Erythropoietin Induces Glycosylphosphatidylinositol Hydrolysis

We showed that erythropoietin induced rapid glycosylphosphatidylinositol (GPI) hydrolysis and tyrosine phosphorylation of phospholipase C (PLC)-γ2 in FDC-P1 cells transfected with the wild-type erythropoietin-receptor. Erythropoietin-induced tyrosine phosphorylation of PLC-γ2was time- and dose-dependent. By using FDC-P1 cells transfected with an erythropoietin receptor devoid of tyrosine residues, we showed that both effects required the tyrosine residues of intracellular domain on the erythropoietin receptor. Erythropoietin-activated PLC-γ2 hydrolyzed purified [3H]GPI indicating that GPI hydrolysis and PLC-γ2 activation under erythropoietin stimulation were correlated. Results obtained on FDC-P1 cells transfected with erythropoietin receptor mutated on tyrosine residues suggest that tyrosines 343, 401, 464, and/or 479 are involved in erythropoietin-induced GPI hydrolysis and tyrosine phosphorylation of PLC-γ2, whereas tyrosines 429 and/or 431 seem to be involved in an inhibition of both effects. Thus, our results suggest that erythropoietin regulates GPI hydrolysis via tyrosine phosphorylation of its receptor and PLC-γ2activation.

The glycosylphosphatidylinositol (GPI) molecules with the general structure lipid-phosphate-inositol glycan have been isolated from plants, bacteria, yeast, parasitic protozoa, and mammalian cells (9 -11). In eukaryotic cells, GPI molecules may be attached to protein moieties thereby providing a membrane anchor. Free GPIs also exist under an uncomplexed form. Both forms of GPI have been shown to participate in signal transduction events (for review see Ref. 12). In different cellular systems GPI turnover is modulated by a variety of hormones, cytokines, and growth factors such as insulin (13,14), insulin growth factor-I (15,16), nerve growth factor (17,18), interleukin-2 (19), thyroid-stimulating hormone (20), or erythropoietin in rat erythroid progenitor cells (21). Then the hydrolysis of glycosylphosphatidylinositol could be another pathway in signal transduction. GPI hydrolysis generates diacylglycerol and the polar group of the lipid, an inositol phosphoglycan (IPG) which can act as a second messenger (for review see Ref. 22).
The dependence on increased receptor phosphorylation and GPI hydrolysis after growth factor stimulation remains to be conclusively defined. However, tyrosine kinase inhibitors were shown to be able to partially block epidermal growth factor (EGF)-stimulated GPI hydrolysis (23), and hydrolysis of GPI in response to insulin is reduced in cells bearing kinase-deficient insulin receptors (24,25).
The nature of the phospholipase C (PLC) responsible for hormone-induced GPI hydrolysis is still unknown. Mammalian GPI-PLC was partially purified from rat liver membranes (26) but has not yet been cloned. PI-PLC from Bacillus cereus was shown to facilitate the release of some PI-glycan-ethanolamineanchored proteins as PI-PLC isolated from trypanosomes (27) and rat liver membranes (26,28). Among eukaryote PI-PLC isoforms, only the ␥ 1 and ␥ 2 types contain SH2 and SH3 domains and become potential candidates for binding to phosphotyrosine residues during Epo receptor activation (29 -31). Epo activation of a PLC was suggested in erythroid progenitor cells (32), and PLC-␥ 1 activation was evidenced in UT7 cells (33), but Epo activation of PLC-␥ 2 has never been described, whereas PLC-␥ 2 activation by macrophage-colony-stimulating factor has been described in FDC-P1 cells (34).
In this report, we show that erythropoietin induces a rapid and transient hydrolysis of GPI and tyrosine phosphorylation of PLC-␥ 2 . Both effects require the presence of the same tyrosine residues on the intracellular domain of Epo receptor. We show that Epo-activated PLC-␥ 2 is able to hydrolyze purified [ (35). A schematic representation of Epo-R wild type (WT) and mutants is done in Fig. 1. The Y 1-2 F 3-4 Y 5-8 Epo-R mutant was kindly provided by Dr. U. Klingmü ller (Zentrum fur Molekulare Biologie, Universitat Heidelberg, Germany).
