Cell surface organization of the erythropoietin receptor complex differs depending on its mode of activation.

During erythroid development erythropoietin (EPO) binds specifically to a receptor primarily present on committed erythroid progenitors, stimulating mitogenic, survival, and differentiative growth response pathways. Other modes of erythropoietin receptor (EPO-R) activation, such as interaction with the env gene Friend virus envelope glycoprotein (F-gp55) of spleen focus-forming virus or specific mutations in the extracellular domain of the EPO-R, give rise to pathological consequences, in vivo and EPO-independent proliferation and differentiation of cultured cells. Activating extracellular receptor mutations result in covalently linked receptor homodimers. These observations and others have led to the proposal that EPO activates the EPO-R by inducing dimer formation on the cell surface. It has been assumed that F-gp55 also induces dimer formation of the EPO-R; however, clear evidence of this is lacking. In addition, EPO and F-gp55 stimulation of the EPO-R elicit different biological responses. To probe whether the cell surface EPO-R is structurally different with these activators, we contrasted the cell surface EPO-R complex formed following receptor activation by EPO, F-gp55, and mutations in the extracellular domain of the receptor. Our results indicate that cell surface forms of activated EPO-R differ, as judged by their differential association with F-gp55 and pattern of associated cell surface proteins. Interestingly, we find that the env gene of an anemic strain of Friend virus, Rauscher virus envelope glycoprotein, does not interact with the EPO-R at the cell surface. Thus, the mode of Rauscher virus envelope glycoprotein-induced erythroblastosis may be distinct from F-gp55-induced erythroblastosis and possibly not involve the EPO-R.

Friend virus, a murine retrovirus, specifically causes erythroleukemia. The disease is acute in its onset and progresses to leukemia following multiple genetic events (1). One to 2 weeks following infection polyclonal erythroblastosis develops. Five weeks or later, but not before, clonal erythroleukemic cell lines can be isolated from the spleen or bone marrow of infected mice (2). Since leukemia induction is rapid, Friend virus is considered an acutely transforming retrovirus, yet it lacks oncogenes derived from the host genome (3), which distinguishes it from most acute transforming retroviruses. Friend virus is a complex of replication-competent Friend murine leukemia virus and a replication-defective Friend spleen focus-forming virus (F-SFFV). 1 The defective F-SFFV virus is responsible for leukemia induction and progression (4,5); the helper virus supplies the capacity to replicate and package the F-SFFV genome and, thus, is interchangeable. Genetic and biochemical studies have established the importance of the env gene to disease induction (6,7). env of SFFV is a hybrid gene composed of dualotropic (mink cell focus-forming virus-like) and ecotropic env gene sequences (8). It directs the synthesis of a 55-kDa glycoprotein in the endoplasmic reticulum; a minor fraction (3-5%) of gp55 is further processed and transported to the cell surface, where it has a molecular size of 65 kDa. The majority of cell surface gp55 (gp55 P ) is present as a disulfide-linked homodimer (9,10), and importantly, cell surface expression is essential for disease induction (11)(12)(13).
Early polyclonal erythroblastosis results from activation of the red cell-specific erythropoietin receptor (EPO-R) following cell surface complex formation between gp55 and the erythropoietin receptor (13)(14)(15)(16)(17). This interaction results in EPO-independent mitogenic signals. Thus, the restricted expression of the EPO-R, not cellular tropism of the virus, is the major determinant of the leukemic phenotype. In cultured cells coexpression of gp55 and EPO-R also results in EPO-independent proliferation of these cells. Despite EPO-independent growth, EPO still binds to EPO-R on the surface of Friend erythroleukemia cells without any detectable change in binding affinity or receptor number. Cell surface EPO protein cross-linking studies have detected a complex of EPO-R and gp55 P (16). In addition, EPO and gp55 stimulation of the EPO-R elicit somewhat different biological responses. Retroviral transfer of either EPO or gp55 into mice both led to acute erythroblastosis, but EPO appeared to promote differentiation over proliferation, whereas gp55 shifted the balance toward survival and proliferation of early erythroid progenitors (18). The molecular basis for this difference and precisely how gp55 activates the EPO-R to signal proliferative signals is not known.
