Cellular Trafficking and Degradation of Erythropoietin and Novel Erythropoiesis Stimulating Protein (NESP)*

Erythropoietin (Epo) is essential for the production of mature red blood cells, and recombinant Epo is commonly used to treat anemia, but how Epo is degraded and cleared from the body is not understood. Glycosylation of Epo is required for its in vivo bioactivity, although not for in vitro receptor binding or stimulation of Epo-dependent cell lines; Epo glycosylation actually reduces the affinity of Epo for the Epo receptor (EpoR). Interestingly, a hyperglycosylated analog of Epo, called novel erythropoiesis-stimulating protein (NESP), has a lower affinity than Epo for the EpoR but has greater in vivo activity and a longer serum half-life than Epo. We hypothesize that a major mechanism for degradation of Epo in the body occurs in cells expressing the Epo receptor, through receptor-mediated endocytosis of Epo followed by degradation in lysosomes, and therefore investigated the trafficking and degradation of Epo and NESP by EpoR-expressing cells. We show that Epo and NESP are degraded only by cultured cells that express the EpoR, and their receptor binding, dissociation, and trafficking properties determine their rates of intracellular degradation. Epo binds surface EpoR faster than NESP (kon = 5.0 × 108 m–1 min–1 versus 1.1 × 108 m–1 min–1) but dissociates slower (koff = 0.029 min–1 versus 0.042 min–1). Surface-bound Epo and NESP are internalized at the same rate (kin = 0.06 min–1), and after internalization 60% of each ligand is resecreted intact and 40% degraded. Our kinetic model of Epo and NESP receptor binding, intracellular trafficking, and degradation explains why Epo is degraded faster than NESP at the cellular level.

Erythropoietin (Epo) 2 and the erythropoietin receptor (EpoR) are required for the production of mature red blood cells (1); recombinant Epo is commonly used to treat anemia in cases of chronic renal failure, cancer, and AIDS. Epo contains one O-linked and three N-linked carbohydrate chains, each having 2-4 branches that often end in a negatively charged sialic acid. These carbohydrate chains are not required for receptor binding in vitro or stimulation of growth of EpoR-expressing cultured cells but are required for the in vivo bioactivity of Epo to increase red blood cell mass (2). Heterogeneous branching of Epo N-linked carbohydrates results in Epo isoforms with different sialic acid contents up to a maximum of 14. Epo isoforms with higher sialic acid content have a lower affinity for EpoR but a longer serum half-life and are more effective for stimulating the production of red blood cells in vivo (3).
How Epo is cleared from the circulation and degraded in the body is not understood (4). In addition, we do not know how the sialic acid content of Epo has such a large effect on its serum half-life and in vivo bioactivity. Epo, both produced endogenously or injected, can be cleared from the circulation by filtration into urine. However, only a small fraction of an injected dose of Epo is excreted intact in urine (5); most of the Epo is degraded in the body, and the degradation products of Epo are excreted in urine (6). Where and how this degradation of Epo occurs is not known, and the clearance of Epo from the circulation and degradation in vivo may occur through more than one mechanism. It is possible that extracellular proteases and enzymes could degrade or modify Epo or that Epo could be taken up through unknown mechanisms and degraded in cells that do not express EpoR. However, there is a strong correlation between the affinity of Epo or Epo analogs for the EpoR and the in vivo lifetime of the ligand. In particular, a hyperglycosylated analog of Epo, called novel erythropoiesis-stimulating protein (NESP), was engineered to contain two additional N-linked carbohydrate chains and a maximum of 22 sialic acids (7). NESP, which is now in clinical use to treat anemia as Aranesp (darbepoetin alfa), has a lower affinity for EpoR and an ϳ3-fold longer serum half-life as compared with Epo (3,8).
We hypothesize that a major mechanism for degradation of Epo and NESP in the body is by cells expressing the Epo receptor, initiated by binding Epo or NESP to surface EpoR, followed by internalization by endocytosis and degradation in lysosomes. In this scenario, the receptor binding and trafficking properties of Epo and NESP should explain their different rates of degradation. To enable us to specifically study trafficking and degradation of Epo and NESP by cells that express the EpoR, we used defined cell culture systems where it was possible to measure the key kinetic parameters. Fig. 1 and our model equations (see "Experimental Procedures") show our kinetic model for cellular trafficking and degradation of Epo and NESP. We have determined all the parameters necessary to describe the cellular trafficking and degradation of Epo and NESP and use these and our model to explain at the cellular level why these ligands are degraded at different rates.
Radiolabeling Ligands-Purified carrier-free recombinant human Epo and NESP suitable for radioiodination (a gift from Amgen Inc., Thousand Oaks, CA), were labeled with Iodine-125 (PerkinElmer Life Sciences) using Iodo-Gen (Pierce Biotechnology). 125 I-Epo and 125 I-NESP had similar specific activities of ϳ4 ϫ 10 6 cpm/pmol, representing an average of one 125 I atom incorporated per molecule of ligand.
