Association of JAK2 and STAT5 with erythropoietin receptors. Role of receptor phosphorylation in erythropoietin signal transduction.

Cytokine receptors act at least partially by associating with Janus tyrosine protein kinases at the conserved box one motif of the receptor. These receptor-associated kinases then activate STAT transcription factors through phosphorylation. We found that the 78-kDa erythropoietin receptor (EPOR), a highly modified form of the 66-kDa receptor which is abundant in HCD57 cells, was phosphorylated on serine residues without EPO stimulation. Coprecipitation experiments showed the 78-kDa EPOR but not the more abundant 66-kDa EPOR was associated with JAK2, a Janus protein kinase, in both the presence and absence of EPO. Solubilized 78-kDa EPOR bound to purified, genetically engineered JAK2 better than the 62-76-kDa receptor proteins, and additional phosphorylation of tyrosine residues further increased the binding of the 78-kDa EPOR to JAK2-agarose beads. STAT5 DNA binding was activated by 10-100-fold lower concentrations of EPO in HCD57 cells than in primary erythroid cells, and STAT5 associated with the EPOR in an EPO-dependent manner. These data suggest that phosphorylation of either serine or tyrosine residues of the EPOR can enhance the association of the receptor with JAK2, possibly increasing the sensitivity to EPO.

teins expressing cytoplasmic domains of the EPOR, and JAK2 was tyrosine-phosphorylated in EPO-treated cells (9). The conserved box one motif of the cytokine receptors (membrane proximal cytoplasmic domain) is required for the association with the Janus tyrosine protein kinases (10 -12); however, the nature of this binding and modifications of either receptor or kinase that affect this interaction have not been reported. Miura and co-workers showed that JAK2 interacted with the EPOR only after EPO binding (7); however, Wakao and coworkers (13) showed JAK2 binding to the EPOR in both the presence and absence of EPO. We have investigated possible explanations for these disparate findings, and report here that the constitutive phosphorylation of the EPOR on serine residues appeared to enhance the association of JAK2 with the EPOR in erythroid cells. In addition, EPO-induced tyrosine phosphorylation of the EPOR appeared to additionally increase the affinity for JAK2. We report more extensive processing of the EPOR in erythroid cells compared to other cells (26 -29).
Recent information suggests that the STAT proteins are substrates of the Janus kinases bound to the cytokine receptors (14 -16). We previously demonstrated that both STAT1 and STAT5 were tyrosine-phosphorylated, translocated to the nucleus, and activated to bind DNA in primary murine erythroid cells activated by EPO (17). Others have also shown that STAT5 was activated in response to EPO and cytokines related to EPO: prolactin, interleukin-2, interleukin-3, interleukin-5, granulocyte-colony stimulating factor, granulocyte-macrophage colony stimulating factor, and factors not related such as epidermal growth factor (13, 18 -24). In this study we demonstrate that in HCD57 erythroleukemia cells, STAT5 is activated to bind DNA at EPO concentrations 100-fold less than previously reported for primary erythroid cells and that STAT5 binds to the activated EPOR.

MATERIALS AND METHODS
Cells-HCD57 cells, obtained from Sandra Ruscetti at the National Cancer Institute in Frederick, MD (25,30), were cultured in the presence of 1.0 unit of EPO/ml in Iscove's modified Dulbecco's medium and 25% fetal calf serum for several days. In most experiments, the cells were further cultured overnight in the same medium devoid of EPO. As previously reported (25), this deprivation of EPO increased cell surface EPOR.
Western Blot Analysis of Phosphotyrosine and EPOR-The 62-78-kDa forms of the EPOR were visualized after immunoprecipitation and Western blotting with anti-receptor, affinity-purified IgG raised against a synthetic peptide corresponding to the 15-terminal amino acids of the receptor (25). The EPOR was detected on x-ray film by a light generating reaction catalyzed by peroxidase linked to a secondary antibody (ECL from Amersham). Tyr(P)-containing proteins were detected by the immunoprecipitation and Western blotting method described before (25) except monoclonal antibodies against Tyr(P) from either Transduction Laboratories (RC20H) or UBI (4G10) were also used. Polyclonal anti-Tyr(P) antibodies from Zymed were used to immunoprecipitate and concentrate the Tyr(P)-containing proteins before they were analyzed by Western blot. In many experiments, the blots were stripped of bound antibody and reprobed one or two more times with other antisera. The blot was stripped as described by the Amersham literature with the ECL kit. Briefly, the bound antibody was released in the presence of 2% SDS and reducing agents at 50°C for 30 min.