There are two strains of Friend virus: polycythemic (FV-P) and anemic (FV-A). Rauscher SFFV is biologically and genetically equivalent to the SFFV component of FV-A virus (19,20). Both FV-P and FV-A cause erythroblastosis and erythroleuke-mia, but the early manifestations of disease and growth factor dependence of infected cells differ. Despite an increase in erythroid cell mass, mice infected with FV-A develop anemia secondary to a concomitant and unexplained increase in plasma volume (21). Infected erythroid progenitors and leukemic cells from these mice remain dependent on EPO for proliferation, whereas these same cells from mice infected with FV-P proliferate in the absence of EPO (22). Genetic differences between the defective FV-P and FV-A (or Rauscher virus) virus env genes are predominantly within the transmembrane domain (8,20). In fact, replacement of the transmembrane domain of FV-A SFFV with that encoded by FV-P converts the FV-A virus to an FV-P virus, highlighting the importance of the transmembrane domain of these proteins to disease phenotype (23). In addition, gp55 of FV-A is either processed to the cell surface less efficiently than FV-P gp55 (10,24) or not expressed at the cell surface (25). In contrast to FV-P gp55 or F-gp55, co-expression of FV-A gp55 or Rauscher gp55 (R-gp55) and EPO-R in the same cells does not lead to EPO-independent proliferation, despite the detection of R-gp55⅐EPO-R complexes in detergentsoluble extracts from these cells (26). Whether R-gp55 P interacts with the cell surface EPO-R has not been determined. Since activation of the EPO-R by F-gp55 initiates unregulated polyclonal erythroblastosis, the mechanism of R-gp55-induced erythroblastosis remains unclear.
The erythropoietin receptor transduces mitogenic (27), differentiative (28), and survival signals (29) following engagement with its physiological ligand EPO, through association with F-gp55 of SFFV, or following acquisition of selective mutations in the extracellular domain (30). Extracellular EPO-R point mutations, for example, R129C, E132C, and E133C, result in the formation of covalently linked receptor homodimers, which are functional in the absence of EPO (31,32). These observations and others (33), including the recent solution of the structure of the EPO-R bound to a peptide agonist (34), have led to the proposal that EPO activates the EPO-R by inducing dimer formation on the cell surface. It has been assumed that F-gp55 also induces dimer formation of the EPO-R; however, clear evidence of this is lacking. Analyses of the structure of the EPO-R-peptide agonist complex suggest that the structure of EPO-R (R129C) may be distinct from wild type EPO-R dimers (34). In addition, in vivo activation of the EPO-R by EPO, F-gp55, or R129C mutations results in different effects on red cell development (18,35). Since cell surface activation of the EPO-R is essential for biological response, we decided to examine and contrast the cell surface EPO-R complex formed following these differing modes of receptor activation. In addition, we wanted to determine whether gp55 of the anemic strain of Friend virus (R-gp55) interacts with the cell surface EPO-R.

EXPERIMENTAL PROCEDURES
Reagents-The SFFV.cEPO-R, SFFV.EPO-R, and Rauscher murine leukemia viruses have been previously described (36). A polycythemic strain of F-SFFV virus and the plasmid pSFe/neo were kindly provided by Dr. J.-P. Li (New York University, New York, NY) (11). To prepare F-SFFV/neo virus, pSFe/neo was transfected into psi-cre (ecotropic) and psi-crip (amphotropic) packaging cell lines by the calcium phosphate method. After selection in G418, the two cell lines were mixed, and the virus was allowed to amplify in culture, as described previously (36). All virus producer cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated calf serum, L-glutamine, and antibiotics. Rabbit polyclonal antisera against the carboxyland amino-terminal 14 amino acids of the EPO-R have been previously described (37). Monoclonal antisera 7C10, directed against SFFV gp55, was kindly provided by Dr. S. Ruscetti (NCI-Frederick) (38). Goat anti-Rauscher murine leukemia virus env gp70 polyclonal antisera was purchased from the NCI (39). Disuccinimidyl suberate (DSS) was obtained from Pierce and prepared as a 100 ϫ stock in Me 2 SO. Wehi-3B cell-conditioned supernatant was used as a source of IL-3. Human EPO was provided by Ortho Pharmaceuticals (Raritan, NJ). Carrier-free pure human EPO was provided by Abbott. G418 (Geneticin) was from Life Technologies, Inc. Lactoperoxidase, hydrogen peroxide, potassium iodide, biotin-x-hydrazide, and sodium periodate were obtained from Sigma. [ 125 I]NaI was from Amersham Corp.
Binding Studies-EPO was iodinated using the iodine monochloride method (40) or Iodogen reagent (Pierce) following the manufacturer's procedure. 125 I-EPO was purified by sequential PD-10 gel filtration (Pharmacia Biotech Inc.) and CM-Sepharose ion exchange chromatography (Sigma). The specific activity determined by self-displacement analysis was 3-6 ϫ 10 6 cpm/pmol (41). For binding studies cells were washed in RPMI 1640 medium supplemented with 10% FBS, resuspended at a concentration of 6 ϫ 10 7 cells/ml, and incubated at room temperature for 30 min to remove residual EPO. Aliquots of 3 ϫ 10 6 cells were incubated with a range of concentrations (25 pM-5 nM) of iodinated EPO at 4°C overnight in binding buffer (RPMI 1640 medium, 10% FCS, and 50 mM HEPES, pH 7.2) in the absence or presence of 100-fold excess unlabeled EPO. Following overnight incubation cells were determined to be greater than 90% viable by trypan blue exclusion analysis. Free EPO was separated from bound EPO by centrifugation through a 100% FBS cushion. The amount of bound and free radiolabeled EPO was determined by counting in a ␥ counter. All points were done in triplicate. The data were graphed according to the method of Scatchard, and K d and cell surface receptor numbers were determined.