Ligand Degradation Assay-Ba/F3 cells were grown in RPMI with 25 mM HEPES, 10% fetal bovine serum (FBS), antibiotics, and WEHI-3B cell conditioned medium as a source of IL-3. Retroviral vector pMX-IRES-green fluorescent protein with hemagglutinin-tagged human EpoR cDNA cloned upstream of the internal ribosome entry site was a gift of Dr. S. N. Constantinescu (Ludwig Institute for Cancer Research, Brussels, Belgium). Samples of cultures were centrifuged to pellet cells, a portion of the medium was counted directly to measure total 125 Iradioactivity and an equal volume of medium was mixed with trichloroacetic acid to measure trichloroacetic acid-soluble radioactivity. Intact (trichloroacetic acid-precipitable) 125 I-ligand in the medium was measured as the difference between total 125 I-radioactivity and trichloroacetic acid-soluble radioactivity. To confirm that trichloroacetic acidprecipitated radioactive ligands were intact, pellets of trichloroacetic acid-precipitated material were washed with acetone, resuspended in protein-gel loading buffer, and subjected to SDS-PAGE. Protein gels were dried onto filter paper and exposed to film for autoradiography.
Equilibrium Binding to Cell Surface EpoR-Cells were incubated in pH 7.3 culture medium with 125 I-Epo or 125 I-NESP for 18 h at 4°C. Bound and free 125 I-ligand were separated by centrifuging cells through a layer of FBS. Nonspecific binding was measured in incubations with 125 I-ligand plus a 200-fold excess of unlabeled Epo as competitor. K D and B max were determined by fitting the model for a single class of non-cooperative binding sites to data obtained over a range of ligand concentrations, using MATLAB software (The Mathworks, Inc., Natick, MA).
Ligand Internalization Assay-Cells were suspended in culture medium at 37°C. To begin the assay an excess of 125 I-Epo or 125 I-NESP was added to the medium. Samples were removed into prechilled tubes in an ice-water bath, where endocytosis is stopped due to the low temperature. Cells were washed with medium at 4°C to remove free 125 Iligand, then surface-bound and internal 125 I-ligand were separated by "acid-stripping" the cells. To acid-strip ligand from the cell surface, cells were suspended in 140 mM NaCl, 25 mM citrate-phosphate buffer, pH 2.6, for 3 min, then pelleted through a layer of FBS (modified from Ref. 9). Cells retained membrane integrity (measured by exclusion of trypan blue dye) in this buffer. Acid-stable radioactivity in the cell pellet is deemed to be internal ligand, and acid-labile radioactivity in the supernatant is deemed to be surface-bound ligand. Nonspecific radioactivity in internal and surface-bound fractions was measured in samples that included 200-fold excess unlabeled ligand as competitor.
Direct Measurement of Surface EpoR Dissociation Rate and Net Association Rate-UT-7/Epo cells (a gift from Dr. Chang-Zheng Chen, Stanford University School of Medicine) were grown in Iscove's modified Dulbecco's medium with 25 mM HEPES, 10% FBS, antibiotics, and 2 units/ml Epo. We obtained a qualitatively similar result for the singlecycle assay comparing UT-7/Epo cells that were grown in 2 units/ml Epo and UT-7/Epo cells that were starved of Epo. Therefore, to increase surface EpoR, and thus obtain data with a greater signal-to-background ratio, our experiments routinely used UT-7/Epo cells that had been deprived of Epo for 18 h in culture medium with FBS.
Direct Dissociation Assay-UT-7/Epo cells were preincubated at 37°C for 10 min in medium with endocytosis inhibitors (0.1% sodium azide ϩ 10 g/ml cytochalasin B). Then, cells were incubated another 10 min with excess 125 I-Epo or 125 I-NESP to allow for binding to surface EpoR. The direct dissociation assay was started by the addition of 200fold excess unlabeled competitor ligand. At selected time points thereafter cells were rapidly separated from the assay medium by pelleting through a layer of FBS. Radioactivity in the cell pellet was counted to measure surface-bound ligand. Nonspecific binding was measured in samples with 200-fold excess competitor ligand included during the initial surface-binding step. The first-order dissociation rate constant, k off , was measured by fitting an exponential curve through triplicate data points.
Direct Association Assay-UT-7/Epo cells were preincubated at 37°C for 5 min in medium with endocytosis inhibitors then the net binding assay started when 125 I-Epo or 125 I-NESP was added to the cells. Cells were separated from the assay medium, and surface-bound ligand was measured as described above. Nonspecific binding was measured in samples with 200-fold excess competitor ligand added together with the 125 I-ligand. The net binding rate, k obs , was determined by fitting the curve LR(t) ϭ LR steady-state (1 Ϫ exp((Ϫk obs )t) to triplicate data points, where k obs ϭ (k on L 0 ϩ k off ) (10).