Enzymatic Deglycosylation and Dephosphorylation of EPOR and Phosphotyrosine-containing Proteins-N-Linked carbohydrate from the immunoprecipitated EPOR was digested away with either 50 or 500 units of N-glycanase/ml for 20 h as described previously (25). Phosphate on all residues of the EPOR and Tyr(P)-containing proteins was eliminated by digestion of immunoprecipitates with 50 units of alkaline phosphatase/ml for 6 h at 37°C in a pH 9.0 buffer as described previously (25,26). In the experiment where deglycosylation and dephosphorylation were both done, alkaline phosphatase treatment of the immunoprecipitated EPOR was followed by dialysis, concentration in a SpeedVac, and then N-glycanase treatment. Undigested controls were treated similarly without enzymes.
In Vivo Labeling of the EPOR with 32 P i and Phosphoamino Acid Analysis-HCD57 cells were labeled with 10 mCi of 32 P i for 1 h in phosphate-free medium and dialyzed serum. The EPOR was immunoprecipitated and analyzed by SDS-PAGE, autoradiography, and Western blotting with anti-EPOR IgG. Afterwards, the radioactive bands comigrating with the 78-kDa EPOR were cut out of the nitrocellulose blot and placed directly in 6 N HCl and heated for 1.5 h at 100°C under nitrogen. Products of this acid hydrolysis were analyzed on two-dimensional high voltage electrophoresis, first at pH 1.9, and then at pH 3.5, on thin layer plates as described by others (31). Phosphoamino acids were visualized by autoradiography and compared to standards visualized by staining with ninhydrin.
Coprecipitation of EPOR, JAK2, and STAT5-These studies were carried out in detergent extracts of HCD57 cells made in 1% digitonin lysis solution as described in the report by Miura and co-workers (7). Briefly, HCD57 cultured without EPO were pretreated with 0.5 mM NaVa 3 O 4 , and either not treated or treated with 10 units of EPO for 10 min at 37°C in the culture medium containing 25% fetal calf serum. The cells were diluted with an excess volume of ice-cold culture medium and washed one time. The pellets were suspended in the ice-cold digitonin lysis buffer for 5 min and insoluble material was removed by ultracentrifugation at 250,000 ϫ g for 30 min. The supernatant was divided into equal aliquots and incubated with anti-EPOR IgG, anti-JAK2 antiserum (UBI, Lake Placid, NY), or anti-Tyr(P) antibody. Following SDS-PAGE and Western blotting the blot was probed with either anti-Tyr(P) monoclonal antibody, anti-JAK2 antiserum, anti-STAT5 monoclonal antibody (Transduction Laboratories), or the anti-EPOR IgG. For a control, nuclear proteins were extracted from parallel cells, by the method described previously (17,24), were also run on the same SDS-PAGE gel and blot and were probed with anti-STAT5 antiserum to indicate the position of activated STAT5 that was translocated to the nucleus in EPO-treated cells. These same techniques were used to measure the effect of EPO concentration on STAT5 tyrosine phosphorylation and nuclear translocation. STAT5 binding to radiolabeled DNA (prolactin inducible element) was done as described previously (17).
In Vitro Binding of EPOR to JAK2 Beads-Purified JAK2 protein expressed in the Sf9 cells-baculovirus system that was conjugated to agarose beads was obtained commercially from UBI. HCD57 cells that were deprived of EPO overnight were pretreated with 0.5 mM NaVa 3 O 4 for 50 min, and the cells were either treated with nothing or 10 units of EPO/ml for 10 min at 37°C. Following this incubation, the cells were then lysed at 4°C in a 1% Triton X-100 solution used previously as the binding buffer to demonstrate the interaction of JAK2 with fusion proteins containing domains of the EPOR (9). The unsolublized material was removed by centrifuging at 25,000 ϫ g for 60 min and the extracts were incubated with 33 l of the JAK2 beads for 4 h at 4°C. The beads were then washed 5 times with ice-cold binding solution to remove any soluble EPOR trapped in the pellet. The JAK2 resin and associated EPOR proteins were then exposed to SDS-PAGE sample buffer at 100°C to release bound protein, and the forms of EPOR bound were analyzed by SDS-PAGE and Western blotting using anti-EPOR IgG.