Cross-linking Analysis-2-5 ϫ 10 7 cells were washed in RPMI 1640 medium supplemented with 10% FBS, resuspended at a concentration of 2-5 ϫ 10 7 cells/ml, and incubated at room temperature for 30 min to remove residual bound EPO. Binding was then done in 5 nM 125 I-EPO for 2-3 h in the presence of 0.2% azide at 22-25°C. For competition studies, the cells were preincubated in a 100-fold excess of unlabeled EPO. Cells were washed three times in cold PBS and resuspended at the same concentration, 0.5 mM DSS was added, and the cells were incubated for 30 min on ice. To quench the cross-linking reactions the cells were washed three times in cold PBS with 0.1 M ethanolamine, pH 8.0, and then solubilized in XL lysis buffer (1.5% Triton X-100, 25 mM HEPES, pH 7.75, 150 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 0.02% sodium azide). Specific antisera were added to clarified soluble extracts followed by protein A-agarose (Sigma). Pelleted beads were washed four times in lysis buffer and then once in 10 mM Tris, pH 6.8, and boiled for 5 min in Laemmli sample buffer containing 2-mercaptoethanol, and products were resolved on a 6% acrylamide gel containing SDS, fixed, dried, and exposed to x-ray film. In some experiments washed immunoprecipitates were denatured, and complex asparagine-linked carbohydrate side chains were removed by endoglycosidase F (New England Biolabs) digestion before resolution on SDS-PAGE.
Biotinylated EPO Binding-EPO was biotinylated by the method of Wognum et al. (42). Briefly, EPO was oxidized using sodium periodate and then purified by PD-10 gel filtration. Oxidized EPO was biotinylated using biotin-x-hydrazide and again purified by PD-10 gel filtration. Aliquots were stored at Ϫ80°C. Cells were washed in RPMI 1640 medium supplemented with 0.5% bovine serum albumin, resuspended at a concentration of 2-5 ϫ 10 7 cells/ml, and incubated at room temperature for 30 min to remove residual EPO. Binding was then done in 5 nM biotin-EPO for 2-3 h in the presence of 0.02% azide at 22-25°C. Cell lysates were prepared in buffer containing 1% Triton X-100, 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 2 mM phenylmethylsulfonyl fluoride, and 10 g/ml aprotinin. From clarified soluble extracts biotinylated EPO-bound proteins were isolated by the addition of streptavidin-agarose. Pelleted beads were washed three times in lysis buffer and once in PBS and boiled for 5 min in Laemmli sample buffer containing 2-mercaptoethanol, followed by SDS-PAGE and transfer to nitrocellulose membranes.
Cell Surface Iodination-Iodination of cell surface proteins was performed using the lactoperoxidase method with some modification (43). Briefly, 5 ϫ 10 7 cells were washed twice in PBS and resuspended in 1 ml of PBS. Lactoperoxidase (20 g) and 1 mCi of [ 125 I]NaI were added. Iodination was then catalyzed by the sequential addition of three aliquots of hydrogen peroxide at 5-min intervals. The reaction was quenched by washing three times in 2 mM potassium iodide. Cell lysates were prepared in buffer containing 0.5% Triton X-100 in PBS. The clarified soluble extracts were immunoprecipitated with specific antisera and protein A-agarose.

RESULTS
The Cell Surface EPO-R Complex-The cell surface EPO-R complex was isolated from erythroid cells expressing endogenous EPO-R (HCD57) and BaF3 cells transfected with EPO-R (BaF3.EPO-R). HCD57 cells are EPO-dependent erythroleukemia cells expressing wild type EPO-R (14), whereas BaF3 cells are IL-3-dependent pro-B cells that following transfection with the EPO-R will proliferate in response to EPO. Saturating amounts of iodinated EPO were added to intact cells and allowed to bind in the presence of azide to block endocytosis of bound EPO. Cells were then chemically cross-linked with DSS and lysed in nonionic detergents, and soluble fractions were immunoprecipitated with antisera against the carboxyl-terminal, cytoplasmic domain of the EPO-R. Antisera directed against the cytoplasmic tail of the EPO-R was used to avoid loss of immunoprecipitable receptor due to possible masking of epitopes following cross-linking of EPO to the extracellular domain of the receptor. The cell surface EPO cross-linked proteins present on erythroid HCD57 cells are shown in Fig. 1A, lane 1. They are the 100-kDa (p70) EPO-R (determined by subtracting one 30-kDa molecule of EPO) and receptor-associated 120-kDa (p90) and 140-kDa (p110) proteins. On BaF3.EPO-R cells (Fig. 1A, lane 2) the 100-kDa p70 EPO-R was detected, as well as the 140-kDa p110 protein. The 120-kDa p90 associated protein was not readily apparent; however, a prominent 110-kDa (p80) EPO cross-linked band was detected, as well as less intense bands of 85 kDa (p55) and 65 kDa (p35).