Cell Cultures-FDC-P1 cells were maintained in ␣-minimum essential medium (␣-MEM) containing 10% fetal calf serum and 3% WEHI conditioned media as a source of interleukin-3. Stable transfections of murine Epo-R cDNA into FDC-P1 cells were done as described previously (35). After transfection, Epo-sensitive cells were maintained in ␣-MEM containing 10% fetal calf serum and 2 units/ml Epo.
Cell Labeling and Stimulation-Cells were grown with 3% WEHI as source of growth factor and metabolically labeled with [ 3 H]ethanolamine (1.5 Ci/ml) or [ 3 H]glucosamine (5 Ci/ml) for 20 h prior to assay. Four hours before the end of this period, cells were starved of growth factor in ␣-MEM containing 1% (v/v) of fetal calf serum. Cells were then washed twice with ␣ medium, 25 mM HEPES, and aliquots of 10 -20 ϫ 10 6 cells were preincubated at 37°C for 15 min before stimulation with 5 units/ml Epo for 0 -30 min. Incubations were stopped with HClO 4 (5% v/v). After centrifugation at 1200 ϫ g, pellets and supernatants were stored at Ϫ20°C before GPI and IPG quantitations.
Glycosylphosphatidylinositol (GPI) and Inositol Phosphate Glycan (IPG) Assays-Lipids were extracted from the resultant pellets and analyzed by sequential acid/base silica gel thin layer chromatography (TLC, Silica Gel 60, Merck) as described previously (14) or by HPTLC (Silica Gel 60 aluminum-backed HPTLC plates, Merck) developed in chloroform/methanol/NH 4 OH/water (45:45:3.5:10, v/v). HPTLC plates were sprayed with EN 3 HANCE (NEN Life Science Products) and fluorographed on Kodak XAR-5 film. To quantitate incorporation of radiolabel into GPI, HPTLC plates were either scanned before fluorography on a Berthold LB 2821 automatic TLC linear analyzer, or 1-cm bands of TLC were scraped and their radioactivity content estimated by scintillation counting. IPG levels from perchlorate supernatants or from the aqueous phase obtained after GPI hydrolysis were determined by chromatography on a Dowex AG 1-X8 (200 -400 mesh) column as described previously (13,21 ]GPI contained in the organic phase was reanalyzed by HPTLC and quantitated on Berthold analyzer before fluorography. IPG contained in the aqueous phase was quantitated as described previously (13,21).
Determination of 2,5-Anhydromannitol-A sample of presumed GPI labeled with [ 3 H]glucosamine from WT Epo-R FDC-P1 cells recovered after the basic thin layer chromatography was hydrolyzed with nitrous acid. Then the deaminated glycan was reduced with sodium borotritide and subjected to solvolysis in 0.5 M methanolic HCl for 12 h at 80°C as described previously (37). Samples were analyzed on Silica Gel 60 HPTLC plates developed for 10 cm in 1-propanol/acetone/water (9:6:5, v/v) (38) and were scanned on a Berthold analyzer. 2,5-[ 3 H]Anhydromannitol control was prepared by treating [ 3 H]glucosamine as above.

