The Carboxyl-terminal Domains of gp130-related Cytokine Receptors Are Necessary for Suppressing Embryonic Stem Cell Differentiation

Cell type-specific responses to the leukemia inhibitory factor (LIF)/interleukin 6 cytokine family are mediated by dimerization of the LIF receptor α-chain (LIFRα) with the signal transducer gp130 or of two gp130 molecules followed by activation of the JAK/STAT and Ras/mitogen-activated protein kinase cascades. In order to dissect the contribution of gp130 and LIFRα individually, chimeric molecules consisting of the extracellular domain of the granulocyte colony stimulating factor receptor (GCSF-R) and various mutant forms of the cytoplasmic domains of gp130 or LIFRα were expressed in embryonic stem (ES) cells to test for suppression of differentiation, or in a factor-dependent plasma cytoma cell line to assess for induction of proliferation. Carboxyl-terminal domains downstream of the phosphatase (SHP2)-binding sites were dispensable for mitogen-activated protein kinase activation and the transduction of proliferative signals. Moreover, carboxyl-terminal truncation mutants which lacked intact Box 3 homology domains showed decreased STAT3 activation, failed to induce Hck kinase activity and suppress ES cell differentiation. Moreover, STAT3 antisense oligonucleotides impaired LIF-dependent inhibition of differentiation. Substitution of the tyrosine residue within the Box 3 region of the GSCF-R abolished receptor-mediated suppression of differentiation without affecting the transduction of proliferative signals. Thus, distinct cytoplasmic domains within the LIFRα, gp130, and GCSF-R transduce proliferative and differentiation suppressing signals.

ities through binding to transmembrane receptors. The type I cytokine receptor superfamily is characterized by conserved cysteine residues and the presence of a tryptophan-serine-Xtryptophan-serine (WSXWS) motif in the extracellular domain (1). Weak sequence similarity between individual members of the type I receptor family extends into the membrane-proximal Box 1 and Box 2 homology regions of the cytoplasmic domain (2,3). Box 1 and Box 2 homology regions associate constitutively with Janus tyrosine kinases (JAKs) and are indispensable for ligand-mediated JAK activation (4). Functional analysis of the cytoplasmic domain of the IL6 signal-transducing subunit, gp130, the ligand-binding subunit of the leukemia inhibitory factor receptor (LIFR␣), the granulocyte CSF receptor (GCSF-R), and the receptor for thrombopoietin (c-Mpl) has revealed an additional region of homology (Box 3) which is specific to gp130, LIFR␣, and GCSF-R but not found in c-Mpl (5,6). Mutational analysis revealed that the Box 3 homology region is required for the induction of some specific biological responses, such as activation of acute-phase protein expression in transfected liver cells, thereby suggesting that distinct regions of the cytoplasmic domain may be required for the induction of specific biological responses (5,6).
The pleiotropic actions of the LIF/IL6 family of cytokines include the regulation of cell growth, apoptosis, protein secretion, and cellular differentiation, often in a cell type-specific manner (7). In some cases, opposing biological responses are triggered by the LIF/IL6 cytokine family such as the induction of differentiation in myeloid leukemia M1 cells and the suppression of differentiation in embryonic stem (ES) cells. Thus, it has been proposed that cellular responses may be the consequence of gene expression programs elicited by an array of different, and sometimes cell type-specific intracellular signaling cascades, generated by the receptors for IL2, GCSF, and granulocyte-macrophage CSF (8 -10). However, in all cell types investigated, binding of LIF to its ligand-binding chain, LIFR␣, results in the formation of a heterodimeric receptor complex with gp130 which in turn triggers the activation of gp130associated JAKs and other cytoplasmic tyrosine kinases (11)(12)(13)(14). In ES cells, ostensibly identical signaling events are initiated following IL6-induced homodimerization of gp130 molecules leading to activation of JAK family members Jak1, Jak2, and Tyk2, as well as of the Src-related tyrosine kinase Hck (14,15). The kinases are thought to undergo autophosphorylation and at least Jak1 has been shown to subsequently * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. phosphorylate tyrosine residues on gp130 and LIFR␣ (12,13). In turn, the phosphotyrosine residues on the cytokine receptors recruit various Src homology (SH2) domain-containing intermediate signal transducing molecules which often result in their subsequent tyrosine phosphorylation and activation. Two such molecules, the tyrosine phosphatase SHP2 and the latent signal transducer and activator of transcription 3 (STAT3), are phosphotyrosine binding partners for gp130 and LIFR␣ (16,17), which contain 6 and 5 tyrosine residues in their respective cytoplasmic domains (15,18). In the case of gp130 and LIFR␣, activation of STAT3 depends on phosphorylation of the most carboxyl-terminal tyrosine residues while a different, more membrane-proximal located phosphotyrosine residue is required for activation of SHP2 (15,18).
