The Erythropoietin Receptor of Rat Erythroid Progenitor Cells CHARACTERIZATION AND AFFINITY CROSS-LINKAGE*

Commercially available ‘2sI-labeled erythropoietin, obtained by genetic engineering from a human gene, was used to characterize receptors for this hormone on the cell surface of rat erythroid progenitor cells. A low number of high affinity binding sites (487 f 32 sites/ cell, Kd = 167 f 14 PM) were found. Nonerythroid cells and erthrocytes did not exhibit specific binding. The high affinity binding was reversible and displaced by unlabeled erythropoietin, but not by other hormones and growth factors. After incubation at 37 “C, nearly 35% of the specifically bound erythropoietin seemed to be internalized, as judged by resistance to acidic buffer treatment. Thus, binding showed characteristics of a hormone-receptor association. lZ6I-Erythropoietin-la-beled cells were treated with the bifunctional reagent dissucinimidyl suberate. Analysis of the cellular extracts by polyacrylamide gel electrophoresis under de- naturing and reducing conditions revealed that erythropoietin can be cross-linked to two molecules of 94 and 78 kDa, respectively. Both labeled bands disap-peared when the cells were labeled in the presence of an excess of unlabeled erythropoietin. Under nonreducing conditions, a cross-linked band of 230-255 kDa was observed. The relationships between these bands are discussed. specific binding capacity of a hormone can be determined directly by using increasing concentrations of its receptor (18). Due to the small number of binding sites per cell in our material (see "Results"), this direct determination was not possible; and we adopted, with a slight modification, the procedure described by Nicola and Metcalf (19). As these experiments necessitated a large number of cells, we used murine erythroleukemic cells. We have previously shown that these cells bind '2'I-erythropoietin (20); and as indicated in this study (see below), this binding seems to be specific. to rat erythroid progenitors.

tion in the absence of erythropoietin.
Recently, methods have been described for the preparation of erythroid progenitors from hemopoietic populations, either normal (11)(12)(13) or infected with the anemic strain of the Friend virus (14). Using the latter material, Krantz and Goldwasser (15) demonstrated the existence of specific binding of [3H]-erythropoietin and found 660 binding sites/cell with a high binding affinity ( K d = 5.2 nM). We describe the preparation from rat fetal livers of cell suspensions essentially made of erythroid progenitors which depend on erythropoietin for proliferation and differentiation (13, 16). These cells were used in this work to test the binding of radioiodinated erythropoietin; we demonstrate the existence of specific and reversible high affinity binding and performed cross-linkage experiments of the hormone with its presumed receptor.

MATERIALS AND METHODS
Chemicals-Pure erythropoietin, produced from an isolated human erythropoietin gene by Kirin-Amgen, Inc. and purchased from Amersham Corp., had a specific activity of 70,000 units/mg. Its labeled derivative, (3-['261]iodotyrosyl)erythr~p~ietin, was also purchased from Amersham Corp.; six batches were used with specific radioactivities at delivery ranging from 400 to 1,200 Ci/mmol. In some experiments, pig nonradioactive erythropoietin (1,010 units/mg of protein; Centre National de Transfusion Sanguine, Paris) was also used and gave results completely superimposable to, and therefore combined with, those obtained with pure human hormone. Bovine insulin was from Behring Diagnostics; pig platelet-derived growth factor was from Bioprocessing LTD.; and mouse epidermal growth factor and phorbol 12-myristate 13-acetate were from Sigma. Bovine platelet-derived growth factor purified as described in Ref. 17, was a generous gift from Dr. J. M. Felix (University of Strasbourg). Rat interleukin 3 obtained by genetic engineering was a generous gift from Dr. A. J.
Hapel (University of Camberra, Camberra, Australia) and was in solution in a culture medium at 200 units/ml (1 unit/ml is the concentration producing half-maximal stimulation in a bone marrow cell proliferation assay). DSS,' Percoll, and RMPI 1640 medium were from Pierce Chemical Co., Pharmacia P-L Biochemicals, and GIBCO, respectively. Inhibitors of proteases were from Sigma.
Preparation of Cells-The preparation of erythroid progenitors from rat fetal livers has been previously described (13,16). Briefly, suspensions of erythroid cells, obtained by mild mechanical dissociation of livers followed by filtration on a nylon sieve, were treated with rabbit antiserum against rat erythrocytes in the presence of 10% (v/ v) guinea pig serum as a source of complement (this procedure produces the lysis of the erythrocytes and of the large majority of the erythroblasts); the remaining intact cells were washed and layered on the top of a discontinuous gradient of Percoll. The cells with a density >LO55 were recovered and washed three times before use. These cells represent roughly 3.8% of the initial cell suspension; over 85% are morphologically undifferentiated, and 65% produce erythroid colonies (mainly erythroid colony-forming units) after culture in semisolid media. Their proliferation and differentiation are strictly erythropoietin-dependent (16).
Rat thymocytes and splenocytes were also prepared by mechanical The abbreviations used are: DSS, disuccinimidyl suberate; SDS, sodium dodecyl sulfate.

