Both proteasomes and lysosomes degrade the activated erythropoietin receptor

The kidney-produced hormone erythropoietin (Epo) is the major regulator of red blood cell production. Epo inhibits apoptosis of the late erythroid progenitors and stimulates their proliferation, allowing the completion of their differentiation program.¹  The Epo receptor (EpoR) is a type 1 transmembrane protein that belongs to the cytokine receptor family. It is synthesized as a 62 kDa precursor that is quickly modified by a high-mannose glycosylation that increases its molecular mass to 64 kDa. The mature EpoR exhibits a 66 kDa molecular mass and a complex endoglycosidase H–resistant glycosylation pattern.^2,3  The cell surface expression of the EpoR appears to be tightly controlled by mechanisms that are only partly understood. The EpoR undergoes dimerization and association with the Jak2 effector kinase during its maturation process, and Janus kinase 2 (Jak2) association with the EpoR is required for EpoR maturation and cell surface expression.⁴  Epo binding induces a conformational change in the receptor complex that activates the associated Jak2 kinases, leading to the initiation of intracellular signaling. Several intracellular relays are then activated, including signal transducer and activator of transcription 5 (STAT5), Ras/mitogen-associated protein (Ras/MAP) kinase, and phosphatidylinositol-3 (PI3) kinase/Akt pathways (reviewed by Constantinescu et al⁵ ). All Epo-dependent responses including proliferation and survival are abrogated when Jak2 activation is disrupted. Simultaneously, mechanisms of down-regulation are turned on, and they lead to the rapid decrease of Epo responsiveness of the target cells. These mechanisms involve the recruitment of the SHP-1 (domain-containing protein-tyrosine phosphatase-1) phosphatase to the receptor⁶  and to the Jak2 kinase,⁷  the production and receptor association of inhibitor proteins of the CIS (Cytokine Inducible SH2-containing protein)/suppressor of cytokine signaling (CIS/SOCS) family, ^8-10  and the internalization of the receptors.¹¹  Moreover, we have shown that the maturation and cell surface expression of newly synthesized EpoRs are blocked by a mechanism involving the proteasomes during Epo stimulation.¹² 

EpoR and the thrombopoietin (TPO) receptor (Mpl) exhibit strong functional similarities. They are homodimeric cytokine receptors that activate Jak2 to initiate intracellular signaling, and their main target cells belong to close differentiation lineages. In contrast to Epo (ErythroPOietin), TPO (ThromboPOietin) induces long-lasting signaling in target cells, and the duration of this signal is required for megakaryocytic differentiation.¹³  It has been recently shown that after TPO binding, 65% to 75% of the TPO-Mpl complexes are internalized and that at least some of the internalized receptors rapidly recycle to the cell surface.¹⁴  Like TPO-Mpl complexes, Epo-EpoR complexes are rapidly internalized, and this process is not dependent on Jak2.¹¹  However, the fate of the internalized Epo-EpoR complexes has not been examined.

The internalization and degradation of ligand-stimulated cytokine receptors have been extensively studied for the growth hormone (GH) receptor and the interleukin-2 (IL-2) receptor. As Mpl and the EpoR, the former is a receptor homodimer associated with the tyrosine kinase Jak2 and the latter is composed of at least 5 proteins, including 2 tyrosine kinases of the Jak family (Jak1 and Jak3) and 3 bona fide receptor chains (IL-2 receptor α [IL-2Rα], IL-2Rβ, and IL-2Rγ). In both cases, endocytosis of the receptors is constitutive although their cognate ligands increase to some extent the rate of receptor disappearance.^15,16  Endocytosis of the GH receptor occurs through clathrin-coated pits and depends on an active ubiquitination system and proteasome activity.¹⁷  The internalized receptor is then routed to the lysosome and degraded together with the hormone. This step also requires the activity of the ubiquitin/proteasome pathway.¹⁸  In contrast, internalization of the IL-2 receptor complex mainly occurs through clathrin-independent mechanisms involving detergent-resistant membrane domains.¹⁹  The subunits of the IL-2 receptor are then differentially sorted: the α chain recycles back to the plasma membrane, while the β and γ chains are sorted through the degradation pathway to the lysosomes.²⁰  IL-2 induces monoubiquitination²¹  or polyubiquitination²²  of the β chain of its receptor. Ubiquitin is clearly involved in IL-2 receptor down-regulation although the steps controlled by the ubiquitin/proteasome system remain controversial. According to Yu and Malek,²³  proteasome inhibitors impair internalization of IL-2–IL-2 receptor complexes. In contrast, Rocca et al have reported that the ubiquitin/proteasome system is not required for internalization of the IL-2 receptor although the β chain of the IL-2 receptor is monoubiquitinated and ubiquitination is necessary for β routing to the lysosomes.²¹ 

