Activation of the Mitogen-activated Protein Kinases Erk1/2 by Erythropoietin Receptor via a Gi Protein βγ-Subunit-initiated Pathway*

We have recently shown that a heterotrimeric Gi protein is coupled to the erythropoietin (Epo) receptor. The Gi protein constitutively associates in its heterotrimeric form with the intracellular domain of Epo receptor (EpoR). After Epo stimulation Gi is released from the receptor and activated. In the present study we have investigated the functional role of the heterotrimeric Gi protein bound to EpoR. In Chinese hamster ovary cells expressing EpoR, the Gi inhibitor pertussis toxin blocked mitogen-activated protein kinase (MAPK) Erk1/2 activation induced by Epo. Epo-dependent MAPK activation was also sensitive to the Gβγ competitive inhibitor βARK1-ct (C-terminal fragment of the β-adrenergic receptor kinase), to the Ras dominant negative mutant RasN17, and to the phosphoinositide 3-kinase (PI3K) inhibitor LY 294002. A region of 7 amino acids (469–475) in the C-terminal end of EpoR was shown to be required for Gi binding to EpoRin vivo. Deletion of this region in EpoR abolished both MAPK and PI3K activation in response to Epo. We conclude that in Chinese hamster ovary cells, Epo activates MAPK via a novel pathway dependant on Gi association to EpoR, Gβγ subunit, Ras, and PI3K. The tyrosine kinase Jak2 also contributes to this new pathway, more likely downstream of βγ and upstream of Ras and PI3K. This pathway is similar to the best characterized pathway used by seven transmembrane receptors coupled to Gi to activate MAPK and may cooperate with other described Epo-dependent MAPK activation pathways in hematopoietic cells.

receptor superfamily, initiates a signal cascade that includes activation of the Jak2 tyrosine kinase, phosphorylation of the EpoR (2)(3)(4), activation of PI3K (5-7), STAT5 (8), and stimulation of the Ras/Raf/mitogen-activated protein kinases (MAPK) pathway (9 -13). The EpoR recruits several SH2-containing proteins through its intracellular phosphorylated tyrosines as well as adaptor molecules (for review, see Ref. 14). We previously developed an expression cloning strategy to characterize new proteins implicated in EpoR signaling. The strategy was based on the reactivity of such components with antibodies produced against proteins reactive with anti-phosphotyrosine (Tyr(P)) antibodies, isolated from the Epo-stimulated hematopoietic cell line UT7. This led us to isolate a cDNA encoding G␤ 2 , the ␤ 2 subunit of the heterotrimeric GTP-binding proteins, and to demonstrate that an heterotrimeric G i protein is constitutively associated with EpoR in hematopoietic cell lines as well as in normal erythroid progenitor cells. The G i -EpoR interaction occurs through the C-terminal end of the cytoplasmic domain of the receptor. After Epo activation the G i protein is activated and released from the receptor, most likely with the concomitant dissociation of G i ␣ and ␤␥ subunits (15).
Defined signal functions of G i include inhibition of adenylyl cyclase, regulation of phospholipase C␤, and activation of K ϩ channels (16). G proteins such as G q and G i also regulate cell growth and differentiation through the stimulation of a large number of complex cascades, leading to the activation of MAPKs. Agonist stimulation of seven transmembrane-spanning G protein-coupled receptors (GPCRs) leads to the exchange of GDP for GTP on the ␣ subunit of the heterotrimeric G protein and the subsequent dissociation of the ␣-GTP and ␤␥ subunits. Although the main functional properties of G proteins were initially thought to be essentially determined by the identity of the ␣ subunit, ␤␥ complexes also regulate a number of effectors. The mechanisms of the Erk1/2 MAPK activation by GPCR has been explored in cell lines that are readily transfected, such as COS-7 cells. The best characterized and probably the main pathway used by GPCR to activate MAPK via the pertussis toxin (PT)-sensitive G i protein depends on the release of free G␤␥ subunits acting on a Ras-dependent pathway (17)(18)(19). MAPK activation is initiated by the G␤␥-mediated tyrosine phosphorylation of Shc and proceeds through a Shc/Grb2/Sos pathway common to both GPCR and tyrosine kinase receptors (19). PI3K activity also participates in G␤␥-mediated MAPK activation upstream of Ras (20 -22) or by stimulating MEK phosphorylation (23). G i may also regulate MAPK activation through ␣ subunits. Indeed, activating mutations have been identified in the G␣ i2 gene, referred to as gip2 (24,25), and MAPK is constitutively activated in gip2-transformed fibroblasts (24).