Single Cycle Assay-UT-7/Epo Cells were preincubated for 5 min at 37°C in medium with endocytosis inhibitors. Then 1 nM 125 I-Epo or 5 nM 125 I-NESP was added, and cells were incubated another 5 min to allow for binding to surface EpoR. Nonspecific binding was measured in parallel samples that included 200-fold excess unlabeled competitor ligand during this surface-binding step. Cells were collected by centrifugation for 1 min at 300 ϫ g at room temperature and washed once with an equal volume of 37°C medium without endocytosis inhibitors. The assay began when half of the washed cells were suspended in 37°C medium without endocytosis inhibitors and the other half of the cells in 37°C medium with inhibitors. Excess competitor ligand (10 nM) was in both assay media to prevent any further binding of 125 I-ligand to surface EpoR after the start of the assay. Thus k on ϭ 0 for 125 I-Epo or 125 I-NESP in this single cycle assay. Samples were collected into prechilled tubes in an ice-water bath. Cells were pelleted at 300 ϫ g at 4°C, a portion of the medium was counted directly to measure total 125 I radioactivity, and an equal volume of the medium was mixed with trichloroacetic acid to measure trichloroacetic acid-soluble 125 I-ligand. Intact ligand in the medium was measured as the difference between total 125 I radioactivity and degraded (trichloroacetic acid-soluble) 125 I-ligand. Then, the cell pellet was washed once with medium at 4°C, and surfacebound and internal cell-associated 125 I-ligand were measured by the acid-stripping procedure.

Kinetic Model of Cellular Trafficking and Degradation of Epo and
NESP-Model equations were solved for the initial conditions of our single cycle assay with MATLAB ode23 solver, using k on ϭ 0 and our experimentally determined values of k off and k in . A direct measurement of internal sorting steps was not possible. Instead, the rate at which intact ligand accumulated in the medium because of resecretion was modeled as the product of an overall first-order process with rate k resec . In single cycle assays, the experimental curve showing net accumulation of intact ligand in the medium is a composite of ligand dissociating directly from surface EpoR and internalized ligand being resecreted intact (see Figs. 1 and 4). The best fit value of k resec for this composite curve, given the measured values of k off and (k off ϩ k in ), was determined. There was no significant difference in the fitted values of k resec for Epo or NESP, and the average of the fitted values of Epo and NESP is k resec ϭ 0.347 min Ϫ1 Ϯ 0.128. To represent the sorting step, k lys was set ϭ 0.204 min Ϫ1 , to make the ratio of internalized ligand sorted for resecretion, k resec /(k resec ϩ k lys ), equal to the average fraction of internalized ligand that is resecreted intact (ϭ 0.63; Table 4). In single cycle experiments degraded (trichloroacetic acid-soluble) ligand began to accumulate after 15-20 min. Then, from 20 to 60 min trichloroacetic acid-soluble ligand accumulated in the medium in a "quasilinear" (10) fashion, with half of the degraded ligand released by 35-40 min (see Fig. 4

, A and B).
To incorporate this observed lag, our model equations require that internalized ligand sorted for degradation at rate k lys pass through a series of intermediate steps with first-order rate constants (Fig. 1). Then, after passing through these "degradation lag" steps, degraded ligand is released into the medium essentially instantaneously (k deg ϭ 3 min Ϫ1 ). Fig.  2 shows that degradation of Epo and NESP by cultured cells requires that they express EpoR. The IL-3-dependent cell line Ba/F3 (which does not express EpoR or respond to Epo) and Ba/F3 cells expressing human EpoR from a retroviral vector (Ba/F3-huEpoR) were cultured for 3 days with excess IL-3 to ensure equal cell growth and either 0.2 nM 125 I-Epo or 125 I-NESP. The amount of intact Epo or NESP in cultures of parental Ba/F3 cells was not decreased over 3 days, indicating that neither Epo nor NESP were degraded (Fig. 2, A and B). SDS-PAGE analysis of acid-precipitated medium revealed only 125 I-proteins the size of intact Epo or NESP (Fig. 2C);  Fig. 1 and our model equations (described under "Experimental Procedures"), but NESP can take the place of Epo as the ligand for EpoR. Epo media (intact) , intact Epo in the medium. Epo⅐EpoR surface , complex of Epo specifically bound to surface EpoR. Epo media (degraded) , degraded Epo in the medium. k on , rate constant for Epo binding to surface EpoR. k off , rate constant for dissociation of Epo from Epo⅐EpoR surface into the medium. k in , rate constant for endocytic internalization of Epo⅐EpoR surface . k resec , overall rate constant for sorting and resecretion of internalized Epo into the medium as intact ligand. k lys , rate constant for sorting internalized Epo into an intracellular degradation pathway. Epo sorted for degradation is subject to additional intracellular trafficking and processing steps, so there is a time lag before Epo is degraded, and the products released from the lysosomal compartment into the medium with rate constant k deg .

