Differential STAT5 Signaling by Ligand-dependent and Constitutively Active Cytokine Receptors*

Many leukemia and cancer cells exhibit constitutive activation of STAT5, which was suggested to provide an anti-apoptotic advantage. Transformation of cytokine-dependent hematopoietic cells, such as Ba/F3 cells to autonomous growth and tumorigenicity equally results in selection for constitutive activation of STAT5. We compared STAT5 signaling between erythropoietin(Epo)-dependent cells and cells that were transformed by oncogenic activation of the erythropoietin receptor (EpoR) by coexpression of the gp55-P envelope protein of the spleen focus forming virus or by expression of the R129C constitutively active EpoR mutant. In transformed cells it was mainly STAT5B that was constitutively activated. In contrast, Epo stimulation activated both STAT5A and STAT5B. In transformed cells, chromatin immunoprecipitation (ChIP) showed STAT5 to be physically bound to promoters of STAT5 target genes, such as BclXL, and to be able to promote transactivation of the BclXL promoter in a constitutive fashion. Sequencing of native sequences after ChIP with anti-STAT5 antibodies in Epo-dependent and -transformed cells indicated that in gp55-transformed cells, STAT5B bound in the chromatin not only to N3 high affinity, but also to low affinity N4 GAS sites. Transactivation for N3 GAS sites in luciferase reporters was specific to gp55 transformation. Because we also found preferential constitutive STAT5B activation after transformation of cells by a truncated form of the G-CSF-R that produces severe neutropenia (Kostmann syndrome) and favors leukemia in humans, we discuss the potential role of STAT5B in oncogenic transformation of hematopoietic cells.

Many leukemia and cancer cells exhibit constitutive activation of STAT5, which was suggested to provide an anti-apoptotic advantage. Transformation of cytokinedependent hematopoietic cells, such as Ba/F3 cells to autonomous growth and tumorigenicity equally results in selection for constitutive activation of STAT5. We comparedSTAT5signalingbetweenerythropoietin(Epo)dependent cells and cells that were transformed by oncogenic activation of the erythropoietin receptor (EpoR) by coexpression of the gp55-P envelope protein of the spleen focus forming virus or by expression of the R129C constitutively active EpoR mutant. In transformed cells it was mainly STAT5B that was constitutively activated. In contrast, Epo stimulation activated both STAT5A and STAT5B. In transformed cells, chromatin immunoprecipitation (ChIP) showed STAT5 to be physically bound to promoters of STAT5 target genes, such as Bcl XL , and to be able to promote transactivation of the Bcl XL promoter in a constitutive fashion. Sequencing of native sequences after ChIP with anti-STAT5 antibodies in Epo-dependent and -transformed cells indicated that in gp55-transformed cells, STAT5B bound in the chromatin not only to N3 high affinity, but also to low affinity N4 GAS sites. Transactivation for N3 GAS sites in luciferase reporters was specific to gp55 transformation. Because we also found preferential constitutive STAT5B activation after transformation of cells by a truncated form of the G-CSF-R that produces severe neutropenia (Kostmann syndrome) and favors leukemia in humans, we discuss the potential role of STAT5B in oncogenic transformation of hematopoietic cells.
Signal transducers and activators of transcription (STATs) 1 are latent cytoplasmic transcription factors that transduce sig-nals from activated cell surface receptors to the nucleus (1,2). Binding of a cytokine to its receptor leads to activation of Janus kinases and phosphorylation of the receptor itself, creating docking sites for SH2 domains containing proteins such as STATs. After recruitment, STAT proteins are phosphorylated, dimerize as homo-or heterodimers, and translocate to the nucleus where they bind to consensus sequences of the chromatin and modulate transcription (2). Association on the chromatin of different STAT complexes with regulating proteins/ transcription factors contributes to the specificity of transcriptional regulation. STAT5 (STAT5A and/or STAT5B) is activated by a broad range of cytokines like prolactin (3), Epo (3)(4)(5), interleukin (IL) 2 (6), or IL3 (3). STAT5A knockout mice exhibit a profound defect in prolactin function (7) and their splenocytes show defective IL2-induced expression of the IL2 receptor ␣-chain (8). STAT5B knockout mice are deficient in growth hormone function (9) and show defects in the pathway of IL2 (10), while STAT5A/B knockout mice harbor defects in prolactin (11), growth hormone (11), IL2 functions (12), T cell proliferation (12), and fetal anemia (13). Constitutive activation of STAT5 has become a hallmark of a number of hematopoietic malignancies (14), after it had been detected in T cells transformed by HTLV-I (15,16), megakaryocytic leukemia (17), chronic myelocytic leukemia (18,19), erythroleukemia (20), acute myelocytic (19,(21)(22)(23), and lymphocytic leukemia (21,24). In addition, constitutive activation of STAT5 has been reported in acute lymphoblastic leukemia and atypical chronic myelogenous leukemia (25,26) as a consequence of the chromosomal translocation TEL/JAK2 (27,28).
