Distinct Roles for Leukemia Inhibitory Factor Receptor α-Chain and gp130 in Cell Type-specific Signal Transduction*

Leukemia inhibitory factor (LIF) induces a variety of disparate biological responses in different cell types. These responses are thought to be mediated through the functional LIF receptor (LIFR), consisting of a heterodimeric complex of LIFR α-chain (LIFRα) and gp130. The present study investigated the relative capacity of the cytoplasmic domains of each receptor subunit to signal particular responses in several cell types. To monitor the signaling potential of LIFRα and gp130 individually, we constructed chimeric receptors by linking the extracellular domain of granulocyte colony-stimulating factor receptor (GCSFR) to the transmembrane and cytoplasmic regions of either LIFRα or gp130. Both chimeric receptors and the full-length GCSFR in expressed in M1 myeloid leukemic cells to measure differentiation induction, in embryonic stem cells to measure differentiation inhibition, and in Ba/F3 cells to measure cell proliferation. Our results demonstrated that whereas GCSFR-gp130 receptor homodimer mediated a GCSF-induced signal in all three cell types, the GCSFR-LIFRα receptor homodimer was only functional in embryonic stem cells. These findings suggest that the signaling potential of gp130 and LIFRα cytoplasmic domains may differ depending upon the tissue and cellular response initiated.

A characteristic feature of cytokines such as leukemia inhibitory factor (LIF) 1 is their ability to regulate a wide range of biological activities (1). The diverse effects of LIF include both stimulation and inhibition of cellular proliferation (2,3) and activation of cell type-specific gene expression (4). LIF also induces macrophage differentiation in M1 myeloid leukemia cells (5), whereas it elicits an opposite effect in embryonic stem (ES) cells, maintaining these cells in an undifferentiated, pluripotent state (6,7).
In addition to functional pleiotropy, the biological actions of LIF and related cytokines, such as interleukin (IL)-6, IL-11, oncostatin M (OSM), cardiotrophin-1, and ciliary neurotrophic factor (CNTF), are largely overlapping. The common activities of the LIF family of cytokines have been attributed in part to the existence of multimeric receptors, which share the affinity converting and signal transducing subunit, gp130 (8 -11). These receptors can be divided into three distinct types (12). First, LIF, OSM and cardiotrophin-1 each use receptors consisting of a heterodimeric complex of gp130 with the LIF receptor ␣-chain (LIFR␣, sometimes known as LIFR␤). In addition, OSM has been shown to signal through an alternative receptor complex, consisting of a heterodimer of a ligand-specific subunit (OSM receptor ␤-chain) and gp130 (13). Second, CNTF binds to a ligand-specific subunit (CNTF receptor ␣-chain), which associates with a heterodimer of LIFR␣ and gp130. In contrast, functional IL-6 and IL-11 receptors are formed by the association of ligand-bound ␣-chains with gp130 homodimers, with no involvement of LIFR␣. Unlike the ligandbinding components of the IL-6, IL-11, and CNTF receptors, which do not contribute to intracellular signaling, LIFR␣ contains an extensive cytoplasmic domain with a structure similar to both gp130 and the GCSFR (14).
The separate contributions of the LIFR␣ and gp130 cytoplasmic domains to LIF-induced signal transduction have not been investigated in detail. Although it has been established that mutant LIFR␣ lacking a cytoplasmic domain is inactive (15), its relative capacity compared with gp130 for triggering diverse biological outcomes is less well established. Both receptor subunits have the ability to associate with and activate the Janus kinases Jak1, Jak2, and Tyk2 as well as several other tyrosine kinases (16), suggesting that common signaling pathways may be triggered by each receptor component (17). Despite their molecular similarity, it is possible that the two receptor chains are not functionally equivalent, with one subunit having a greater potential to transduce particular responses. Differences in the signal transduction pathways triggered by LIFrelated cytokines have been implicated by previous studies (18,19), in which enforced expression of the SCL transcription factor in M1 cells reduced the ability of these cells to differentiate in response to LIF and OSM (signaling through LIFR␣-gp130 heterodimers) but not IL-6 (signaling through gp130 homodimers).