Activation of the EPO-R by F-gp55 versus EPO Results in a Different Pattern of Cell Surface EPO-R-associated Proteins-
To determine whether activation of the EPO-R by EPO or F-gp55 resulted in different cell surface EPO-R complexes, BaF3 cells expressing EPO-R and EPO-R and F-gp55 were generated. EPO equilibrium binding studies and Scatchard analysis were performed (Table I). The K d for EPO and the number of cell surface EPO binding receptors were not significantly different between cells. EPO protein cross-linking studies were performed, and the EPO-R complex was isolated by immunoprecipitation with antisera against the EPO-R (Figs. 1B and 2A) or antisera against retroviral env gp70, which cross-reacts with F-gp55 (Figs. 1C and 2A).
On BaF3 co-expressing EPO-R and F-gp55 the 100-kDa band detected following immunoprecipitation with antisera against EPO-R and F-gp55, as no EPO cross-linked proteins were detected in anti-env gp70 immunoprecipitations of cells expressing F-gp55 alone (Fig. 1C, lane 2) or cells expressing EPO-R alone (Fig. 1C, lane 3). This suggests an intimate association between EPO, EPO-R, and F-gp55 at the cell surface, and that the 190-kDa complex contains at least EPO⅐EPO-R⅐F-gp55 P .
BaF3 cells are immortalized IL-3-dependent pro-B cells, and although EPO-R expression in these cells allows them to proliferate and partially differentiate in the presence of EPO (28), the EPO-R is not expressed on developing B cells, and B-cell progenitors do not respond to EPO in vivo or in ex vivo cultures. In addition, the pattern of cell surface EPO-R-associated proteins detected by EPO cross-linking appeared to differ between BaF3.EPO-R and erythroid HCD57 cells. Thus, we wanted to determine and contrast the cell surface EPO-R complex on erythroid progenitors when activated by EPO or F-gp55.  (Table I), and that antibodies against gp70 did not co-immunoprecipitate cell surface EPO⅐EPO-R (see Fig. 4). Thus, endogenously expressed EPO-R also forms a complex with F-gp55 on the cell surface, and the reduction in detectable p90 and p110 EPO-R-associated proteins following interaction of EPO-R with F-gp55 ( Fig  also more intense than that detected on BaF3.EPO-R cells (Fig.  1B, lane 5); however, it was not as diffuse as that detected on BaF3.EPO-R:F-gp55 cells, and the intensity of the EPO-Rassociated 140-kDa proteins was unchanged. In contrast to BaF3.EPO-R:F-gp55 cells, on BaF3.EPO-R (R129C):F-gp55 cells there was no 190-kDa EPO cross-linked complex detected in EPO-R immunoprecipitations. Surprisingly, neither the 100-kDa band nor the 190-kDa band was detected in anti-env gp70 immunoprecipitations (Fig. 1C, lane 6). The trace amount of 100-kDa complex detected on these cells (Fig. 1C, lane 6) was most likely due to induction of erythroid differentiation and low level endogenous wild type EPO-R expression in BaF3 cells expressing EPO-R (R129C) (28, 44) 2 (also see Fig. 4, lanes 3 and  4). These results suggest that in contrast to wild type EPO-R, EPO-R (R129C) does not appear to form a cell surface complex with F-gp55, as determined by this method. Similar to results obtained with BaF3.EPO-R (R129C), erythroid HCD57 cells expressing EPO-R (R129C) had much more detectable 100-kDa EPO-R, with little change in the detectable intensity of EPO-R-associated proteins when contrasted with HCD57 cells (Fig. 2B). Also, similar to results obtained with BaF3.EPO-R (R129C):F-gp55 cells, on the erythroleukemic cell line A1.8 (A1.8 expresses EPO-R (R129C) but not wild type EPO-R; Ref. 36) transfected with F-gp55, the 190-kDa EPO⅐EPO-R⅐F-gp55 complex was not detected, and the pattern of cell surface EPO-R-associated proteins was unchanged (see Fig. 4, lanes 1 and 3). On these cells wild type EPO-R is not induced following expression of EPO-R (R129C); thus, there was no evidence of trace detectable 100-kDa EPO⅐F-gp55 complex (see Fig. 4, lane 3), as was present on BaF3.EPO-R (R129C):F-gp55 cells (Fig. 1C, lane 6).