In this study, we investigate the mechanisms of Epo and EpoR internalization and degradation. We show that Epo strongly increases the rate of receptor internalization and degradation that involves both proteasomes and lysosomes. Lysosome targeting of Epo and EpoR requires Jak2 activity and not proteasome activity. Moreover, Jak2 activation induces polyubiquitination of the EpoRs, leading to the degradation at the cell surface of the C-terminal part of the intracellular domain of the receptor by proteasomes.

Highly purified recombinant human Epo (specific activity 120 000 U/mg) was a generous gift from Dr M. Brandt (Roche Laboratories, Basel, Switzerland). The protease inhibitors N-Ac-Leu-Leu-norLeucinal (LLnL) and lactacystin were obtained from Calbiochem (Merck Biosciences, Darmstadt, Germany). Other inhibitors were purchased from Sigma (St Louis, MO). Anti-EpoR antibodies used in immunoprecipitation experiments (C-236) were produced by immunizing rabbits with a recombinant protein composed of the full intracellular domain of the human EpoR fused to glutathione S-transferase (GST). Experiments using various EpoR mutants have shown that these antibodies recognize epitopes included in the first 60 amino acids of the intracellular domain of the EpoR, because they immunoprecipitate full-length EpoRs and EpoRs truncated after amino acid 310. Anti-GST control antibodies were prepared using the same protocol. Anti-EpoR antibodies from Santa Cruz Biotechnology (catalog no. SC-695) (Santa Cruz, CA) were used for immunoblot analysis. These antibodies (C-20) recognize the last 20 amino acids of the EpoR. Antiubiquitin antibodies (catalog no. PW 8810) and anti-Jak2 antibodies (catalog no. SC-278) were from Affiniti Research Products (Exeter, United Kingdom) and Santa Cruz Biotechnology, respectively. Bissulfosuccinimidyl suberate (BS3) was from Pierce (Rockford, IL). Deglycosylation enzymes and proteinase K were from Roche (Mannheim, Germany). Human UT-7 cells²⁴  were cultivated in α minimum essential medium (αMEM) containing 10% fetal calf serum complemented with 2 U/mL Epo. Before each experiment, the cells were growth factor deprived by overnight incubation in Iscove Dulbecco MEM containing 0.1% bovine serum albumin treated with AG 501-X8 ion exchange resin Bio-Rad (Hercules, CA) 25 μg/mL iron-loaded human transferrin.

Whole cell extracts, immunoprecipitations, and Western blots were performed as previously described.²⁵  Extracts from 5 × 10⁵ cells were used in each assay for the Western blot experiments. Western blots were quantified using the ImageJ 1.32 software after densitometric scanning of the films (http://rsb.info.nih.gov/ij/).