Degradation of Epo and NESP by Cells Is Mediated by Surface EpoR-
Epo migrated faster during SDS-PAGE than NESP because of the smaller amount of attached carbohydrate. In contrast, the amount of intact Epo or NESP in cultures of Ba/F3-huEpoR cells decreased over time, showing that Epo and NESP were degraded by Ba/F3-huEpoR cells (Fig. 2, A and B). Epo was degraded at a faster rate than was NESP by Ba/F3-huEpoR cells; after 3 days 71% of Epo had been degraded and only 21% of NESP (Fig. 2, A and B). We obtained similar results with higher (1 nM) and lower (0.05 nM) initial ligand concentrations (data not shown). These results show that degradation of Epo and NESP by cells requires that they express EpoR.
Presumably EpoR mediates degradation of Epo and NESP by binding the ligands at the cell surface, after which they are internalized and degraded in lysosomes. To test for specific surface binding, cells were incubated 18 h with labeled ligand at 4°C, a temperature where endocytosis does not occur, so all cell-associated Epo or NESP was bound to the cell surface. Parental Ba/F3 cells incubated with 0.5 nM 125 I-Epo bound only 0.14 fmol of 125 I-Epo/10 6 cells, and this amount was unchanged when 200-fold excess unlabeled Epo was included as com-petitor in the incubation (Table 1). This binding is deemed nonspecific. In contrast, Ba/F3-huEpoR cells bound 3.19 fmol of 125 I-Epo/10 6 cells, and the addition of a 200-fold excess of competitor reduced binding to 0.10 fmol of 125 I-Epo/10 6 cells, the same amount of 125 I-Epo bound to parental Ba/F3 cells in the presence or absence of unlabeled competitor (Table 1). Specific 125 I-Epo binding, the difference between total and nonspecifically bound 125 I-Epo, was 3.09 fmol of 125 I-Epo/10 6 cells or 1860 molecules of radiolabeled Epo/cell. In a separate experiment, an analysis of equilibrium binding at 4°C of a range of concentrations of 125 I-Epo to Ba/F3-huEpoR cells revealed ϳ2400 surface Epo binding sites/cell (B max ϭ 3.93 fmol/10 6 cells (95% confidence interval: 3.69, 4.17); K D ϭ 0.109 nM (95% confidence interval: 0.075, 0.143)) (data not shown).
To test for internalization of Epo or NESP, cells were incubated with 125 I ligand at 37°C, a temperature where endocytosis occurs; surfacebound and internal 125 I ligand were measured by acid stripping (see "Experimental Procedures"), which releases surface-bound ligand. Parental Ba/F3 cells incubated with 1 nM 125 I-Epo or 125 I-NESP at 37°C had only a small amount of radioactivity in surface-bound and internal fractions (ϳ100 -400 cpm), which did not change during the assay, was not changed by addition of excess competitor, and was the same obtained with Ba/F3-huEpoR cells incubated at 37°C with 125 I ligand and excess unlabeled competitor (data not shown). Therefore, parental Ba/F3 cells showed no specific surface binding or internalization of Epo or NESP at 37°C.
In contrast, Ba/F3-huEpoR cells rapidly bound Epo or NESP at the cell surface and after a few minutes internal ligand appeared (Fig. 2, D  and E). The amount of 125 I-Epo specifically bound to the surface of Ba/F3-huEpoR cells reached a maximum after 5 min and was constant thereafter (Fig. 2D). After a lag of 5 min internal 125 I-Epo began to increase and approached a steady level after 60 min that was greater than the amount of surface-bound Epo (Fig. 2D). Presumably these levels of surface and internal ligand represent a steady state between the internal ligand sorted for degradation and for resecretion and external ligand binding to the surface and internalization.
In a similar manner, 125 I-NESP first bound to surface EpoR and then was internalized by Ba/F3-huEpoR cells (Fig. 2E). In comparison to Epo, NESP had a slower net rate of surface association, and smaller amounts of NESP were bound and internalized (note different y axis scales in Fig.  2, D and E), which is a reflection of the lower affinity of NESP for surface EpoR.
Together, these experiments showed that degradation of Epo or NESP by Ba/F3-huEpoR cells was mediated by surface EpoR. Furthermore, Ba/F3-huEpoR cells specifically bound Epo or NESP at the cell surface and then internalized the surface-bound ligand, whereas parental Ba/F3 cells did not specifically bind or internalize these ligands.