RESULTS
Two Forms of EPOR, 72 and 78 kDa, Were Phosphorylated on Tyrosine Residues after EPO Treatment-We previously demonstrated the 78-kDa form of the EPOR and an unidentified 95-kDa protein were phosphorylated on tyrosine residues following treatment of HCD57 erythroid cells with EPO (25). This experiment was repeated with monoclonal anti-Tyr(P) antibod-ies instead of the polyclonal antiserum. Fig. 1 shows an experiment comparing the effects of increasing EPO concentration on Tyr(P) proteins in HCD57 cells determined with a monoclonal anti-Tyr(P) (Fig. 1, upper panel) and the polyclonal antiserum used previously (25) (lower panel). The polyclonal anti-Tyr(P) was effective in recognizing a phosphorylated 95-kDa protein and weakly detecting the 78-kDa EPOR, but the monoclonal antiserum strongly recognized the EPOR and number of proteins ranging from 47 to 200 kDa. We have determined that the predominant band at 78 kDa in EPO-treated cells is the phosphorylated EPOR as previously reported (Fig. 2). This was shown by the appearance of the 78-kDa band in the anti-EPOR immunoprecipitate (Fig. 2, lane D) and the corresponding anti-Tyr(P) immunoprecipitate when reprobed with anti-EPOR IgG (lane G). However, prolonged exposure to film revealed another Tyr(P)-containing band of 72 kDa in the anti-EPOR immunoprecipitate. Reprobing this blot with anti-EPOR IgG revealed that the 72-kDa band, immunoprecipitated by anti-Tyr(P), was also a form of the EPOR. This 72-kDa band reacted with both anti-COOH and anti-NH 2 anti-EPOR IgG, indicating that it was not a truncated form of receptor. Therefore, both a major 78-kDa EPOR and less abundant 72-kDa EPOR were phosphorylated in a EPO-dependent manner.
The Largest Form of EPOR, 78 kDa, Was Glycosylated and Phosphorylated on Serine Residues in the Absence of EPO-Previous studies showed that the Tyr(P)-containing 72-kDa form of the EPOR resulted from the shift of a 66-kDa EPOR to 72 kDa because of phosphate interference in SDS-PAGE (7, FIG. 1. Effect of EPO concentration on Tyr(P)-containing proteins in HCD57 cells. HCD57 cells were deprived of EPO overnight and then were either untreated (lane A) or treated with the indicated concentration of EPO for 10 min at 37°C. The cells were then lysed in a lysis buffer containing 1% Triton X-100, the extract was divided into two aliquots, and Tyr(P)-containing proteins were isolated by immunoprecipitation with a polyclonal anti-Tyr(P) antiserum. Following SDS-PAGE and blotting, the blot was probed with anti-Tyr(P) monoclonal antibody (upper panel A), or probed with the anti-Tyr(P) polyclonal antiserum (lower panel B). 26 -28). However, the 78-kDa EPOR was present before EPO treatment. Fig. 3 shows that the Tyr(P)-containing 78-kDa EPOR was shifted to 72 kDa by enzymatic dephosphorylation. Control experiments in which the alkaline phosphatase activity was inhibited with vanadate (not shown) verified that the 6-kDa shift of the EPOR in apparent molecular mass was due to phosphate on the protein and was not the result of contaminating proteinase activity.
Because the 78-kDa EPOR was also present in cells not treated with EPO, the possibility that the 78-kDa EPOR was phosphorylated before EPO-induced tyrosine phosphorylation was tested. Fig. 4, lane 7, shows that sequential deglycosylation and dephosphorylation completely converted all the forms of EPOR (62-78 kDa) found in HCD57 cells in the absence of EPO into the 62-kDa unprocessed peptide (lane 7). This indicates that both N-linked glycosylation and phosphorylation is contributed to the larger molecular weight forms of the receptor because deglycosylation alone generated a subset of receptors with a molecular mass of 68 kDa (Fig. 4, lanes 4 and 6) and dephosphorylation alone shifted the total receptors from 78 to 72 kDa. These data are consistent with the 78-kDa EPOR shifted in migration on SDS-PAGE both due to extensive glycosylation (10-kDa shift) and phosphorylation in the absence of EPO (6-kDa shift) such that EPO-induced phosphorylation on tyrosine residues did not additionally retard the migration of the protein on SDS-PAGE.
As a control in the experiment shown in Fig. 4, HCD57 cells were exposed to EPO for 24 h or longer such that the 70 -78-kDa EPOR were down-regulated through destruction of the cell surface receptors (Fig. 4, lane 1). The remaining 64 -69-kDa receptors in these cells were completely converted to the 62-kDa unprocessed protein by N-glycanase, which demonstrated that the digestion with N-glycanase was complete (lanes 1 and 3). As a further control, 10-fold more N-glycanase did not diminish the 68-kDa material that remained after enzymatic deglycosylation (compare lane 6 with lanes 3 and 4). The combined digestion with alkaline phosphatase and N-glycanase did not generate fragments of EPOR that would be expected if a contaminating proteinase had cleaved 6 kDa from the receptor in Fig. 3, indicating the absence of proteolytic artifacts.