Epo and its receptor (EpoR) are crucial for red cell formation and are especially required for the proliferation, survival, and differentiation of erythroid progenitors into mature red blood cells (29 -31). Epo binding to the EpoR leads to strong activation of STAT5 (4,5,(32)(33)(34). Among STAT5regulated genes, the anti-apoptotic Bcl XL plays a major role in erythroid cell development (13,(35)(36)(37), while cytokineinduced SH2 protein (Cis) (38,39) and suppressors of cytokine signaling, SOCS1 and 3 (40), are the most important negative regulators of the pathway. Prolonged activation of STAT5 after Epo stimulation is characteristic of benign erythrocytosis, induced by hypersensitivity to low levels of Epo, because of a receptor heterozygous truncation (41). However, such hypersensitivity to Epo or transgenic expression of Epo has not been associated with induction of leukemia (42). Pathologically, the murine EpoR can be activated by coexpression of the envelope protein gp55 of the polycytemic (P) strain of the spleen focus forming virus (SFFV) (43). gp55-P (gp55) is able to bind to the EpoR (44) via specific interactions between the transmembrane domains (45) and induce proliferation of Epo-dependent cell lines in an Epoindependent way (46 -48). Expression of gp55-P triggers an expansion of erythroid progenitors, which upon subsequent acquisition of genetic defects leads to erythroleukemia (49). In vitro, STAT5 tyrosine phosphorylation of gp55-transformed cells is constitutive (50,51). Yet another mechanism of activation of the EpoR in a ligand-independent manner is represented by the R129C point mutation in the extracellular domain. This mutation leads to disulfide-linked dimeric receptor (52), which renders cells ligand-independent for their growth (53). Retroviral expression of the constitutively active EpoR R129C mutant in mice leads to leukemias of different lineages (54). Constitutive STAT5 DNA binding was detectable in the nucleus of cells expressing the EpoR R129C (55). We investigated STAT5 signaling in cells stimulated by Epo or transformed to autonomous growth by gp55 coexpression or by the R129C point mutation. We found that in cytokineindependent cells the constitutively active pool of nuclear STAT5 is essentially STAT5B and could be resolved from another pool, which is responsive to Epo stimulation (STAT5A and STAT5B). In transformed cells, STAT5 was physically bound to the STAT binding sites in the promoters of known STAT5 target genes, such as Bcl XL and able to promote transactivation of the Bcl XL promoter. Chromatin immunoprecipitation and sequencing of native sequences bound by STAT5 (STAT5A and STAT5B), STAT5A, or STAT5B, and transactivation assays of different known STAT5-targeted promoters showed that in gp55-transformed cells, STAT5B binds in the chromatin not only to ␥-interferon-activated sites (GAS, N3 high affinity), but also to low affinity N4 GAS sites. This binding to N4 sites of STAT5 in the gp55-transformed cells was not seen after Epo treatment. Among the STAT5 target genes we found that Cis, which is a negative regulator, is regulated differently than Bcl XL . To assert the biological relevance of this finding, we studied STAT5 activation in Ba/F3 cells transformed with a truncated form of the granulocyte colony-stimulated factor receptor (GCSFR) involved in severe neutropenia (Kostmann syndrome) (56 -58) and acute myeloid leukemia. We found that in Ba/F3 cells transformed by the truncated GCSFR STAT5B and not STAT5A was also preferentially constitutively activated. We suggest that these biochemical and transcriptional differences are important for oncogenic transformation of hematopoietic cells.

EXPERIMENTAL PROCEDURES
Cell Culture-The Epo-dependent mouse erythroleukemia HCD57 cells (48) were routinely cultivated in Iscove's modified Dulbecco's medium (IMDM, Invitrogen Life Technologies, Inc.), 20% (v/v) fetal calf serum (FBS, Perbio), 1 unit/ml Epo (from Amgen), 10 l/liter ␤-mercaptoethanol. Ba/F3 cells expressing the EpoR were obtained from the mouse IL3-dependent pro-B Ba/F3 cells (59), after transfection with pcDNA3 coding for EpoR and vector coding for the antibiotic resistance gene and selection for antibiotic resistance (clone IH) or after transduction with a bicistronic retrovirus encoding for the EpoR and GFP, and selection for high levels of EpoR by cell sorting for GFP expression (pool Ba/F3-EpoR) (45). Cells were routinely cultivated in RPMI 1640 (Invitrogen, Life Technologies, Inc.), 10% FBS, and 1 unit/ml Epo. Ba/F3 IH55 (IH55) and HCD57-gp55 cells were obtained after transfection of IH and HCD57 cells, respectively, with a gp55-P-encoding construct and selection for growth in the absence of Epo. EpoR R219C (R129C) cells were obtained after transfection of the Ba/F3-EpoR cells with the EpoR R129C mutant and selection for growth in the absence of Epo. The HA-tagged wild-type GCSFR and HA-tagged GCSFR (containing a cytoplasmic Q7183 Stop mutation) cDNAs were cloned into the pMX-IRES-GFP bicistronic retroviral vector (60). Retroviral supernatants were generated by transient transfection of BOSC cells and used to infect Ba/F3 cells growing in IL3. Cells were FACS-sorted for similar GFP levels. All cell types were cultivated at 37°C, in an 8% CO 2 , 92% air atmosphere. For stimulation with Epo or GCSF, cells were washed, starved for 4 h in medium without FBS, with 0.5% bovine serum albumin (Roche Applied Science), resuspended in a minimum volume of starvation medium for 20 min at 37°C for equilibration before stimulation.
Immunoblotting-Protein samples (50 g) were migrated on a 7.5% SDS-polyacrylamide gel and transferred onto a nitrocellulose membrane (Amersham Biosciences). The membrane was blocked with TBST (Sigma) containing 5% bovine albumin (Sigma) for 1 h at room temperature and subjected to immunoblotting. The STAT5 antibodies raised against the C terminus of the murine STAT5B (C17, Santa Cruz Biotechnology), STAT5A, STAT5B (kind gifts of Dr. Warren Leonard, Laboratory of Molecular Immunology, NHLBI, National Institutes of Health), or the tyrosine 694-phosphorylated STAT5 (Cell Signaling Technology) were used at a dilution of 1:1000 in an overnight incubation at 4°C. Anti-rabbit IgG horseradish peroxidase-linked whole antibody (Amersham Biosciences) was used at a dilution of 1:4000. Signals were revealed with the Super Signal West Pico Chemiluminescent Substrate (Pierce). For reprobing, blots were incubated in a stripping solution containing 62.5 mM Tris (Invitrogen) pH 6.8, 2% SDS (Sigma), 0.1 M ␤-mercaptoethanol at 60°C, washed extensively in water, blocked, and subjected to immunoblot.