The MAPKs Erk1/2 play important role in Epo-induced proliferation, differentiation, and apoptosis (26 -28). In hematopoietic cells Epo activates the Ras/Raf/MEK/MAPK pathway by recruitment of Grb2 either directly or indirectly via adaptor molecules such as Shc (29,30) or SHP-2 (31,32). This interaction allows the guanine nucleotide release factor Sos, constitutively bound to Grb2, to convert Ras to an active GTP-bound form. Furthermore, C3G, an other guanine nucleotide exchange factor, through its interaction with CrkL, has also been found participating in Epo-mediated MAPK signaling pathway via Ras (33) and Rac (34). PI3K and protein kinase C have also been implicated in Raf/MAPK activation by Epo (35,36).
Our demonstration that heterotrimeric G i proteins associate with EpoR and that receptor activation leads to the activation and dissociation of G i from the receptor led us to investigate the downstream targets of G i . We considered the possibility that Epo may activate MAPK via the G i protein bound to EpoR-like seven-transmembrane receptors. In this study, we characterized a new Epo-activated MAPK pathway dependent on G i . This pathway, similar to the best characterized MAPK pathway used by GPCR coupled to G i , is the main pathway for Epo-dependent MAPK activation in CHO cells. It proceeds through the ␤␥ subunit of the G i bound to EpoR and involves the activation of Jak2, Ras, PI3K, and MEK. Our data also demonstrate that G i plays a crucial role for Epo-induced PI3K activation in CHO cells.

EXPERIMENTAL PROCEDURES
Materials-Purified recombinant human Epo (specific activity 120,000 units/mg) was a kind gift of Dr. M. Brandt (Roche Molecular Biochemicals). Rabbit polyclonal antibodies against G␤ (M-14), Akt 1/2 (H-136), and Erk (C-16) were purchased from Santa Cruz Biotechnology. Anti-Jak2 antibodies (06-255) were supplied from Upstate Biotechnology. Polyclonal antibodies specific to the phospho-forms of Erk (Thr-202/Tyr-204) and Akt (Ser-473) were supplied from Cell Signaling Technology. Anti-phosphotyrosine (anti-Tyr(P)) monoclonal antibodies 4G10 and PY72 were produced from hybridoma cell lines kindly provided by B. Drucker and B. Sefton. Anti-MalE antiserum and anti-EpoR antiserum against a fusion protein between glutathione S-transferase and the cytoplasmic portion of the human EpoR were described previously (37). Anti-HA peptide monoclonal antibody was purchased from Roche Molecular Biochemicals. Peroxidase-conjugated anti-rabbit antibodies were purchased from Amersham Biosciences and Cell Signaling Technology. MEK1 inhibitor PD98059 was supplied from New England Biolabs, PI3K inhibitor LY 294002 and PMA were from Sigma, pertussis toxin was from Alexis, and protease inhibitors were from Roche Molecular Biochemicals. Amylose resin was supplied from Biolabs. The minigene encoding the C terminus of bovine ␤ARK1 (␤ARK-ct) (38) was kindly provided by R. Jockers at Institut Cochin, Paris, France. The expression plasmid pcDNA3/HA-RasN17 was a gift from A. Eychène and the pXM/EpoR-W282R from J. Ihle.
The murine EpoR mutant Ϫ41 was described previously (15). To generate the ⌬V-L EpoR mutant lacking residues Val-469 to Leu-475 within the intracellular domain, mutation was introduced into the full-length EpoR by PCR. The PCR product was digested with HindIII/ XbaI and subcloned into the same restriction sites of a modified pCDNA3 expression vector (15). The fidelity of all constructs was confirmed by sequencing.