To examine the nature of non-tyrosine phosphorylation on the EPOR in the unstimulated state more directly, the EPOR was immunoprecipitated from 32 P i -labeled HCD57 cells untreated or treated with EPO. As illustrated in Fig. 5, radioactive bands comigrating with the 78-kDa EPOR were seen in cells treated with both EPO and untreated cells (lanes A and B). The radioactivity was increased after EPO treatment by 2-3-fold compared to untreated cells, but the band was not shifted on SDS-PAGE. A control Western blot with anti-Tyr(P) antiserum showed that no Tyr(P) was detected in the 78-kDa band in the absence of EPO (Fig. 5, lane F). These radioactive bands were cut out of the blot, the proteins were hydrolyzed, and radioactive phosphoamino acids were analyzed by twodimensional electrophoresis as shown in Fig. 6. The band corresponding to the EPOR from untreated cells contained radioactive phosphoserine and no radioactivity was detected in Tyr(P). Treatment of the HCD57 cells with EPO resulted in the  A, C, F, and H). The cells were then lysed in a lysis buffer containing 1% Triton X-100, the extract was divided into two aliquots, and either Tyr(P)-containing proteins were isolated by immunoprecipitation with anti-Tyr(P) antiserum (lanes A, B, G, and H) or the EPOR and associated proteins were recovered by immunoprecipitation with anti-EPOR IgG (lanes C-F). Following SDS-PAGE and blotting, the blot was probed with anti-Tyr(P) monoclonal antibody (A-D) or stripped of bound antibody and reprobed with anti-EPOR antiserum (E-H). appearance of 32 P in Tyr(P) in the band at 78 kDa. EPO treatment also appeared to increase the 32 P incorporated into phosphoserine residues of the EPOR in this experiment. These studies indicate that the 78-kDa EPOR was shifted in molecular mass by the constitutive serine phosphorylation as well as N-linked carbohydrate, such that additional phosphorylation on tyrosine residues do not retard the protein on SDS-PAGE.
JAK2 Associated with EPOR in the Presence and Absence of EPO, and STAT5 Associated with the EPOR Only in EPOtreated Cells-Proteins coprecipitating with the EPOR were examined in HCD57 cells lysed in 1% digitonin, as shown in Fig. 7. Proteins coprecipitating with JAK2 were also analyzed. The anti-Tyr(P) blot (lanes A-D) showed that the tyrosinephosphorylated 78-kDa EPOR was immunoprecipitated from EPO-treated cells by anti-EPOR IgG and JAK2 (135 kDa) was coprecipitated with EPOR (lane B). Tyrosine-phosphorylated JAK2 was immunoprecipitated from EPO-treated cells with anti-JAK2 antiserum and the 78-kDa EPOR was coprecipitated with JAK2 (lane D). In addition, a 95-kDa Tyr(P)-containing band was present in both immunoprecipitates from EPOtreated cells but was severalfold more abundant in the anti-EPOR immunoprecipitate. This blot was stripped of bound anti-Tyr(P) antibody and reprobed with anti-JAK2 antiserum (lanes E-H). JAK2 was associated with the EPOR in both the absence and presence of EPO (lanes E and F). Furthermore, approximately half of the JAK2 protein in the cell appeared to be associated with the EPOR. The blot was then stripped of bound JAK2 antibody and reprobed with anti-STAT5 monoclonal antibody. The 95-kDa Tyr(P)-containing protein coprecipitated with the EPOR in EPO-treated cells appeared to be STAT5 because a 95-kDa band was seen in anti-EPOR immunoprecipitates when probed with anti-STAT5, but only in immunoprecipitates from EPO-treated cells (lane J). A many fold weaker 95-kDa band was also seen in the anti-JAK2 immunoprecipitates from EPO-treated cells, but this is likely STAT5 bound to the coprecipitated EPOR in these immune complexes. For comparison, nuclear extracts from parallel HCD57 cells not treated or treated with EPO (Fig. 7, lanes M and N, respectively) were run on the same gel and probed with anti-STAT5 antibody. STAT5 was translocated to the nucleus in an EPOdependent fashion as we have demonstrated previously in other erythroid cells (17). This authentic STAT5 comigrated with the 95-kDa Tyr(P)-containing band and STAT5 coprecipitated with the EPOR. It was estimated that approximately 2% of the STAT5 activated, as quantified by the amount translocated to the nucleus in EPO-treated cells, was bound to the EPOR in this experiment.