Immunoprecipitation-10 7 cells were washed in cold phosphate-buffered saline and resuspended for 20 min at 4°C in cold lysis buffer containing 50 mM Tris, 150 mM NaCl, 1 mM EDTA, 15% glycerol, 1% Nonidet P-40 solution, and protease inhibitors. After centrifugation, 2 l of anti-STAT5A or anti-STAT5B antibody were added, and cell extracts were incubated overnight at 4°C. For precipitation, 60 l of protein A-Sepharose (Amersham Biosciences) were added, and extracts were incubated for 3 h. After washes with the lysis buffer, proteins were resuspended in Laemmli SDS-PAGE buffer and denatured 10 min at 100°C.
Chromatin Immunoprecipitation (ChIP)-2 ϫ 10 8 cells were starved and stimulated for 15 min. Then, chromatin immunoprecipitation of STAT5 binding sites was performed as described (61) with protein A-Sepharose instead of Staph A cells and with the STAT5-C17 (Santa Cruz Biotechnology), STAT5-N20 (Santa Cruz Biotechnology), STAT5A, STAT5B antibodies or a non-immune rabbit serum. 10 l of chromatin supernatants were saved before ChIP as INPUT (no antibody, positive control). A MOCK sample (no chromatin, negative control) containing dilution buffer instead of chromatin was also included, and sample volumes were adjusted with the dilution buffer. At the end of the protocol, precipitated DNA samples were resuspended in 30 l, while total INPUT samples (positive controls) were diluted 200 times. PCR was performed using primers amplifying part of the Bcl XL promoter containing the STAT5 binding sites: 5Ј-gtc gcc gga gat aga ttt ga-3Ј and 5Ј-aag gtc tga gtc cgg gtt ct-3Ј or part of the Cis promoter containing the STAT5 binding sites 5Ј-gag ctc ggt cgg tct aga tgc tcgtct-3Ј and 5Ј-ctc gag aac agc ttg gaa gga caa gg-3Ј. For the cloning step, DNA fragments were blunt-ended with mung bean nuclease (Promega), and oriented double-stranded linkers were linked to their extremities. Then a PCR amplification was performed with Taq polymerase (Takara) using primers related to the linker, and amplified DNA fragments were cloned into the pCR-II-Topo vector (Topo TA cloning kit, Invitrogen). Positive colonies were sequenced and analyzed for the presence of STAT consensus sites. As a control for the frequency of STAT binding sites, chromatin was prepared from Ba/F3 cells, sonicated, and DNA was purified to be cloned as the ChIP samples, in the pCR-II-Topo vector.
Dual Luciferase Assays-The luciferase reporter constructs were pL-HRE (62, 63) and a pGL2-TK promoter construct into which test promoter sequences were introduced. The TK promoter luciferase construct was kindly provided by Dr. Frederic Lemaigre, Hormone and Metabolic Research Unit, Université Catholique de Louvain, Brussels. 8 ϫ 10 6 cells were starved for 4 h, transiently transfected by electroporation with 40 g of pBcl XL -TK, pLHRE, pCis-TK, N4N3-TK, N4-TK, or N3-TK, and 10 g of pRLTK, seeded into 24-well plates at 1 ϫ 10 6 cells/well, and stimulated with 1 unit/ml Epo for 2 h. For the pBcl XL -TK plasmid, the part of the Bcl XL promoter containing the 2 consensus sites for STAT5 was obtained by PCR using the same primers as in ChIP, cloned into the pCR-II-Topo vector, and subcloned into the pGL2-TK promoter vector. For the pCis-TK, the part of the Cis promoter containing the two double STAT5 binding sites was obtained by PCR using the primers used for ChIP and cloned directly into the pGL2-TK promoter vector. For the N4N3-TK plasmid, the part of the ␤-casein promoter containing the two STAT5 binding sites was obtained by PCR with the following primers: 5Ј-gagctcccttaaccagcttctgaattgc-3Ј and 5Ј-ctcgaggtcctatcagactctgtgaccg-3Ј, cloned into the pGL2-TK and named N4N3-TK. For the N4-TK plasmid, only the part of the ␤-casein promoter containing the STAT5 binding site TTCTTGGGAA was cloned using the primers 5Ј-gag ctcggtcggtctagatgctcgtc-3Ј and 5Ј-ctcgagcgtgattagaaaatggtttctttc-3Ј. For the N3-TK plasmid, only the part of the ␤-casein promoter containing the STAT5 binding site TTCTTGGAA was cloned using the primers 5Ј-gagctcgaaagaaaccattttctaatcacg-3Ј and 5Ј-ctcgaggtcctatcagactctgtgaccg-3Ј. The pLHRE-Luc (62) and pRLTK (Promega) have been previously described. The luciferase activity was evaluated with the dual luciferase reporter assay system (Promega) using the pRLTK signal inducing as internal control.
Reverse Transcriptase-PCR-RNA were prepared from 5 ϫ 10 6 cells, starved, and stimulated with 25 units/ml Epo for 2 h using TRIzol reagent (Invitrogen). Then cDNA was prepared from 1 g of total RNA and used in a PCR amplification with primers amplifying part of the Bcl XL transcript: 5Ј-ctg aat cgg aga tgg aga cc-3Ј and 5Ј-tgg gat gtc agg tca ctg aa-3Ј, part of the Cis transcript: 5Ј-cac att cct cgt gcg aga ca-3Ј and 5Ј-cag agc tgg aac ggg tac tg-3Ј, and part of the ␤-actin transcript: 5Ј-atg gat gac gat atc gct gc-3Ј and 5Ј-gct gga agg tgg aca gtg ag-3Ј. PCR products were separated on a 2% agarose gel.
Whole Cell Extracts-10 7 cells were washed in cold phosphate-buffered saline and resuspended in immunoprecipitation lysis buffer with protease inhibitors for 20 min at 4°C. After centrifugation, proteins were resuspended in Laemmli SDS-PAGE buffer and denatured 10 min at 100°C.