Transfections-Stable transfectants of FDCP-1 myeloid cell line expressing the wild type and mutated forms of EpoR were obtained by electroporation, as previously described (8). CHO cells were transfected with expression plasmids coding for full-length EpoR, ⌬V-L, or W282R EpoR mutants using LipofectAMINE Plus TM Reagent (Invitrogen) according to the manufacturer's instruction, selected with 20 g/ml puromycin, and cloned by limiting dilution. EpoR expression was determined using immunoblot analysis of cellular extracts and 125 I-Epo binding on whole cells. Cell lines expressing a similar number of 125 I-Epo binding sites of either wild type or mutated receptors were selected for the study. Transient transfections of CHO cells were carried out by the same methodology when 80% confluence was reached using 1 g of plasmid/well (10 cm 2 ), and 24 h post-transfection the cells were harvested and used in experiments as described. LipofectAMINE transfection of CHO cells consistently resulted in transfection efficiencies greater than 60% (data not shown).
Cell Culture and Treatments-The human erythroleukemic cell line UT7 was maintained in ␣-minimal essential medium supplemented with 5% fetal bovine serum, penicillin, streptomycin, 2 mM L-glutamine, and 2 units/ml Epo. FDCP-1 myeloid cells expressing EpoR were grown in ␣-minimal essential medium supplemented with 5% fetal bovine serum and 2 units/ml Epo. The cells were starved of Epo by replacing Epo with 3% WEHI-conditioned medium as a source of interleukin 3 1-2 days before use in the experiments. CHO cells were grown in Dulbecco's minimal essential medium/nut mix F-12 (Ham's) medium supplemented with 7% fetal bovine serum. Before stimulation, cells were washed and serum-starved 3-4 h or overnight when indicated by incubation in Ham's medium in the presence of 0.4% bovine serum albumin. The cells were stimulated with 10 units/ml Epo or 10 ng/ml PMA. In some experiments PD98059 (30 M) or LY 294002 (50 M) were added 30 or 15 min, respectively, before stimulation, and pertussis toxin (50 ng/ml) and PMA (0.6 g/ml) were added 16 h before stimulation.
Immunoprecipitations and MalE Pull-down Assays-Cells (1 ϫ 10 7 ) were washed in phosphate-buffered saline containing 50 M Na 3 VO 4 and lysed on ice in 150 mM NaCl, 50 mM Tris-HCL (pH 7.5), 1% Brij 98, 1 mM Na 3 VO 4 , 2 mM EGTA, 10% glycerol, and protease inhibitors (lysis buffer). Insoluble material was removed by centrifugation at 15,000 ϫ g for 10 min at 4°C. For immunoprecipitation, lysates were incubated with anti-EpoR antiserum 1/250 for 1 h at 4°C. Protein G-Sepharose (Amersham Biosciences) was added, and the incubation was pursued for 1 h at 4°C with rotation. Samples were washed 3 times with lysis buffer and 2 times with 150 mM NaCl, 50 mM Tris-HCL (pH 7.5), 0.1% Brij 98, 1 mM Na 3 VO 4 . For MalE fusion protein pull-down assays, lysates were incubated with 5 g of MalE fusion protein prebound to 20 l of amylose resin beads for 2 h at 4°C with rotation. Then, beads were collected and processed as described above for immunoprecipitation. Proteins were eluted by boiling in SDS sample buffer, resolved on SDS-PAGE, and subjected to immunoblotting as described previously (3,39).
Determination of MAPK and Akt Activation-Erk and Akt activation were studied using phospho-specific antibodies. Cells lysates were prepared as described above using lysis buffer containing 1% Nonidet P-40. The protein content was normalized by the micro-BCA assay (Pierce), and equal amounts of proteins (10 to 50 g) were separated by 10% SDS-PAGE. Proteins were transferred to nitrocellulose membranes and analyzed in immunoblot.
ADP-ribosylation Assay-To detect G␣ i its ADP-ribosylation was followed by measuring incorporation of [ 32 P]ADP-ribose in the presence of pertussis toxin in vitro as previously described (15).