Another approach to determine whether EPO-R (R129C) and F-gp55 P do or do not associate at the cell surface made use of biotinylated EPO to specifically isolate cell surface EPO-R complexes in the absence of chemical cross-linking. The presence of 2 G. Longmore, unpublished observation.

FIG. 3. Biotinylated EPO detection of cell surface EPO-R complex. Biotinylated EPO (5 nM) was added to intact cells (2 ϫ 10 7 cells)
without chemical cross-linking. Cells were lysed in nonionic detergents; soluble biotinylated EPO-bound complexes were isolated with streptavidin-agarose, separated on 10% SDS-PAGE in the presence of 2-mercaptoethanol, and transferred to nitrocellulose membranes; and immunoblotting was performed with monoclonal antisera against gp55.  gp55). Immunoprecipitated products were isolated with protein A-agarose, washed, and denatured in SDS. To one-half of each sample endoglycosidase F digestion was performed overnight. The other half underwent mock digestion overnight. The next morning the products were separated on 6% SDS-PAGE in the presence of 2-mercaptoethanol and dried, and autoradiography was performed. Lanes 1-4, A1.8.F-gp55 erythroid cells; lanes 5-8, BaF3.EPO-R cells; lanes 9 -12, BaF3.EPO-R:F-gp55 cells. Arrowheads on the left, mobility of EPO⅐EPO-R complex before (upper arrowhead) and following (lower arrowhead) digestion with endo-F. Arrows on the right, mobility of EPO⅐F-gp55 complex before (upper arrow) and following (lower arrow) digestion with endo-F. Molecular mass standards in kDa are on the right. F-gp55 in these complexes was then determined by immunoblotting with monoclonal antisera against F-gp55 (Fig. 3). F-gp55 was present in streptavidin-biotinylated EPO products isolated from cells co-expressing EPO-R and F-gp55 (Fig. 3, lane 4) but not EPO-R alone (Fig. 3, lane 3) or F-gp55 alone (Fig. 3, lane 2). Also, on BaF3.EPO-R (R129C):F-gp55 cells (Fig.  3, lane 6) and erythroid A1.8:F-gp55 cells (Fig. 3, lane 8) no association between EPO-R (R129C) and F-gp55 was detected.

The 100-kDa EPO Cross-linked EPO-R Detected on EPO-R: F-gp55 Cells Is Composed of EPO⅐EPO-R and EPO⅐F-gp55: The Increased Intensity of the 100-kDa Band Detected on EPO-R (R129C) Cells Is Due to the Presence of Preformed Cell Surface
Dimers-Co-expression of F-gp55 and EPO-R results in a diffuse 100-kDa EPO cross-linked band of increased intensity (Fig. 1B, lane 4), equivalent in intensity to the band detected when preformed cell surface EPO-R (R129C) dimers are present (Fig. 1B, lane 5). This could result from induced dimerization of the EPO-R by F-gp55 or the presence of EPO⅐EPO-R and EPO⅐F-gp55 P co-migrating complexes co-immunoprecipitated with antisera against the EPO-R (37). To distinguish between these two possibilities we made use of the distinctive electrophoretic mobility shifts of the EPO⅐EPO-R and EPO⅐F-gp55 P complexes when asparagine-linked carbohydrate side chains are removed. EPO was bound to cells and cross-linked, and soluble extracts were immunoprecipitated with antisera against the EPO-R or env gp70. Washed immunoprecipitates were denatured and either digested with endoglycosidase F, which cleaves asparagine-linked carbohydrate side chains, or mock treated before separation of bound products on SDS-PAGE under reducing conditions (Fig. 4). Arrowheads on the left identify the mobility of untreated EPO⅐EPO-R (Fig. 4, upper arrowhead, lane 1) and endo F-treated EPO⅐EPO-R (Fig. 4,  lower arrowhead, lane 2). Likewise, arrows on the right identify the shift in mobility of EPO⅐F-gp55 following treatment with endo F. On cells co-expressing EPO-R and F-gp55 the 100-kDa band detected in EPO-R immunoprecipitates was more intense and broader (Fig. 4, lane 9 versus lane 5). Endo F digestion revealed that the 100-kDa band in Fig. 4, lane 9, is composed of two proteins, which migrated with the mobility of EPO⅐EPO-R and EPO⅐F-gp55 cross-linked complexes. In addition, following treatment with endo F the amount of detectable EPO⅐EPO-R detected on cells co-expressing F-gp55 and EPO-R did not differ from cells expressing EPO-R alone (Fig. 4, lane 10 versus lane  6). In contrast to anti-EPO-R immunoprecipitations, anti-gp70 immunoprecipitates of EPO cross-linked BaF3.EPO-R:F-gp55 cells did not co-immunoprecipitate cell surface EPO⅐EPO-R (Fig. 4, lanes 11 and 12). The same antisera does co-immunoprecipitate EPO-R and F-gp55 when total cell membranes (including the endoplasmic reticulum) are analyzed (37). On cells co-expressing EPO-R (R129C) and F-gp55 no detectable EPO⅐F-gp55 complex was present in EPO-R and gp70 immunoprecipitates (Fig. 4, lanes 1 and 2 and 3 and 4, respectively).