JAK2 Selectively Coprecipitated the 78-kDa EPOR-The forms of EPOR that coprecipitated with JAK2 were examined. The blot used in Fig. 7 did not give a clean result when stripped and reprobed with anti-EPOR IgG. Therefore, a new blot was prepared from the same cell extracts, and this blot was probed with anti-EPOR IgG as show in Fig. 8. In addition to immunoprecipitates using anti-EPOR (Fig. 8, C and D) and anti-JAK2 antibodies (E and F), anti-Tyr(P) antibodies recovered Tyr(P)containing EPOR (A and B). All these immunoprecipitates were Western blotted and probed for EPOR proteins. Confirming the data from Fig. 2, the 78-kDa form of the EPOR was phosphorylated on the tyrosine residues in EPO-treated cells (Fig. 8, lanes A and B), a minor 72-kDa form was seen with longer film exposures, and there was no apparent shift in molecular mass of the EPOR forms ranging from 62 to 78 kDa with EPO treatment (lanes C and D). Interestingly, only the 78-kDa form of the EPOR was detected associated with JAK2 as shown in Fig. 8, lanes E and F. As expected from the data in Fig. 7, the 78-kDa EPOR was coprecipitated with JAK2 in both After washing at 4°C to remove unincorporated 32 P i , the cells were lysed in lysis buffer (containing 1% Triton X-100) and centrifuged at 25,000 ϫ g for 1 h. The nuclei were discarded and the EPOR was recovered by immunoprecipitation with anti- EPOR IgG (lanes A-D). Tyr(P)-containing proteins were immunoprecipitated from a parallel culture of HCD57 not labeled with 32 P i but either treated with EPO (E) or not treated (F). Following SDS-PAGE and Western blotting, the blot was exposed to film to detected the 32 P-labeled EPOR (A and B) and then probed with anti-EPOR IgG (C-F).
FIG. 6. The 78-kDa EPOR protein was phosphorylated on serine residues in the absence of EPO and became additionally phosphorylated on tyrosine residues after the cells were treated with EPO. The bands from lanes A and B of Fig. 5 were cut out of the blot and subjected to partial acid hydrolysis and phosphoamino acid analysis as described under "Materials and Methods." Panel A shows the position of phosphoserine, phosphothreonine, and Tyr(P) standards run on the thin layer plate show in panel B and visualized after staining with ninhydrin. Panel B is an autoradiogram of the hydrolyzed 78-kDa EPOR band from EPO-treated cells, and panel C is the phosphoamino acid analysis of the 78-kDa band from untreated cells.

untreated and EPO-treated cells.
The 78-kDa EPOR Selectively Associated with Genetically Engineered JAK2 Protein Compared to the 62-76-kDa Forms of EPOR-To rule out the possibly that the unique association of JAK2 with the 78-kDa EPOR resulted from circumstances other than the direct interaction of these proteins, solubilized HCD57 cellular extracts, 1% Triton X-100, were incubated with purified JAK2 conjugated to agarose beads. The forms of EPOR in the detergent extract that associated with these JAK2 beads were analyzed by Western blotting with anti-EPOR IgG. As shown in Fig. 9, in untreated HCD57 cells, the 78-kDa EPOR was the predominant receptor form that bound to the JAK2 beads while the most abundant 66-kDa from of EPOR was bound to a lesser extent. In contrast to the earlier coprecipitation experiments showing equal association of JAK2 with the EPOR in the presence and absence of EPO, the binding of the 78-kDa EPOR to the JAK2 beads was increased 10-fold in extracts from EPO-treated cells compared to untreated cells. This experiment was carried out in a stringent 1% Triton X-100 detergent solution rather than mild digitonin lysis buffer such that JAK2, STAT5, and probably other associated proteins would be dissociated from the EPOR (shown by the experiment in Fig. 2). This reduced the possibility that JAK2 or other proteins binding the EPOR facilitated or interfered with the binding of the receptor to the JAK2 beads. We greatly reduced the possibility that the increased affinity of the 78-kDa EPOR from EPO-treated cells resulted from a conformational change of the EPOR induced by EPO binding rather than EPO-induced tyrosine phosphorylation by determining that virtually all the EPO bound to EPOR was released during this experiment. These results are consistent with the hypothesis that the posttranslational modification of the EPOR by either constitutive phosphorylation on serine residues or EPO-induced phosphorylation on tyrosine residues enhanced the interaction of the EPOR with JAK2.