Epo labeling using IODO-GEN (Pierce, Rockford, IL) and Epo binding were done as previously described.^26,27  A saturating ¹²⁵I-Epo concentration (1 nM; 2 U/mL) was used for these experiments to suppress the effects of putative modifications of EpoR affinity induced by the different inhibitors used. Nonspecific binding, determined using a 250-fold excess of unlabeled Epo, was less than 5% in each case. All reported data represent specific binding. In most experiments, we performed an acidic wash to separate cell surface–bound and internalized Epo. After incubation with ¹²⁵I-Epo, the cells were washed twice at 4°C to remove unbound ligand. They were incubated in 0.5 mL acidic buffer (NaCl 150 mM, sodium acetate 50 mM, pH 3.5) for 3 minutes at 4°C. The pH was then adjusted to 7.4 using 1 M Tris (tris(hydroxymethyl)aminomethane)–HCl, pH 9, and the cell suspension was centrifuged. The radioactivities of the supernatant (cell surface–bound Epo) and of the cell pellet (internalized Epo) were determined. When ¹²⁵I-Epo was bound to the cells at 4°C to inhibit Epo internalization, more than 95% of cell-bound ¹²⁵I-Epo was recovered in the acidic wash supernatant using this method. Each experiment was performed at least 3 times with similar results; 2 × 10⁶ cells were used in each assay for the ¹²⁵I-Epo binding experiments.

To cross-link ¹²⁵I-Epo to internalized receptors, the cells were incubated with ¹²⁵I-Epo, washed to remove unbound Epo, and solubilized with 25 mM HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid), 150 mM NaCl, 5 mM EGTA (ethyleneglycoltetraacetic acid), and 1% Nonidet P-40, pH 8.00. The lysate was cleared by centrifugation and incubated for 20 minutes in ice with 2 mM BS3. The water-soluble cross-linker BS3 was used to achieve the high concentration of cross-linker required for these experiments. The reaction was stopped by adding ethanolamine, pH 8.00, to a final concentration of 0.1 M, and the EpoR complexes were immunoprecipitated using C-236 anti-EpoR antibodies. Immunoprecipitated material was denatured by boiling in 1% sodium dodecyl sulfate (SDS) and 50 mM dithiothreitol (DTT). DTT was removed by chromatography through G25 Sephadex spun columns. Recovered material was then diluted 10-fold with solubilization buffer and subjected to another immunoprecipitation using C-236 anti-EpoR antibodies. Lastly, nonprecipitated ¹²⁵I-Epo–cross-linked proteins were recovered using anti-Epo antibodies; 50 × 10⁶ to 100 × 10⁶ cells were used in each assay for the cross-linking experiments.

The half-life of the cell surface EpoR in the absence of Epo was determined by blocking protein synthesis with cycloheximide and measuring cell surface EpoRs at several time points thereafter by short incubations with ¹²⁵I-Epo. Disappearance of EpoR from the cell surface in unstimulated cells followed a first-order kinetics showing an EpoR half-life of 3 hours ± 14 minutes (mean ± SD of 3 independent experiments). In contrast, Epo stimulation led to the disappearance of more than 75% of cell surface EpoRs within 30 minutes, with kinetics that cannot be fitted by a simple first-order mechanism (Figure 1A). During Epo stimulation, ¹²⁵I-Epo transiently appeared inside the cell, with a maximum internalization level after 20 minutes of incubation (Figure 1B). The rates of disappearance from the cell surface and intracellular accumulation of Epo reflect the combination of internalization and recycling. Monensin has been shown to allow receptor internalization but to block the traffic of intracellular vesicles, leading to the inhibition of both routing to the lysosomes and recycling to the plasma membrane.²⁸  The initial rate of Epo accumulation inside the cell was not increased in monensin-treated cells, indicating that few, if any, Epo-EpoR complexes recycled back to the cell surface after internalization. After 30 minutes of incubation, intracellular radioactivity decreased in control cells, suggesting that Epo was degraded. In contrast, radioactivity did not decrease in monensin-treated cells, suggesting that Epo degradation could occur in lysosomes.