Cell Titer Glo TM Luminescent Cell Viability Assay-1.5 ϫ 10 5 cells/ well were washed and placed in RPMI 10% FBS with or without GCSF for 2 days, in a 96-well plate. Cell numbers were counting with the Cell Titer Glo TM Luminescent cell viability assay (Promega).

STAT5B Is Constitutively Translocated and Active in the
Nucleus of gp55 and R129C-transformed Cells-Cytokine-de-pendent hematopoietic cells, such as the Ba/F3 (59) are not tumorigenic, but upon their transformation to autonomous growth (cytokine-independence) they become tumorigenic (53) and can be used as a model for some of the steps involved in oncogenesis. Tyrosine-phosphorylated transcriptionally active STAT5 was previously detected in the nucleus of gp55-and R129C-transformed Ba/F3 cells (50,51,55).
We first measured by immunoblotting with anti-phosphotyrosine 694 STAT5, anti-STAT5B, anti-STAT5A (kind gifts of Dr. Warren Leonard, Laboratory of Molecular Immunology, NHLBI, National Institutes of Health), and anti-STAT5 C17 antibodies the presence of STAT5 in the nucleus of starved cells that had been stimulated or not with Epo (Fig. 1A). STAT5 was found to be translocated in the nucleus and phosphorylated upon stimulation with Epo in all cell lines but also in nonstimulated transformed cells. The same result was observed when anti-STAT5B or anti-STAT5 C17 was used. The anti-STAT5 C17 antibody detects both STAT5A and STAT5B proteins and can discriminate between non-phosphorylated (faster mobility) and phosphorylated STAT5B (slower mobility) and can also detect the phosphorylated STAT5A at a slower mobility than that of phosphorylated STAT5B. Utilization of specific anti-STAT5A antibodies showed STAT5A to be phosphorylated and translocated in the nucleus upon Epo stimulation in IH, Ba/F3EpoR, and IH55, but we could not detect its constitutive activity in IH55 cells. There was a very low activation of STAT5A in R129C cells.
Next, immunoprecipitation in whole cell lysates was performed with either an anti-STAT5A or an anti-STAT5B-specific antibody. Immunoblotting was performed using the anti-P-Tyr 694 STAT5 antibody (Fig. 1B). Immunoprecipitation with the anti-STAT5B antibody demonstrated STAT5B activation in all cell lines upon Epo addition. Transformation to autonomous growth via coexpression of gp55 and EpoR or via expression of R129C resulted in significant constitutive levels of STAT5B in the nucleus. Immunoprecipitation with the anti-STAT5A antibody revealed that STAT5A is highly phosphorylated in an Epo-dependent manner, but no constitutive STAT5A activation was detectable in non-stimulated gp55 transformed cells and only marginally in non-stimulated R129C (apparent in overexposing blots in Fig. 1B). Thus, it seems that transformation to Epo independence leads to constitutive activation of mainly STAT5B. This effect may be due to a specific activation of STAT5B by constitutively active receptors, to selection for FIG. 1. Constitutive nuclear translocation of STAT5B in hematopoietic cells transformed to growth factor independence. A, cells were starved for 4 h and stimulated with 10 units/ml Epo for 15 min. Nuclear extracts were prepared as described, and 50 g of nuclear proteins were subjected to immunoblotting using anti-P-Tyr 694 STAT5 antibody. The same membrane was reprobed after treatment with ␤-mercaptoethanol with anti-STAT5A, anti-STAT5B, and anti-STAT5 C17 antibodies. IH and Ba/F3-EpoR cells are two different stable transfectants (cell populations) of Ba/F3 cells expressing the murine EpoR. IH55 cells were obtained after retroviral transduction of IH cells with a gp55-P-encoding construct, and selection for growth in the absence of Epo. R219C cells was obtained after retroviral transduction of the Ba/F3-EpoR cells with the R129C mutant and selection for growth in the absence of Epo. B, cells were starved for 4 h and stimulated with 1 unit/ml Epo for 15 min. STAT5 proteins were immunoprecipitated either with anti-STAT5A or anti-STAT5B antibody, subjected to immunoblotting using anti-P-Tyr 694 STAT5 antibody. The same blot was reprobed with anti-STAT5A, anti-STAT5B, and anti-STAT5 C17 antibodies.
STAT5B activation by the selection to autonomous growth, or could result from a quantitative difference in the ability of EpoR to activate STAT5A and STAT5B. The latter was not the case, because low Epo concentrations were potent in activating STAT5A (data not shown).
We evaluated STAT5 binding activity to native chromatin in living cells by ChIP with anti-STAT5 antibodies. Cytokine-dependent and -independent cells were treated with 100 units/ml Epo or with culture medium for 15 min. Chromatin extracts were prepared, and STAT5-immunoprecipitated DNA fragments (with anti-STAT5 C17 antibody) were purified and used as template in PCR amplification with primers designed to amplify the part of the Bcl XL promoter responsive to STAT5 ( Fig. 2A). In IH cells, STAT5 binding onto the Bcl XL promoter was not detected without previous Epo stimulation, whereas in IH55 and R129C cells, STAT5 was found to be physically bound to the promoter constitutively. In agreement with our data on activation and nuclear localization of STAT5B in transformed cells, the same ChIP results were obtained using anti-STAT5B antibody or the C17 antibody. Concerning the ChIP with the anti-STAT5A antibody, using the same number (40) of PCR cycles for DNA amplification, we could detect DNA amplification only with the input samples. We therefore had to increase the number of PCR cycles (to 45 cycles), which led to DNA amplification in all tested cells including the non-stimulated IH cells, but the intensity of the signal was stronger in Epostimulated cells.