13985
dissociation followed by filtration on a nylon sieve (50-pm mesh). Splenocytes were isolated by centrifugation on Percoll. Circulating rat erythrocytes and leukocytes were also prepared by centrifugation on a density gradient. Murine erythroleukemic (Friend) cell line 745 was maintained by reseeding every 3-4 days in RPMI 1640 medium containing 10% (v/ v) heat-inactivated horse serum. The cells were sampled for use at the end of the phase of exponential growth.
Binding of Radioactive Erythropoietin-Rat erythroid progenitor cells (1-10 X lo6) were incubated in 100 pl RPMI 1640 medium containing 10% heat-inactivated bovine fetal serum and increasing concentrations (at least X10, and generally x50) of radioactive erythropoietin in the presence or absence of an excess of the non-labeled hormone (durations and temperatures of incubations are given under "Results"). Incubations were ended by the addition Of 4 ml of ice-cold RPMI 1640 medium; the cells were recovered after 5 min of centrifugation at 180 X g and washed three more times, and we verified that the last supernatant fraction was not radioactive. Radioactivity bound was determined by y-counting. As usual, high affinity (specific) bound radioactivity was calculated as the difference between total (absence of nonlabeled erythropoietin) and nonspecific (presence of nonlabeled erythropoietin) bound radioactivities. Determination of Cell Surface-bound lz61-Erythropoietin-Rat erythroid progenitor cells, labeled and washed as described above, were resuspended in 1 ml ice-cold acidic buffer (50 mM glycine/acetic acid, pH 3.0) containing 150 mM NaC1. After 5 min at 0 "C, they were centrifuged, and the radioactivities of the pellet and of the supernatant fraction were measured.
Determination of the Total Specific Binding Capacity of "'I-Erythropoietin-The maximal specific binding capacity of a hormone can be determined directly by using increasing concentrations of its receptor (18). Due to the small number of binding sites per cell in our material (see "Results"), this direct determination was not possible; and we adopted, with a slight modification, the procedure described by Nicola and Metcalf (19). As these experiments necessitated a large number of cells, we used murine erythroleukemic cells. We have previously shown that these cells bind '2'I-erythropoietin (20); and as indicated in this study (see below), this binding seems to be specific. We shall therefore consider that the conclusions drawn from this study using Friend cells can be applied to rat erythroid progenitors.
Tubes containing lo7 cells and 10,000 cpm of '"1-erythropoietin in 100-pl culture medium were incubated for 2 h at 37 "C in the absence (test medium) or presence (evaluation of nonspecific binding) of an excess of nonlabeled hormone; in parallel, 100-pl samples of culture medium containing 10,000 cpm of '251-erythropoietin were incubated without cells (control medium). At the end of this incubation, the radioactivity specifically bound to the cells was determined as above described; test and control media were recovered, and the radioactivity/ml of the recovered control medium was adjusted to that of the recovered test medium. A second binding experiment was then performed using, per lo7 cells, 100 pl of the recovered test medium or 100 p1 of the adjusted control medium, with or without an excess of nonlabeled erythropoietin; in parallel, 100-p1 samples of the adjusted control medium were incubated without cells. At the end of this second incubation, the radioactivities specifically bound to the cells were determined, and the radioactivity/ml of the adjusted control medium incubated without cells was adjusted to that of the second recovered test medium. Binding was tested again, using the second recovered test and the second adjusted control media. The use of adjusted control media allowed the elimination of eventual modifications of the binding affinity due to the incubation procedure.
Comparison of the Binding Affinities of Radioactive and Native Erythropoietin-This comparison was done by self-displacement analysis (18,19). Rat erythroid progenitors were incubated in 100 p1 of medium containing 10,000 cpm of '251-erythropoietin and increasing amounts of either labeled or nonlabeled hormone, in the presence or absence of a large excess of nonlabeled hormone. For each point, the radioactivity specifically bound (B) was determined, and the free radioactivity remaining in the medium (F) was measured. The B/F ratio was plotted as a function of the logarithm of the amount of erythropoietin (labeled or not) added (19).
Affinity Cross-linkage--Rat erythroid progenitors were incubated for 3 h at 20 "C in the presence of 400-800 PM '251-erythropoietin. They were then washed and resuspended in ice-cold phosphatebuffered saline at a density of 5-10 X lo6 cells/ml. Cross-linkage was performed by adding 20 pl of DSS in dimethyl sulfoxide, with the final DSS concentrations varying from 40 p M to 1 mM. The reaction mixture was kept at 0 "C for 1 h, with intermittent shaking. The cells were then recovered, washed four times in phosphate-buffered saline, and resuspended at a density of 5 X lo6 in 50 pl of 10 mM Tris-HC1 buffer, pH 6.8, containing 1% Triton X-100 in the presence or absence of the following mixture of protease inhibitors: 2 mM phenylmethylsulfonyl fluoride + 10 pM leupeptin + 2 mM 0-phenanthroline + 2 mM EDTA + lo3 units/ml aprotinin. After 20 min of incubation at 20 "C, the samples were centrifuged (20 min, 20,000 X g). The supernatant fractions were incubated for 5 min at 95 'C in the presence of 2% SDS and 5% B-mercaptoethanol and then submitted to SDSpolyacrylamide gel electrophoresis (21). After migration (16 h, 2.5 V/ cm), the proteins were stained with Coomassie Blue, and the gels were dried for autoradiography (1-3-wk exposure at -80 'C; Kodak X-Omat AR5 film, two Du Pont New England Nuclear Cronex Hi Plus intensifying screens).