These results were confirmed by Western blot experiments using an antibody (C-20) that recognizes the last 20 amino acids of the intracellular domain of the EpoR. This antibody revealed 2 forms of EpoR in UT-7 cell extracts (bands 1 and 2 in Figure 2A). In Epo-starved cells, band 1 was most generally detected as a doublet with an apparent molecular mass of 66 kDa. After Epo stimulation, most of the lower band of the doublet shifted to the upper band (Figures 2C, 3B, and 6B), most likely because of the high level of EpoR phosphorylation induced by ligand binding.³⁰  Band 1 disappeared after endoglycosidase F treatment of immunoprecipitated EpoRs but not after digestion by endoglycosidase H (Endo H) (Figure 2A) and was sensitive to proteinase K treatment of intact cells (Figure 2A). These results show that band 1 corresponds to the mature, cell surface form of the EpoR. Band 2 (around 64 kDa) was sensitive to Endo H treatment and resistant to proteinase K treatment of intact cells. It represents the maturing, endoplasmic reticulum form of the EpoR. Band 3 (around 62 kDa) was only detected after endoglycosidase treatment and probably corresponds to the unglycosylated form of the EpoR. In resting UT-7 cells, the half-life of the maturing form (band 2) was 15 minutes and the half-life of the mature form (band 1) was around 3 hours (Figure 2B). This latter value is in perfect agreement with that obtained from the experiments depicted in Figure 1A. In Epo-stimulated cells, the half-life of the maturing form did not significantly change (Figure 2C), whereas that of the mature form strongly decreased, confirming the results obtained in ¹²⁵I-Epo binding experiments. Thus, most of the mature EpoRs are quickly degraded during Epo stimulation. Cell surface expression of the EpoR has been shown to be controlled by Jak2 association.⁴  The intracellular pool of Jak2 was not significantly modified by cycloheximide during the incubation in the presence or absence of Epo, showing that the half-life of Jak2 was much longer than that of the EpoR and that disappearance of the mature form of the EpoR was not due to the lack of Jak2.

Figure 1B shows that ¹²⁵I-Epo was quickly internalized after EpoR binding. However, Epo dissociation from its receptor is also very rapid at 37°C. In a 2-step protocol in which Epo is bound at 4°C and internalization triggered by a shift to 37°C after removal of unbound ligand, most ¹²⁵I-Epo dissociates from the cell surface before internalization (data not shown). Thus, to test for Epo degradation after internalization, UT-7 cells were first incubated for 30 minutes at 37°C with iodinated Epo to accumulate radiolabeled hormone inside the cells and washed to remove free Epo. The cells were resuspended in fresh medium, and the appearance of Epo degradation products in the culture medium was tested by trichloroacetic acid (TCA) precipitation. As shown in Figure 3A, most internalized ¹²⁵I-Epo was degraded after 1 hour of incubation in control cells. We then tested the effect of methylamine, an inhibitor of lysosomal degradation,³¹  and lactacystin, a specific proteasome inhibitor. None of these inhibitors significantly modified ¹²⁵I-Epo internalization after 30 minutes of Epo stimulation (legend to Figure 3A). ¹²⁵I-Epo degradation was inhibited by methylamine but not by lactacystin. Therefore, we conclude that internalized Epo was degraded by the lysosomes and that proteasomes are not involved in Epo degradation.