Time Course and EPO Concentration Dependence of EPOR, JAK2, and STAT5 Activation and Effect of the Phosphatase
Inhibitor, Orthovanadate-The time course and concentration of EPO needed for phosphorylation for total cellular proteins, EPOR, JAK2, and STAT5 were determined. As shown in Figs. 1 and 10, the tyrosine phosphorylation of the 78-kDa EPOR was detected with 0.01 unit of EPO/ml (less than the physiological concentration of 20 -40 milliunits/ml of plasma) and was maximum at 1 unit of EPO/ml in a 10-min treatment, time of peak kinase activity. Fig. 10 also shows that the 135-kDa band that coprecipitates with the EPOR, shown to be JAK2 in Fig. 7, is substantially tyrosine-phosphorylated at 100 milliunits of EPO/ml and near maximally phosphorylated at 1 unit of EPO/ml. The tyrosine phosphorylation of STAT5, nuclear translocation of STAT5, and activation of STAT5 DNA binding activity were more sensitive to low concentrations of EPO than the phosphorylation of EPOR or JAK2, becoming maximal at 100 milliunits of EPO/ml. Indeed, direct comparison of the EPO dose-response in HCD57 cells compared to primary proerythroblasts from mice infected with the anemia strain of Friend virus, FVA cells, showed that the activation of STAT5-like DNA binding activity was 10 -100-fold more sensitive to EPO in the HCD57 cells (shown in Fig. 11).  (lanes B, D, F, H, J, L, and N) or were not treated (lanes A, C, E, G,  I, K, and M). The cells were lysed in a detergent solution containing 1% digitonin, and insoluble material was removed by centrifugation at 250,000 ϫ g for 30 min. The lysate was divided into two aliquots and either anti-EPOR IgG (lanes A, B, E, F, I, and J) or anti-JAK2 antiserum was used for immunoprecipitation of the desired protein and associated proteins.  (lanes A and B), anti-EPOR IgG (lanes C and D), and anti-JAK2 antiserum (lanes E and F). After SDS-PAGE and Western blotting, the blot was probed with anti-EPOR IgG. Arrows indicate the 62-78-kDa range of EPOR forms present in these cells. FIG. 9. The binding of EPOR to purified, genetically engineered JAK2 protein conjugated to beads. HCD57 cells were deprived of EPO overnight, pretreated with 0.5 mM NaVa 3 O 4 for 50 min, and then either treated with 10 units of EPO/ml for 10 min (lanes 2 and 4) or were untreated (lanes 1 and 3). The cells were then lysed in a lysis buffer containing 1% Triton X-100 and the extract was divided into two aliquots, one aliquot was incubated with anti-EPOR IgG to immunoprecipitate the 62-78-kDa forms of EPOR present (lanes 1 and 2) and the other was incubated with the JAK2 beads as described under "Materials and Methods" (lanes 3 and 4). The total EPOR present in the cells and the EPOR forms that bound to the JAK2 beads were determined by SDS-PAGE and Western blotting using the anti-EPOR IgG. As a control, the JAK2 beads alone without cellular extract were simultaneously processed and analyzed to control for the nonspecific interaction of the antiserum with the proteins released from the JAK2 beads when boiled in sample buffer (lane 5).
Our unpublished experiments 2 found that less than 10% of the 78-kDa EPOR molecules were phosphorylated on tyrosine in EPO-treated HCD57 cells; however, when the cells are pretreated with 0.5 mM orthovanadate, EPO induced phosphorylation on tyrosine residues of nearly all the 78-kDa EPOR molecules (Figs. 2 and 8). Fig. 12 shows that pretreatment with vanadate both increased the tyrosine phosphorylation of the 78-kDa EPOR and most other phosphorylated proteins by 10 -20-fold and changed the kinetics of phosphorylation following EPO induction, but did not significantly change the population of Tyr(P)-containing proteins (compare Figs. 1 and 3 without vanadate to Figs. 2 and 12 with vanadate treatment). In the absence of vanadate, the Tyr(P) content of the 78-kDa EPOR in EPO-treated HCD57 cells was constant for 60 min as previously reported (25). However, in the presence of vanadate, the tyrosine-phosphorylated 78-kDa EPOR was diminished after 10 min of treatment with EPO in parallel with the EPO-dependent down-regulation of the cell surface EPOR previously reported (25). Since vanadate is an inhibitor of phosphatase activity, it is surprising that treatment of HCD57 cells with it accelerated the disappearance of Tyr(P). However, this decline in Tyr(P)-containing proteins likely reflects the EPO-dependent destruction of the signaling molecules and other means to deactivate signaling more than dephosphorylation by phosphatase. JAK2, STAT5, and other Tyr(P)-containing proteins also declined after 10 min.