Cellular Trafficking and Degradation of Epo and NESP-To understand why Epo was degraded faster than NESP we determined the rate constants and key parameters that describe this ligand-receptor system (Fig. 1). To make these measurements it was first necessary to distinguish between loss of surface-bound ligand due to direct dissociation from EpoR at 37°C and loss of surface-bound ligand due to internalization. Each process decreases the amount of surface-bound 125 I ligand, but moves the ligand to a different location (Fig. 1). To measure k off , ligand dissociation was measured as the loss of ligand from the cell surface at 37°C in the presence of endocytosis inhibitors, which were used to block the internalization of ligand. Ba/F3 cells began to lose membrane integrity (as measured by loss of trypan blue dye exclusion) after ϳ20 min in the presence of endocytosis inhibitors, so we searched for another cell line to measure Epo binding and trafficking rate con-stants. UT-7/Epo is an Epo-dependent cell line that expresses human EpoR from the endogenous gene, and endocytosis of Epo or NESP is inhibited in these cells by the combination of 0.1% sodium azide and 10 g/ml cytochalasin B (7,(11)(12)(13). UT-7/Epo cells retained viability and membrane integrity for at least one h at 37°C in the presence of these endocytosis inhibitors. Both 125 I-Epo and 125 I-NESP were degraded in cultures of UT-7/Epo cells using the assay described in Fig. 2 (data not shown). Endocytosis of Epo by UT-7/Epo cells was effectively inhibited by the combination of sodium azide and cytochalasin B, allowing us to label surface EpoR with 125 I ligand at 37°C (supplemental Fig. S1, A and  B; Fig. 4

, A and B, right panels).
We measured the dissociation rate constant (k off ) for Epo and NESP from cell-surface EpoR by a direct dissociation assay (see "Experimental Procedures") and independently during single cycle assays (see "Experimental Procedures"; Fig. 4C and Table 2). There was day-to-day variation in k off measured between experiments, resulting in overlapping standard deviations for Epo and NESP k off (Table 2). However, among five direct comparisons of Epo and NESP where dissociation was measured on the same day, the k off for NESP was consistently faster than for Epo, paired t test (two-tailed), t ϭ 4.3578; p ϭ 0.0121 (Table 2). Averaging these k off measurements with those obtained for Epo and NESP measured in separate single cycle experiments on different days, Epo k off ϭ 0.029 Ϯ 0.009 min Ϫ1 and NESP k off ϭ 0.042 Ϯ 0.012 min Ϫ1 . Thus NESP dissociates from surface receptors ϳ40% faster than Epo.
To determine the association rate (k on ) for Epo and NESP to cellsurface EpoR, we measured the net binding rates, k obs (10), of 125 I-Epo (at 0.2 and 0.5 nM) and 125 I-NESP (at 0.5 and 1 nM) to UT-7/Epo cells at 37°C in the presence of endocytosis inhibitors ( Fig. 3 and Table 3). Using our experimentally determined values of k off and measuring binding at two different initial concentrations of Epo and NESP, we found that Epo k on ϭ 5.0 ϫ 10 8 M Ϫ1 min Ϫ1 and NESP k on ϭ 1.1 ϫ 10 8 M Ϫ1 min Ϫ1 (Fig. 3 and Table 3). The 5-fold difference between Epo and NESP k on explains why the net binding curves measured with 0.2 nM 125 I-Epo and 1 nM 125 I-NESP overlapped each other (Fig. 3, right panel).
To assess whether our experimentally determined on and off rates are valid, we calculated EpoR dissociation constants at 37°C, K D , as k off /k on ratios. Derived dissociation constants were Epo K D ϭ 0.058 nM and NESP K D ϭ 0.40 nM (Table 3). By comparison, equilibrium binding measurements at 4°C of 125 I-Epo to Ba/F3-huEpoR cells yielded K D ϭ 0.109 nM (95% confidence interval: 0.075, 0.143) and to UT-7/Epo cells yielded K D ϭ 0.178 nM (95% confidence interval: 0.119, 0.238), whereas equilibrium binding of 125 I-NESP at 4°C to UT-7/Epo cells yielded K D ϭ 1.279 nM (95% confidence interval: 0.9336, 1.624). Considering that the single rate constants k off and k on were measured at 37°C and our equilibrium binding analysis was measured at 4°C, there is good agreement between our 37°C k off /k on ratios and our 4°C K D values. This is independent verification that our measured k off and k on rates are valid.

TABLE 1 Total and non-specific binding of 125 I-Epo and 125 I-NESP to Ba/F3 and Ba/F3-huEpoR cells
Cells were incubated 18 h at 4°C with the indicated combinations of 125 I-labeled ligands with or without 200-fold excess unlabeled competitor. Cells were separated from the media by pelleting through a layer of FBS, and total cell-associated 125 I-ligand was determined in a ␥ counter. In the key experiment we followed the fate of Epo and NESP through a single cycle of endocytosis, after first binding to surface EpoR (see "Experimental Procedures" and Fig. 4). At the beginning of the experiment, as expected, almost all 125 I-labeled ligand was surface-bound, there was only a small amount of intact ligand in the medium, and there was little non-acid-strippable ligand (internal) associated with the cells (Fig. 4).