The LIF/IL6 cytokine family has been shown to induce tyrosine phosphorylation and DNA binding of a subset of STAT proteins, namely STAT1, STAT3 via a receptor binding dependent mechanism, and possibly STAT5a via a direct interaction with JAK kinases (19 -21). In gp130 and LIFR␣, tyrosine phosphorylation and activation of STAT3 appears dependent on the presence of phosphotyrosine residues with a glutamine at position ϩ3 relative to the tyrosine (YXXQ) thereby contrasting the sequence requirement for the activation of SHP2 which requires a valine at position ϩ3 (YXXV) (15). The relative specificity of STAT activation by individual cytokine receptors prompted the notion that individual STAT family members determine the specificity of signal transduction from a particular cytokine receptor. Support for this view has recently been provided by genetic deletion studies in mice which point to a critical role for STAT1 in the innate immunity elicited by interferon-␣ and -␥ (22,23). By contrast, lack of STAT3 expression results in early embryonic lethality (24), while STAT5a-deficient mice lack terminally differentiated mammary glands (25). In vitro, activation of STAT3 appears essential for IL6-mediated differentiation of mouse leukemia M1 cells (18,26) and the prevention of apoptosis in factordependent Ba/F03 cells (27).
In the present study, we attempt to identify the cytoplasmic region of individual cytokine receptor chains required for either ligand-dependent suppression of ES cell differentiation or promotion of cellular proliferation. To this end, we expressed chimeric receptor constructs in which the transmembrane and cytoplasmic domains of gp130, LIFR␣, or GCSF-R were fused to the extracellular domain of the GCSF-R. This approach permits GCSF-dependent homodimerization of chimeric receptor subunits independent of the corresponding endogenous receptor chains (5,6,19,20,27). Our results demonstrate that carboxyl-terminal truncations or mutation of specific tyrosine residues within Box 3-containing receptor chains have no effect on the transduction of proliferative signals, while reducing the capacity to suppress ES cell differentiation. This response correlates with a decrease in Hck kinase activity and STAT3 phosphorylation. A critical role for STAT3 is supported by our finding that the presence of STAT3 antisense (AS)-oligonucleotides impairs the differentiation-inhibiting activity of fulllength gp130 and LIFR␣ chains on ES cells in vitro.

EXPERIMENTAL PROCEDURES
Biological Reagents and Cell Cultures-Recombinant human LIF (ES-GRO) was purchased from Amrad (Melbourne, Australia) and recombinant human GCSF and murine IL6 were kindly provided by Amgen (Thousand Oaks, CA) and R. Simpson (Ludwig Institute, Melbourne, Australia), respectively. The monoclonal antibodies used for Western blotting of STAT3 and anti-phosphotyrosine (4G10) were obtained from Upstate Biotechnology Inc. (Lake Placid, NY). Immunoprecipitation assays were carried out with polyclonal antisera directed against SHP2 (from Santa Cruz Biotechnology, Santa Cruz, CA), STAT3 and STAT5a (from Transduction Laboratories; Lexington, KY), and Hck (gift from C. Lowell, University of California, San Francisco, CA).
Full-length chimeric receptor constructs containing the extracellular domain of the murine GCSF-R ligated to the transmembrane and cytosolic domain of gp130 and LIFR␣ have been described previously (20). The full-length G/mpl and the truncated chimeric receptor were obtained by fusing the extracellular domain of the GCSF-R via an introduced BamHI site (20) with fragments encoding the transmembrane and cytoplasmic domains of murine c-Mpl (28), gp130, and the LIFR␣, respectively. These fragments were generated by polymerase chain reaction and contained at their 5Ј-end a silent BamHI restriction site over the second last amino acid of the extracellular domain and at their 3Ј-end translational stop codons at amino acid positions 220, 156, and 130 in gp130 (counting from the most membrane-proximal amino acid of the cytoplasmic domains) and position 136 in LIFR␣ yielding construct G/gp130 ⌬220 , G/gp130 ⌬156 , and G/gp130 ⌬130F , and G/LIFR␣ ⌬136 (Table I). In the G/gp130 ⌬130F construct, the tyrosine (Y) residue in the putative STAT3 binding sequence Y 126 XXQ in Box 3 was mutated to phenylalanine (F). In order to facilitate subcloning of isolated fragments into the mammalian expression vectors, the 3Ј-polymerase chain reaction oligonucleotides also contained a XbaI restriction site. All newly generated chimeric receptor cDNA constructs were sequenced throughout their coding regions on an Applied Biosystem 373A DNA sequencing system and subcloned into the XbaI site of the mammalian expression vector 6 -16PGK-IRESneo designed to give rise to dicystronic mRNAs (20). For analysis of the full-length GCSF-R, Tyr 3 Phe substitutions were introduced by site-directed mutagenesis at Tyr 74 , Tyr 114 , and at both tyrosine residues (Tyr 74 , Tyr 144 ) as described previously (29). The GCSF-R constructs were cloned into the XbaI site of the mammalian expression vector pEF-BOS (29) and co-transfected with the resistance marker plasmids PGKneo (14). High performance liquid chromatography-purified phosphothioate AS oligonucleotides directed against the amino-terminal region of mouse STAT3 (5Ј-GCCAG-GAACTGCCGCAGC-3Ј) or consisting of a randomized sequence of a similar nucleotide composition (5Ј-AGTCCAGGCCCAGGTCCG-3Ј) were purchased from Bresatec (Australia). All ES cell lines were derived from the HCK-hprt modified E14TG2a cell line which contains a hypoxanthine-guanine phosphoribosyltransferase minigene under the control of an hck promoter fragment which is transcriptionally inactive in differentiated ES cells (20). ES cell lines were maintained free of feeder cells in ES cell medium (14) supplemented with 1000 units/ml LIF which was substituted during experiments by the indicated concentration of LIF or GCSF. The IL6-dependent murine plasmacytoma cell line 7-TD1 was maintained in Dulbecco's modified Eagle's medium supplemented with 10% FCS, 0.24 mM Lasparagine, 0.55 mM L-arginine, 50 M 2-mercaptoethanol, and 5 pM IL6 which was replaced during experiments by the indicated concentration of GCSF.