RESULTS
Kinetics of High Affinity Binding of '251-Erythropoietin 251-Erythropoietin was bound with high affinity by the erythroid progenitors from rat fetal liver, and binding was proportional to the cell concentration (Fig. 1). As illustrated in Fig. 2 A , in the presence of 140 p~ labeled hormone, equilibrium was reached at 37 "C after 1-h of incubation; and the radioactivity specifically bound remained constant for at least 4 h. When an excess (5 nM) of nonlabeled hormone was added after 1 h of incubation, the radioactivity specifically bound decreased, and 90% of it was displaced after 4 h of chase. By plotting the logarithm of the bound radioactivity as a function of chase duration (Fig. 2B), two phases were evidenced an initial rapid exchange during which roughly 70% of the specifically bound radioactivity was released (tu& = 30 min) and a slower and prolonged dissociation (tlh = 3-4 h). Dissociation kinetics were also studied by transferring cells, prelabeled for 1 h, in a ligand-free medium at 37 "C; the results were essentially the same (not shown). The cells were also able to bind the hormone at 20 "C, and the radioactivity specifically bound at this temperature was essentially the same as that observed at 37 "C for the same hormone concentration, but equilibrium was reached after 3 h incubation (not shown). At 0 "C, specific binding was very slow, for the values obtained after 6 h of incubation were only 30% of those observed at 37 "C after 1 h of incubation (not shown); we did not try longer incubations.
Characteristics of Specific Binding of '251-Erythropoietin As shown in Fig. 3, at least 85% of the labeled erythropoietin could be specifically bound. Moreover, even when more  than 40% of the hormone was removed by preceding incubations, the radioactivity specifically bound did not differ significantly from that obtained with the corresponding adjusted control medium. In what follows, we shall consider that the maximal binding capacity of the labeled erythropoietin is 100%; even if this value is overestimated by 15%, such an error does not affect the results significantly.
Comparison of the binding affinities of the labeled and nonlabeled hormones by self-displacement yielded two straight and paralled displacement curves, suggesting that the affinities are very similar (Fig. 4).
The number of binding sites and the affinity were evaluated by Scatchard plots (22). Fig. 5 presents the results of a typical experiment. As judged by the linearity of the representation, the binding sites belong to one cateeorv: their number Der cell

11)
, only the nonlabeled erythropoietin was able to modify significantly the specific binding '251-erythropoietin to rat erythroid progenitors. A similar specificity was observed when using Friend cells.

