Coupling of heterotrimeric Gi proteins to the erythropoietin receptor.

To identify new proteins involved in erythropoietin (Epo) signal transduction, we purified the entire set of proteins reactive with anti-phosphotyrosine antibodies from Epo-stimulated UT7 cells. Antisera generated against these proteins were used to screen a lambdaEXlox expression library. One of the isolated cDNAs encodes Gbeta2, the beta2 subunit of heterotrimeric GTP-binding proteins. Gbeta and Galpha(i) coprecipitated with the Epo receptor (EpoR) in extracts from human and murine cell lines and from normal human erythroid progenitor cells. In addition, in vitro Gbeta associated with a fusion protein containing the intracellular domain of the EpoR. Using EpoR mutants, we found that the distal part of the EpoR (between amino acids 459-479) was required for Gi binding. Epo activation of these cells induced the release of the Gi protein from the EpoR. Moreover in isolated cell membranes, Epo treatment inhibited ADP-ribosylation of Gi and increased the binding of GTP. Our results show that heterotrimeric Gi proteins associate with the C-terminal end of the EpoR. Receptor activation leads to the activation and dissociation of Gi from the receptor, suggesting a functional role of Gi protein in Epo signal transduction.

From the ‡INSERM, U 363 and §CNRS-UPR 0415, Institut Cochin de Génétique Moléculaire, 75014 Paris, France To identify new proteins involved in erythropoietin (Epo) signal transduction, we purified the entire set of proteins reactive with anti-phosphotyrosine antibodies from Epo-stimulated UT7 cells. Antisera generated against these proteins were used to screen a EXlox expression library. One of the isolated cDNAs encodes G␤ 2, the ␤ 2 subunit of heterotrimeric GTP-binding proteins. G␤ and G␣ i coprecipitated with the Epo receptor (EpoR) in extracts from human and murine cell lines and from normal human erythroid progenitor cells. In addition, in vitro G␤ associated with a fusion protein containing the intracellular domain of the EpoR. Using EpoR mutants, we found that the distal part of the EpoR (between amino acids 459 -479) was required for G i binding. Epo activation of these cells induced the release of the G i protein from the EpoR. Moreover in isolated cell membranes, Epo treatment inhibited ADP-ribosylation of G i and increased the binding of GTP. Our results show that heterotrimeric G i proteins associate with the Cterminal end of the EpoR. Receptor activation leads to the activation and dissociation of G i from the receptor, suggesting a functional role of G i protein in Epo signal transduction.
Activation of the EpoR 1 elicits multiple intracellular signals that ultimately lead to cell division and differentiation of erythroid progenitor and precursor cells. One primary signaling event following receptor activation is the phosphorylation of certain cellular proteins on tyrosine residues (1,2). Interaction of Epo with its receptor results in the activation of the cytoplasmic tyrosine kinase JAK2 and in the phosphorylation of the intracellular domain of EpoR (3)(4)(5). Tyrosine phosphorylated residues on EpoR then constitute binding sites for other intracellular proteins that become eventually tyrosine phosphorylated. Other tyrosine kinases are also activated in response to Epo such as c-Fes (6), Lyn (7), and Syk (8). Tyrosine phosphorylated proteins may be recruited directly to the EpoR via their Src homology domains or indirectly through adaptor or scaffold proteins. Several tyrosine kinase substrates phosphorylated upon Epo activation have been identified including STAT5 transcription factor (9), SHP-2 tyrosine phosphatase (10), Shc (11), phospholipase C-␥1 (12), Vav (13), c-Cbl (14), IRS-2 (15), GAB-1 (16), and CrkL (17). Yet the relations between all the components involved in Epo signaling as well as the identification of their respective targets are only partially elucidated, and EpoR may activate other additional signal transduction pathways.
In an attempt to characterize substrates for epidermal growth factor (EGF) receptor, Fazioli et al. (18) developed an expression cloning strategy for cDNAs encoding EGF receptor substrates. The approach relied on batch purification of an entire set of putative substrates, achieved by immunoaffinity chromatography using anti-phosphotyrosine antibodies (19,20). Antisera generated against the entire pool of purified proteins were subsequently used for the screening of cDNAs expression libraries. We applied this methodology to identify new cDNAs encoding signaling proteins involved in Epo activation, either tyrosine phosphorylated proteins or proteins bound to these proteins. In the present work we report that one of these cDNAs encodes G␤ 2 , the ␤ 2 subunit of heterotrimeric GTPbinding proteins, or G proteins.