Cell Lines-Stable transfected ES cell lines were obtained by electroporating 15 g of the linearized 6 -16PGK-IRESneo-based expression constructs or 15 g of the linearized pEF-BOS-based constructs together with 1.5 g of the selectable marker plasmid PGKneo (500 farads, 270 V, Bio-Rad Genepulser) into 1.5 ϫ 10 7 ES cells. Receptor expression level on individually expanded G418-resistant (175 g/ml Geneticin, Life Technologies, Inc.) colonies were carried out by binding of radiolabeled GCSF as described (20). Cell lines with a comparable number of receptors (800 -1400 receptor/cell) were used for experiments.
The proliferative capacities of the chimeric gp130 and LIFR␣ receptor constructs were assessed in stable transfected 7-TD1 cells which were obtained following electroporation (960 farads, 270 V) of 5 ϫ 10 5 cells with 15 g of linearized plasmid. Wild-type and mutant GCSF-R expressing IL3-dependent Ba/F03 cell lines were obtained essentially as described (20) by electroporating (960 farads, 270 V) 15 g of linearized receptor plasmid together with 1.5 g of the selectable marker plasmid PGKneo. Receptor expression on clonally derived neomycin-resistant (20 g/ml) Ba/F03 cell lines was assessed by flow cytometry (FACScan, Becton Dickinson) as described (29).
Cell Culture Assays-The extent to which ES cell differentiation was prevented through GCSF-dependent formation of mutant receptor homodimers was determined as described by morphological appearance (14,51) as well as by selection against differentiated cells in the presence of HAT-containing (200 M hypoxanthine, 0.8 M aminopterin, 32 M thymidine) culture medium and measuring the conversion of MTT (3-(4,5-dimethlydiazol-2-yl)-2,5-diphenyltetrazolium bromide; Sigma, 0.5 mg/ml final concentration) to a product with maximal optical absorbance at 540 nm (20). AS oligonucleotide experiments were carried out by incubating 24-h cultures (inoculated at 5000 cells/cm 2 ) for 10 min with streptolysin-O-containing (5 units/ml, Sigma) permeabilization buffer (150 mM KCl, 37.5 mM NaCl, 6.25 mM MgCl 2 , 0.8 mM EGTA, 1 mM CaCl 2 , 1.24 mM ATP, 12.5 mM PIPES, pH 7.5) supplemented with 3 M AS oligonucleotide. Cultures were then rinsed carefully several times with ES cell medium and incubated for 2 days in the presence of 3 M AS oligonucleotide in ES cell medium supplemented with the indicated concentration of LIF. At this time, the medium was replaced with fresh medium containing 3 M AS oligonucleotide and the cultures were incubated for an additional 3 days.
The GCSF-dependent survival/proliferation of G418-resistant, chimeric receptor expressing 7-TD1 cells was assessed by determining cell numbers following an 8-day exposure of cultures (inoculated in 6-well multiculture dishes at a concentration 2 ϫ 10 4 cells/well) to either GCSF (50 ng/ml) or 5 pM IL6 as a control. The mitogenic activity of GCSF-R-expressing Ba/F03 cells was determined by thymidine incorporation. For this purpose, the cells were inoculated in 96-well multiculture dishes at a density of 10 4 cells/well, stimulated for 48 h with GCSF (10 ng/ml) in RPMI 1640 medium supplemented with 10% FCS. The cultures were pulsed for 4 h with 0.5 Ci/well [methyl-3 H]thymidine (NEN Life Science Products Inc.) prior to harvesting and incorporated radioactivity was measured using a Betaplate liquid scintillation counter (Pharmacia Biotech Inc.).
Electrophoretic mobility shift assays using the high affinity c-sis, factor inducible-binding site m67 were carried out as described previously (20, 30) with 16 g of extracts prepared from ES cultures stimulated for 15 min with LIF (500 units/ml), GCSF (10 ng/ml), or saline and, where indicated, with STAT1 or STAT3 antisera. The DNA-binding complexes were then subjected to electrophoresis and visualized using a PhosphorImager (Molecular Dynamics).