In the present study, we investigated the relative potential of LIFR␣, gp130, and GCSF receptor chains to signal cell typespecific responses. For this purpose, we expressed chimeric receptor constructs that comprised the extracellular domain of GCSFR and the transmembrane and cytoplasmic regions of either LIFR␣ or gp130. This approach enabled us to drive GCSF-dependent homodimerization of chimeric receptor subunits independently of endogenous receptor chains. The validity of this strategy had been established by previous studies (20,21), including those demonstrating that human hepatoma cells expressing similar chimeric receptors acquired GCSF responsiveness (15,22). Our results demonstrate that the signaling potential of LIFR␣, gp130, and GCSFR homodimers varies in different cell types, with LIFR␣ homodimers playing an active role in suppressing ES cell differentiation but having a reduced potential to induce macrophage differentiation compared with either gp130 or GCSFR homodimers.

EXPERIMENTAL PROCEDURES
Cell Culture and Cytokines-M1 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FCS). Ba/F3 cells were cultured in RPMI 1640 medium supplemented with 10% FCS and 10% WEHI-3B D Ϫ conditioned medium as a source of IL-3. The ES cell lines, derived from the parental line E14TG2a, which contains a null mutation in the hypoxanthine-phosphoribosyl transferase gene (hprt Ϫ ; Ref. 23), were passaged in ES cell medium (DMEM containing 15% FCS, 0.1 mM 2-mercaptoethanol, and 1000 units/ml LIF) in the absence of feeder cells. Recombinant murine LIF was produced in Escherichia coli and purified as described previously (24). Purified recombinant human GCSF, used for all biological assays, was the gift of AMGEN, and purified recombinant IL-3 was purchased from PeproTech Inc. (Rocky Hill, NJ). Mouse GCSF was produced in Pichia pastoris (25), purified, and then iodinated for use in binding assays.
DNA Constructs-Chimeric receptors were constructed by cloning the HindIII-XbaI fragment of the murine GCSFR (26) into pBLUE-SCRIPT (SK)ϩ. A silent mutation was then introduced at nucleotide ϩ2045, to create a BamHI restriction enzyme site near the start of the transmembrane region (pBS/mGR(Bam)). BamHI sites were also introduced into murine gp130 at position ϩ1851 and in human LIFR␣ (14) at position ϩ2489 by nucleotide substitution in the region immediately preceding the transmembrane domain of each molecule. Expression constructs were generated by annealing the BamHI site of the GCSFR extracellular domain with the transmembrane/cytosolic domain of either gp130 or LIFR␣ (construct 1) and ligating the chimeric receptor cDNA into the SalI site of the expression vector 6P-IRESneo-BS, driven by a PGK promoter (27). An additional GCSFR-LIFR␣ construct (construct 2) was generated by polymerase chain reaction amplification of a fragment of the human LIFR␣ cDNA containing the transmembrane and cytoplasmic regions. Polymerase chain reaction primers contained an in-frame BamHI site at the 5Ј end (5Ј-ACGTGGATCCATCTGACGT-GGGATTAATTATTGCCATT-3Ј) and an XbaI site at the 3Ј end (5Ј-AGCTTCTAGACTGTTAATCGTTTGGTTTG-3Ј). This fragment was then inserted into BamHI-XbaI digested pBS/mGR(Bam). The GCSFR-LIFR␣ fragment was released by digestion with HindIII and XbaI, and the ends were filled in with Klenow (Promega) and ligated into the expression vector pEF-BOS (28) using BstXI adaptors (Invitrogen). Although the two GCSFR-LIFR␣ constructs differed slightly, because construct 1 contained a Pro 3 Leu substitution at amino acid 599 and a deletion of amino acid 601 at the 3Ј end of the GCSFR extracellular domain, no difference between the constructs was observed in a colony assay in the presence of GCSF following stable transfection of M1 cells. Similarly, ES cells were transfected with both constructs, and expression of either construct was found to confer a GCSF-responsive phenotype in these cells. The sequence of all chimeric constructs was verified using an ABI PRISM Dye Terminator Cycle Sequencing Kit (Perkin-Elmer) and a model 373 Automated DNA Sequencer (Perkin-Elmer). cDNA encoding the full-length human GCSFR (a gift from S. Nagata, Osaka Bioscience Institute) was cloned into the XbaI site of the pEF-BOS expression vector. The pHCK-hprt plasmid was created by substituting the PGK promoter in the PGK-hprt expression cassette of pGEM-7Z with an XhoI-BamHI fragment of the murine hck promoter, encompassing the region between Ϫ645 and ϩ240 relative to the major transcription initiation site (29).