Taken together, these results demonstrate that F-gp55 and the EPO-R interact at the cell surface as expected (16); however, compared with wild type EPO-R activation by EPO, the amount of EPO-R-associated proteins was dramatically decreased when the EPO-R was complexed with F-gp55, whereas R129C mutations in the receptor did not result in changes in pattern or amounts of EPO-R-associated EPO cross-linked proteins. The inability to detect the 190-kDa EPO⅐EPO-R⅐F-gp55 and 100-kDa EPO⅐F-gp55 complexes on cells co-expressing EPO-R (R129C) and F-gp55 indicate that preformed EPO-R dimers do not interact with F-gp55. Thus, the structure of the cell surface EPO-R complex differs when activated by EPO, F-gp55, or R129C mutations.
To determine whether the loss of Arg 129 , Glu 132 , or Glu 133 or the acquisition of a cysteine residue at these sites and subsequent receptor dimer formation was responsible for the inability to interact with F-gp55, we generated BaF3 cells expressing EPO-R (129 -133A). This receptor isoform has alanine residues at positions 129 -133. Cells co-expressing EPO-R (129 -133A) and F-gp55 were also generated. Like EPO-R cells, EPO-R (129 -133A) cells were EPO-responsive, and co-expression of EPO-R (129 -133A) and F-gp55 resulted in EPO-independent growth. In contrast to R129C and E133C cells, yet similar to cells expressing wild type EPO-Rs, on EPO-R (129 -133A):F-gp55 cells surface 190-kDa EPO⅐F-gp55⅐EPO-R and 100-kDa EPO⅐F-gp55 complexes were present (Fig. 6B, lanes 3 and 7). The amount present was roughly equivalent to that detected on EPO-R:F-gp55 cells (Fig. 6B, lanes 3 versus 1 and 7 versus 5). These results support the conclusions that: 1) preformed EPO-R dimers are precluded from interacting with F-gp55; and 2) the acquisition of cysteine (and dimer formation), not the loss of Arg 129 , Glu 132 , or Glu 133 , is responsible for this result.
Rauscher gp55 Is Expressed at the Cell Surface but Does Not Form a Detectable Complex with the EPO-R-Variant forms of gp55 have been shown to associate with the EPO-R in intracellular membranes yet do not activate EPO-independent proliferation (26). It is not clear whether these variant env genes are expressed at the cell surface and, if so, whether they interact with the cell surface EPO-R complex. Rauscher gp55 is a viral variant of the Friend gp55 env gene product. Rauscher virus also induces erythroleukemia, but the early phase of disease differs from Friend virus, and EPO-independent proliferation of infected erythroid progenitors has not been observed. The env gene product M1-gp55 is a variant of F-gp55 that has acquired four amino acid changes in the mink cell focus-forming virus portion of F-gp55 and does not induce EPO-independent proliferation in cells containing EPO-R. Both variant env cDNAs were transfected into EPO-dependent BaF3.EPO-R cells. Neither resulted in EPO-independent proliferation, as expected. EPO-R was present in anti-env gp70 immunoprecipitations from whole cell detergent-soluble extracts of both cells (Fig. 6A, lanes 3 and 4), indicating that intracellular R-gp55⅐EPO-R and M1-gp55⅐EPO-R complex formation could be detected, as expected (26). For F-gp55 to activate the EPO-R cell surface, complex formation must occur (12,13,17). To determine whether R-gp55 P and M1-gp55 interact with the EPO-R at the cell surface, EPO protein cross-linking studies were performed (Fig. 6B). Neither EPO-R:M1-gp55 cells (Fig.  6B, lanes 2 and 6) nor EPO-R:R-gp55 cells (Fig. 6B, lanes 4 and  8) cells had detectable cell surface 190-kDa EPO⅐EPO-R⅐gp55 or 100-kDa EPO⅐gp55 complexes in immunoprecipitations of the EPO-R (Fig. 6B, lanes 2 and 4) or env gp70 (Fig. 6B, lanes  6 and 8). Thus, either these two forms of gp55 do not interact with cell surface EPO-R, or during biosynthesis they are not processed and transported to the cell surface.