After cells were surface-labeled, washed, and suspended at 37°C without endocytosis inhibitors (Fig. 4, A and B, left panels), the amount of internal 125 I-Epo or 125 I-NESP increased over the first 20 min and then decreased from 20 to 60 min. Accumulation of internal ligand shows that surface-bound Epo or NESP was internalized by endocytosis; subsequent loss of internal ligand shows that internalized ligand was released to the medium. With endocytosis inhibitors present during the single cycle assay (Fig. 4, A and B, right panels) the amount of internal 125 I-Epo or 125 I-NESP did not increase over 60 min, showing that the endocytosis inhibitors blocked internalization of surface-bound Epo or NESP.
Surface-bound 125 I-Epo or 125 I-NESP was lost at a faster rate in the absence of endocytosis inhibitors than when inhibitors were present  a For each experiment, Epo and NESP dissociation was measured on the same day, with the only variable being whether the ligand was 125 I-Epo or 125 I-NESP, and all samples and measurements were handled in parallel. b Labeled Epo or NESP was bound to UT-7/Epo cells at 37°C in the presence of endocytosis inhibitors then the loss of 125 I ligand from the cell surface was measured in medium containing inhibitors and excess unlabeled Epo to prevent rebinding of 125 I-ligand. The first-order dissociation rate constant, k off , was measured by fitting an exponential curve through the data. c Paired t test (two-tailed): t ϭ 4.3578; p ϭ 0.0121 a UT-7/Epo cells were preincubated at 37°C for 10 min with endocytosis inhibitors (0.1% sodium azide ϩ 10 g/ml cytochalasin B) then 125 I ligand was added. At a range of incubation times, cells were collected and rapidly separated from the medium by pelleting through a layer of FBS, then cell-associated radioactivity was measured. The net binding rate, k obs , was determined by fitting the curve LR(t) ϭ LR steady-state (1 Ϫ exp((Ϫk obs )t) to the triplicate data points. b k off was determined both in direct dissociation assays and in single cycle assays. Paired t-test (two-tailed): t ϭ 4.3578; p ϭ 0.0121 (see "Results" and Table 2). c k on ϭ (k obs Ϫ k off )/͓L͔. (Fig. 4, A and B, compare left and right panels; Fig. 4C). In the absence of endocytosis inhibitors, 125 I-Epo or 125 I-NESP was lost from the surface through two processes, direct dissociation from EpoR into the medium and through internalization, at a combined rate of (k off ϩ k in ) (Fig. 1). In the presence of endocytosis inhibitors, 125 I-Epo or 125 I-NESP was lost from the surface only through direct dissociation from EpoR into the medium, with a first-order rate constant k off .

TABLE 3 Kinetic rate constants for Epo and NESP binding to cell surface EpoR
Thus we were able to determine the rate constants for internalization (k in ) of surface-bound Epo or NESP as the difference between the rates of loss of surface-bound ligand in the absence and in the presence of endocytosis inhibitors, or k in ϭ (k off ϩ k in ) Ϫ k off (Fig. 4C). In the absence of endocytosis inhibitors, 125 I-Epo was lost from the surface with a combined rate of (k off ϩ k in ) ϭ 0.09 min Ϫ1 , and in the presence of endocytosis inhibitors, 125 I-Epo dissociated from surface EpoR with rate k off ϭ 0.032 min Ϫ1 (Fig. 4C, left panel). The difference between these rates yielded k in ϭ 0.058 min Ϫ1 . In a similar manner we measured NESP (k off ϩ k in ) ϭ 0.112 min Ϫ1 and NESP k off ϭ 0.048 min Ϫ1 (Fig. 4C, right  panel), yielding NESP k in ϭ 0.064 min Ϫ1 . Thus the rates of internalization of surface-bound Epo and NESP are the same, ϳ0.06 min Ϫ1 .
Another important aspect of the endocytic cycle of Epo and NESP was revealed by this single cycle experiment. Some of the internalized Epo or NESP was returned to the medium as intact ligand, and some was degraded. In the presence of endocytosis inhibitors, as labeled ligand dissociated from surface EpoR with rate k off , there was a corresponding accumulation of intact (trichloroacetic acid-precipitable) 125 I-labeled ligand in the medium (Fig. 4, A and B, right panels). With internalization blocked by inhibitors, as expected, there was no release of trichloroacetic acid-soluble radioactivity into the medium (Fig. 4, A and B, right  panels). This suggests that internalization is required for degradation of the surface-bound 125 I-Epo or 125 I-NESP, consistent with the experiments using Ba/F3 cells (Fig. 2).