Identification of the Sequence in the EpoR Required for G i
Binding-In a previous report we demonstrated that constitutive association between heterotrimeric G i protein and the EpoR occurred through the C-terminal end of the receptor (15). To characterize more precisely the EpoR domain involved in G i binding, we first generated recombinant fusion proteins between MalE and various EpoR sequences contained in this region (Fig. 1A). Fusion proteins were incubated with UT7 cell lysates, and bound proteins were analyzed by immunoblotting with anti-G␤ antibodies (Fig. 1B). Recombinant fusion protein containing the EpoR amino acids 458 -483 allowed G␤ binding in vitro, in agreement with previous data obtained in vivo using EpoR mutants (15). A similar binding was detected with the MalE-EpoR fusion peptide 465-477, whereas G␤ did not bind to EpoR peptides 433-459 and 458 -470. Western blotting with anti-MalE antibody showed that the amounts of recombinant proteins bound to amylose resin used in the pull-down assays were similar. These results indicate that amino acids 465-477 are sufficient to bind G␤ in vitro. This region does not include tyrosine residues, which fits well with the constitutive binding of G i to EpoR (15).
To evaluate the role of the identified sequence (465-477) in vivo we wished to delete this sequence within the EpoR. To avoid perturbation of receptor structure and particularly of Tyr-464 and Tyr-479 located in the vicinity of G i binding region, we constructed an EpoR mutant deleting a shorter region encompassing amino acids 469 -475 (⌬V-L). The EpoR expressed in FDCP-1 cells was immunoprecipitated from cell lysates, and G␣ i associated with EpoR was detected by in vitro ADP-ribosylation assay (Fig. 1C). In cells transfected with fulllength mouse EpoR, G␣ i was coprecipitated with anti-EpoR antibodies. However, G␣ i was not precipitated with EpoR when mutants Ϫ41 (15) or ⌬V-L were expressed. The absence of G i binding to the EpoR in cells expressing the ⌬V-L mutant is unlikely to be due to a general signaling defect of the mutated receptor since the receptor kept the ability to bind the tyrosine kinase Jak2 (Fig. 1C). These results indicate that the sequence V 469 PDSEPL 475 located in the C-terminal domain of the EpoR is necessary for G i /EpoR association in vivo. This region does not bind any known EpoR effector.
Epo Stimulates a G i -dependent Erk1/2 Activation Pathway in CHO Cells Expressing the EpoR-A number of receptors that couple to heterotrimeric G proteins, including G i -coupled receptors, have been shown to stimulate MAPK activation (17,18,38). Therefore, we explored the possibility that G i may participate in MAPK activation after Epo stimulation. The sensitivity of MAPK activation to the G i inhibitor, pertussis toxin, was first examined in hematopoietic cell lines expressing an endogenous or exogenous EpoR as well as in normal human erythroid progenitor cells. A weak and inconstant inhibition of Epo-induced MAPK phosphorylation was observed with these cells (data not shown). We previously observed that Epo in-duced a stronger G protein activation in cell membranes isolated from epithelial CHO cells expressing EpoR than in UT7 hematopoietic cell line (15). This led us to hypothesis that a G i -dependent Epo-activated MAPK pathway may be more easily detected in these cells.
The role of G i in the activation of Erk1/2 by Epo was investigated in CHO cells stably transfected with the full-length EpoR cDNA (CHO-ER WT). Cells were stimulated with Epo for various times, and the lysates were subjected to immunoblotting with phospho-specific anti-Erk antibodies as shown in Fig.  2. Epo induced a transient Erk1/2 activation. Preincubation of the cells with 50 ng/ml pertussis toxin overnight completely inhibited Epo-induced Erk activation. Total Erk content was unchanged under all conditions as shown by immunoblotting with anti-Erk antibodies recognizing both the phosphorylated and the unphosphorylated forms of Erk 1 and 2. Toxin treatment had no effect on cell viability (data not shown). These results establish a role for a pertussis toxin-sensitive G protein in Epo-induced Erk activation and show that a G i -dependent pathway is the main pathway used by Epo to activate MAPK in CHO cells.