To distinguish between these two possibilities, cell surface proteins were labeled with radioactive iodine using lactoperoxidase. Since BaF3 cells contain endogenous retroviruses, env gp70 was iodinated and immunoprecipitated from all cell lines (Fig. 7, lanes 5-8). On cells co-expressing EPO-R and F-gp55 (Fig. 7, lanes 2 and 6) but not on cells expressing EPO-R alone (Fig. 7, lanes 1 and 5), a 65-kDa F-gp55 P cell surface form is readily detected using both antisera. Cells co-expressing EPO-R and R-gp55 also express cell surface R-gp55 P (Fig. 7,  lanes 4 and 8). In contrast, no detectable M1-gp55 was present on the cell surface (Fig. 7, lanes 3 and 7). Chemical biotinylation of cell surface proteins also did not label any immunologically distinguishable M1-gp55 (not shown). Thus R-gp55, an env gene biologically equivalent to the anemic strain of Friend virus, interacts with the EPO-R in intracellular membranes but not at the cell surface, despite the presence of R-gp55 at the cell surface. DISCUSSION During erythroid development EPO binds specifically to a receptor primarily present on committed erythroid progenitors in the bone marrow and spleen, stimulating mitogenic, survival, and differentiative growth pathways (for review, see Ref. 45). Other modes of EPO-R activation give rise to pathological consequences. Interactions between the env gene F-gp55 of the Friend erythroleukemia virus and the EPO-R or specific mutations in the extracellular domain of the EPO-R can induce EPO-independent erythroblastosis in mice, which ultimately progresses to erythroleukemia (15,36). In addition, EPO and F-gp55 stimulation of the EPO-R elicit different biological responses (18). Mice infected with a virus-expressing EPO-R (R129C) develop acute erythroblastosis in a manner similar to mice infected with an EPO virus (35). We contrasted the cell surface EPO-R complex formed following receptor activation by EPO and F-gp55 and mutations in the extracellular domain of the receptor.
EPO protein cross-linking experiments performed herein detect the same complement of EPO-R-associated proteins on cells expressing EPO-R or EPO-R (R129C). The intensity of the 100-kDa EPO⅐EPO-R complex, however, is much greater on EPO-R (R129C) cells. Most likely this is due to the presence of preformed covalent EPO-R dimers on the cell surface. The amounts of 90and 110-kDa proteins detected did not differ. Interestingly EPO-R (R129C) did not form detectable associations with F-gp55 at the cell surface. Most likely this is due to preformed dimers, as receptor isoforms with mutations changing Arg 129 to alanine behave like wild type receptors, and other constitutively dimeric EPO-Rs (e.g. E132C and E133C) also did not interact with F-gp55. Previous studies had suggested that on the cell surface not all EPO-R (R129C) molecules were dimeric; in fact a minority appeared to be dimers (31). If the presence of preformed dimers was the sole factor precluding interaction with F-gp55, then we would have expected to detect some EPO-R⅐F-gp55 complexes. That we did not could mean that either all cell surface EPO-R (R129C) isoforms are dimeric or that monomeric EPO-R (R129C) also does not interact with F-gp55. Although we cannot distin-guish between these two possibilities, that hematopoietic progenitor cells expressing EPO-R (R129C) remain responsive to EPO (46,47) would suggest that some monomeric EPO-Rs (R129C) are present on the surface of cells and that EPO-R (R129C) per se cannot interact with F-gp55. Alternatively, in the presence of EPO the cell surface structure of EPO-R (R129C) dimers is altered, leading to an augmented signal. Analyses of the recently published crystal structure of the EPO-R and a peptide agonist suggest that the Arg 129 residue is not at the dimer interface. This observation plus our results indicate that either the structure of EPO⅐EPO-R is different from the peptide-EPO-R complex or, more likely, that EPO-R (R129C), although functional, has a structure distinct from wild type EPO-R when activated by EPO.
Whether the cell surface EPO-R (R129C) complex signals in a manner distinct from EPO⅐EPO-R has not been well studied. 32D cells expressing EPO-R (R129C) are not dependent on tyrosine phosphorylation of Shc and activation of mitogenactivated protein kinase for proliferation in cultures lacking added growth factors; however, tyrosine phosphorylation of Shc and mitogen-activated protein kinase activation were responsive to EPO in these cells (48), suggesting that the Shc-grb2mitogen-activated protein kinase pathway may not be required for EPO-induced mitogenesis.
Activation of the EPO-R by F-gp55 resulted in a dramatic alteration in the pattern of cell surface EPO-R-associated proteins cross-linked to EPO. The amounts of 90-and 110-kDa EPO-R-associated proteins were significantly decreased. This change was most pronounced when the analysis was performed on erythroid HCD57 cells, as opposed to the pro-B cell BaF3.EPO-R line. The implication of this alteration is unclear, since the function of the cell surface EPO-R-associated proteins is unknown. The 90-and 110-kDa EPO-R-associated proteins are cross-linked to EPO only in the presence of the EPO-R. Thus, at the concentration of EPO used for these experiments, neither appears to bind EPO directly. EPO protein cross-linking studies on EPO-R-expressing COS and L cells did not detect the 110-kDa protein. Neither of these cells proliferate in response to EPO. Hematopoietic cells that respond to EPO all have detectable 110-kDa protein. It has been suggested that the 110-kDa protein may be a heretofore undescribed second subunit of the EPO-R.