G proteins traditionally associate with G protein-coupled receptors (GPCRs) that contain seven membrane-spanning domains. G proteins function as intermediates that couple cell surface receptors to intracellular effectors. Heterotrimeric G proteins are made of three polypeptides: an ␣ subunit that binds and hydrolyzes GTP, and ␤␥ subunits that form a functional monomer. Receptor activation induces the exchange of GDP for GTP on the G ␣ subunit. Once GTP is bound, the ␣ subunit dissociates both from the receptor and from ␤␥. The free ␣ and ␤␥ subunits each activate target effectors. However, a number of single-spanning transmembrane receptors such as receptors for EGF (21), insulin and insulin-like growth factor (IGF)-I and IGF-II (22)(23)(24)(25), fibroblast growth factor (26), and T lymphocyte receptors (27,28) have been reported to activate G proteins. In some cases a physical association, in addition to a functional coupling, has also been demonstrated between a single-spanning membrane receptor and G proteins (29 -32).
Heterotrimeric G proteins could be important intermediates in the signal transduction of hematopoietic cytokines. Changes in the expression level and GTPase activity of G␣ 16 , a member of the G q family of G proteins uniquely expressed in hematopoietic cells, may modulate cellular proliferation or differentiation in T lymphocytes and in MB-02 erythroleukemia cells (33,34). Pertussis toxin (PT) modifies the response to several hematopoietic growth factors. PT catalyzes the ADP-ribosylation of the G i family of G proteins and uncouples G proteins from surface receptors. PT inhibits the signal transduction and/or proliferation induced by interleukin (IL)-1, IL-3, granulocyte-macrophage colony-stimulating factor (GM-CSF) and colonystimulating factor-1 (CSF-1) (35)(36)(37)(38) in hematopoietic cells. Expression of a dominant negative mutant of G␣ i2 also inhibits cell proliferation in response to CSF-1 in BAC 1.2F5 macrophage cell line (39). In erythroid precursor cells a pertussis toxin-sensitive G protein identified as G␣ i2 is required for the regulation of voltage-independent calcium channels by Epo (40,41). The increase in [Ca 2ϩ ] i appears to be a stage of differentiation specific and restricted to differentiating erythroblasts (42).
In the present study we demonstrate the constitutive association of G i with the EpoR in hematopoietic cell lines as well as in erythroid progenitors. We show that the C-terminal region of EpoR is required for G i protein binding. In addition, Epo activates G protein in cell membranes and induces the release of G i bound to the EpoR in hematopoietic cells. Thus, EpoR appears to be physically and functionally coupled to G proteins.

EXPERIMENTAL PROCEDURES
Antibodies and Reagents-Anti-phosphotyrosine (anti-Tyr(P)) monoclonal antibodies 4G10 and PY72 were produced from hybridoma cell lines kindly provided by B. Drucker (Portland, OR) and B. Sefton (La Jolla, CA), and were affinity purified by chromatography on phosphotyramine. Anti-EpoR antiserum used for immunoprecipitation was produced against a fusion protein between glutathione S-transferase and the cytoplasmic portion of human EpoR. Peroxydase-conjugated antirabbit antibodies were purchased from Amersham Pharmacia Biotech. Antibodies specific for human EpoR used for immunoblotting were purchased from Santa Cruz (sc-695) and anti-JAK2 antiserum from Upstate Biotechnology Inc. (catalog number 06-255). The fusion protein between the maltose-binding protein (MalE) and the cytoplasmic region of EpoR was described previously (43). Purified recombinant human Epo (specific activity, 120,000 units/mg) was a gift of Dr. M. Brandt (Roche Molecular Biochemicals). Pertussis toxin was purchased from Alexis, ATP was from Amersham Pharmacia Biotech, and GTP and GTP␥S were from Sigma.
DNA Constructs and Expression Vectors-The murine EpoR mutant F1-Y58 that contains a deletion between Glu 377 and Tyr 431 was described previously (9). A panel of EpoR deletion mutants was produced from polymerase chain reaction-amplified fragments. In mutants Ϫ41, Ϫ24, Ϫ20, and Ϫ5, stop codons were inserted just after codons 442, 459, 463, and 478, respectively (see Fig. 6). All receptor constructs were subcloned into a modified pCDNA3 expression vector where the cytomegalovirus promotor was changed to Rous sarcoma virus and the neomycin resistance gene was replaced by puromycin. The fidelity of all constructs was confirmed by sequencing.