Stable Transfection of Cell Lines-M1 and Ba/F3 cells were transfected with plasmids expressing either full-length GCSFR, GCSFR-LIFR␣ (construct 2), or GCSFR-gp130 by electroporation, essentially as described (11). Receptor constructs were cotransfected with the pPGK-PuropA expression vector (kindly provided by Prof. S. Cory), and transfected cells were selected with 20 g/ml puromycin (Sigma). For all transfected cell lines, expression of the receptors was determined by assessing the ability of cells to bind 125 I-GCSF.
HCK-hprt ES cells were obtained by transfection of parental hprt Ϫ ES cells with the expression plasmid pHCK-hprt. Cells were selected in HAT medium (100 M hypoxanthine, 0.4 M aminopterin, 16 M thymidine). Individual resistant colonies were expanded and tested for HAT sensitivity after a 5-day induction of cellular differentiation in the absence of LIF. ES cells expressing chimeric constructs were obtained by electroporation of HCK-hprt cells with the 6P-IRESneo-BS-based expression constructs (Bio-Rad Gene Pulser; 270 V, 500 microfarad), followed by selection for 7 days in 175 g/ml Geneticin (Life Technologies, Inc.). ES cells were transfected with the GCSFR-LIFR construct 1. ES cell lines expressing the full-length GCSFR were obtained by coelectroporation with the plasmid PGKneo (30). Several independently derived M1, Ba/F3, and ES cloned cell lines expressing each receptor construct were selected for further analysis.
M1 Cell Colony Assays-To quantitate the differentiation of transfected M1 clones in response to cytokine, 300 cells were cultured in 35-mm Petri dishes containing 1 ml of DMEM supplemented with 20% FCS, 0.3% agar, and 0.1 ml of serial dilutions of LIF or GCSF. Cultures were incubated at 37°C in a humidified incubator containing 10% CO 2 for 7 days. The dishes were then scored for the percentage of differentiated colonies, as judged by colonies with a halo of dispersed cells. The total number of colonies was also determined to assess the degree to which proliferation had been extinguished by the addition of the cytokine.
ES Cell Assays-The extent to which cytokine-mediated signaling prevents ES cell differentiation was determined by both morphology (as described previously; Refs. 31 and 32) and MTT staining of undifferentiated, proliferating cells after HAT selection. For the MTT assay, cells were seeded in quadruplicate cultures at 1500 cells/cm 2 in gelatinized 24-well multiculture dishes (Nunc, Kamstrup, Denmark). Cells were grown for 6 days at the indicated concentration of GCSF in the absence of LIF, in medium supplemented with 2 ϫ HAT. At this time, Ͼ95% of the morphologically differentiated cells had died. The cultures were then supplied with 0.5 mg/ml MTT and incubated for 3 h at 37°C, after which the aspirated cultures were air-dried. The reduced MTT dye was solubilized in Me 2 SO, and the optical absorbance was measured at 560 nm and expressed as a percentage of the maximal absorbance measured in undifferentiated cultures maintained in 2.5 ng/ml LIF.
Proliferation Assays-The survival/proliferation of Ba/F3 cells in response to cytokine was measured in Lux 60 microwell HL-A plates (Nunc Inc., Roskilde, Denmark). Cells were washed three times in DMEM containing 20% newborn calf serum and resuspended at a concentration of 2 ϫ 10 4 cells/ml in the same medium. Aliquots of 10 l of cell suspension were placed in the culture wells with 5 l of serial dilutions of 1 ng/ml IL-3 or 100 ng/ml GCSF. After a 2-day culture at 37°C in a humidified incubator containing 10% CO 2 , viable cells were counted using an inverted microscope.