DISCUSSION
This study extends our previous discovery and preliminary characterization of the 78-kDa form of EPOR in erythroid cells (25). The hypothesis that these serine-phosphorylated and highly glycosylated 78-kDa EPOR molecules transduced an intracellular signal better than the less modified EPOR molecules was tested. Supporting data are as follows. 1) the 78-kDa form of EPOR was selectively coprecipitated with JAK2. 2) The 78-kDa EPOR was phosphorylated on tyrosine residues in EPO-treated cells while only a minor 72-kDa form of EPOR containing 5% or less of the Tyr(P) of the 78-kDa EPOR was also observed. 3) Approximately half of the total cellular JAK2 was associated with this 78-kDa EPOR in both the absence and presence of EPO. 4) In vitro binding experiments showed that the soluble, serine-phosphorylated 78-kDa EPOR bound to genetically engineered JAK2 better than other forms of EPOR; 2 S. T. Sawyer and K. Penta, unpublished data.  B-F). The cells were then lysed in a lysis buffer containing 1% digitonin as described in the legend to Fig. 7, and the EPOR and coprecipitating proteins were isolated by immunoprecipitation with a polyclonal anti-EPOR antiserum. Following SDS-PAGE and Western blotting, the blot was probed with anti-Tyr(P) monoclonal antibody. however, EPO-induced tyrosine phosphorylation greatly increased the ability of the soluble, 78-kDa EPOR molecule to interact with immobilized JAK2.
A central role of JAK2 in signal transduction through the EPOR is now clear. Expression of a dominant negative JAK2 mutant protein abolished EPO-dependent proliferation and survival in responsive cells, proving an essential role of JAK2 in EPO signal transduction (32)(33)(34). One study found that JAK2 was coprecipitated with the EPOR only after EPO treatment of 32D cells (7), but others found JAK2 associated with the EPOR in the absence of EPO (13). Here, we sought to determine which forms of EPOR interacted with JAK2 and if the binding required EPO. Although our in vitro data show that the interaction of soluble molecules of EPOR and JAK2 was increased after tyrosine phosphorylation of the EPOR, the data from the intact cells showed that the 78-kDa EPOR was associated with JAK2 in the presence and absence of EPO. The findings of this report support the hypothesis put forth by Witthuhn and co-workers (9), that JAK2 may be bound to the EPOR in the unstimulated state and may be activated by the interaction with another JAK2 protein after EPO-dependent homodimerization of the EPOR (29). However, the results presented here also suggest that this interaction may depend on post-translational modification of the EPOR. Additional glycosylation in the 78-kDa EPOR might facilitate the interaction of JAK2. However, because the N-linked carbohydrate is added at a single site in the extracellular domain of the EPOR (35) and carbohydrate-deficient mutant EPOR molecules transduced a signal (36), the constitutive phosphorylation of the EPOR on serine seems to be the factor that affects the association with JAK2. The finding that EPO-induced tyrosine phosphorylation also increased binding of EPOR molecules to immobilized JAK2 is consistent with serine phosphorylation increasing the affinity of EPOR for JAK2.
We report here the EPO-dependent tyrosine phosphorylation, nuclear translocation, and activation of DNA binding of STAT5 in HCD57 cells as well as the direct association of STAT5 protein with the EPOR. Recently two genes coding for STAT5A and STAT5B have been identified (20). The monoclonal anti-STAT5 antibody used here recognized both A and B forms, and preliminary experiments showed that both STAT5A and STAT5B were equally expressed in tissues and equally activated by a number of cytokines (20). Thus, we likely have detected the EPO-dependent activation of both STAT5A and STAT5B.
The interaction of the STAT5 protein with the activated EPOR may confer specificity in phosphorylation of STAT5 by JAK2. Indeed we did not detect STAT5 binding to JAK2. The SH2 domain of the STAT5 may bind to a Tyr(P) residue of EPOR, as is the mechanism used by some STAT proteins to bind to other receptors (14 -16, 37). When Tyr 343 in the EPOR is mutated, activation of STAT5 occurred only at very high concentrations EPO (38). This result was interpreted as the possible interaction of STAT5 with the Tyr 343 of the EPOR. However, STAT5 can also interact with either the EPOR (13,38) or granulocyte macrophage-colony stimulating factor receptor (20) in which all the Tyr residues have been deleted. This probably means that docking of STAT5 to the EPOR can occur through mechanisms other than the binding of the SH2 domain of STAT5 to a Tyr(P) residue of the EPOR but does not preclude the interaction of STAT5 with Tyr(P). Homodimerization of STAT5 proteins through the binding of a Tyr(P) residue (Tyr 694 ) to the SH2 domain of a partner STAT is required for the nuclear translocation and activation of DNA binding activity (14 -16, 39). Therefore, the affinity of the STAT5 protein for another STAT dimerization partner is likely greater than the affinity of STAT5 proteins for the EPOR. The small percentage (1-5%) of the total STAT5 activated (assumed to be the STAT5 transported to the nucleus) that bound to the EPOR is likely STAT5 that is not yet phosphorylated and/or phosphorylated STAT that had not been approached by a dimerization partner.