In the experiments performed without endocytosis inhibitors, essentially all the ligand that accumulated in the medium during the first 15-20 min of the endocytic cycle was intact. From 20 min onward degraded 125 I-ligand was also released into the medium (Fig. 4, A and B,  left panels). These results indicate that following a lag of 15-20 min, a portion of the Epo or NESP that was internalized from the surface had been degraded intracellularly and was released into the medium as tri- chloroacetic acid-soluble 125 I-ligand. It is noteworthy that the quantity and rate of the net decrease of internal ligand from 20 to 60 min is similar to the quantity and rate of the net increase of degraded ligand in the medium from 20 to 60 min (Fig. 4, A and B, left panels). Thus after 20 min of endocytosis, most of the 125 I-ligand that remained inside the cell had already been sorted to the lysosome for degradation. Furthermore, intact 125 I-ligand was released into the medium at a faster rate from cells that internalized ligand than from cells incubated in the presence of endocytosis inhibitors (Fig. 4, A and B). This suggests that surfacebound ligand returns intact to the medium through a combination of direct dissociation from surface EpoR and resecretion of internalized ligand.
Using our values of Epo k off and k in , the expected ratio between dissociation and internalization of surface-bound Epo is calculated from k off /(k off ϩ k in ) ϭ 0.032 min Ϫ1 /0.09 min Ϫ1 ϭ 0.35 (Fig. 4, A and C). In other words, 35% of surface-bound 125 I-Epo dissociated directly into the medium during this single cycle endocytosis assay, whereas 65% was internalized. If the entirety of the 65% of the original surface-bound Epo that was internalized was subsequently degraded, only 35% of 125 I-Epo accumulated in the medium at the end of the experiment (60 min) would have been intact. However, in the experiment shown in Fig. 4A (left panel), 75% of the 125 I-Epo that accumulated in the medium over 60 min was intact. This means that 40% of the original surface-bound 125 I-Epo (ϭ 75% intact Epo in the medium at 60 min Ϫ 35% Epo directly dissociated from surface receptors) was internalized and then resecreted intact into the medium. Thus the fraction of internalized 125 I-Epo that was resecreted as intact Epo in this experiment was 0.61 (ϭ 40% resecreted/65% internalized). The remaining 39% of the internalized 125 I-Epo was degraded and released to the medium. This accounts for the trichloroacetic acid-soluble radioactivity that accumu-lated in the medium by 60 min. In other words, 39% of the 65% of surface-bound 125 I-Epo that became internalized, or equivalently 25% of the initial surface-bound 125 I-Epo, appeared in the medium as products of degradation.
Assays with NESP were analyzed as described above for Epo, and Table 4 summarizes the key parameters determined in three independent repetitions of our single cycle assays with Epo or NESP. Epo and NESP were internalized at the same rate, k in ϭ 0.060 min Ϫ1 . In addition, ϳ60% of internalized Epo or NESP was resecreted intact. Overall, this means that once Epo or NESP has bound to surface EpoR on cells, a larger fraction of Epo will be internalized than NESP because of the faster dissociation rate of NESP. Because similar fractions of internalized Epo and NESP are degraded and are resecreted intact, there will be more Epo than NESP subject to degradation in each round of endocytosis that follows the binding of ligand to surface EpoR.
The only kinetic difference between Epo and NESP was their rates of receptor binding (k on ) and dissociation (k off ). There is good agreement between our model curves and the data from single cycle assays (Fig. 5). This overall good fit suggests that our kinetic model correctly describes the trafficking and degradation of Epo and NESP; Epo is degraded faster than NESP because of a faster k on and slower k off .