To ascertain the functional significance of the G i -EpoR interaction on MAPK activation we studied the effect of disrupting it. When stably expressed in CHO cells the EpoR ⌬V-L mutant, which is unable to associate with G i (Fig. 1C), could not activate Erk after Epo stimulation (Fig. 3A). In contrast, PMA treatment led to MAPK activation, showing that these cells retained the ability to activate Erk. To evaluate if the ⌬V-L deletion would perturb early events of EpoR activation, Jak2 was immunoprecipitated and subjected to analysis using anti-phosphotyrosine antibodies. Fig. 3B illustrates that the EpoR ⌬V-L mutant had the ability to associate with the Jak2 tyrosine kinase and that both proteins became phosphorylated after Epo stimulation. Indeed, when Jak2 was immunoprecipitated, the phosphorylated forms of Jak2 and EpoR were detected in immune complexes isolated from cells expressing EpoR WT as well as EpoR ⌬V-L. Thus, the ⌬V-L deletion has no general effect on Epo signal transduction. Tyrosines located in the vicinity of the Val-469 -Leu-475 region constitute potential binding sites for components implicated in Epo-dependent MAPK activation such as CrkL (40) and Grb2 (32) for Tyr-460 and Tyr-464, respectively. Therefore, we wanted to determine whether the ⌬V-L deletion did not alter the phosphorylation of these residues and their ability to bind their effectors. After Epo activation, phosphorylated EpoR ⌬V-L bound GST-CrkL and GST-Grb2 as efficiently as phosphorylated EPOR WT did (data not shown). This strongly suggests that the absence of CHO-ER cells expressing WT EpoR were preincubated overnight at 37°C in the absence (PTϪ) or presence (PTϩ) of 50 ng/ml pertussis toxin. Subsequently the cells were stimulated with 10 units/ml Epo for the indicated times and lysed. The soluble fraction was separated by SDS-PAGE and transferred to nitrocellulose membranes. Determination of MAPK activity was assessed by immunoblotting with phosphospecific anti-Erk antibodies (anti-pErk). The same blot was also probed with antibodies to Erk1/2 (anti-Erk) to confirm that equal amounts of enzyme were present in each lane.
MAPK activation in EpoR ⌬V-L is due to a defect in G i binding and not to a perturbation of the phosphorylation of proximal tyrosine residues. Altogether, these data show that the EpoR region that couples G i to EpoR is required for Epo-induced Erk activation in CHO cells and confirm the results obtained with pertussis toxin treatment.
Epo-mediated MAPK Activation in CHO-ER Cells Is Dependent on G␤␥-Given the sensitivity of Epo action to pertussis toxin, we first examined whether Epo-mediated Erk1/2 activation may be secondary to inhibition of adenylyl cyclase by G i . Indeed, in some cells a rise in cAMP has been shown to inhibit the Ras/MAPK pathway at the level of Ras/Raf interaction (41). Therefore, a G i -mediated fall in cAMP level may relieve Raf inhibition by protein kinase A and positively regulate MAPK. Pretreatment of CHO-ER WT cells with the cell-permeable cAMP derivative dibutyryl-cAMP, which mimics the effect of cAMP, failed to prevent Epo-induced MAPK activation (data not shown). This suggests that MAPK activation does not proceed through ␣ i -mediated inhibition of adenylyl cyclase.
␤␥ subunits derived from PT-sensitive heterotrimeric G proteins were shown to mediate Ras-dependent MAPK activation (17,18,42). MAPK activation through G i -coupled receptors such as the ␣ 2A -adrenergic receptor and the lysophosphatidic acid receptor is sensitive to inhibition by the C-terminal fragment of the ␤-adrenergic receptor kinase (␤ARK1-ct). ␤ARK1-ct interacts with free ␤␥ and acts as a competitive inhibitor of G␤␥-mediated signals but does not affect G␣ i -mediated signaling, allowing distinguishing between ␣ i and ␤␥ pathways (38). To determine whether the PT-sensitive phosphorylation of Erk1/2 stimulated by Epo was mediated by G␤␥, we studied the effect of Epo on CHO-ER WT cells transiently transfected with the ␤ARK1-ct peptide. As shown in Fig. 4A phosphorylation of endogenous Erk1/2 following Epo stimulation was significantly reduced compared with cells transfected with control plasmid, although the total Erk amount was not modified. We conclude that Epo-induced MAPK activation in CHO-ER cells occurs through the release of ␤␥ subunits of heterotrimeric G i protein.
Epo-mediated MAPK Activation in CHO-ER Cells Is Dependent on JAK2, Ras, and MEK-To determine whether Epoinduced JAK2 tyrosine kinase activation was required for MAPK activation, the EpoR mutant (W282R) defective in JAK2 binding and kinase activation (43) was stably expressed in CHO cells. In cells expressing the EpoR W282R mutant, Epoinduced MAPK activation was lost (Fig. 4C). Thus, JAK2 activation is required for MAPK activation in these cells. The requirement for MEK kinase, an upstream activator of MAPK, was also evaluated using the MEK inhibitor PD98059. As shown in Fig. 4B, Epo-induced MAPK activation was abolished when the cells were pretreated with the MEK inhibitor before Epo stimulation.