The transactivation capacity of STAT5 was measured using a dual reporter luciferase assay. To this end, cells were starved, transiently transfected with a STAT5 activity reporter plasmid (pBcl XL -TK or pLHRE), and an internal control plasmid, seeded into wells and stimulated with 1 unit/ml Epo for 2 h (Fig. 2B). The region of the Bcl XL promoter responsive to STAT5 (pBcl XL -TK) (13) and the synthetic construct containing multiple STAT5 consensus sites (pLHRE) are both recognized and activated upon Epo stimulation in IH (Fig. 2B) and Ba/F3-EpoR (data not shown) cells and are constitutively activated in IH55 (Fig. 2B) and R129C cells (data not shown). Although similar levels of phosphorylation of STAT5 were detected in Fig. 1, Epo-induced transcriptional activity in IH cells was lower than the constitutive transcriptional activity in IH55 cells. This difference might be due to starvation of cells since, during starvation from Epo, signaling by the EpoR induced by binding to gp55 cannot be blocked. Alternatively, other signaling pathways may converge in IH55 cells on STAT5-dependent transcriptional activation. Taken together, these results indicate that STAT5 is constitutively activated in the nucleus of gp55-and R129C-transformed cells.
Constitutively Active STAT5 Exhibits a Longer Half-life in the Nucleus of gp55-transformed Cells Compared with Ligandactivated STAT5-The constitutive localization of activated STAT5B in the nucleus of gp55 or R129C-transformed cells could be due either to continuous activation of STAT5, whenever a new STAT5 protein is translated, and/or by stabilization of the activated STAT5 in the nucleus by inhibition of dephosphorylation, recycling, and/or degradation. To discriminate between these hypotheses, a kinetic measurement of STAT5 expression in the nucleus was performed using inhibitors of activation of the Jak/STAT pathway and protein synthesis. Staurosporine addition blocks cytokine receptor signaling (64,65), and cycloheximide blocks new receptor protein synthesis. Cells were starved and stimulated or not with a pulse of Epo for 15 min. Epo was removed, and cells were chased by treating with staurosporine and cycloheximide for 45, 105, or 225 min. In this way, the STAT5 activated by the Epo pulse can be followed in the absence of newly synthesized or activated STAT5 molecules. Immunoblotting experiments revealed the levels of phosphorylated (activated) and non-phosphorylated (inactive) STAT5B after activation and chase STAT5 (Fig. 3A). In Epo-stimulated IH cells, the activated STAT5B form disappears as early as 2 h after Epo stimulation (Fig. 3B). Interestingly, in non-stimulated IH55 cells, the activated form of STAT5 exhibited a longer half-life in the nucleus. In R129C cells, disappearance of activated STAT5 is slightly decreased at

FIG. 2. Transcriptional activity of STAT5 in hematopoietic cells transformed to growth factor independence.
A, cells were starved and stimulated or not with 100 units/ml Epo for 15 min. Chromatin extracts were prepared and subjected to a STAT5 chromatin immunoprecipitation as described under "Experimental Procedures" using the anti-STAT5 C17 (ChIP STAT5 C17), the anti-STAT5A (ChIP STAT5A), the anti-STAT5B (ChIP STAT5B) antibodies, or a non-immune rabbit serum (ChIP control). Finally, purified immunoprecipitated DNA was used as template for a PCR amplification of part of the Bcl XL promoter responsive to STAT5 (see the murine Bcl XL promoter sequence, STAT5 consensus sites are in bold, oligonucleotides used for PCR are underlined). INPUT represents positive controls and MOCK the negative control as established under "Experimental Procedures." B, cells were starved, electroporated with the Bcl XL -TK or the plHRE-Luc plasmids and the internal control pRL-TK plasmid and stimulated with 1 unit/ml Epo or left untreated for 2 h. Cells were lysed and subjected to a luciferase assay. The STAT5 reporter activity is given relative to the internal plasmid activity. Luciferase ratio (rlu) represents the mean of three independent experiments Ϯ S.D. 60 and 120 min when compared with the control Epo-stimulated Ba/F3 EpoR (data not shown) but the levels of remaining active STAT5 were equivalent after 240 min. These results lead us to hypothesize that the constitutively active nuclear STAT5B pool is more stable and might be biochemically different from the Epo-activated STAT5. Next we asked whether this constitutive activation might have an effect on STAT5 function as a transcription factor.
Effects of STAT5 Constitutive Activation on Target Gene Expression-We first compared gene expression of different known STAT5 targeted genes in normal and transformed cells that were stimulated with Epo or left untreated for 2 h. In all cell types, Epo stimulation resulted in the increase of the STAT5 target gene expression (Fig. 4A, left panel) and, as expected, constitutive STAT5 activation is linked to constitutive activation of target genes such as Bcl XL (13) and Cis (39) as well as SOCS3, ␤-casein, and c-Myc (not shown). Second, gene expression was also assessed in Epo-dependent cells maintained continuously in Epo in culture, or in ligand-independent cells maintained in normal medium with fetal calf serum without starvation. The Bcl XL transcript appears to be expressed at higher levels in transformed cells than in steady-state liganddependent cells (Fig. 4A, right panel). This result was also obtained for SOCS3, ␤-casein, and c-Myc (not shown). Regulation of Cis appears to contrast with regulation of Bcl XL . In non-stimulated transformed cells, there is detectable low basal induction of the mRNA for Cis but the Cis transcript is induced to lower extents in all transformed cells (in the absence of ligand) than in parental normal cells stimulated with Epo (Fig.  4A). Under steady-state conditions, it appears that expression of Cis is slightly blunted as well, when transformed cells are compared with parental cells maintained continuously in medium containing 1 unit/ml Epo. In order to discriminate between a transcriptional or a post-transcriptional regulation of the Cis mRNA expression, we compared transcription activity of the corresponding promoters in IH and IH55 cells (Fig. 4B). Only the segments of the promoters that contain STAT5-binding sites were cloned into the pGL2-TK luciferase reporter construct. The transactivation activities of the Bcl XL (Fig. 4B), and the ␤-casein (data not shown) promoters were increased in cytokine-dependent cells upon Epo stimulation, and this increase was significantly enhanced in gp55-transformed cells. In contrast, Epo stimulation resulted in strong induction of the activity of the Cis promoter in cytokine-dependent cells, which was significantly blunted in transformed cells (IH55). Comparable results were obtained in R129C-transformed cells and in HCD57-gp55 cells (data not shown). Thus, we hypothesize that in transformed cells, transcription through the Cis promoter is regulated differently from the other STAT5-targeted promoters such as the Bcl XL or ␤-casein promoters, probably because of a difference in STAT5 binding and/or activity on this promoter. Such differential regulation may be important for the ability of the signal promoted via STAT5 constitutive activation to support cell growth in the absence of Epo since Cis is a negative regulator of EpoR signaling (38,39). Whether selection for decreased Cis induction by Epo is caused by the intrinsic STAT5-DNA binding properties or by other STAT5-associated or independent proteins binding to the Cis promoter remains unclear. By chromatin immunoprecipitation with anti-STAT5 C17 antibodies, we could see no significant difference in physical binding of STAT5 to the two promoters (Bcl XL and Cis) (Fig. 4C).