Evidence for Glycosylphosphatidylinositol in Epo-R-transfected FDC-P1 Cells-GPI identification was done with cells metabolically labeled for 20 h with [ 3 H
]ethanolamine, a radiolabeled precursor of GPI. Polar lipids were then extracted and resolved by sequential acid/base thin layer chromatography (TLC) or by HPTLC. During the first acid TLC, GPI did not migrate and was clearly separated from phosphatidylethanolamine which migrated at 9 cm (not shown). The material around the origin (2 cm) was recovered and analyzed on a second TLC or on HPTLC which was run in an alkaline solvant system. As shown in Fig. 2, A-C, a single peak of [ 3 H]glycolipidic fraction was recovered between phosphatidylcholine (PC) Characterization of GPI was confirmed by testing the sensitivity of the radiolabeled product to several GPI-hydrolyzing conditions. The recovered [ 3 H]ethanolamine glycolipid fraction was first treated with nitrous acid that hydrolyzes the glycosidic bond between inositol and glucosamine. After deamination by nitrous acid, about 65% of the labeled material was recovered in the aqueous phase, and 10.5% of the radioactivity was recovered in the aqueous phase in control incubations. Moreover, we have shown 2,5-[ 3 H]anhydromannitol production after having deaminated/reduced and methanolyzed the recovered [ 3 H]glucosamine glycolipid fraction (Fig. 3, A-C). We further investigated the cleavage of the purified recovered 3 Hlabeled lipidic peak from WT Epo receptor-transfected FDC-P1 After treatment, about 21% of the labeled material was recovered in the aqueous phase versus control 5%. These values are in good agreement with previously published results showing the sensitivity of GPI from various tissues to B. cereus PI-PLC (24,39). Altogether, our results show that the lipidic products extracted from FDC-P1 WT Epo receptor-transfected cells and migrating between PIP and PC corresponded to GPI.
Epo-induced GPI Hydrolysis in Epo-R-transfected FDC-P1 Cells-GPI hydrolysis was assessed by measuring both GPI or inositol phosphate glycan (IPG) levels (its hydrolysis product) from cells metabolically labeled with [ 3 H]ethanolamine. Epo (5 units/ml) induced a rapid and transient GPI hydrolysis with a parallel IPG release within 1 min in FDC-P1 cells transfected with WT Epo-R. Then the IPG level rapidly decreased and returned to control values after 30 min of Epo exposure. No GPI hydrolysis was detected after Epo stimulation of non-transfected FDC-P1 cells (Fig. 4).
Epo Induced the Tyrosine Phosphorylation of PLC-␥ 2 in Epo-R-transfected FDC-P1 Cells-To assay the possible involvement of PLC-␥ 2 in Epo-induced GPI hydrolysis, we first investigated its tyrosine phosphorylation in WT Epo-R FDC-P1 cells upon Epo stimulation. As shown in Fig. 5A, Epo induced the tyrosine phosphorylation of a 142-kDa protein immunoprecipitated with anti-PLC-␥ 2 antibodies. This effect was maximal after 1-2 min of Epo stimulation and then decreased after 5 min of stimulation. Reprobing the blot with anti-PLC-␥ 2 antibodies confirmed that the 142-kDa protein was indeed PLC-␥ 2 . Epo-induced PLC-␥ 2 tyrosine phosphorylation was dosedependent as shown in Fig. 5B and the Epo effect was maximal at 0.1 units/ml.
Tyrosine Residues of the Epo-R Are Required for Epo-induced PLC-␥ 2 Tyrosine Phosphorylation-In order to know whether tyrosine residues of the Epo-R play a significant role in Epoinduced PLC-␥ 2 tyrosine phosphorylation, we used interleukin-3-dependent FDC-P1 cells transfected with the wild-type (WT) Epo-R and with Epo-R devoid of tyrosine residue (ZERO) (Fig. 1). Then, cells were stimulated for 1 min with Epo 5 units/ml. Cellular extracts were immunoprecipitated with anti-PLC-␥ 2 antibodies and analyzed by Western blotting. As shown in Fig. 6, in contrast to FDC-P1 cells expressing WT Epo-R, Epo did not induce the tyrosine phosphorylation of PLC-␥ 2 in FDC-P1 cells ZERO, suggesting that the tyrosine residues of the intracellular domain of the Epo-R were required for Epoinduced tyrosine phosphorylation of PLC-␥ 2 . Reprobing the blot with anti-PLC-␥ 2 antibodies showed that the same amounts of PLC-␥ 2 had been immunoprecipitated from each sample. Moreover, JAK2 immunoprecipitations carried on the same cellular extracts followed by an anti-phosphotyrosine blot showed that these cells were stimulated by Epo.