For Hck in vitro kinase experiments, 500 g of cell lysate were immunoprecipitated with an Hck-specific antiserum 1077 (1:500 dilution, gift of C. Lowell, University of California, San Francisco, CA) and the tyrosine autophosphorylation reaction was carried out in the presence of 10 Ci of [␥-32 P]ATP (3000 Ci/mmol, Bresatec, Australia) as described previously (14,51).

The Membrane-distal Cytoplasmic Receptor Domains Are Required for the Regulation of ES Cell Differentiation-In order to
test for the independent signaling capacity of the two components of the native heterodimeric LIF receptor complex, we exploited the fact that undifferentiated ES cells do not express GCSF-R. Cell lines expressing chimeric receptor constructs, in which the intracellular domain of either gp130 or LIFR␣ was fused to the extracellular domain of the GCSF-R, were established (Table I). This strategy permits GCSF-dependent dimerization and activation of chimeric receptor chains without the interference of the endogenous LIFR␣ and gp130 receptor chains. Clonally derived ES cell lines, expressing comparable numbers of chimeric and endogenous receptor chains (800 -1400 per cell) as assessed by binding of radiolabeled GCSF and LIF (Ref. 20, and data not shown) were chosen for subsequent analysis. The ability of the introduced receptor chains to suppress ES cell differentiation was determined in two independent assay systems. The first takes advantage of the striking morphological differences between the densely packed colonies of undifferentiated ES cells and the flattened morphology of differentiated cells (14,51). Homodimerization of the fulllength cytoplasmic domain of gp130 (G/gp130) or LIFR␣ (G/ LIFR␣), respectively, prevented ES cell differentiation in a GCSF-dependent manner (Fig. 1a). Moreover, GCSF also prevented differentiation of ES cells transfected with the fulllength GCSF-R (Fig. 1c), but had no effect on mock-transfected wild-type cells (20). Signaling through the endogenous LIFR␣/ gp130 receptor heterodimer, formed in response to stimulation with LIF, prevented differentiation of approximately 95% of colonies in all ES cell lines investigated, indicating that the presence of chimeric receptors did not interfere with activation of the endogenous receptor chains. Since the scoring of cell morphology is potentially subject to investigator bias, we also carried out a colony assay based on selection against differentiated cells. This sensitive "killing assay" (Fig. 1, b and d) facilitated direct measurement of the proportion of undifferentiated ES cells after pulsing the cultures with the MTT dye. For this purpose, we exploited the HCK-hprt ES cell line which harbors a hprt minigene under control of the murine hck promoter that is transcriptionally inactivated in differentiating ES cells (20,51). Thus, differentiating cells are killed when incubated continuously in HAT-containing medium, while undifferentiated, HAT-resistant cells proliferate and metabolize the MTT dye to a product which can be measured photometrically.
Since the YXXQ motifs in the COOH-terminal region of gp130 are required for the induction of differentiation in myeloid M1 cells in response to IL6 (18,26), we next tested the biological effect of carboxyl-terminal truncations of the cytoplasmic tail of gp130 and LIFR␣. Ligand-dependent suppression of differentiation was fully maintained by the G/gp130 ⌬220 receptor retaining two out of four YXXQ motifs, but was significantly reduced with the G/gp130 ⌬156 chimera in which only the single YXXQ motif within Box 3 is retained (Figs. 1, a and c). Extending the truncation further to the membrane-distal border of Box 3 in conjunction with a phenylalanine substitution of the fourth YXXQ motif (F 126 XXQ) within Box 3 abolished the capacity of the G/gp130 ⌬130F receptor to suppress differentiation. Similarly, a truncated LIFR␣ with a deletion of the Box 3 region and lacking two out of three YXXQ motifs (G/LIFR␣ ⌬136 ), was no longer capable of preventing ES cell differentiation (Fig. 1, b  and d). Furthermore, GCSF-dependent activation of a chimeric receptor containing the full-length cytoplasmic domain of the thrombopoietin receptor c-Mpl, which lacks the Box 3 homology region but contains one YXXQ motif, was also unable to maintain the undifferentiated ES cell phenotype.
We next investigated the potential of wild-type and truncated chimeric receptor homodimers to mediate a proliferative response in the IL6-dependent cell line 7-TD1. Unlike the traditionally employed, factor-dependent pro-B cell line Ba/F03, the plasmacytome cell line 7-TD1 shows marked induction of proliferation upon activation of gp130 mediated signaling. As revealed in Fig. 2, both full-length, as well as the most truncated receptors, mediated a GCSF-dependent increase in cell number which was independent of STAT3 phosphorylation as assessed by anti-phosphotyrosine blotting of cell lysates. These data, therefore, suggest that the carboxyl-terminal membranedistal regions of gp130 and LIFR␣ which contain the Box 3 homology region as well as several YXXQ motifs play a critical role in the signal transduction pathways required to suppress differentiation of ES cells in vitro. By contrast, the carboxylterminal regions of gp130 and the LIFR␣ are dispensable for the transduction of proliferative signals in 7-TD1 cells which appears to be independent of STAT3 phosphorylation.