Protein kinase C is activated by Epo and may be implicated in MAPK activation by Epo (36) and also by G i coupled to classical GPCR (44). We examined whether protein kinase C might contribute to Epo-induced MAPK activation in CHO-ER WT cells. Down-regulation of protein kinase C induced by prolonged exposure of cells to PMA inhibited MAPK activation in response to PMA but not to Epo (Fig. 5B). Therefore, Epostimulated MAPK activation is independent of a PMA-sensitive protein kinase C. The best characterized pathway used by G i -coupled receptors to activate MAPK requires p21 ras activity (45,46). To investigate the role of p21 ras in Epo-mediated MAPK activation in CHO-ER WT cells, the effects of the dominant negative mutant RasN17 on EpoR-stimulated MAPK activation was assessed. As shown in Fig. 5A, RasN17 expression inhibited Epo-stimulated phosphorylation of Erk. In contrast RasN17 expression did not alter the level of Epo-stimulated Jak2 tyrosine phosphorylation (see Fig. 8A). Taken together these results suggest that in CHO cells expressing EpoR, Epo acts through Jak2 to activate Ras and Erk1/2 MAPK.
Functional Role of PI3K in G i -dependent Epo-induced MAPK Activation-PI3K has been implicated in the ␤␥-dependent  Fig. 2. B, Jak2 was immunoprecipitated, and phosphorylated Jak2 and EpoR were identified by immunoblotting with anti-phosphotyrosine (pTyr) antibodies. activation of MAPK by G i -coupled receptors (21,22). Therefore, we next examined whether PI3K also contributed to Epo-induced MAPK activation in CHO-ER WT cells. Pretreatment of the cells with LY 294002, a PI3K inhibitor, blocked Epo-stimulated activation of the PI3K, as detected by antibodies specific to the phospho-form of the PI3K effector Akt (Fig. 6A). LY 294002 also strongly decreased the activation of MAPK by Epo without modification of the total Erk content (Fig. 6A). These data show that PI3K contributes to Epo-dependent MAPK activation in CHO-ER WT cells. Because MAPK activation is essentially G i -dependent in these cells, they also suggest that PI3K is a component of the G i -dependent Erk1/2 pathway. To determine whether G i is required for PI3K activation, Epoinduced Akt phosphorylation was compared in CHO-ER WT expressing the full-length receptor or the ⌬V-L mutant defective in G i interaction and MAPK activation. As shown in Fig.  7A the deletion mutant had lost the ability to activate Akt in response to Epo. As expected, expression of the ␤␥ competitor ␤ARK1-ct also inhibited Akt phosphorylation by EpoR WT (Fig.  7B).
The requirement for Ras and Jak2 in PI3K activation was then evaluated. Expression of RasN17 in CHO-ER WT prevented Akt activation after Epo stimulation (Fig. 8A). Furthermore the EpoR mutant defective in Jak2 binding and activation (W282R) failed to activate Akt (Fig. 8B), showing that Jak2 tyrosine kinase and Ras are upstream of PI3K activation. Taken together, these data demonstrate that the ␤␥ subunit of G i plays a crucial role in Epo-mediated PI3K activation and that PI3K is a component of the Epo-activated G i -dependent MAPK activation pathway downstream of Jak2 and Ras.

DISCUSSION
In this study we showed that Epo can activate the Erk1/2 MAPK pathway via a new G i -dependent pathway sensitive to pertussis toxin. The activation of MAPK by EpoR is mediated by the ␤␥ subunits and not the ␣ subunit of G i . Indeed, MAPK phosphorylation in response to Epo was inhibited by the coexpressed ␤␥ competitor ␤ARK1 polypeptide. The marked inhibition observed with these agents suggests that the ␤␥ pathway is the main pathway for Epo-mediated MAPK activation in CHO cells expressing the EpoR. Classical G protein-coupled receptors, (17,18,42,47) and some tyrosine kinase receptors, such as the receptors for insulin-like growth factor-1 and fibroblast growth factor (48,49), activate Erk1/2 through mechanisms involving ␤␥ subunits derived from pertussis toxin-sensitive G proteins in neuronal cells, epithelial cells, and fibroblasts. To our knowledge our study represents the first demonstration of a functional role of G␤␥ subunits in cytokine receptor signaling. Signaling from G i to MAPK more often involves ␤␥ subunits of heterotrimeric G proteins acting on a Ras-dependent pathway (42,45). Likewise, Epo-induced MAPK activation is consistent with this general mechanism, since it is dependent on Ras as shown by its inhibition by a dominant negative mutant of Ras (RasN17).