Constitutively Active STAT5 and Epo-activated STAT5 Interact Differently with the Chromatin and Exert Different Transcriptional Effects-The optimal binding sites for STAT5A and STAT5B have been previously defined (66). Specifically, STAT5 is able to bind to the STAT consensus binding sites TTC NNNN GAA (N4) with low affinity and to TTC NNN GAA (N3) with high affinity (67,68). We evaluated the binding activity of STAT5 to native chromatin in living cells by ChIP with anti-STAT5 C17, anti-STAT5A, and anti-STAT5B antibodies followed by cloning of the precipitated DNA fragments. Clones were sequenced and analyzed for the presence of potential STAT binding N4 and N3 sites (per 1000 bp) (Fig. 5A and Supplementary Data). As a control for the frequency of STAT binding sites, we cloned in parallel random DNA fragments from Ba/F3 cells. Examination of the NCBI data base showed that 47.4% of our positive fragments were located in coding regions of DNA or at less than 10 kb from a coding region, while 19.5% of the fragments were located in non-coding region, arbitrarily considered at more than 10 kb from a coding region. For 33.1% of the fragments, we could not find any location. However, for our ChIP assays we took the overall frequency of binding to chromatin, as we evaluated the global biochemical affinity of STAT5 for the chromatin. Using anti-STAT5 C17 and anti-STAT5 N20 antibodies (no difference was observed between them in Western blot or in ChIP/cloning), the frequency of N3 sites compared with control was increased in Epo-stimulated cells and also in non-stimulated transformed cells. Strikingly, the frequency of N4 sites was clearly increased uniquely in non-stimulated IH55 cells. This increase in binding to N4 sites could not be observed anymore after Epo stimulation. Interestingly, the same increase in N4 frequency was observed when anti-STAT5B but not anti-STAT5A antibody was used for ChIP. The low frequency of GAS sequences detected by ChIP with anti-STAT5A antibodies in gp55-transformed cells may reflect the presence of STAT5A in heterodimeric complexes with STAT5B, or low levels of constitutive STAT5A activation that are below our limits of detection by Western blot of nuclear extracts or phosphorylated STAT5 isoforms.
To evaluate the relevance of this finding, luciferase reporter assays were performed using part of the ␤-casein promoter. The ␤-casein promoter contains a low affinity STAT5 binding site (TTC TTGG GAA) and a high affinity STAT5 binding site (TTC TTG GAA). To test whether STAT5 proteins activated by ligand or by oncogenic activation of the EpoR exhibit different effects, the transcriptional activity of STAT5 consensus sites, N4 and N3 alone or N3 and N4 together (N4N3) were tested (Fig. 5B). In IH cells, the basal transcriptional activity from N4, N3, and N4N3 sites was low without previous Epo stimulation (respectively, 7.87, 8.96, 11.04% of the maximal activity recorded in same cells transfected with N3N4-TK and stimulated with Epo), whereas in comparison in IH55 cells, the transcriptional basal level was significantly increased (21.6, 56.2, 63.57%, respectively of the maximal activity recorded in same cells transfected with N3N4-TK and stimulated with Epo). Notably, the transcriptional activity of STAT5 on N4 GAS sites was increased in IH55 cells. Comparable results were noted in HCD57 and HCD57-gp55 cells (Fig. 5B). Remarkably, Ba/F3 or HCD57 cells transformed by gp55-P exhibited significant activity on the single N4 site, which is in agreement with the higher frequency of N4 sites precipitated in the ChIP assays from gp55-P transformed cells (Fig. 5A). In R129C-transformed cells, the transcriptional constitutive level is higher for the N3 site alone or in combination with the N4 site, but not for the isolated N4 site (Fig. 5C). In IH55 cells there were higher levels of N3 transactivation as well, but the maximal effects induced by Epo were similar between the ligand-dependent (IH or HCD57 cells) and transformed cells (IH55 or HCD57-55 cells). Interestingly, in transformed cells, the activity of single N4 or N3 sites was not further stimulated by . RNA was prepared with the TRIzol reagent, and 1 g was used to prepare cDNA. PCR amplifications were performed for Bcl XL , Cis, and actin as control. B, cells were starved, electroporated with the Bcl XL -TK or the Cis-TK plasmids and the internal control pRL-TK plasmid, and stimulated with 1 unit/ml Epo for 2 h. Cells were lysed and subjected to a luciferase assay using the dual luciferase reporter assay system. The enzymatic activity is given relative to the activity registered for the non-stimulated IH cells. Results are the mean of three independent experiments Ϯ S.D. C, cells were starved and stimulated or not with 100 units/ml Epo for 15 min. Chromatin extracts were prepared and subjected to a STAT5 chromatin immunoprecipitation as described under "Experimental Procedures" using the anti-STAT5 C17 antibody (ChIP STAT5 C17) or a non-immune rabbit serum (ChIP control). Purified immunoprecipitated DNA was used as template for a PCR amplification of part of the Cis promoter responsive to STAT5 (See the murine Cis promoter sequence, STAT5 consensus sites are in bold, oligonucleotides used for PCR are underlined). INPUT represents positive controls and MOCK, negative control, as established under "Experimental Procedures." Epo, whereas the activity of N3N4 sites was sharply increased by Epo stimulation, suggesting that there are major differences between Epo-induced EpoR signaling and signaling by constitutively active receptors.