Cell Lines and Stimulation-The human leukemic cell line UT7 (44) was maintained in ␣-minimum essential medium supplemented with 5% fetal calf serum, penicillin, streptomycin, 2 mM L-glutamine, and 2 units/ml Epo. TF-1 (45) and MO7E cell lines (46) were cultured in ␣-minimum essential medium supplemented with 10% fetal calf serum and 2.5 ng/ml GM-CSF. TF-1-ER (47) and MO7E-ER (8) cell lines were obtained after infection of TF-1 and MO7E with an amphotropic virus encoding a murine EpoR and were cultured with Epo. FDCP-1 myeloid cells were grown in ␣-minimum essential medium supplemented with 5% fetal calf serum and 3% WEHI conditioned medium as a source of IL-3. After transfection with EpoR expression vectors and selection in Epo, cells were grown in the presence of Epo. 32D myeloid cells expressing wild-type EpoR and W282R mutant (48) were generously provided by G. D. Longmore (St. Louis, MO). Erythroid progenitors were purified from human umbilical cord blood cells as described previously (49). Briefly CD34ϩ cells were cultured for 7 days in serum-free conditions in the presence of interleukin-3, interleukin-6, and stem cell factor. CD36ϩ cells were then purified and expanded for 2 additional days in the presence of the same cytokines plus 2 units/ml Epo. CHO-ER cell line, a kind gift of E. Goldwasseur (Chicago, IL), was cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
For Epo triggering experiments, exponentially growing UT7 or TF-1 cells were washed and incubated for 16 -18 h in Iscove's modified Dulbecco's medium supplemented with 0.4% bovine serum albumin and 20 g/ml iron saturated transferrin, in serum-free conditions. Normal erythroid progenitors were washed and incubated for 4 h in serum-free Iscove's modified Dulbecco's medium in the presence of 5% bovine serum albumin, 50 g/ml insulin, and 1 mg/ml transferrin. FDCP-1 cells expressing EpoR and MO7E-ER cells were Epo starved by replacing Epo, respectively, with WEHI conditioned medium and GM-CSF one to 2 days prior to stimulation. The cells (1 ϫ 10 7 /ml) were stimulated with 10 units/ml Epo at 37°C as described in the text.
Immunoaffinity Chromatography-Cells (1 ϫ 10 9 ) were lysed on ice with buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 1 mM orthovanadate, 2 mM EGTA, 30 mM disodium pyrophosphate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptine, 10 g/ml aprotinine, 10 g/ml EG4, and 1 g/ml pepstatine. Anti-Tyr(P) columns were prepared by covalently cross-linking affinity purified PY72 monoclonal antibody to Sepharose beads (Hi Trap TM NHS-activated Sepharose from Amersham Pharmacia Biotech). Solubilized proteins were applied to the anti-Tyr(P) column (1 ml) using a fast protein liquid chromatography system. The column was washed successively with 10 column volumes of lysis buffer and 30 volumes of washing buffer (same as lysis buffer with 0.1% Nonidet P-40 instead of 1%). Then elution was done with the washing buffer supplemented with 40 mM phenyl phosphate. Fractions were collected and analyzed for protein content by silver staining and for Tyr(P) content by immunoblotting. The peak of tyrosine phosphorylated proteins, which coincided with the peak of eluted proteins, was pooled. Protein quantification was carried out with the micro BCA assay (Pierce).
Rabbit Immunization and Antisera Analysis-Anti-Tyr(P) reactive proteins (1 ϫ 10 10 cell equivalents) in complete Freund's adjuvant were used to immunize two New Zealand White rabbits as follows. Two intradermic injections (50 g) were followed by footpad injections with 25 g of proteins. Bleeds were collected 15 days after each boost and screened for the production of antibodies. Because expression cloning relies on the detection of denatured antigens, immune antisera were tested by immunoblotting on UT7 cell lysates. The rabbit antiserum from the rabbit with the highest antibody titer and the lowest reactivity of the preimmune serum was selected for further screening of the cDNA library. This serum was also tested for its ability to precipitate proteins known to be tyrosine phosphorylated following Epo stimulation and was shown to contain anti-Shc antibodies (data not shown). IgG were purified and antibodies reacting with bacterial proteins were removed by absorption on column of bacterial proteins cross-linked to Sepharose beads (Hi Trap NHS-activated Sepharose from Amersham Pharmacia Biotech).