In striking contrast with Epo, the EpoR was only partially protected by methylamine, as assessed by Western blot experiments using the C-20 antibody (Figure 3B-C). Moreover, as we reported previously,¹²  blocking proteasome activity with lactacystin only slightly protected EpoR from degradation. A strong inhibition of EpoR degradation was observed only when both lysosome and proteasome activities were blocked by a combination of methylamine and lactacystin or by LLnL, which inhibits proteasome and lysosome proteases such as cathepsin B. These results suggested that EpoR could be degraded by both lysosome proteases and the proteasomes and that each degradation process could be independent of the other. However, because our blotting antibody (C-20) recognizes the last 20 amino acids of the receptor, we could not exclude that only the C-terminal part of the receptor was degraded by the proteasomes. To test this hypothesis, we tested how ¹²⁵I-Epo remained associated with the EpoR when Epo degradation was inhibited by methylamine. To do this, UT-7 cells were incubated for 90 minutes with ¹²⁵I-Epo in the presence of methylamine. At this time point, most specifically bound ¹²⁵I-Epo was internalized and protected from degradation by methylamine (Figure 3A). As a control, cells were incubated for 8 minutes with ¹²⁵I-Epo. At this time point, no significant degradation of Epo (Figure 1B) or EpoR (Figure 2C) was detectable. The cells were then solubilized, and ¹²⁵I-Epo-EpoR complexes were immunoprecipitated using C-236 anti-EpoR antibodies. Similar proportions of radioactivity were immunoprecipitated from cells incubated for 8 minutes with Epo alone (51% ± 11%) and from cells incubated for 90 minutes with Epo and methylamine (48% ± 3%). These results show that most ¹²⁵I-Epo molecules that were protected from degradation by lysosome inhibition remained associated with the receptors, strongly suggesting that only the C-terminal part of the EpoR carrying the epitope recognized in Western blot experiments was degraded in methylamine-treated cells.

Because the simultaneous inhibition of proteasomes and lysosomes fully protected EpoR from degradation (Figure 3B), whereas lysosome inhibitors alone protected the N-terminal part of the EpoR only, we hypothesized that the C-terminal part of the EpoR was degraded by the proteasomes. None of the available antibodies recognizing the extracellular domain of the EpoR are able to detect endogenous human EpoRs in Western blots. To overcome this problem and to probe whether the C-terminal part of the EpoR was indeed degraded, we used chemical cross-linking to covalently link ¹²⁵I-Epo to its receptor. UT-7 cells were incubated for 30 minutes with ¹²⁵I-Epo, washed to remove unbound radioactivity, and solubilized. Solubilized extracts were then cross-linked using BS3. In a first step, EpoRs were immunoprecipitated and precipitates were denatured by boiling in SDS. Precipitates were resolubilized and first submitted to control immunoprecipitation using anti-GST antibodies and then to an immunoprecipitation using C-236 anti-EpoR antibodies. Lastly, unprecipitated ¹²⁵I-Epo–cross-linked proteins were recovered by immunoprecipitation with an anti-Epo antibody. Several protein bands were detected by autoradiography after the first anti-EpoR immunoprecipitation (Figure 4, lane 1). None of these bands were reprecipitated by a nonspecific antibody (anti-GST antibody; Figure 4, lane 2). Only bands A and B corresponding to the association of ¹²⁵I-Epo (34 kDa) with 71 and 40 kDa proteins were reprecipitated by C-236 anti-EpoR antibody (Figure 4, lane 3) and thus corresponded to the receptor itself. The other bands were recovered after anti-Epo precipitation. The highly radioactive band with an apparent molecular mass of 34 kDa corresponded to non–cross-linked ¹²⁵I-Epo, whereas the other bands should correspond to ¹²⁵I-Epo cross-linked to EpoR-associated proteins that remain to be identified. When C-20 anti-EpoR antibody was substituted for C-236 in the second immunoprecipitation, band A but not band B was detected (data not shown). These experiments strongly suggest that band A corresponded to the full-length EpoR, whereas band B could be a truncated form of the EpoR. We then tested the effects of protease inhibitors on the accumulation of these proteins (Figure 4, right panel). When cells were pretreated with methylamine, band B intensity significantly increased with a simultaneous decrease of band A. In contrast, when cells were incubated with LLnL, an inhibitor of both proteasome and lysosome activity, no band B was detected, whereas band A was strongly increased. Lastly, a pulse-chase experiment using unlabeled Epo indicated that the intensity ratio of band B–band A increased during incubation time (data not shown). These results strongly suggest that band B could be a degradation product of the EpoR produced by proteasome action. From the apparent molecular mass observed in crosslinking experiments (40 kDa after subtracting the 34 kDa of Epo), we estimate that more than half of the intracellular domain was removed from the EpoR by proteasomal degradation.