Correlation between Epo-induced GPI Hydrolysis and PLC-␥ 2 Tyrosine Phosphorylation-Treatment of recovered [ 3 H]GPI with Epo-activated PLC-␥ 2 bound on protein G beads from immunoprecipitates of WT or ZERO-FDC-P1 cells resulted in an hydrolysis of about 50.2 and 9.8%, respectively (Fig. 8A). Analysis of the resultant products showed a greater IPG content released in the aqueous phase in samples treated with Epo-activated PLC-␥ 2 from WT FDC-P1 cells than from ZERO FDC-P1 cells (Fig. 8A). To the contrary, analysis of [ 3 H]GPI remaining in the organic phase is greater in sample hydrolyzed with Epo-activated PLC-␥ 2 from ZERO FDC-P1 than from WT FDC-P1 cells (Fig. 8, B and C). This demonstrated that PLC-␥ 2 tyrosine phosphorylation is required for GPI hydrolysis. Moreover, as described previously, Epo induced a rapid IPG release within 1 min in FDC-P1 cells transfected with WT Epo-R, and no change of IPG levels was observed in FDC-P1 cells ZERO after Epo stimulation (Fig. 9). IPG levels were also determined upon Epo stimulation (5 units/ ml, 1 min) in metabolically labeled FDC-P1 cells expressing various Epo-R mutants. These experiments show an excellent correlation between Epo-induced PLC-␥ 2 tyrosine phosphorylation and Epo-induced GPI hydrolysis. Indeed, after Epo stimulation, IPG levels were increased in Y 1 and in F 1 Y 2 mutants which only retains a single tyrosine at positions 343 and 401, respectively. These cells differ from FDC-P1 ZERO by only one tyrosine. Nevertheless, IPG level in Y 1 or Y 2 mutants was about half the level obtained in WT-Epo-R FDC-P1 transfected cells. IPG level was greatly increased after Epo stimulation in F 1 Y 5-8 mutant, but IPG increase was also lower than in WT-Epo-R FDC-P1 transfected cells. In Y 1-2 F 3-4 Y 5-8 , IPG level obtained after Epo stimulation was greater than in WT-Epo-R FDC-P1-transfected cells. In F 1 Y 3-4 and Y 1-6 mutants, Epo induced an inhibition of IPG basal level (Fig. 9). DISCUSSION Our results show that FDC-P1 cells contain GPI molecules that are hydrolyzed to produce IPG after Epo stimulation. After TLC or HPTLC analysis, GPI migrated as a single peak between PC and PIP as described previously in Epo-sensitive cells (21) and in other cells (14,37,36). The material of this peak exhibited all the characteristics of GPI. Indeed, it could be labeled with [ 3 H]inositol and [ 3 H]glucosamine in addition to [ 3 H]ethanolamine. Moreover, the GPI structure of this lipid was confirmed by its partial sensitivity to PI-PLC from B. cereus and nitrous deamination of the free amino group on glucosamine. Presence of glucosamine in 3 H-recovered glycolipidic fraction was also confirmed by 2,5-anhydromannitol obtained after deamination/reduction and methanolysis of this molecule. Altogether these results indicate that the labeled glycolipid fraction corresponds to GPI. It has been previously shown that GPI displays various degrees of sensitivity to bacterial PI-PLCs cleavage (14,19,23,24,36,39). Epo stimulation of Epo receptor-transfected FDC-P1 cells metabolically labeled with [ 3 H]glucosamine induced the same level of GPI hydrolysis with release of water-soluble IPG forms. In contrast to GPI molecules involved in insulin signal transduction in H35 hepatoma cells which does not contain ethanolamine (14,40), our results show that GPI molecules containing ethanolamine are involved in Epo signal transduction.