Activation of Intermediate Signaling Molecules-The heterodimeric LIFR␣⅐gp130 receptor complex is physically and functionally associated with the Src-related tyrosine kinase Hck and the JAK/STAT signaling cascade in ES cells (14,51). We therefore investigated the capacity of the chimeric receptors to activate p56/59 Hck and to phosphorylate STAT3, the predominant STAT protein activated in response to the LIF/IL6 family of cytokines. GCSF stimulation of cells expressing the full-length G/gp130, G/LIFR␣, or the truncated G/gp130 ⌬220 receptors led to increased in vitro autophosphorylation of the two Hck isoforms (p56/59 Hck ) in cell lysates immunoprecipitated with an anti-Hck antiserum comparable to the increase observed after LIFdependent stimulation of the endogenous gp130⅐LIFR␣ complex (Fig. 3). By contrast, the truncated G/gp130 ⌬156 , G/gp130 ⌬130F , nor G/LIFR␣ ⌬136 receptors were capable of mediating significant activation of Hck. Furthermore, Hck kinase activity remained unaffected after stimulation of the G/mpl receptor, indicating that the most carboxyl-terminal sequences of the cytoplasmic tails of gp130 and the LIFR␣ are required for ligand-dependent Hck activation. Immunoprecipitation of STAT3 proteins followed by Western blotting with anti-phosphotyrosine antibody showed a prominent increase in tyrosine phosphorylation of STAT3 in response to GCSF in ES cells expressing the full-length G/gp130, G/LIFR␣, or the truncated G/gp130 ⌬220 receptor (Fig. 3). Weak ligand-dependent phosphorylation of STAT3 was registered following stimulation of chimeric receptor with a single YXXQ motif (G/gp130 ⌬156 , G/LIFR␣ ⌬136 , G/mpl), while no response was observed in G/gp130 ⌬130F expressing cells. Reprobing of the blots with anti-STAT3 antibodies demonstrated comparable amounts of STAT3 protein loaded.
Since activation of the most extensively truncated receptors (G/gp130 ⌬130F and G/LIFR␣ ⌬136 ) still induced proliferation of 7-TD1 cells (Fig. 2), we investigated ligand-dependent activation of the mitogen-activated protein kinase (MAPK). GCSF treatment of 7-TD1 cells led to an increase in MAPK activity as assessed by the phosphorylation-dependent mobility shift of p42/44 MAPK of the full-length and the truncated receptors (Fig.  4a, top). A similar change in MAPK mobility was observed in serum-starved ES cells expressing chimeric receptors following LIF or GCSF stimulation (Fig. 4a, bottom), confirming that gp130 and LIFR␣-dependent activation of MAPK occurs independently of YXXQ motifs (31). Furthermore, ligand-dependent activation of MAPK in ES cells correlated with an increase in tyrosine phosphorylation of the phosphatase SHP2 (Fig. 4b) which has been suggested as a molecular mechanism by which gp130 connects to the Ras/MAPK pathway (27). Thus, liganddependent activation of SHP2 and MAPK correlates with the receptors' capacity to stimulate proliferation, however, activation of this pathway in isolation does not suppress ES cell differentiation.
STAT3 Is Required for the Suppression of ES Cell Differentiation-The heterodimeric LIFR␣⅐gp130 receptor complex in ES cells is known to induce binding of STAT1 and STAT3 containing protein complexes to the high affinity sis-inducible SIE element (17,20). Based on the retardation of these complexes by STAT antibodies, the most prominent SIF-A complex consists of STAT3 homodimers whereas the less abundant SIF-B and SIF-C complexes represent STAT3/STAT1 heterodimers and STAT1 homodimers, respectively (19,20). Consistent with decreased or absent STAT3 phosphorylation following GCSF stimulation of the G/LIFR␣ ⌬136 , G/mpl, or G/gp130 ⌬130F receptors, we observed reduced STAT-DNA binding in cell lines expressing G/LIFR␣ ⌬136 or G/mpl, when compared with the pattern obtained after activation of the fulllength G/gp130 and G/LIFR␣, respectively (Fig. 5). Activation of the G/gp130 ⌬130F receptor failed to induce STAT binding activity all together.