The involvement of proteasomes in EpoR degradation led us to test for the ubiquitination of the receptors. To this end, UT-7 cells were incubated for 105 minutes with LLnL, and Epo was added to cell samples 10, 30, 60, and 90 minutes before the end of incubation with LLnL. Control cells were incubated with LLnL alone for 105 minutes. Whole cell lysates were analyzed by Western blots using anti-EpoR antibodies (Figure 5A). Immunoprecipitates of EpoRs were also analyzed by Western blots using antiubiquitin antibodies. Smears in the high molecular mass region of the gels were observed in immunoblots of anti-EpoRs from whole cell lysates of Epo-stimulated cells. These smears were also recognized by antiubiquitin antibodies in immunoprecipitates of anti-EpoRs. These results show that the EpoRs are polyubiquitinated in Epo-stimulated cells. Sequential immunoprecipitations using anti-EpoR and antiphosphotyrosine antibodies clearly showed that the tyrosine-phosphorylated EpoRs were ubiquitinated (Figure 5B). These experiments were performed in the presence of LLnL to inhibit the degradation of ubiquinated proteins. In the absence of LLnL, EpoR ubiquitination was also detectable in Epo-stimulated cells, although very transiently (data not shown).

To examine whether stimulated EpoRs were ubiquinated at the cell surface, Epo was bound to UT-7 cells, and the cells were incubated with anti-Epo antibodies before solubilization. Immune complexes were then analyzed by Western blot using anti-EpoR antibodies (Figure 5C). Solubilized extracts from Epo-stimulated and unstimulated cells were also immunoprecipitated using C-236 anti-EpoR antibodies as controls. The maturing form of the EpoRs (Figure 5C, band 2) was observed in EpoR precipitates but not in Epo precipitates, showing that no Epo exchange occurred during solubilization or antibody recovery. These results clearly show that the cell surface, Epo-bound EpoRs were ubiquitinated. Methyl β cyclodextrine (MBCD) has been shown to inhibit cell surface protein internalization.^32,33  Accordingly, ¹²⁵I-Epo did not internalize and remained at the cell surface of UT-7 cells preincubated with 5 mM MBCD (data not shown). Cell surface blockage of the EpoR by MBCD did not prevent its ubiquitination or its tyrosine phosphorylation (data not shown) or the degradation of its C terminus (Figure 5D-E). Lactacystin inhibited the degradation of EpoR in MBCD-treated cells. These results show that both ubiquitination and proteasomal degradation of the EpoR occurred at the cell surface.

To test whether Jak2 activation was required for EpoR ubiquitination, we used the tyrphostin AG490, which has been shown to specifically inhibit Jak2 activity.³⁴ Figure 6A shows that blocking Jak2 activity nearly completely abolished Epo-induced EpoR ubiquitination. AG490 also strongly inhibited the Epo-induced degradation of the mature form of EpoR (Figure 6B) and totally inhibited the appearance of the 40 kDa EpoR in cross-linking experiments (data not shown). Intracellular ¹²⁵IEpo accumulation was slowed in AG490-treated cells. However, monensin increased ¹²⁵I-Epo accumulation in AG490-treated cells to values similar to those observed in control cells (Figure 6C). Because monensin inhibits the recycling of internalized hormone-receptor complexes but does not increase the internalization rate (Figure 1B), the apparent slowdown of ¹²⁵I-Epo internalization was most likely due to some recycling of the complexes to the cell surface. The results presented in Figure 6C also suggested that ¹²⁵I-Epo was not degraded in AG490-treated cells; this was confirmed by TCA precipitation experiments: no TCA-soluble radioactivity was detected in AG490-treated cells incubated up to 2 hours with ¹²⁵I-Epo (data not shown). Overall, these results strongly suggest that Epo-EpoR complexes were efficiently internalized when Jak2 activity was blocked but that internalized complexes were not routed to the lysosomes and that a part of them recycled to the plasma membrane.