Bacterial PI-PLCs can partially mimic the effects of the extracellular ligand on GPI through the generation of diacylglycerol and IPG (13,41,42), suggesting that the formation of IPG from GPI is due to the activation of a phospholipase C. A GPI-specific PLC has been purified and cloned from T. brucei (43) and peanuts (44). It is able to hydrolyze GPI-anchored proteins and phosphatidylinositol. However, this enzyme is not active on free GPI phospholipids (45). A mammalian GPI-PLC has been partially purified from rat liver membranes (26) but has not been cloned. The presence of two different GPI-PLC activities has also been reported in brain membranes (46). Our results strongly suggest that Epo induced the hydrolysis of GPI through the activation of PLC-␥ 2 . Indeed, PLC-␥ 2 immunoprecipitated from Epo-stimulated FDC-P1 cells was able to hydrolyze GPI molecules purified from these cells. The time course of Epo-induced PLC-␥ 2 tyrosine phosphorylation displays a good parallelism with the kinetics of Epo-induced GPI hydrolysis. Moreover, we observed a full correlation between the ability for mutated Epo receptors to activate PLC-␥ 2 and to induce IPG release. PLC-␥ 2 is expressed mainly in hematopoietic cells (47). In FDC-P1 cells, macrophage-colony-stimulating factor induces activation of PLC-␥ 2 and its tyrosine phosphorylation (34). Here, we observed that Epo stimulation of WT FDC-P1 cells induced a rapid tyrosine phosphorylation and dephosphorylation of PLC-␥ 2 since maximal tyrosine phosphorylation was detected between 1 and 2 min and decreased after this time. Although Epo-induced PLC-␥ 2 has not been previously reported, it has been shown that Epo activates PLC-␥ 1 in the UT-7 human cell line (33). The mechanism of PLC-␥ 1 activation by Epo was not reported. Our results indicate that the tyrosine residues of the Epo receptor are involved in the activation of PLC-␥ 2 . Indeed, Epo receptors devoid of tyrosine residue on their intracellular domain did not activate PLC-␥ 2 . In contrast, Epo induced the activation of PLC-␥ 2 in FDC-P1 cells expressing Epo receptors with either the first (Tyr 343 ) or second tyrosine (Tyr 401 ) residue albeit to a reduced level compared with cells expressing normal Epo receptors. Nevertheless, our results indicate that the main activation site for PLC-␥ 2 is located in the C-terminal part of the receptor, suggesting that the seventh or eighth tyrosine residue could be involved in this activation. Yet, we could not show that PLC-␥ 2 is associated with the activated Epo-R and coimmunoprecipitate with it as for PLC-␥ 1 in UT7/Epo cells (33). This was probably due to the rapid dissociation of the SH2 domains of PLC-␥ 2 from the phosphorylated tyrosine residues of the Epo receptor (48). Previous results have indicated that an intact tyrosine kinase activity is required for GPI hydrolysis in response to growth factors in Chinese hamster ovary cells carrying normal human insulin receptors. These cells hydrolyze up to 70% of their GPI within 2 min after the addition of 0.1 nM insulin, whereas cells expressing a mutant cDNA (Lys-1018 to Ala) that encodes a receptor lacking kinase activity (49) hydrolyzes only 20 -30% in response to insulin (24). More recently, Clemente et al. (23) had indicated that tyrosine kinase inhibitors partially block GPI hydrolysis in parallel to the inhibition of both EGF receptor autophosphorylation and EGF-induced cell proliferation.
Our results also indicate that the tyrosine residues Tyr 429 and/or Tyr 431 are involved in deactivation of PLC-␥ 2 phosphorylation and in GPI hydrolysis inhibition. Indeed, Epo-induced PLC-␥ 2 tyrosine phosphorylation and GPI hydrolysis in mutants F 1 Y 3-4 are lower than in controls without Epo but are greater in mutants Y 1-2 F 3-4 Y 5-8 (deprived of Tyr 429 and Tyr 431 ) than in WT FDC-P1 cells. Then tyrosines Tyr 429 and/or Tyr 431 appear to be negative regulators of GPI hydrolysis. The Tyr 429 was identified as the protein tyrosine phosphatase SHP1 in the cytoplasmic domain of the Epo-R (50). Thus, our results suggest that PLC-␥ 2 could be a substrate for Epoactivated SHP1.