Constitutive Nuclear Translocation of STAT5B in Ba/F3 Cells Transformed with a Truncated Form of the GCSFR Involved in the Kostmann Syndrome-To assert the biological relevance of these findings, we tested STAT5 expression in Ba/F3 cells transformed by the expression of a truncated form of the GCSFR. Truncated mutants of the GCSFR at amino acids 718 and 731 were reported to be related to the transformation to secondary myelodysplastic syndrome and acute myelocytic leukemia in severe congenital neutropenia (Kostmann syndrome) (56 -58). The truncated forms of GCSFR are constitutively active and transduce a strong proliferative signal that is unable to promote neutrophil differentiation (69). Their signal transduction is also characterized by an increased STAT5/ STAT3 activation ratio (70). The HA-tagged wild-type GCSFR and HA-tagged GCSFR containing a cytoplasmic Q7183 Stop mutation (GCSFR-T718) cDNAs were cloned into the pMX-IRES-GFP bicistronic retroviral vector (60), and retroviral supernatants were used to infect Ba/F3 cells growing in IL3. Cells were FACS-sorted for similar GFP levels. Because of the truncation of 96 amino acids in the C terminus, GCSFR-T718 is shorter than wild-type GCSFR (Fig. 6A). As reported previously (71), we also found that the cell surface levels of the truncated GCSFR are higher than those of the wild-type GCSFR (Fig. 6B) and that the expression of the truncated GCSFR in Ba/F3 cells leads to cell proliferation in the absence of cytokine (Fig. 6C). Western blot analysis of nuclear extracts showed that STAT5B and not STAT5A were constitutively translocated to the nucleus in these transformed cells (Fig. 6D). Moreover, immunoprecipitation experiments with specific anti-STAT5A and anti-STAT5B antibodies followed by detection of the tyrosine phosphorylation of STAT5 revealed that STAT5B, but not STAT5A was constitutively activated in cells transformed by truncated GCSFR (Fig. 6E). Strikingly, addition of GCSF to cells expressing either the wild-type GCSFR or the truncated form of GCSFR resulted in activation of both STAT5A and STAT5B (Fig. 6, D and E). DISCUSSION Constitutive activation of STAT proteins has emerged as a hallmark of cancer (14). In hematopoietic malignancies, persis- Chromatin extracts were prepared and subjected to chromatin immunoprecipitation as described using anti-STAT5 (C17 and N20), anti-STAT5B, or anti-STAT5A antibodies. Double-stranded linkers were linked to the purified immunoprecipitated DNA. After an amplification step by PCR, DNA fragments were cloned into the pCR-II-Topo vector. Positive colonies were sequenced and analyzed for the presence of STAT5 binding sites: TTC NNN GAA (N3) and TTC NNNN GAA (N4). Total number, total length, and size average of all analyzed colonies are also given. B, cells were starved, electroporated with the N3N4-TK, the N4-TK, or the N3-TK plasmids and the internal control plasmid pRL-TK, and stimulated with 1 unit/ml Epo for 2 h. Cells were lysed and subjected to a luciferase assay. Luciferase ratio (rlu) represents the mean of three independent experiments Ϯ S.D. C, values (ratio between transcriptional activities of transformed/ligand-dependent cells) were normalized for each pair of cell lines IH55/IH*, HCD57-gp55/HCD57**, and R129C/Ba/F3-EpoR*** to the ratio of transcriptional activity on N3N4 sites for the same set of cell line pairs in the presence of Epo. For each pair of transformed/normal cells the ratio of transcriptional activity on N3N4 sites was considered 1. The absolute values for each cell line are shown in B.
tently activated STAT transcription factors, such as STAT5 and STAT3 participate in oncogenesis presumably through stimulation of expression of anti-apoptotic factors such as Bcl XL and cell cycle regulators (14). Transgenic overexpression in mice of the wild-type STAT5A or STAT5B genes within the lymphoid compartment leads to thymic T cell lymphoblastic lymphoma (72), suggesting that regulation of levels of STAT5 expression might also be a target of oncogenesis.
Treatment of cytokine-dependent or transformed Ba/F3 cells with Epo results in activation of both STAT5 isoforms, STAT5A and STAT5B. Surprisingly, in the same cells that were rendered cytokine-independent for growth, it was essentially STAT5B that was constitutively active. This result was obtained for both cells transformed by the R129C and by coexpression of EpoR and gp55-P and also for cells transformed with the oncogenic truncated GCSFR. This mutant induces severe neutropenia and favors acute myeloid leukemia in humans. This result is not because of a higher sensitivity of STAT5B to Epo activation, because at low Epo concentrations both STAT5A and STAT5B were activated with comparable sensitivity (data not shown). For STAT5B the activation rapidly plateaus after 0.25 units/ml Epo, while the levels of STAT5A activation continue to be increased proportional to the Epo concentration. It is thus tempting to speculate that STAT5B might be the anti-apoptotic STAT5, while STAT5A may have targets that need higher levels of activation. Were this to be true, selection for autonomous growth at low levels of oncogenic signaling would be expected to select preferentially for STAT5B constitutive activation. Further studies will be required to test, in patient-derived samples, whether the reported constitutive STAT5 activity may in fact be mainly STAT5B. Our data showing preferential STAT5B activation in GCSFR-T718 transformed would support this hypothesis.