Internalization
Hormones bound to cell-surface receptors are generally eluted with acidic buffers (23,24). Here, over 90% of the radioactivity specifically bound after 6 h of incubation of the cells a t 0 "C in the presence of the labeled hormone could be recovered by this method. After 30 min of incubation a t 37 "C, only 63 f 3% (mean f S.D. n = 3) were eluted and 62 f 5% (mean & S.D., n = 7) after 2 h incubation.

Affinity Cross-Linkage
After cross-linkage, the analysis of the cellular extract by polyacrylamide gel electrophoresis after treatment by @-mercaptoethanol revealed the existence of three labeled bands (Fig. 6), with the major one corresponding to the free hormone. The efficiency of the cross-linkingprocedure was rather low, as already observed by others (25). The two labeled bands of high molecular masses were observed only after DSS treatment and were not observed when nonlabeled erythropoietin was in excess during the cell incubation; their molecular masses, determined by comparison with markers stained with Coomassie Blue after either simple (7.5% polyacrylamide) or gradient  1 lanes (A-C), 0.2 ( l a n e D) and 0.04 ( l a n e E ) mM DSS or no DSS ( l a n e F). After washing, the cells were extracted with Triton X-100 in the absence ( l a n e A ) or presence (lanes B-F) of protease inhibitors (see "Materials and Methods"). Extracts corresponding to 5.106 cells were boiled for 5 min in the presence of 2% SDS and 5% 8-mercaptoethanol and subjected to electrophoresis on a 7.5% polyacrylamide slab gel. Unlabeled high and low molecular mass markers sets from Pharmacia P-L Biochemicals revealed by Coomassie Blue staining were used.
inhibitors were omitted during cell lysis, the two labeled bands were attenuated, with the 112-kDa band more than the 128-kDa one (Fig. 6, lane A ) .
In the absence of reductant, the cross-linked material migrated as a single band of higher molecular mass (Fig. 7). By comparison with unreduced molecular mass markers on adjacent lanes, this molecular mass was estimated to be 230-255 kDa. It should be noticed that, under these conditions of electrophoresis, a fraction of the radioactivity did not penetrate the gel; and attempts to dissociate these aggregates with high concentrations of urea or butanol or with acetone precipitation (-20 "C) were unsuccessful. In one experiment, the two bands of lower molecular mass (usually seen in the presence of reductant) were also observed, together with the 230-255-kDa band (not shown).