Electrophoretic Mobility Shift Assays-Assays were performed as described previously (33), using the high affinity c-sis-inducible factor binding site m67 (34). Protein extracts were prepared from M1 cells incubated with saline, 10 ng/ml LIF, or 100 ng/ml GCSF for 10 min at 37°C. For ES cells, cultures consisting of approximately 8 ϫ 10 7 undifferentiated ES cells were starved overnight in ES cell medium free of serum and LIF before being stimulated for 10 min at 37°C with saline, 10 ng/ml LIF, or 100 ng/ml GCSF. For certain experiments, protein samples were preincubated with antibodies specific for either STAT1 (Transduction Laboratories), STAT3 (Santa Cruz Biotechnology Inc., CA), or STAT5A (specific for the C terminus, a gift from Dr. A. Mui, DNAX Research Institute, Palo Alto, CA) as described (33).
Binding Assays-Binding assays were performed essentially as described (35). Approximately 2 ϫ 10 6 cells in 40 l of RPMI 1640 medium containing 20 mM Hepes, pH 7.4, and 10% FCS were incubated for 3 h on ice with varying amounts of 125 I-GCSF in the presence or the absence of 100-fold excess of unlabeled GCSF. Cell-associated and free 125 I were then separated by rapid centrifugation of the cell suspension through 200 l of FCS. The amount of 125 I in the cell pellet and the supernatant was quantitated in a ␥-counter. Scatchard analysis of saturation binding isotherms were performed using the computer program LIGAND (36).

RESULTS AND DISCUSSION
Roles of LIFR␣, gp130, and GCSFR Cytoplasmic Domains in M1 Cell Differentiation-Stimulation of M1 myeloid cells with LIF or IL-6 induces macrophage differentiation and an inhibition of cellular proliferation (5). M1 cells were transfected with either the GCSFR-LIFR␣, GCSFR-gp130, or wild type GCSFR constructs, and the expression of these receptors was confirmed by the ability of the transfected cells to bind 125 I-GCSF (Table  I). The capacity of the transfected cells to differentiate in response to GCSF was assessed by semi-solid agar colony assays. Untransfected parental M1 cells failed to respond to GCSF, and all cell lines expressing the chimeric receptors responded normally to LIF (data not shown). M1 cells expressing either full-length GCSFR (Fig. 1A) or GCSFR-gp130 (Fig. 1C) responded to GCSF in a similar manner, with complete differentiation and clonal extinction at higher concentrations of cytokine. In contrast, we were unable to detect a GCSF-induced response by cells expressing GCSFR-LIFR␣ receptors, because neither differentiation nor clonal suppression of M1 cells expressing these receptors was evident (Fig. 1B). The expression of characteristic macrophage markers, including Fc␥ receptor types I and II, Mac-1, and the macrophage colony-stimulating factor receptor, was also assessed in these cells in an effort to determine whether any aspects of differentiation were induced in response to GCSF. Flow cytometric analysis demonstrated that none of these markers were up-regulated by GCSF stimulation of these cells (data not shown). Furthermore, unlike cells expressing GCSFR, no change in expression of the flk-2 receptor was detected in cells expressing GCSFR-LIFR␣ in response to GCSF, as assessed by flk-2 ligand binding assays (data not shown). Thus, by all criteria examined, no evidence for the induction of differentiation through GCSFR-LIFR␣ receptors in M1 cells could be demonstrated. Collectively, these data suggest that homodimerization of GCSFR or gp130 cytoplasmic domains, but not LIFR␣ cytoplasmic domains, is sufficient to induce macrophage differentiation in M1 cells.

Roles of LIFR␣, gp130, and GCSFR Cytoplasmic Domains in ES Cell Differentiation-In previous studies, the degree of cellular differentiation of ES cells in vitro had been quantitated primarily by morphological inspection of individual cells and colonies.
To utilize a chemical selection protocol allowing for the survival of undifferentiated ES cells only, we used the observation that the murine hck gene undergoes transcriptional down-regulation following the induction of ES cell differentiation in vitro (M.E., unpublished observation). Thus, HCK-hprt ES cell lines, which contain an hprt minigene driven by an 865bp proximal hck promoter fragment, are resistant to HAT selection if they remain in an undifferentiated state. The proportion of undifferentiated, proliferating cells in an ES cell culture can then be determined by optical absorbance measurement of the reduced mitochondrial stain, MTT.