EXlox Library Screening-Absorbed IgG were used to screen a commercial (Novagen RD Systems Europe) 16-day-old murine embryo cDNA library in EXlox vector, according to the manufacturer's instructions. Screening conditions were chosen to get an optimal signal-tonoise ratio and no reactivity with preimmune serum in immunoblot. This was obtained by initially testing the reactivity of the selected serum with a Shc clone previously isolated from the same library, a kind gift of J. Finidori (Paris, France). Briefly 2 ϫ 10 6 recombinant plaques were initially screened with IgG (3 g/ml) in 25 mM Tris-HCl, pH 7.5, 0.1% Tween, 140 mM NaCl, 3 mM KCl containing 5% (w/v) low fat powder milk. Colorimetric detection was carried out with anti-rabbit IgG conjugated to alkaline phosphatase and the bromochloroindolyl phosphate/nitro blue tetrazolium substrate (Promega). A second screen confirmed the reactivity of positive phages. Selected phages were plaque purified by conventional methods, and autosubcloning in plasmid vector was generated with the loxP-cre system. The insert cDNA sequence was determined using PerkinElmer Life Sciences automatic sequencing. Nucleotidic and protein data bases were screened with the BLAST program.
Preparation of Cell Membranes-After washing UT7 cells in phosphate-buffered saline, the cells were suspended in hypotonic buffer (10 mM Tris-HCl, pH 7.4, 10 mM KCl, 2 mM EDTA) in the presence of protease inhibitors and homogenized with a Dounce pestle. After addition of 0.25 M sucrose, nuclei and unbroken cells were removed by centrifugation at 375 ϫ g, and a membrane enriched fraction was obtained by centrifuging the supernatant at 150,000 ϫ g for 45 min. The resulting membrane pellet was stored at Ϫ80°C in 25 mM Hepes, pH 7.5, 150 mM NaCl, 5 mM EGTA, 10% glycerol or resuspended in appropriate buffer and used immediately for [ 35 S]GTP␥S binding. CHO-ER membranes were prepared as described previously (50).
ADP-ribosylation Assay-ADP-ribosyltransferase activity was measured by following the incorporation of [ 32 P]ADP-ribose (51). To assess the activity present in immunoprecipitates, immune complexes bound to protein G-Sepharose were washed in 50 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 1 mM ATP, and 0.1 mM GTP. Immunoprecipitated proteins were suspended in 50 l of the same buffer containing 2 Ci of adenylate [ 32 P]NAD (PerkinElmer Life Sciences; 800 Ci/mmol) and 7 g/ml activated pertussis toxin. The toxin (17 g/ml) was preactivated immediately before use for 1 h at room temperature in 50 mM Tris-HCl, pH 7.5, containing 62.5 mM dithiothreitol. ADP-ribosylation was carried out at 37°C for 1 h, and the reaction was stopped by adding SDS sample buffer. Samples were boiled for 5 min, and the proteins were separated on a 10% SDS-polyacrylamide gel followed by transfer to nitrocellulose and autoradiography.
To assess the effect of Epo on pertussis toxin-induced ADP-ribosylation in cell membranes, membranes (50 g) were initially incubated in 35 l of buffer A (25 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA, 100 M GTP, 10% glycerol) with or without Epo for 10 min at 37°C. Other additions or deletions are as noted under "Results" and in the figure legends. Then membranes were combined with 35 l of 2ϫ buffer B (100 mM Tris-HCl, pH 7.5, 2 mM EDTA, 2 mM ATP, 0.2 mM GTP). After addition of adenylate [ 32 P]NAD and activated pertussis toxin in 1ϫ buffer B, ADP-ribosylation was performed as described above.
GTP␥S Binding Assay-The [ 35 S]GTP␥S binding was measured as described (52)

Screening of a Bacterial Expression Library with Antibodies Specific for Anti-Tyr(P) Reactive Fractions of Epo-stimulated
Cells-To purify proteins that become tyrosine phosphorylated upon Epo stimulation and their associated proteins, our initial concern was to get a high level of tyrosine phosphorylation. UT7 cells were selected for their high surface EpoR expression (ϳ7000 receptors/cell) and their ability to proliferate in response to Epo. An optimal system was obtained by stimulating the cells with 10 units/ml Epo for 10 min at 37°C and adding pervanadate (100 M vanadate and 50 M H 2 O2) for the last 2 min of stimulation. The low pervanadate concentration used did not modify the basal level of tyrosine phosphorylated proteins in the absence of Epo but allowed a significant increase of Epo-induced tyrosine phosphorylation (data not shown). Epostimulated cells were solubilized, the soluble fraction was loaded on anti-Tyr(P) column, and anti-Tyr(P)-reactive fractions were eluted with phenyl phosphate. As shown in Fig. 1A, most of tyrosine phosphorylated proteins present in Epo stimulated cells (lane 1) were purified on the anti-Tyr(P) column (lane 4). Comparison of the silver-stained protein profiles of the anti-Tyr(P) reactive fractions with the anti-Tyr(P) Western blotting profiles of the same fractions revealed that the purified proteins contained several