Receptor-dependent engagement of particular STAT family members is one of the mechanisms by which different biological outcomes may be achieved through JAK-STAT signal transduction pathways (4). Consistent with this finding, stimulation of the interferon-␣ receptor, which activates STAT1 and STAT2, fails to suppress ES cell differentiation (14). We therefore focused on possible functions of STAT3 by attempting to reduce the intracellular protein levels of STAT3 by exploiting phosphothioate AS oligonucleotides directed against the aminoterminal sequences of murine STAT3. A 5-day treatment of undifferentiated ES cells with 3 M STAT3-AS oligonucleotides specifically decreased the STAT3 protein level by more than 90% when compared with cells exposed to a control AS oligonucleotide consisting of a randomized sequence with a similar nucleotide composition (Fig. 6a). Furthermore, the effect of the STAT3 AS oligonucleotide was specific, since in the same cells the protein level of STAT5a was not affected. The addition of STAT3 AS oligonucleotides significantly decreased the proportion of undifferentiated wild-type ES cells in LIF-treated cultures, while the randomized control AS oligonucleotide was without effect (Fig. 6b). Addition of STAT3 AS oligonucleotides also decreased the proportion of undifferentiated colonies in GCSF-treated cultures of cells expressing either G/gp130 or G/LIFR␣ (Fig. 6b and data not shown). We also tested the effect of the STAT3 AS oligonucleotide on the mitogenic response of 7-TD1 cells stimulated with IL6. Fig. 6c shows that inhibition of STAT3 protein production by STAT3 AS oligonucleotides had no effect on IL6-induced DNA synthesis. By contrast, the IL6 effect was abolished in cultures treated with the specific MAPK kinase (MEK) inhibitor PD098059, consistent with activation of the Ras/MAPK pathway via the membrane-most proximal SHP2 phosphotyrosine-binding sites in gp130 and LIFR␣. Taken together, our data obtained in ES and 7-TD1 cells suggest a selective functional involvement of STAT3 in differentiation suppressing signal(s) generated by the heterodimeric LIFR␣⅐gp130 complex or their respective homodimeric receptor counterparts. Tyrosine Substitution in GCSF-R Impairs the Regulation of ES Cell Differentiation-Our limited survey of cytokine receptors suggests that activation of the Box 3-containing cytoplasmic GCSF-R domain can suppress ES cell differentiation (Fig.  1). Since the tyrosine residues within the cytoplasmic tail of the GCSF-R contribute to specific signaling pathways (31), we focused on the two tyrosine residues which showed the most prominent impairment of GCSF-mediated differentiation in the myeloid M1 cell line (29,30). Stimulation of a GCSF-R mutant with a phenylalanine substitution of the membrane most proximal tyrosine residue (Y74F) suppressed ES cell differentiation to a similar extent as stimulation of the wild-type GCSF-R (Fig. 7). By contrast, substitution of the tyrosine residue within Box 3 (Y144F) decreased the capacity to inhibit differentiation in a GCSF-dependent manner by approximately 80%. The capacity to suppress differentiation was entirely lost in ES cells expressing GCSF-R containing the double substitution (Y74F,Y144F). In all cases, stimulation through the endogenous LIFR␣⅐gp130 receptor complex suppressed differentiation of more than 90% of ES cell colonies expressing the various versions of the GCSF-R constructs.
We also tested whether the tyrosine residues in GCSF-R required for signaling in ES cells contribute to propagation of mitogenic signals. For this purpose, we introduced the various GCSF-R expression constructs into the factor-dependent Ba/ F03 cell line and screened for surface expression of GCSF-R protein by flow cytometry (data not shown) and representative cell lines were chosen which showed comparable levels of receptor expression. With all GCSF-R constructs tested, we were able to derive GCSF-dependent Ba/F03 cell lines. Moreover, GCSF-dependent [ 3 H]thymidine incorporation was recorded in Ba/F03 cells expressing wild-type, Y74F, and Y144F GCSF-R, respectively, while some of the cell lines containing the Y74F,Y144F double substitution responded slightly less than wild-type receptor expressing cells (Fig. 8). These results suggest that Tyr 144 within the GCSF-R Box 3 region is absolutely required for the regulation of ES cell differentiation, but is dispensable for the transmission of a mitogenic signal. Cultures were subsequently stimulated for 10 min with 20 ng/ml GCSF (G), 500 units/ml LIF (L), or saline (c) and cell lysates were prepared. Autophosphorylation activity of p56/59 Hck (left panel) was determined by immunoprecipitating Hck from 500 g of ES cell lysates followed by an in vitro autophosphorylation assay in the presence [␥-32 P]ATP. The reaction products were separated by SDS-PAGE and visualized by autoradiography. The two bands represent the two protein isoforms p56 Hck and p59 Hck . Tyrosine phosphorylation of STAT3 was assessed (right panel) by immunoprecipitating 500 g of ES cell lysates with anti-STAT3 antibodies, separated by SDS-PAGE, transferred to nitrocellulose and probed with anti-phosphotyrosine (Tyr(P)) antibodies. The membranes were stripped and reprobed to assess for equal amounts of STAT3 transferred.

FIG. 4. Ligand-dependent activation of the MAPK pathway.
Confluent cultures of ES cells or 7-TD1 cells expressing the indicated chimeric receptors were starved for 12 h of LIF (ES cells) or IL6 (7-TD1 cells) in medium containing 1% FCS. Cultures were then stimulated for 10 min with 20 ng/ml GCSF (G), 500 units/ml LIF (L), or saline (c) and cell lysates were prepared. a, MAPK mobility shift in 7-TD1 cells (upper panel) and ES cells (lower panel). Sixty g of cell lysate was separated by SDS-PAGE and transferred to nitrocellulose and blotted with Erk-1 antibodies. The phosphorylated, retarded forms of p42/44 MAPK are indicated by arrowheads. b, tyrosine phosphorylation of SHP2. Five hundred g of ES cell lysates were immunoprecipitated with anti-SHP2 antibodies, separated by SDS-PAGE, transferred to nitrocellulose, and blotted with anti-phosphotyrosine antibodies (Tyr(P)). Blots were stripped and reprobed to assess for equal amounts of SHP2.