Maintenance of the pluripotency of ES cells in vitro normally requires signaling initiated by the LIF family of cytokines (6,7,37). ES cell lines were transfected with either full-length GCSFR or the chimeric receptor constructs, and expression of the receptors was confirmed by the ability of transfected cells to bind 125 I-GCSF (Table I). In these cell lines, signaling through the endogenous LIFR␣-gp130 heterodimer was normal, because culture in LIF maintained the cells in an undifferentiated state (Fig. 1, D, E, and F). The capacity of GCSF to substitute for LIF and retard differentiation of the transfected cells was assessed. The dose-response curve obtained with GCSF stimulation was similar for both the cell morphology (Fig. 1, D, E, and F) and MTT (Fig. 1, G, H, and I) assays, with a half-maximal effect at approximately 63 pM GCSF. Thus, signaling through homodimerized cytoplasmic domains of LIFR␣, gp130, or GCSFR maintained up to 80% of all ES cell colonies in an undifferentiated state (Fig. 1, D-I). In contrast, parental ES cells did not respond to GCSF in either assay, indicating the absence of endogenous GCSFR expression in untransfected cells (data not shown). These data suggest that homodimerization of the cytoplasmic domains of either LIFR␣ or gp130 is sufficient to mediate a signal that maintains the pluripotentiality of ES cells in vitro. Futhermore, these data indicate that in addition to the LIFR, the GCSFR can transmit a signal that maintains the undifferentiated state of ES cells.
Roles of LIFR␣, gp130, and GCSFR Cytoplasmic Domains in Triggering Proliferation-The potential of LIFR␣, gp130, and GCSFR cytoplasmic domains to signal a mitogenic response was determined after introduction of the receptor constructs into the IL-3-dependent cell line, Ba/F3. Transfected cells were analyzed in a proliferation assay. As shown in Fig. 1 (J, K, and L), the full-length GCSFR was able to transduce a proliferative/survival signal comparable to that elicited through the endogenous IL-3 receptor. Ba/F3 cells expressing GCSFR-gp130 showed a weaker proliferative or survival response to GCSF, whereas no GCSFdependent proliferation or survival was observed for cells expressing the GCSFR-LIFR␣ chimera (Fig. 1, J, K, and L). The reduced ability of gp130 homodimers to generate a GCSF response parallels the transient response of Ba/F3 cells expressing gp130 to stimulation by IL-6 and soluble IL-6 receptor (38). This may be due to reduced expression of LIFR␣-gp130 signaling intermediates in Ba/F3 cells compared with M1 or ES cells, both of which express endogenous LIF receptors. This hypothesis is supported by preliminary results in the IL-6-dependent plasmacytoma cell line, 7TD1, that suggest that the GCSFR-LIFR␣ is active in these cells (data not shown).
Similar STAT Complexes Are Formed by Stimulation of Chimeric Receptors-Considering the differing potential of LIFR␣, gp130, and GCSFR cytoplasmic domains to mediate cell-type specific responses, we were interested to compare the activation pattern of the Jak/STAT signaling pathway in response to receptor dimerization. To characterize the STAT complexes induced in response to GCSF, extracts of M1 and ES cells transfected with various receptor constructs were examined in electrophoretic mobility shift assays, using the high affinity c-sis-inducible factor binding sequence, m67, as a probe (34). All DNA-protein interactions were specifically competed by an excess of unlabeled m67 probe (data not shown). Three main complexes were induced in M1 cells stimulated with LIF through the activation of the endogenous LIFR ( Fig. 2A). A similar pattern was observed upon GCSF stimulation of M1 cells expressing GCSFR-gp130 ( Fig. 2A) and for GCSF stimulation of M1 cells expressing wild type GCSFR (Fig. 2B). In contrast, no m67-binding complexes were formed in response to GCSF in M1 cells expressing GCSFR-LIFR␣ nor in parental M1 cells (Fig. 2A). Hence, activation of m67-binding complexes correlated with the expression of active cell surface receptors, suggesting that STAT molecules may act as downstream effectors of the response.  ND  4524  97  GCSFR-LIFR  3030  110  1060  280  963  303  GCSFR-gp130  2600  1100  850  430  763  350 a ND, not determined.