proteins that were not tyrosine phosphorylated (data not shown). Proteins recovered from the column represented about 0.06% of solubilized proteins, a value in good agreement with EGF receptor substrates previously isolated with a similar procedure (20). Polyclonal antisera were generated using the entire pool of anti-Tyr(P) reactive proteins as immunogen (see "Experimental Procedures"). We next used these antibodies for immunological screening of a murine embryo cDNA library in EXlox vector and identified several positive plaques. One of the cDNA encoded the G␤ 2 subunit of heterotrimeric G proteins and is the subject of the present study. It contained the entire coding sequence of the G␤ 2 subunit of heterotrimeric G proteins and the 5Ј noncoding se-quence (53). We first wanted to exclude the possibility that the G␤ 2 clone was isolated because of the cross-reactivity of the antibodies used for expression cloning with a protein present in the immunogen but different from G␤ 2 . As shown in Fig. 1B the G␤ protein was detected by immunoblotting with anti-G␤ antibodies in UT7 cell lysates as well as in anti-Tyr(P) reactive proteins purified from the soluble fraction of Epo-stimulated cells. As expected G␤ was not present in a preparation of purified proteins isolated from nonstimulated cells. Thus, the isolation of the G␤ cDNA with the expression cloning strategy suggests that an heterotrimeric G protein could potentially be tyrosine phosphorylated or associated with a tyrosine-phosphorylated protein upon Epo activation.
G␤ Subunit Associates with EpoR-When G␤ was immunoprecipitated from Epo-stimulated UT7 cells, we never detected a tyrosine phosphorylated form of the protein by immunoblotting with anti-Tyr(P) antibodies. This suggests that G␤ is not phosphorylated upon Epo activation. Because G␤ was present in anti-Tyr(P) reactive fractions (Fig. 1B), we hypothesized that G␤ may have been copurified because of its association with another tyrosine-phosphorylated protein upon Epo stimulation. We then wanted to determine whether G␤ was associated with the EpoR, one of the highly phosphorylated proteins upon Epo activation (4). UT7 cells were incubated or not with Epo at 37°C for various times, and after cell solubilization, the EpoR was precipitated. As shown in Fig. 2A, immunoblotting with anti-G␤ antibodies revealed the presence of G␤ when cell ly- sates were precipitated with anti-EpoR antiserum but not with preimmune serum. The amount of G␤ coprecipitated with EpoR decreased following Epo activation, suggesting that G␤ is constitutively associated with EpoR and dissociates from the activated receptor. We extended our analysis to determine whether the interaction between G␤ and EpoR could also be evidenced in vitro. Proteins binding to the cytoplasmic region of the EpoR were isolated from UT7 cell lysates, using a recombinant fusion protein between MalE and the cytoplasmic region of EpoR or a control MalE protein bound to amylose resin (Fig.  2B). When bound proteins were analyzed by immunoblotting with anti-G␤ antibodies, G␤ was detected only when MalE-EpoR fusion protein was used. We conclude that G␤ binds to EpoR both in vivo and in vitro and that the interaction between G␤ and EpoR occurs through the intracellular region of EpoR.
G␣ i Subunit Associates with EpoR-G proteins are associated to seven transmembrane receptors in an heterotrimeric form where the ␣ subunit is associated to ␤/␥. Because EpoR is not a "classical" G protein-coupled receptor, we investigated whether only G␤ binds to the EpoR or whether the ␣ chain of G proteins was also associated to the EpoR. Several forms of G␣ were identified in UT7 cells by immunoblotting including G i , G s , and G q . Our initial attempt to detect G␣ coprecipitated with the EpoR by immmunoblotting was unsuccessful because of the high nonspecific background in the region of G␣ migration. We used another method to investigate whether G␣ was bound to the EpoR, assuming that G i/ would be a good candidate because a pertussis toxin-sensitive G protein has been shown to be required for Epo-dependent calcium activation in erythroid precursors (40). The ␣ subunit was detected by following the in vitro incorporation of [ 32 P]ADP-ribose in the presence of preactivated pertussis toxin. This toxin catalyzes the ADP-ribosylation of ␣ i/o/t subunits of G proteins. ␣ t expression is restricted to the nervous system, and ␣ o is not expressed in erythroid precursors or erythroleukemia cells (40,54). This was confirmed in UT7 cells. Indeed we did not detect ␣ o protein by immunoblotting with anti-G␣ o antibodies. In initial experiments we observed that a 41-kDa band was ADP-ribosylated in vitro in UT7 membranes. Labeling was linear for up to 1 h and pertussis toxin-dependent (data not shown). Therefore, this band is presumptively referred to as G␣ i , the ␣ subunit of the inhibitory guanine nucleotide regulatory protein.