The levels of cell surface expression of Epo-R have been already published for the cell lines used in this study (35) except for the mutant Y 1-2 F 3-4 Y 5-8 which expressed 880 Ϯ 50 Epo receptors. Most Epo-R mutants used in our experiments have been previously shown to be able to mediate Epo-induced Stat5 activation (35). Moreover, the ZERO Epo-R mutant which did not activate Stat5 (35) mediates Epo-induced JAK2

FIG. 8. Correlation between Epo-induced GPI hydrolysis and PLC-␥ 2 activation. A, [ 3 H]GPI from WT FDC-P1 cells recovered after TLC
was incubated for 2 h at 37°C with Epo-activated PLC-␥ 2 from WT FDC-P1 cells or PLC-␥ 2 from Epo-R mutant ZERO as described under "Experimental Procedures." Control reaction was obtained with protein G beads in buffer alone. Values were corrected by subtraction of values obtained from control incubations in the absence of PLC-␥ 2 (14%). Results are presented as the means of triplicate experiments and are expressed as the percentage of the initial label recovered in the aqueous phase after treatment and extraction. Analysis of hydrolysis products was done on one representative experiment; IPG was measured in the aqueous phase (inset), and [ 3 H]GPI remaining in the organic phase was quantitated by Berthold analysis (B) or by fluorographic analysis (C). O, origin; F, solvent front; PIP, phosphatidylinositol monophosphate; PC, phosphatidylcholine.
FIG. 9. Epo-induced GPI hydrolysis in FDC-P1 cells transfected with various Epo-R mutants. FDC-P1 cells expressing either the wild-type Epo receptor or Epo-R mutants were labeled with [ 3 H]ethanolamine (1.5 Ci/ml) for 20 h, deprived of growth factor for 4 h, and then stimulated for 1 min with 5 units/ml Epo (ϩ) or with vehicle alone (Ϫ). IPG release was determined as described under "Experimental Procedures" in WT cells and Epo-R mutants. Data are the means of three independent experiments. activation (Fig. 6). These results show that all Epo-R mutants used in our experiments were activated by Epo. Although the levels of the cell surface expression are not identical, it should be noted that in WT FDC-P1 cells which expressed about 2100 Epo receptors or in Y 1-2 F 3-4 Y 5-8 , we could detect a good level of phosphorylation of PLC-␥ 2 and GPI hydrolysis, whereas in mutant ZERO (without tyrosine residue) or in F 1 Y 3-4 which expressed about 5700 and 7700 Epo receptors, respectively, we could not observe any PLC-␥ 2 activation nor any GPI hydrolysis. From our results, we conclude that the levels of both Epo receptors are not correlated to GPI hydrolysis.
A low level of PLC-␥ 2 tyrosine phosphorylation in immunoprecipitates of control cells was often observed. Nevertheless, no tyrosine phosphorylation of the Epo receptor was observed in these cells (not shown). This low level of PLC-␥ 2 could be attributed to unidentified molecules present in the fetal calf serum required to maintain cell survival during Epo starvation.
The GPI/IPG pathway in Epo signal transduction has already been described in rat erythroid progenitor cells. Erythropoietin-stimulation of these cells increases GPI hydrolysis with a parallel increase in IPG levels, and purified erythroid rat IPG partially mimicked Epo-proliferative effects on erythroid colonies (21). We have also shown that IPG induced Raf-1 and mitogen-activated protein kinase (p44 form) activation, and we also suggested that protein kinase C could be involved in this activation (51). Data presented here suggest a direct link between Epo receptor tyrosine phosphorylation leading to activation of PLC-␥ 2 and GPI hydrolysis and provide complementary information on the GPI/IPG pathway in Epo-proliferative effects.