Activation of growth factor receptors by their cognate ligands is followed by desensitization processes that largely contribute to control both amplitude and duration of intracellular signaling. Both parameters have been shown to play key roles in cellular responses to extracellular signals,^35,36  and their deregulation is associated with severe human pathologies including cancer.^37-40  In contrast to tyrosine kinase receptors such as the epidermal growth factor (EGF) receptor whose desensitization mechanisms have been extensively studied (recently reviewed by Shtiegman and Yarden⁴¹ ), down-regulation mechanisms used by cytokine receptors are poorly understood. In the present study, we have studied the internalization and degradation of the EpoR in erythroid cells. For these experiments, we have used the Epo-dependent UT-7 human cell line that constitutively expresses the EpoR. Intracellular signaling activated by Epo has been shown to be very similar in UT-7 cells and in normal erythroid progenitors that are the true target cells of Epo.⁴²  However, internalization studies require numbers of cells difficult to obtain with normal progenitors. Because UT-7 cells express endogenous EpoRs, they constitute a valuable model for EpoR trafficking studies because artifacts due to overexpression of transfected receptors are not supposed to occur. Accordingly, our results show that more than half of the EpoRs in resting UT-7 cells show a complex glycosylation pattern and are present at the cell surface, whereas it has been reported that most EpoRs accumulate inside EpoR-transfected cells.^2,3  Therefore, the artifactual presence of a large intracellular pool of EpoR in transfected cells that is prone to interfere with trafficking studies is avoided by the use of cell lines expressing endogenous receptors.

In the present paper, we show that Epo binding to its receptor induces ubiquitination, internalization, and degradation of the EpoR. Figure 7 summarizes our results and describes the down-regulation mechanisms that are turned on by Epo binding to its receptor. In sharp contrast with the GH receptor (GHR), for which internalization is constitutive¹⁶  (and GHR degradation is only slightly increased by GH),⁴³  internalization and degradation of the EpoR are strongly increased by ligand binding. It has been shown previously that EpoRs unable to bind and activate Jak2 were efficiently internalized after Epo binding, suggesting that Jak2 activation was dispensable for EpoR internalization.¹¹  In agreement with these results, we observed that blocking Jak2 activity did not inhibit Epo-induced EpoR internalization. Because EpoR ubiquitination was inhibited when Jak2 activity was blocked, we conclude that EpoR ubiquitination is not required for internalization. Overall, these results suggest that Epo binding to its receptor induces a conformational change of the EpoR that is sufficient to promote EpoR internalization.

In contrast to internalization, Epo and EpoR degradation is dependent on Jak2 activity, and when Jak2 activity was blocked, some Epo-EpoR complexes recycled back to the cell surface. Lysosome inhibitors such as methylamine (Figure 3), ammonium chloride, or chloroquine (data not shown) protected internalized Epo from degradation, indicating that Epo is degraded in the lysosomes. The cathepsin B–specific inhibitor⁴⁴  CA-074-Me largely inhibited ¹²⁵I-Epo degradation, showing that cathepsin B is involved in Epo degradation (data not shown). In contrast, blocking proteasome activity using lactacystin did not inhibit Epo degradation.

The EpoRs are rapidly ubiquitinated after Epo binding. This process occurs at the cell surface and requires Jak2 activation. The E3 ligase proteins responsible for EpoR ubiquitination are not identified. We have observed that c-Cbl associates with the tyrosine-phosphorylated form of the EpoR (F.V., unpublished observation, December 2003). However, overexpression in UT-7 cells of a mutated form of c-Cbl devoid of E3 ligase activity or blocking c-Cbl synthesis using small interfering RNA (siRNA) did not significantly decrease Epo-induced ubiquitination of the EpoR (data not shown). SOCS-3 also associates with the activated EpoR¹⁰  and could belong to a ubiquitination complex. However, the kinetics of SOCS-3 induction after Epo stimulation is poorly compatible with the rapid ubiquitination of the activated EpoR. Moreover, overexpression in UT-7 cells of a SOCS-3 mutant deleted of the SOCS domain did not decrease Epo-induced ubiquitination of the EpoR (data not shown), confirming that SOCS-3 was probably not involved in EpoR ubiquitination. Recently, the ubiquitin ligase Rul (p33) has been shown to associate to the EpoR and to induce its ubiquitination.⁴⁵  However, at least in COS cells, p33-induced ubiquitination of the EpoR is constitutive and not regulated by Epo stimulation of the cells. Whether Rul mediates Epo-induced ubiquitination of the EpoR in Epo-responsive cells has to be determined.