The EpoR was immunoprecipitated from the soluble fraction of UT7 cells, and the ADP-ribosyltransferase activity present in the precipitates was measured. Fig. 3 illustrates that G␣ i coprecipitated with the EpoR in resting UT7 cells, and cell stimulation with Epo decreased the amount of ADP-ribosylated ␣ i associated with the EpoR. On the contrary, the amount of EpoR remained constant. The G protein-EpoR complex was recovered in the 150,000 ϫ g supernatant of solubilized cells and thus corresponded to solubilized complexes and not to membrane fragments (data not shown). These data show that G␣ i is constitutively bound to the EpoR, more likely as an ␣/␤/␥ heterotrimer because G␣ i is a better substrate for PT in its heterotrimeric form (51,55). They also suggest that the G protein dissociates from the EpoR upon activation. To exclude the possibility that the G protein was precipitated by anti-EpoR antiserum because the EpoR and the G protein share a common epitope, the association was also studied both in cells that express an exogeneous EpoR and in cells that do not express EpoR. In MO7E cells expressing a murine EpoR, ADPribosylated ␣ i was coprecipitated with EpoR but no 41-kDa band was detected in cells that do not express EpoR (Fig. 4A).
Having established an association between G protein and EpoR in cell lines that express endogenous or exogenous receptor, we wanted to determine whether G protein binding to EpoR holds true in normal human erythroid progenitors. CD36ϩ red cell progenitors were isolated after culture of CD34ϩ cells from umbilical cord blood (49) and then deprived of growth factors and stimulated or not with Epo. The data shown in Fig. 4B demonstrate that an association between ␣ i and EpoR also exists in normal erythroid progenitors and that Epo significantly reduces the amount of G i associated with the EpoR.
G i Binds to the C-terminal End of EpoR Cytoplasmic Region-To identify the region of the EpoR involved in G protein binding, we first examined G i association to EpoR in TF-1 cells (Fig. 5). This human erythroleukemia cell line was previously with anti-EpoR antiserum (EpoR) or preimmune serum (C). Immunoprecipitates were subjected to in vitro ADP-ribosylation as described under "Experimental Procedures." The proteins were separated by SDS-PAGE and transferred to nitrocellulose membrane, and G␣ i was visualized by autoradiography. EpoR protein was detected by immunoblotting.
shown to overexpress an abnormal EpoR caused by a deletion of the 96 C-terminal amino acids, together with a minor expression of full-length EpoR (47,56). TF-1 cells were solubilized, and the G␣ i protein coprecipitated with EpoR was detected by in vitro ADP-ribosylation in the presence of pertussis toxin. Very little G␣ i was detected in anti-EpoR immunoprecipitates either before or after Epo stimulation, suggesting that these cells have a defect in G␣ i binding to EpoR. When the cells were infected with a virus encoding a normal murine EpoR (TF-1-ER), G␣ i binding to the EpoR was restored. This suggests that the C-terminal end of EpoR is necessary for G␣ i binding to EpoR.
To examine this possibility in greater detail we next studied the association of G␣ i with wild-type murine EpoR or C-terminal deletion mutants (Fig. 6A) expressed in FDCP-1 cells. In FDCP-1 cells transfected with the murine EpoR deletion mutants Ϫ41 and Ϫ24, no G␣ i was found associated with EpoR, and a weak binding was detected in mutant Ϫ20 (Fig. 6B). The absence of binding was not due to a decrease in EpoR surface expression because FDCP-1 cells transfected with wild-type EpoR, Ϫ41, Ϫ24, and Ϫ20 EpoR C-terminal deletion mutants expressed, respectively, 695, 580, 685, and 1820 125 I-Epo binding sites on the cell surface. G␣ i binding to the Ϫ5 EpoR deletion mutant and to EpoR mutant F1-Y58 deleted between amino acids 377 and 431 was similar to the binding to the wild-type receptor. We controlled that a similar fraction of EpoR bound 125 I-Epo was precipitated with anti-EpoR antibodies for the different EpoR mutants. In addition G protein expression, as detected by anti-G␤ immunoblotting was similar in FDCP-1 cells transfected with either wild-type EpoR or EpoR deletion mutants (data not shown). These results show that the region between amino acids 459 and 479 in the C-terminal end of EpoR is necessary for heterotrimeric G protein binding to the EpoR. It cannot be excluded that amino acids 432-458, also present in F1-Y58 mutant, contribute to the binding.