Analysis of STAT3 phosphorylation indicated reduced tyrosine phosphorylation following activation of the GCSF-R mutants Y144F when compared with the pattern observed with the wild-type and the Y74F receptor, respectively (Fig. 9). Furthermore, we could not detect significant STAT3 phosphorylation following stimulation of ES cell lines expressing GCSF-R (Y74F,Y114F). Similar results were obtained in electrophoretic mobility shift assays using the SIE probe (data not shown). Thus, the decreased STAT3 phosphorylation and DNAbinding observed with the GCSF-R(Y74F,Y144F) correlates with an inability to prevent differentiation of ES cells. DISCUSSION In this study we demonstrate that the membrane-proximal region of the cytoplasmic region of gp130 and LIFR␣ are required for regulation of mitogenic activity while the membranedistal region with an intact Box 3 homology motif is critical for suppressing differentiation of ES cells in vitro. This observation correlates with ligand-dependent activation of Hck kinase and STAT3 phosphorylation and, unlike the transduction of mitogenic signal, is selectively impaired in the presence of STAT3 AS oligonucleotides. Therefore, our data support findings by Niwa et al. (32) who suggested a critical role for STAT3 in maintaining the pluripotent phenotype of ES cells in vitro.
Many cytokine receptors, including those for the ILF/IL6 family, initiate divergent intracellular signaling pathways through domains that are specifically required for regulation of proliferation, differentiation, or apoptosis (33)(34)(35). Signaling by most type I cytokine receptors depends minimally on the Box 1 motif which is required for the association with JAKs. Transduction of mitogenic signals from the GCSF-R (36) and gp130 (27) depends on the membrane-proximal region extending over Box 1 and Box 2 domains. In particular, phosphorylation of SHP2 via the YXXV motif, which is located distal to the Box 2 and proximal to the Box 3 motif in gp130, is absolutely required for mitogenesis in Ba/F03 cells (27). Here, we show that the truncated G/gp130 ⌬130F and G/LIFR␣ ⌬136 receptors, which retain the YXXV motif, also promote growth of 7-TD1 cells. Phosphorylated SHP2 binds Grb2 (13,27,37) which is constitutively associated with the nucleotide exchange factor Sos thereby providing a direct link between the YXXV motif and transient activation of the Ras/MAPK pathway. By contrast, sustained activation of the Ras/MAPK pathway in ES cells in response to LIF and IL6 (14) may also occur through direct association of Shc with phosphorylated gp130 (38) or through constitutive association of the two isoforms of Hck with gp130 (51). We extend these findings here by identifying the carboxyl-terminal regions of gp130 and LIFR␣ as sites required for the induction of ligand-dependent Hck activation. While the molecular nature of this interaction remains to be established, the absence of conspicuous cysteine residues in gp130 and LIFR␣ suggests a different mechanism to that identified between the Src family kinase Lck and the T-cell co-receptors CD4/CD8 (39). The present study suggests a correlation between gp130/LIFR␣-medi- ated Hck activation and suppression of ES cell differentiation. Furthermore, GCSF-R dependent activation of the Src-related kinase Lyn (40) and the observation that overexpression of constitutive active Lyn partially suppressed ES cell differentiation, 2 suggests that Src kinases may amplify signals that sup-press ES cell differentiation.
Specificity of the JAK-STAT pathway is controlled at the level of JAK kinase engagement (14,41), SH2 domain dependent and independent interaction of STAT proteins with specific cytokine receptors, STAT dimerization, and the resulting specificity in DNA binding (42)(43)(44). Receptor phosphotyrosine-independent STAT5 activation and cell type-specific differences in STAT expression patterns may therefore account for the fact that the truncated G/gp130 ⌬130F receptor induced growth of the plasmacytome cell line 7-TD1 but only induced a transient mitogenic signal in the pro-B cell line Ba/F03 (27). In fact, STAT5 activation is required for maximal proliferation in response to IL2 receptor activation (45) and we found in 7-TD1 cells weak tyrosine phosphorylation of STAT5a with all truncated chimeric gp130 and LIFR␣ constructs (data not shown).