DISCUSSION
The lifetime of Epo and NESP in the circulation is the key factor that determines the bioactivity of these ligands in vivo. Although 5-fold more NESP than Epo is required to achieve equivalent receptor binding and activation (Table 3 and Ref. 14), NESP has increased and prolonged bioactivity to increase red blood cells in vivo (3,7,15). Because NESP is cleared from the body more slowly than Epo, this suggests that how long signaling is sustained in vivo, and not receptor occupancy per se, is the most important property for obtaining an in vivo cellular response from a limited amount of hormone. Little is known about how Epo and NESP are cleared from the body or the nature of their main sites of degradation in vivo (4). Our experiments with cultured cells show that Epo and NESP are degraded only by cells expressing EpoR, through binding to surface EpoR, internalization by endocytosis, and degradation in lysosomes. As detailed in the Introduction, clearance of Epo and NESP in vivo may occur through more than one mechanism. Our data indicates that intracellular degradation of Epo and NESP in vivo will occur in cells that FIGURE 5. Fitting trafficking and degradation kinetic model to single cycle endocytosis data. Curves were calculated by solving the differential equations describing our kinetic model of Epo and NESP endocytosis, resecretion and degradation (see Fig. 1 and model equations under "Experimental Procedures") and plotted along with the data points from the single cycle assays shown in Fig. 4. The data for cell-associated ligand have been normalized as the fraction of cell-associated (surface ϩ internal) ligand at time 0. The data for ligand in the medium have been normalized so the net accumulation of total 125 I radioactivity at 60 min ϭ 1, and the small amount of intact ligand in the medium at time 0 has been subtracted from all medium samples. The differential equations were solved using the initial conditions of the time 0 data for surface-bound and internal ligand and with 0 ligand in the medium and k on ϭ 0; Epo k off ϭ 0.029 min Ϫ1 ; NESP k off ϭ 0.042 min Ϫ1 ; k in ϭ 0.060 min Ϫ1 ; k resec ϭ 0.347 min Ϫ1 ; k lys ϭ 0.204 min Ϫ1 ; k lag1 -k lag6 ϭ 0.2 min Ϫ1 ; k deg ϭ 3 min Ϫ1 (see "Experimental Procedures"). A, data and curves for Epo. B, data and curves for NESP. express surface EpoR, and thus EpoR-expressing hematopoietic cells will be a site of degradation of Epo and NESP in the body. EpoR is expressed on erythroid progenitor cells and also on non-hematopoietic cells, such as neurons, glia, myoblasts, and endothelial cells (16 -21). We do not know whether or not Epo is degraded by non-hematopoietic cells that express the EpoR. We used cultured Ba/F3 and UT-7/Epo cells as model systems. The use of these defined cell culture systems allowed us to test the role of the EpoR in mediating cellular degradation of Epo and NESP and to measure the key kinetic parameters that explain Epo and NESP receptor binding, intracellular trafficking, and degradation. Only cells that expressed EpoR degraded Epo and NESP, and NESP was degraded more slowly than Epo. Once bound to surface EpoR, Epo and NESP were internalized at the same rate (k in ϭ 0.06 min Ϫ1 ). Following internalization there was no difference in the sorting of Epo and NESP for resecretion or degradation; 60% of internalized Epo or NESP was resecreted intact, and the balance was degraded. Thus the increased sialic acid content of NESP does not change how it is internalized or sorted for degradation or resecretion; Epo and NESP differ only in the way they bind to surface EpoR. Epo dissociates slower than NESP from surface EpoR. At 37°C Epo dissociates with a half-time of 24 min and NESP with a 17-min half-time. Consequently, after binding to surface EpoR more Epo than NESP will be internalized and subject to degradation and less will dissociate directly into the medium. In addition, Epo binds to EpoR ϳ5 times faster than NESP.
This mechanism by which additional sialic acids extend the half-life of Epo is different from that by which the half-life of mutated analogs of G-CSF and IL-2 are extended. In these cases a larger fraction of endocytosed mutant cytokine is resecreted compared with the wild-type hormone (22,23). Essentially all of the G-CSF receptor is sorted for degradation after endocytosis (23); a larger fraction of the G-CSF mutant is resecreted because of a lower affinity for the G-CSF receptor at the slightly acidic pH of the endosome, although the affinity is normal at the neutral pH of the cell surface. The mutant IL-2 has a higher relative affinity at endosomal pH than wild-type IL-2 for the IL-2 receptor ␣ subunit, which is constitutively recycled to the surface, than to the ␤ subunit, which is degraded in lysosomes (22). In contrast, NESP is degraded slower than Epo due only to differences in binding to and dissociation from cell surface EpoR.
The consequence for signaling is that more NESP than Epo is required for equivalent EpoR signaling in acute stimulation of cells (14), presumably directly reflecting a lower equilibrium constant for receptor occupancy and the lower k on . Importantly, binding of Epo or NESP to EpoR is rapidly followed by phosphorylation and activation of Janus kinase 2 (JAK2) bound to the EpoR cytosolic domain. This leads to phosphorylation of tyrosines in the EpoR cytoplasmic domain and subsequent activation of downstream signaling pathways including the STAT5, Ras, and phosphatidylinositol 3-kinase pathways (24). The termination of EpoR signaling takes longer than activation, because it involves recruitment and activation of tyrosine phosphatases that dephosphorylate and inactivate JAK2 (25), down-regulation of surface EpoRs (26 -28), and induction and binding of the CIS/SOCS (cytokineinducible SH2-containing protein/suppressor of cytokine signaling) family of negative regulators of cytokine signaling (29,30). Therefore, signaling initiated from a cell surface Epo⅐EpoR complex likely continues for several minutes even after Epo or NESP have dissociated. The faster dissociation of NESP than Epo from surface EpoR likely would not materially affect the duration of signaling from the NESP⅐EpoR complex, and the longer biological half-life of NESP would then enable it to signal many more times in its lifetime than Epo.