Epo Induces the Release of G Protein from the EpoR-A decrease in G protein association with EpoR following Epo addition was constantly observed in the different hematopoietic cells studied. The kinetic of G protein binding to EpoR was analyzed in UT7 cells stimulated or not with Epo for various times at 37°C. The amount of G i coprecipitated with the EpoR in the soluble fraction was determined by measuring the 32 P incorporated in the 41-kDa protein following in vitro ADPribosylation (Fig. 7). Epo induced a rapid decrease in the amount of G␣ i bound to the EpoR. About 50% of the binding was lost after 10 min of incubation, and then the amount of binding decreased more slowly. We conclude that heterotrimeric G i is constitutively associated with EpoR and Epo induces the dissociation of EpoR-bound G i . To determine whether Epo-induced JAK2 tyrosine kinase activation is required for G i release from the EpoR, we investigated whether the decrease in EpoR-bound G i was observed in 32D cells expressing EpoR W282R mutant. This mutant has lost the ability to bind JAK2 and activate the kinase in response to Epo (Refs. 5 and 48 and data not shown). As shown in Fig. 8 Epo activation induced G i release in cells expressing the normal as well as the mutant receptor. Thus JAK2 activation is not required for the release of EpoR-bound G i . G Protein Activation by Epo-G protein-coupled receptors activate G proteins resulting in GDP exchange for GTP and heterotrimer dissociation in ␣ and ␤␥ subunits. To investigate whether the G proteins associated with EpoR are activated upon receptor stimulation with Epo, we first monitored alter- ations of pertussis toxin-catalyzed ADP-ribosylation in isolated cell membranes. This assay was used to assess G␣ i protein subunit conformational changes, because the ␣ subunit serves as a good substrate for the toxin only when G proteins are in the trimeric form (51,55). Results showed that preincubation of UT7 cell membranes with Epo inhibited subsequent ADP-ribosylation of the 40-kDa substrate (Fig. 9A, lanes 4 and 5). The action of Epo is dependant on the Mg 2ϩ concentration (Fig. 9A,  lanes 1-4), which is also essential for G i protein trimer dissociation (57). GTP␥S, the nonhydrolyzable GTP analog that causes ␣␤␥ subunit dissociation and irreversibly activates G i , inhibited pertussis toxin-catalyzed ADP-ribosylation independently of Epo. These data show that Epo induces an alteration in G␣ i and suggest that Epo activation leads to the heterotrimer dissociation.
GPCR-catalyzed activation of G proteins is associated with an enhancement of their GTP binding. We have therefore examined the effect of Epo on the nonhydrolyzable GTP analog GTP␥S binding to cell membranes. In the presence of Epo, the amount of nucleotide that specifically bound was increased by 18% in UT7 cell membranes and by 56% in CHO-ER cell membranes (Fig. 9B). The increase in [ 35 S]GTP␥S binding observed in Epo-activated CHO-ER cells is similar to the increase detected in HEK293 cells expressing a GPCR, the Mel1a melatonin receptor, following activation with melatonin (Ref. 52 and data not shown). DISCUSSION In the present investigation, we provide the first demonstration of the physical association between the heterotrimeric G protein of the G i family and the erythropoietin receptor, both in hematopoietic cell lines and in human normal erythroid progenitor cells. The G i protein associated with the EpoR is more likely in an heterotrimeric conformation. Indeed both G␤ and G␣ i , identified as a 41-kDa ADP-ribosylated band, coprecipitated with the EpoR. By utilizing natural and engineered truncated EpoR mutants, we showed that the intracellular C-terminal end of the EpoR is required for G protein association. The association between the G protein and EpoR is constitutive. Classical G protein-coupled receptors are believed to associate with G proteins following ligand activation (58), although they may in some cases be preassociated. Actually, the association between a G i protein and a single-spanning transmembrane receptor has previously been reported. A peptide corresponding to the intracytoplasmic sequence of ␤-1,4-galactosyltransferase, the sperm receptor for the mouse egg, bound a heterotrimeric G protein that contained the G␣ i subunit (30). G␣ i was found to copurify with the ␤ chain of insulin receptor isolated from adipocyte plasma membranes (31). A transient association between G␣ i and the EGF receptor also occurred in rat hepatocytes after ligand activation (29,59), and G␣ i and G ␤ were recently shown to constitutively bind the IGF-I receptor (60). A region of 18 amino acids in the C terminus of EpoR is necessary for G i binding to EpoR. A 14-residue sequence in the IGF-II receptor, with several basic residues, presents structural similarity with the terminal portion of the third cytoplasmic loops of most G-coupled receptors. This sequence activates and directly interacts with G i proteins (61, 62), but the EpoR cytoplasmic domain does not contain such a G protein-binding motif. Another possibility is that the interaction between G i and EpoR occurs through an adaptor protein bound to the EpoR. The tyrosine kinase Jak2 and the docking protein IRS2 constitutively bind to the EpoR membrane proximal region (5,15). Because G protein binding to EpoR involves the C-terminal end of the receptor, these proteins are unlikely to play a role of adaptor between G protein and EpoR, and another component whose nature is unknown could be required. The ability of the EpoR mutant W282R to bind G i in the absence of Jak2 binding also excludes Jak2 as a potential adaptor.