Our data also demonstrate that PI3K activation by Epo is dependent on G i ␤␥ subunits and is required for Erk activation downstream of Ras in CHO cells. Indeed (i) Epo-induced Akt phosphorylation was inhibited by expression of the ␤ARK1-ct or RasN17, (ii) the EpoR mutant defective in G i binding lost the ability to phosphorylate both Erk1/2 and the PI3K effector Akt, and (iii) the PI3K inhibitor LY 294002 inhibited MAPK activation by Epo. Because MAPK activation by Epo is mainly G i -dependent in CHO cells, these data also suggest that the ␤␥-de- pendent PI3K activation by Epo is upstream of MAPK and does not form a separate pathway. Similarly, PI3K was shown to be involved in G i -dependent MAPK activation by platelet-derived growth factor in airway smooth muscle cells (50,51). PI3K has a critical role linking G protein-coupled receptors and G␤␥ to the MAPK-signaling pathway. Different mechanisms could be implicated in this process. Previous reports implicate PI3K␣ or -␤ (21, 52) and PI3K␥ (22) in the activation of MAPK by classical G i -coupled receptors upstream of Sos and Ras activation (21,22). PI3K␥ could represent the link between ␤␥ and a downstream tyrosine kinase required for Ras-dependent MAPK activation (22). However, this mechanism does not apply here since we show that PI3K is downstream of Ras. PI3K could also contribute to the G i -initiated MAPK pathway without Ras requirement (23).
Several candidate tyrosine kinases have been proposed as intermediates between ␤␥ and Ras/MAPK activation including c-Src family kinases (53), Syk (54), and Pyk2 (55). Alternatively, G i can activate MAPK through transactivation of the epidermal growth factor tyrosine kinase receptor (56). Activated c-Src or epidermal growth factor receptor mediate Ras/ MAPK activation through tyrosine phosphorylation of Shc and its subsequent association with Grb2 (53,56). An essential component for G i -dependent Erk activation by EpoR is the tyrosine kinase Jak2. Indeed CHO cells expressing the W282R EpoR mutant defective in Jak2 binding and activation have lost the ability to activate MAPK. Epo-induced PI3K activation was also inhibited in these cells, suggesting that PI3K is located downstream of Jak2 in this pathway. Ras is probably positioned downstream of Jak2, since RasN17 expression had no effect on Jak2 phosphorylation. Jak2 contribution to MAPK activation is independent of G i binding to EpoR and subsequent G i dissociation from the receptor. Indeed the ⌬V-L mutation in EpoR, which disrupts G i binding to EpoR, did not alter Jak2 phosphorylation by Epo and its binding to the receptor. Furthermore in 32D cells, we previously showed that Epo-induced G i dissociation from the EpoR is not affected when the Jak2deficient EpoR mutant is expressed (15). Therefore, Jak2 is more likely acting downstream of ␤␥ release and upstream of Ras and PI3K. In addition to Jak2, the tyrosine kinases Syk and Lyn have been implicated in EpoR activation (39,57), but their contribution to G i -dependent MAPK activation remains to be determined.
