPCs and subsequent differentiation by modifying the expression and activity of lineage-specific transcription factors.
In conclusion, we here found that FIP1L1-PDGFR␣ can enhance eosinophil development from HSCs/HPCs through the MEK/ERK and p38 MAPK cascades by controlling the expression and activity of lineage-specific transcription factors. Furthermore, as far as we explored, this is the first report providing evidence that LTK has an ability to convert the lineage of committed progenitor cells. Further studies based on these findings would undoubtedly provide more useful information to understand the pathophysiology of various hematologic malignancies caused by LTKs.