Epo induces the release of G i from the EpoR. Hallak and co-workers (60) have recently reported that heterotrimeric G i is constitutively associated with IGF-I receptor and that IGF-I also induces the release of the G␤␥ subunit from the IGF-I receptor. Jak2 tyrosine kinase is probably not required for G i protein release from the EpoR because this process still occurs in 32D cells that express the EpoR mutant W282R defective in Jak2 activation. It cannot be excluded that other tyrosine kinases involved in Epo activation such as Lyn, Syk, or c-Fes (6 -8) may be required for G i release. Although Jak2 plays a pivotal role in EpoR signaling (63,64), G i protein release from EpoR could participate in a new Jak2-independent pathway.
Following activation of heterotrimeric G proteins by classical GPCR, both the ␣ and the ␤␥ subunits dissociate from the receptor and activate target effectors (58,65). Epo inhibited PT-catalyzed ADP-ribosylation of G␣ i in UT7 cell membranes in a Mg 2ϩ -dependant manner, suggesting that the ␣ subunit also dissociates from the ␤␥ subunit complex of the heterotrimeric G protein after Epo activation. Similar data were obtained in membranes activated with other single-spanning transmembrane receptors, such as insulin and IL-1 (66,67), suggesting that other receptors that GPCR can modify G protein conformation and potentially induce trimer dissociation.
In addition to demonstrating the physical association of G i with the EpoR, our data show that Epo increases GTP␥S binding in cell membranes, providing evidence that EpoR can activate G proteins. Some studies have suggested that Epo may increase adenylate cyclase activity and cAMP levels (40,68), whereas others did not report any modification of cAMP (69). However, to our knowledge, no studies have reported a decrease in cAMP levels in response to Epo stimulation. In erythroid precursors a PT-sensitive G␣ i2 has been demonstrated to regulate an Epo-modulated Ca 2ϩ channel (40,41), and G protein activation by GTP␥S mimics the rise in [Ca 2ϩ ] i (41). These data support an important role for heterotrimeric G proteins in the signaling pathways of Epo. Epo-stimulated changes in [Ca 2ϩ ] i were not detected in nonhemoglobinized or poorly hemoglobinized early erythroblasts with a large proliferative capacity. In contrast, in hemoglobinized late erythroblasts with a reduced proliferative capacity, Epo increased [Ca 2ϩ ] i (42), suggesting that regulation of [Ca 2ϩ ] i may be restricted to rather mature precursors. Recently, tyrosine 460 in the intracytoplasmic domain of EpoR was shown to be critical for Epo-stimulated Ca 2ϩ influx, in CHO cells as well as in Ba/F3 hematopoietic cells transfected with EpoR mutants (70). Interestingly the C-terminal end of EpoR required for G i binding to EpoR includes the tyrosine 460 (Fig. 6). Although in erythroid precursors the role of G i protein in Epo-induced Ca 2ϩ activation has been clearly established, transducers that bind to Tyr 460 have not yet been identified (70). Future investigations should allow the identification of the components that participate in G protein and calcium activation by Epo.
Heterotrimeric G i proteins could be important intermediates in cell proliferation mediated by hematopoietic growth factors. Indeed PT inhibits cell growth induced by IL-1 (38), IL-3 (35), CSF-1 (35), and GM-CSF (36). Expression of a dominant neg-ative G␣ i2 inhibits cell proliferation mediated by CSF-1 receptor in macrophages (39) and cell proliferation and transformation in NIH3T3 cells transfected with the oncogenic form of the CSF-1 receptor v-fms (71). PT inhibits Epo-stimulated erythroid colonies formation in rat erythroid progenitor cells from fetal liver (72). This suggests that G i proteins could be involved in Epo-dependent cell growth. Further studies will be required to elucidate the functional significance of G protein coupling to EpoR.
In conclusion, the results presented here show that G i protein is physically associated with EpoR. Epo activates G protein in cell membranes and induces the release of G i bound to the receptor in hematopoietic cells. Our data strongly suggest that G i proteins are important components in Epo signaling.