On BaF3.EPO-R cells used in this study the amount of p90 EPO cross-linked EPO-R-associated protein we detected was variable and less than others have observed (49). In addition, we routinely detected an 80-kDa EPO cross-linked EPO-R-associated band on these cells. The intensity of the p80 protein detected was not affected by interactions between F-gp55 and the EPO-R. This suggests that p80 may represent a processed form of the cloned EPO-R present in the BaF3 clone we used in our studies or possibly a proteolytic product of p90. Alternatively in BaF3 cells, as opposed to erythroid cells and other hematopoietic cell lines, it may represent a distinct surface protein present in the functional EPO-R complex. The function, if any, of this protein in EPO-R signaling in our BaF3 cells is unclear.
Whether EPO-R⅐F-gp55 complexes signal in a manner distinct from EPO⅐EPO-R or EPO-R (R129C) is not clear. EPOinduced tyrosine phosphorylation of the EPO-R, and activation of the Janus kinase and STAT signal transduction pathways appear to be essential for EPO-induced mitogenesis and possible erythroid differentiation (for review, see Ref. 50). In erythroid cells EPO activates Janus kinase 2 and STAT5. The pattern of tyrosine-phosphorylated substrates differs for EPOstimulated EPO-R versus gp55-stimulated EPO-R in BaF3 cells (51). In HCD57 cells grown in EPO or infected with SFFV and growing EPO independently, similar patterns of activated STAT proteins were observed, except that in HCD57.SFFV cells these factors were constitutively activated (52).
During murine erythroid development the production of early burst-forming unit erythroid progenitors is dependent on stem cell factor (SCF). SCF has been shown to retard differentiation and to enhance proliferation, or survival, of early progenitors, whereas EPO was required primarily to complete erythroid differentiation (53). Janus kinase 2 has been reported to associate with c-kit (the receptor for SCF) and to become phosphorylated in response to stem cell factor (54). Also, SCF induces serine phosphorylation of STAT3, and when cells are co-stimulated with SCF and other cytokines, DNA binding activity of STAT3 is induced (55). Finally, in HCD57 cells the SCF receptor c-kit and the EPO-R may interact, allowing for "cross-talk" between the the major growth factors regulating erythropoiesis (56). Mice infected with F-SFFV exhibit a more pronounced erythroid progenitor proliferative, or survival, response as opposed to a more differentiative response when the EPO or EPO-R (R129C) genes are virally transduced. Whether activation of the EPO-R by F-gp55 could affect SCF-mediated signaling pathways is a possibility that needs to be tested.
The anemic strain of Friend virus also induces erythroblastosis in infected mice. In contrast to mice infected with the polycythemic strain of SFFV, erythroid progenitor cell cultures and the culture of erythroleukemic cell lines established from mice infected with the anemic strain of SFFV require EPO for growth (22). In BaF3.EPO-R cells transduced with R-gp55 (analogous to the env gene of F-SFFV-A) an EPO-R⅐R-gp55 complex was detected within intracellular membranes (26). This interaction did not convert these cells to EPO-independent proliferation, however. We have gone on to demonstrate that cell surface EPO-R⅐R-gp55 complex formation was not detected on these cells. Since cell surface interactions between F-gp55 and the EPO-R are required for mitogenic signals, the absence of detectable interaction between R-gp55 and the EPO-R at the cell surface could explain the absence of EPO-independent proliferation in erythroid progenitors from anemic SFFV-infected mice. The lack of detectable cell surface EPO-R⅐R-gp55 interaction was not due to the absence of cell surface expression of R-gp55. However, it is possible that the level of surface R-gp55 expression was not enough to form detectable surface EPO-R⅐R-gp55 complex. Alternatively the structure of the fully processed cell surface R-gp55 protein may not be capable of forming a functional EPO-R complex. The presence of detectable intracellular membrane EPO-R⅐R-gp55 complexes may then simply reflect overexpression of both EPO-R and R-gp55 in BaF3 cells, predominantly in the endoplasmic reticulum (37). The level of cellular EPO-R present in erythroid progenitors in vivo is much less than that present in BaF3.EPO-R cells. If R-gp55 does not functionally interact with the EPO-R, how then does Rauscher virus infection lead to erythroblastosis? There is controversial evidence as to whether the env genes of SFFV or mink cell focus-forming virus can interact with other cytokine receptors (57). Possibly R-gp55 activates erythroblastosis through interactions with other cytokine receptors relevant to erythroid development (e.g. c-mpl or thrombopoietin receptor, c-kit or SCF receptor, and IL-11).