The result of the experiment shown in Fig. 2 clarifies the controversial observation that a 78-kDa form of EPOR was phosphorylated on tyrosine residues in EPO-treated erythroid cells (25) while others observed only a 72-kDa form of phosphorylated EPOR in cells expressing the EPOR cDNA (7, 26 -28). This observation of both 72-and 78-kDa forms of Tyr(P)-containing EPOR eliminates the possibility that the 72-and 78-kDa forms are the same EPOR assigned different molecular weights in different laboratories.
This report also clarifies the unusual observation that the Tyr(P) content of EPOR molecules was constant over time while the majority of cell surface receptors were destroyed in EPO-treated cells (25). In the absence of vanadate (a tyrosine phosphatase inhibitor) only a minority of cell surface receptors were phosphorylated after EPO treatment; however, exposure to vanadate resulted in virtually every cell surface 78-kDa EPOR becoming phosphorylated on tyrosine residues. Thus, the tyrosine-phosphorylated EPOR molecules were then downregulated in the presence of EPO in parallel with EPO binding sites and total cell surface receptors. In the absence of vanadate, there appears to be a steady state equilibrium of tyrosinephosphorylated EPOR molecules in EPO-treated cells that reflect the dephosphorylation by phosphatase, destruction of occupied EPOR, continual phosphorylation by JAK2, and possibly other events. The physiological consequences of this is not clear. However, HCD57 cells were increased in sensitivity to EPO by vanadate treatment (40) which suggests that the level of receptor phosphorylation is directly related to signal transduction.
Although vanadate can further increase the sensitivity of HCD57 cells to EPO, these cells are remarkably sensitive to EPO in the absence of vanadate, being 10 -100-fold more sensitive than most cells (40). This increased sensitivity is apparent in this study of the activation of STAT5 DNA binding activity in HCD57 cells by 10 -100-fold less EPO compared to the activation of STAT5 in primary FVA erythroblasts (Fig.  11). Cells expressing EPOR cDNA may be manipulated to become hypersensitive to EPO through truncation of the COOH-terminal domain of the receptor (41) or with the disruption of the ability of a tyrosine phosphatase to interact with the FIG. 12. Time course of Tyr(P)-containing proteins following EPO treatment of control and vanadate-pretreated HCD57 cells. HCD57 cells were deprived of EPO overnight and then either untreated or pretreated with the 0.5 mM Na 3 VO 4 for 50 min at 37°C. Then 10 units of EPO/ml was added and at the indicated times after the addition of EPO, the cells were immediately cooled to ice bath temperature, lysed in a detergent solution, and Tyr(P) containing proteins were analyzed by immunoprecipitation and SDS-PAGE and Western blotting with a monoclonal anti-Tyr(P) antibody as described under "Materials and Methods." EPOR at Tyr 429 (42,43). However, the 10 -20-fold increase in EPOR tyrosine phosphorylation seen in Fig. 12 after vanadate treatment is not consistent with a loss of phosphatase activity interacting with the EPOR in HCD57 cells, and we have sequenced the cDNA coding for the EPOR, analyzed the receptor mRNA, and studied receptor proteins in HCD57 cells without finding any mutations or evidence that truncated EPOR molecules are expressed.
In the primary FVA erythroid cells, the 78-kDa EPOR was modified and associated with JAK2 as described here for HCD57 cells, with two notable exceptions. 3 First, the number of EPOR molecules processed into the serine-phosphorylated 78-kDa form was only 20% of the level found in HCD57 cells while the total numbers of receptors per cell were similar; and second, the fraction of total JAK2 associated with the EPOR was much less than the 50% level found in HCD57 cells (5% or less). We suspect that the increased sensitivity to EPO found in HCD57 cells compared to FVA cells and other cells may result from the increased post-translational modification of EPOR molecules to the 78-kDa species which has increased affinity for JAK2. If most EPOR molecules are already associated with the JAK2 kinase as in HCD57 cells, the occupancy of a small number of receptors may transduce a stronger intracellular signal.
In summary, these data demonstrate a major role of the most post-translationally modified form of EPOR in signal transduction. The 78-kDa EPOR had a higher affinity for the JAK2 protein kinase than other cellular forms of the EPOR, and was the preferred substrate for the JAK2 kinase that was bound to this receptor in the unactivated state. It is likely that the constitutive phosphorylation of this 78-kDa EPOR on serine residues enhanced the ability of this receptor to interact with JAK2. The increased affinity of the tyrosine-phosphorylated 78-kDa EPOR for JAK2 may contribute to prolonged signal transduction following binding of EPO. The strong possibility that serine phosphorylation of the EPOR influences its association with JAK2 suggests that there may be as yet unknown physiological control of EPO-mediated signal transduction through serine protein kinases. This report also shows that STAT5 associated with the EPOR only after EPO binding. This supports the notion that STAT proteins must first bind to activated cytokine receptors before becoming substrates for the Janus protein kinase(s) associated with the receptor.