Based on the severity of the phenotype observed in mice with deletions for either STAT3 or gp130, STAT3 appears to be a crucial signaling intermediate for a number of growth factor and cytokine receptors in addition to those for the LIF/IL6 cytokine family. STAT3 plays an important role in self-renewal of lymphocytes (46) and is critically involved in transducing some of the biological responses of the LIF/IL6 cytokine family, including IL6-induced terminal differentiation, growth arrest, and macrophage differentiation of myeloid leukemia M1 cells (18,26) and prevention of apoptosis in factor-dependent Ba/F03 cells (27). Similarly, phenylalanine substitutions or deletions of all YXXQ motifs in gp130 or LIFR␣, respectively, abolish the ligand-dependent transduction of differentiation promoting signals in M1 cells and the induction of acute-phase proteins in hepatoma cells (5). However, the presence of a single YXXQ motif retained ligand-dependent differentiation in M1 cells, but not ␤-fibrinogen expression in HepG2 cells (5). In response to IL6, STAT3 activation occurs rapidly and transiently in ES and HepG2 cells but persists for more than 24 h in M1 cells (26). The sustained nature of STAT3 activity may be important in causing growth arrest and terminal differentiation, as in the case of sustained MAPK activation in nerve growth factorinduced PC12 cell differentiation (47) and gp130-mediated prevention of ES cell differentiation (14). Gp130 and LIFR␣-dependent STAT3 activation requires at least one YXXQ consensus motif (51). Two of these motifs in gp130 (pY 126 RHQ and pY 173 FKQ) mediate activation of STAT3 while the other two motifs (pY 265 LPQ and pY 275 MPQ) are capable of mediating the activation of STAT3 and STAT1 (48). Thus, substitution of the tyrosine residue in the truncated G/gp130 ⌬130F receptor blunts the ligand-dependent formation of the SIF-A, SIF-B, and SIF-C DNA-binding complexes (Fig. 5 and Ref. 19). STAT1 appears to be a minor component of the total STAT proteins activated by gp130. It is therefore difficult to assess the role of STAT1 in ES cell directly, because (as observed in M1 cells) 2 M. Ernst, unpublished observation.

FIG. 8. [ 3 H]Thymidine incorporation into Ba/F03 cells expressing mutant GCSF-R.
Clonally derived Ba/F03 cell lines expressing the indicated GCSF-R mutants were plated in 96-well plates and stimulated for 48 h with 10 ng/ml GSF. Triplicate cultures were labeled with [ 3 H]thymidine for the last 4 h and incorporated acid-precipitable radioactivity was determined in a ␤-scintillation counter. Similar results were obtained with at least one more independently derived cell line for each receptor construct, mean Ϯ S.D.
FIG. 9. GCSF-dependent STAT3 phosphorylation and in ES cells expressing mutant GCSF-R. Confluent cultures of ES cells, expressing the indicated GCSF-R construct, were starved for 12 h of LIF in medium containing 1% FCS before stimulating for 10 min with 20 ng/ml GCSF (G), 500 units/ml LIF (L), or saline (c). Cell lysates were immunoprecipitated with anti-STAT3 antibodies, separated by SDS-PAGE, transferred to nitrocellulose membranes, and blotted with antiphosphotyrosine antibodies (Tyr(P)). Blots were stripped and reprobed to assess for equal amounts of STAT3. The results for one representative cell line transfected with the indicated receptor construct are shown. STAT1 activity is weak and only detected when STAT3 is highly activated. STAT3-dependent induction of M1 cell differentiation and suppression of Ba/F03 cell apoptosis appears to be independent of the number and position of these motifs in gp130 (18,26,27). In ES cells, however, a single YXXQ motif outside of the Box 3 motif (G/LIFR␣ ⌬136 , G/mpl) failed to suppress differentiation, while the single YXXQ motif within the Box 3 homology domain (G/gp130 ⌬156 ) partially retained the activity in ES cells. These findings are reminiscent of observations that only Box 3 containing receptors, but not c-Mpl, induced transcriptional activation of acute-phase protein genes (5).
The presence of the common Box 3 motif in the cytoplasmic domain of GCSF-R, LIFR␣, and gp130 is likely to contribute to the remarkably similar pattern of biological activities transduced in a number of cell systems (4,5,20,29). However, GCSF-Rmediated STAT3 phosphorylation and M1 cell differentiation is less tightly correlated with individual tyrosine residues when compared with gp130 and LIFR␣ (29) and the Y74F substitution of the only YXXQ motif in the human GCSF-R did not have a significant effect on STAT3 phosphorylation and GCSF-mediated inhibition of ES cell differentiation. By contrast, the Y 114 LRC 3 FLRC substitution within the Box 3 sequence decreased STAT3 phosphorylation and GCSF-mediated suppression of differentiation. Thus, the regulation of ES cell differentiation by the cytoplasmic domain of the Box 3-containing gp130, LIFR␣, and GCSF-R appears to depend largely on STAT3, however, the mechanisms by which GCSF-R mediates STAT3 activation may be cell type-specific and more complex than SH2 domain-dependent binding to the YXXQ motif. In fact, GCSF-R-mediated STAT3 activation is not observed in neutrophils (40), while STAT3 and STAT5 activation have been observed with versions of the GCSF-R, the ␤-chain of the granulocyte-macrophage CSF receptor, and c-Mpl which contain substitutions of all cytoplasmic tyrosine residues (29,49,50).
While distinct, phosphotyrosine-mediated intracellular signaling cascades appear to be the predominant mechanism by which mitogenic signals are dissociated from those not affecting cell growth, we present evidence that seemingly identical signaling mechanisms may regulate signals that suppress as well as promote cellular differentiation. Further studies are required in order to establish whether differences in the strength of activation of a particular signaling pathway (as observed here for STAT3) may help to distinguish between the diverse biological outcomes mediated by these signaling intermediates.