Ubiquitination contributes to the control of several cellular processes including targeting proteins for proteasomal degradation, internalization, and sorting of internalized proteins to late endosomes and lysosomes. Polyubiquitination most often leads to protein degradation by the proteasome, whereas internalization and intracellular routing of proteins generally involve monoubiquitination.⁴⁶  Polyubiquitination of the EpoR is probably responsible for the proteolysis of the EpoRs by proteasomes (Figure 4) that removes a significant part of the intracellular domain. A similar mechanism of degradation has been reported for the GHR⁴⁷  as well as the shared β subunit (β_c) of the IL-3, IL-5, and granulocyte-macrophage colony-stimulating factor (GM-CSF) receptors.⁴⁸  Agonistic ligation of these receptors induces proteasomal degradation of the intracellular domain of β_c, leading to a signalization-deficient receptor form containing only 25 to 35 intracellular amino acids. Although the extent of EpoR degradation by proteasomes cannot be accurately deduced from our cross-linking experiments, our results suggest that more than half of the intracellular domain is degraded by these proteases. Because tyrosine residues are located in this region of the receptor, this mechanism could hasten the desensitization process. In contrast to β_c, whose proteasomal degradation of the intracellular domain is necessary for internalization and lysosome routing,⁴⁸  inhibiting proteasome activity did not significantly modify the degradation of Epo and EpoR by the lysosomes, showing that initial degradation of this part of the EpoR is not required for lysosome targeting and degradation. Epo and EpoR targeting to the lysosomes and their degradation correlate with EpoR ubiquitination, strongly suggesting that EpoRs behave similarly to several receptors, such as EGF receptors (EGFRs),⁴⁹  interferon-α receptors,⁵⁰  or IL-2 receptors,²¹  whose ubiquitination controls targeting to the lysosome.

After internalization, both Epo and the EpoR are degraded by the lysosomes; and few, if any, EpoRs are recycled back to the membrane. This is in sharp contrast with the TPO receptor that promptly recycles onto the cell surface.¹⁴  These properties could at least partly explain why intracellular transduction by the TPO receptor is sustained for several hours,¹³  whereas Epo signaling is of short duration.¹²  Prolonged activation of TPO signaling has been shown to be required for megakaryocytic differentiation by maintaining a high level of Erk activation.¹³  In contrast, it is suggested that erythroid differentiation could require a decreased level of intracellular signaling. Indeed, it has been reported that activation of Foxo3A ⁵¹  and caspases⁵²  was necessary for terminal erythroid differentiation. These activations are inhibited by intracellular signaling relays such as the phosphatidylinositol-3 kinase pathway, which is strongly activated in Epo-stimulated cells and is required for erythroid cell proliferation.⁴²  Thus, we can propose that decreasing Epo receptor signaling could lower cell proliferation and facilitate terminal differentiation. Because the cell surface expression of newly synthesized EpoRs is inhibited by a mechanism involving proteasomes in stimulated cells (our previous results: Verdier et al¹² ), both this mechanism and the degradation of activated receptors should lead to the progressive disappearance of cell surface receptors; and, accordingly, reticulocytes are devoid of EpoRs.⁵³ 

We are grateful to Dr Véronique Duprez and Dominique Duménil (Cochin Institute, Paris, France) for critical reading of the manuscript and helpful comments and Odile Muller for technical expertise.