Four documented pathways have been implicated in Epomediated regulation of the Erk cascade in various hematopoietic cell lines or erythroid progenitors expressing the EpoR. Three of them, the Shc/Grb2/Sos pathway (11), the Grb2/Sos pathway (32), and the CrkL/C3G pathway (33), lead to MAPK activation via Ras. The fourth one involves PI3K and might require an atypical protein kinase C (36). All these pathways are initiated through effector binding to phosphorylated tyrosines located in the intracellular domain of the receptor. Particularly, tyrosines 460 and 479 have been shown to be involved in CrkL/C3G and in PI3K-dependent Erk activation by the EpoR, respectively (36,40), and tyrosine 464 has a consensus Grb2 binding motif (32). In CHO cells, MAPK activation by Epo is unlikely to occur through PI3K binding to tyrosine 479 since PI3K activation is inhibited by the ␤␥ competitor ␤ARK-ct. The region of the EpoR required for G i binding in vivo and G i -dependent MAPK phosphorylation by Epo (amino acids 469 -475) does not include tyrosines. Thus, G i -dependent Erk activation may involve a different mechanism. The possibility that the absence of Erk activation in the ⌬V-L EpoR deleted of the G i binding domain resulted from a conformational change affecting the nearby tyrosines is highly unlikely. Indeed Epo-induced CrkL and Grb2 interaction is preserved in the ⌬V-L mutant. Furthermore, pertussis toxin and ␤ARK-ct also inhibit Epo-induced MAPK activation.
In hematopoietic cells, G i binds to the EpoR through the same EpoR domain required for G i -dependent Erk activation in CHO cells, and G i is activated and dissociates from the receptor after Epo stimulation (15). However, in these cells, pertussis toxin has little effect on Epo-dependent Erk activation. As described above, there are several different pathways leading to MAPK activation by Epo in hematopoietic cells. The G i / PI3K/MAPK pathway may be only one of them, explaining the relative inefficiency of the toxin in such cells. The relative contribution of the different pathways leading to MAPK activation by the EpoR and their interconnections remain poorly documented and need to be explored further. In contrast with hematopoietic cells, the G i -dependent pathway is the main pathway involved in Epo-dependent MAPK activation in CHO epithelial cells. Epo-induced EpoR tyrosine phosphorylation is lower in CHO than in hematopoietic cells. 2 CHO cells may mainly use the G i -dependent pathway because of the inefficiency of MAPK pathways initiated by phosphorylated tyrosines of the receptor.
In erythroid precursors G i regulates an Epo-modulated Ca 2ϩ channel (58,59). This regulation is mediated by G␣ i2 , and tyrosine 460 in EpoR is critical for Epo-stimulated Ca 2ϩ influx in CHO and in Ba/F3 hematopoietic cells (60). Tyrosine 460 is not required for G i interaction with the EpoR in vitro, and G i -dependent Erk activation involves the ␤␥ subunits of G i . These data suggest that different G i -dependent pathways lead to calcium (through ␣ i2 ) and MAPK (through ␤␥) activation by EpoR. Epo function in non-hematopoietic cells is not unique to epithelial cells. Indeed, myoblasts, neural cells, and endothelial cells have been shown to express EpoR (61)(62)(63). Epo has antiapoptotic function in neural cells and in brain (64,65) and is implicated in brain development and neurogenesis (66,67). Epo has also been implicated in angiogenesis (68) and may regulate muscle cell proliferation and differentiation (62). Al-2 P. Mayeux, unpublished data.
FIG. 8. PI3K activation is dependent on Jak2 and Ras and Jak2 activation is independent on Ras. A, Ras is required for PI3K activation but not for Jak2 activation. CHO-ER cells WT were transfected with 0.75 g of HA-RasN17 or empty vector. 24 h later the cells were serum-starved for 4 h, stimulated with Epo (ϩ) or not (Ϫ) for 5 min, and lysed. Jak2 was immunoprecipitated (IP), and phosphorylated Jak2 was identified by immunoblotting with anti-phosphotyrosine (PY). Total Jak2 in immunoprecipitates was also determined by anti-Jak2 antibodies (top). Phosphorylated Akt (pAkt), total Akt, phosphorylated Erk (pErk), total Erk, and HA-RasN17 present in cell lysates were visualized by immunoblotting as indicated (bottom). B, Jak2 is required for PI3K activation. CHO-ER cells expressing WT EpoR or EpoR mutant W282R were serum-depleted, stimulated with Epo (ϩ) or not (Ϫ) for 5 min, and lysed. Phosphorylated and total Akt, phosphorylated EpoR (REpo), and phosphorylated and total Erk present in cell lysates were analyzed by immunoblotting. though the intracellular pathways activated by Epo in nonhematopoietic cells are still poorly documented, Epo was shown to activate MAPK and Jak2 in neural and in muscle cells (62,65,69). Further investigations would allow the determination of the function and contribution of the G i -dependent pathway to MAPK activation pathway in hematopoietic and non-hematopoietic cells.