The molecular nature of the STAT complexes was further investigated by the addition of antibodies specific for individual STAT proteins. The pretreatment of M1 cell extracts with antibodies specific for STAT1 supershifted the two minor lower bands to slower migrating complexes, indicating that these bands contain STAT1 (Fig. 2B). Similarly, the upper two bands were supershifted by the addition of an antibody recognizing STAT3 (Fig.  2B). The most slowly migrating DNA complex was also affected by binding of an anti-STAT5A antibody, suggesting that this complex comprises a STAT3-STAT5A heterodimer. The pattern and composition of m67-binding complexes induced by either GCSF, via the introduced receptors, or LIF stimulation, through endogenous LIF receptors, were identical for M1 cells expressing either GCSFR-gp130 or wild type GCSFR (Fig. 2B). The same signaling molecules were activated by either LIF or GCSF stimulation of ES cells expressing GCSFR-gp130 (Fig. 2C). Furthermore, identical complexes were induced by GCSF stimulation of ES cells signaling through GCSFR-LIFR␣ (Fig. 2C). The level of STAT activation induced by this chimeric receptor was comparatively weaker, despite its comparable capacity to retard differentiation (Fig. 1, D-I).
It is clear from this study that despite a high degree of sequence similarity, the signaling potential of the cytoplasmic domains of LIFR␣, gp130, and GCSFR differ depending on the cell type as well as the type of biological response initiated. Differences in the signal transduction pathways triggered by gp130 homodimers and gp130-LIFR␣ heterodimers have previously been suggested by studies in which overexpression of SCL in M1 cells inhibited the response to LIF and OSM, which use gp130-LIFR␣ heterodimers, but did not affect signaling through the IL-6 receptor, which uses gp130 homodimers (18,19). LIFR␣ alone may not be sufficient for transducing a differentiative signal in M1 cells nor for induction of a proliferation/survival signal in Ba/F3 cells, because homodimerized LIFR␣ cytoplasmic domains were unable to signal these responses independently of gp130. However, homodimers of either LIFR␣ or gp130 cytoplasmic domains were able to deliver the signal to block differentiation in ES cells. In addition, a previous report has demonstrated that either of these cytoplasmic domains can induce an acute phase response in hepatoma cells and mediate signaling in neuronal cells when activated as homodimers (22).
The cell-specific differences in activity described here are unlikely to be due to differences in the level of receptor expression, because Scatchard analyses of binding isotherms indicated a similar level of expression of the chimeric receptors in each cell type and a comparable binding affinity for GCSF (Table I). Because none of the three cell lines express detectable levels of endogenous full-length GCSFR, dimer formation between the extracellular domains of endogenous and introduced receptor subunits should not occur. Furthermore, cooperativity between the cytoplasmic domains of introduced receptor subunits and endogenous LIFR␣ or gp130 is unlikely, given that the GCSFR-LIFR␣ receptor is unable to signal in M1 cells, which contain both endogenous LIFR␣ and gp130.
The observed differences in signaling potential of the homodimeric cytoplasmic domains may be due to either quantitative or qualitative differences in the signaling molecules activated. First, different biological responses may require a different threshold of STAT activation to be attained, with certain responses requiring a higher level of activation. The activation of STAT molecules by homodimeric LIFR␣ cytoplasmic domains may fail to reach the required threshold in M1 or Ba/F3 cells for a biological response to occur, perhaps due to a reduced efficiency of the LIFR␣ chain to activate STATs or a reduced pool size of signaling intermediates. The number of available STAT molecules in ES cells may be greater than that in either M1 or Ba/F3 cells. Although the LIFR␣ chain may be less efficient at activating STATs, the increased availability of STATs, coupled with a lower threshold requirement for biological activity in these cells, could explain the ability of the GCSFR-LIFR␣ receptor to signal in ES cells.
Alternatively, different signaling intermediates may be activated by the different receptor subunits. Activation of pathways necessary for M1 cell differentiation and Ba/F3 proliferation may require gp130 cytoplasmic sequences, whereas both LIFR␣ and gp130 may be able to interact with the specific cytoplasmic intermediates required for signal transduction in ES, hepatoma, and neuronal cells. Verification of this model would require the identification of the relevant signaling path-ways. In either case, the results described here clearly show that LIFR␣ homodimers are less efficient than either gp130 or GCSFR homodimers in signaling biological responses. Further investigation of the signaling molecules activated by LIFR␣ and gp130 is needed to fully document the different signaling potentials of these receptor subunits.