DISCUSSION
Weiss et al. (26) have shown, with fluorescent antibodies, that only 1-2% of the rat bone marrow hemopoietic cells bind erythropoietin. This very low percentage prevents biochemical evaluation of erythropoietin binding by whole suspension of bone marrow. Mouse erythroid progenitor cells transformed by the anemic strain of Friend virus are erythropoietindependent for their differentiation, and they bind erythropoietin specifically (15). In this work, rat normal erythroid progenitors prepared from fetal liver by a technique (13,16) derived from the procedure of Cantor et al. (27) bind this hormone specifically, with a high affinity, and reversibly. Iodinated erythropoietin has been claimed to be inactive (10); and as emphasized by Nicola and Metcalf for granulocyte colony-stimulating factor (19), preservation of the biological activity depends largely on the conditions of iodination. The labeled preparations we tested were active, but this activity was perhaps due to the presence of nonlabeled molecules since the highest specific radioactivity used (1200 Ci/mol) corresponds statistically to less than one atom of iodine/erythropoietin molecule and there are 4 tryosines potentially available for iodination/molecule (17). It was therefore important to show by self-displacement analysis that the affinities of the native and labeled hormones were the same. Among the cells we tested, only the rat erythroid progenitors (Table I) (Table I) or neoplastic (K 562, HL 60, L1210) (20), did not bind the hormone. Mature erythrocytes (rat or human) had no receptors for erythropoietin, a result a t variance with the data of Baciu et al. (28); but the technique used by these authors does not seem adequate for distinction between high and low affinity binding.
Nonlabeled erythropoietin could readily displace the specifically bound labeled hormone, but the other hormones or growth factors tested could not ( Table 11). The binding sites are therefore likely to be erythropoietin receptors. Phorbol esters, like phorbol 12-myristate 13-acetate, have been shown to inhibit epidermal growth factor (29,30) or macrophage colony-stimulating factor (31) binding to their receptors; we found that 10" M phorbol 12-myristate 13-acetate did not alter erythropoietin binding (Table 11), despite the fact that a 50% variation in the Kd would have been easily detected a t the concentration of erythropoietin used (200 pM).
The data from the Scatchard plot suggest that the receptors belong to a single category, whereas the kinetics of chase experiments demonstrated two patterns of dissociation of the bound hormone (tM = 30 min and 3-4 h, respectively). A plausible explanation is that these two patterns correspond to the dissociation of the membrane-bound hormone and of the internalized hormone-receptor complex.
The number of erythropoietin receptors per rat erythroid progenitor cell was rather small (485 k 32), in agreement with the data of Krantz and Goldwasser (15) concerning murine erythroid cells transformed with the anemic strain of Friend virus. As previously mentioned, the large majority of the population submitted to erythropoietin binding was of the erythroid colony-forming unit type; among the 35% of the cells unable to form colonies, one-half were cells of the white lineages, and its is likely that they do not possess erythropoietin receptors. The number of binding sites per progenitor cell (erythroid colony-forming units) may therefore be underestimated. Binding affinity was roughly 30 times higher in our experiments than that described by Krantz and Goldwasser (15). This discrepancy might be due to differences in the biological materials, but this explanation seems difficult to reconcile with the fact that both cell types were biologically reactive to the same range of hormone concentrations (10, 16) and that we found similar Kd values in mouse erythroleukemic cells (20) and rat normal erythroid progenitors. More likely, the discrepancy reflects different properties of the labeled ligands. In our system, half-maximal biological stimulation was obtained a t around 40 PM erythropoietin (16); at this concentration, roughly 20% of the receptors are occupied. Maximal stimulation was observed a t around 200 PM erythropoietin, a value close to Kd for specific binding.
Internalization was apparently modest (only some 35% after 2 h a t 37 "C), and SDS-polyacrylamide gel electrophoresis of the bound radioactivity did not reveal substantial hormone degradation (not shown). Nevertheless, as the full biological activity of the iodinated hormone was not proven, it is possible that its internalization and/or degradation was not representative of that of native erythropoietin.
The bifunctional reagent DSS has been often used to crosslink hormones and receptors (32)(33)(34)(35)(36)(37). In our system, two bands were revealed by electrophoresis. These bands, which were not observed in the absence of DSS or in the presence of a large excess of nonlabeled erythropoietin, may represent cross-linked hormone-receptor complexes. The difference between their molecular masses (16 kDa) cannot be attributed to the cross-linkage of a supplementary molecule of the ligand (34 kDa). The fact that the 112-kDa band seems more affected by the presence of proteases than the 128-kDa band precludes a degradative filiation between them. By analogy with what was described for the insulin receptor (38), we suppose that specific binding of erythropoietin occurs only on the 94-kDa subunit for the following reasons. 1) Scatchard plots suggest that there is only one category of binding sites; and 2) as in the case of the insulin receptor, cross-linkage is stronger on one subunit (the 94-kDa protein). According to this hypothesis, the 78-kDa subunit should be close enough to the binding site of the 94-kDa subunit to allow some cross-linkage with specifically bound erythropoietin. In the absence of reductant, a single band of cross-linked material was observed; as some proteins behave anomalously in their nonreduced form, the estimated molecular mass of this band (230-255 kDa) may be inaccurate. This band was never observed in the presence of /3-mercaptoethanol and might represent a disulfide-bonded association of the two proteins (94 and 78 kDa, respectively, when hormone-free) observed under reducing conditions. However, as a large fraction of the radioactivity did not enter the gel under nonreducing conditions, we cannot draw any definite conclusion. In one experiment, the cross-linked sub-

Rat Erythropoietin Receptor
units were observed in the absence of P-mercaptoethanol, together with the 230-kDa form; this may be an artifact during the preparation of the cellular extracts, but it should be noted that free (nonreduced) subunits of the insulin receptor were observed in the membrane of various cells and that their existence may have a physiological significance (39).