We found that in the nucleus, the half-life of constitutively activated STAT5 in cells transformed upon expression of gp55-P was significantly longer than the half-life of Epo-activated STAT5 in the same cells or in cytokine-dependent cells. That ligand-activation apparently induces a similar rapid rise and then decrease of STAT5 activation in the nucleus of transformed cells suggests that the pools of STAT5 activated by the constitutive signal and by ligand are different and functional. The increase in stability of STAT5 that we have observed could be explained by inhibition of the negative regulation of STAT signaling i.e. by inhibition of phosphatases, of SOCS-, PIAS-, or ubiquitin-mediated degradation or by differential post-transcriptional STAT modifications such as glycosylation in the N-terminal region (73). Furthermore, this increase in stability of nuclear STAT5 may also promote effects of STAT5 that may be independent of DNA binding, as has been shown for other STATs, such as STAT1 (74).
The constitutively active STAT5 in cytokine-independent cells (gp55-P-or R129C-transformed cells) binds to the Bcl XL promoter and transactivates it (Fig. 2). As Bcl XL is a major anti-apoptotic gene in erythroid cell development (75), constitutive expression of Bcl XL would be relevant for the transformation process. Another important STAT5-targeted gene is Cis. Cis binds to the phosphorylated Tyr 401 of the EpoR (38,76) interfering with STAT5 binding onto the receptor, representing a powerful negative feedback of the Epo signaling pathways. A decrease in Cis expression would represent an advantage for the growth of transformed cells. Under steady-state conditions, expression of Cis was consistently lower in transformed cells than in their counterparts growing in Epo (Fig. 4A). Similarly, for assays performed in starved cells, Epo induced significantly FIG. 6. Constitutive nuclear translocation of STAT5B in Ba/F3 cells transformed with a truncated form of the GCSFR. A, whole cell extracts were prepared and subjected to immunoblotting using anti-HA antibody. B, cell surface expression of wild-type and truncated receptors was analyzed by flow cytometry using an anti-HA antibody and R-phycoerythrin-conjugated donkey F(abЈ)2 anti-mouse IgG secondary antibody. Fluorescence of truncated GCSFR is expressed as % of HA staining for the wild-type HA-GCSFR, which was considered 100%. Negative control is represented by wild-type GCSFR only labeled with the secondary antibody. C, cells were washed and placed in RPMI 10% FBS with or without GCSF for 2 days. Living cells were counting with the Cell Titer Glo TM Luminescent cell viability assay. Luminescence represents the mean of three independent experiments Ϯ S.D. D, cells were starved for 4 h and stimulated with 100 ng/ml GCSF for 15 min. Nuclear extracts were prepared as described and 50 g of nuclear proteins were subjected to immunoblotting using anti-P-Tyr 694 STAT5 antibody. The same membrane was reprobed with anti-STAT5A, anti-STAT5B, and anti-STAT5 C17 antibodies. E, cells were starved for 4 h and stimulated or not with 100 ng/ml GCSF for 15 min. STAT5 proteins were immunoprecipitated either with anti-STAT5A or anti-STAT5B antibody, subjected to immunoblotting using an anti-P-Tyr 694 STAT5 antibody. The same blot was reprobed with anti-STAT5A, anti-STAT5B and anti-STAT5 C17 antibodies. lower levels of Cis in transformed cells than in ligand-dependent cells. This expression profile contrasted with those of Bcl XL and of ␤-casein, which were induced to higher levels in transformed cells than in ligand-dependent cells. Ongoing studies in our group are aimed at determining which segment of the Cis promoter is able to transfer this down-regulation to a chimeric Bcl XL -Cis promoter in transformed hematopoietic cells.
Chromatin immunoprecipitation studies indicated that STAT5B was found at an increased frequency to be bound to N4 low affinity GAS sites in cells transformed by expression of gp55-P. Transactivation studies indeed suggest that this was specific to gp55-transformation since only in cells transformed by gp55 (IH55 or HCD57-55 cells), but not by the R129C or the oncogenic GCSFR mutant, we detected transactivation via the single ␤-casein N4 site. It is possible that in non-stimulated gp55-transformed cells there is a link between the extended half-life of STAT5B in the nucleus and transactivity of N4 sites by this pool of STAT5. These results open the possibility that in transformed cells, STAT5 might not only activate some of the STAT5 target genes that are normally regulated by Epo, but also promoter sequences containing the consensus STAT5 sites that are not normally or less activated through Epo stimulation. The increased activity of the activated STAT5B on N4 sites in gp55-transformed cells may reflect also the very low ratio between activated STAT5A (not detectable) and STAT5B. It has been reported that the N4 site is bound by STAT5B homodimers but not by STAT5A homodimers, while STAT5A was able to bind (as STAT5B did) to the ␤-casein N3 site (68). It is therefore possible that in cells transformed by the R129C the very low levels of constitutive activation of STAT5A may have prevented targeting of single N4 sites or favor binding to N3 sites instead, like in gp55-transformed cells stimulated with Epo. Alternatively, this may depend on the selection for ligand-independent growth, namely a weak signal would select for STA5B activation at high levels, while a stronger signal may select for weaker but more diverse signals, which would include STAT5A and other pathways activated.
As in many leukemias and lymphomas there is constitutive activation of STAT5 (14), it will be interesting to test whether, in such cells, single N4 or N3 sites may be transcriptionally activated. Subgroups of a specific type of leukemia might differ in their abilities to transactivate different GAS sites in a manner analogous to cells transformed by coexpression of gp55-P and EpoR. Such functional differences may be added to gene expression monitoring (77) to further classify leukemias (discovery of subclasses) and other forms of cancer where STATs are constitutively activated.