Erythropoietin receptors associate with a ubiquitin ligase, p33RUL, and require its activity for erythropoietin-induced proliferation.

The proliferation and survival of hematopoietic cells is strictly regulated by cytokine growth factors that act through receptors of the Type I cytokine receptor family, including erythropoietin (Epo) and its receptor, EpoR. Mitogenic signaling by these receptors depends on activation of Jak tyrosine kinases. However, other required components of this pathway have not been fully identified. In a screen for proteins that interact with EpoR and Jak2, we identified a novel member of the U-box family of ubiquitin ligases. This receptor-associated ubiquitin ligase, RUL, co-precipitated with EpoR from mammalian cells and mediated ubiquitination of EpoR. Also, endogenously expressed RUL was rapidly and transiently phosphorylated on serine after cytokine treatment of factor-dependent hematopoietic cells. Expression of ubiquitin ligase-deficient mutants of RUL inhibited Epo-induced expression of c-myc and bcl-2, two immediate-early genes normally associated with Epo-induced cell growth. Consistent with that finding, expression of mutant RUL also inhibited Epo-dependent proliferation and survival of factor-dependent cells. Together, these observations suggest that RUL is a required component of mitogenic signaling by EpoR. We also show that RUL is phosphorylated in response to growth factors that act through non-cytokine receptors, suggesting that RUL may function as a common regulator of mitogenesis.

The proliferation and survival of hematopoietic cell lineages is strictly governed by cytokine growth factors that utilize structurally related receptors of the Type I cytokine receptor family. The hematopoietic lineage in which a specific cytokine acts is determined by the expression pattern of its receptor. For example, erythropoietin (Epo), 1 acting through its receptor, EpoR, is largely restricted to supporting the growth of developing erythroid cells, whereas IL-3 promotes the growth of a broad range of immature myeloid cell lineages expressing its receptor. Cells of these lineages exhibit an absolute dependence on their corresponding growth factors and undergo rapid G 1 phase cell cycle arrest and apoptotic death in the absence of that growth factor. Thus, these cytokine receptors regulate signaling pathways that directly promote cell cycle progression and inhibit apoptosis.
Type I cytokine receptors have universally been shown to mediate their activities through association with, and activation of, Jak family tyrosine kinases (1,2). EpoR activates Jak2 through an interaction with its cytoplasmic domain proximal to the plasma membrane (3,4). All Epo-dependent responses, including proliferation, are abrogated by disrupting Jak2 activation (4,5). EpoR can also activate other well characterized signaling intermediates, including components of the Ras pathway, Stat transcription factors, and PI 3-kinase (1). In several instances, functional outcomes have been assigned to activation of these signaling pathways. For example, Epo-induced activation of Stat5 has been reported to enhance expression of the anti-apoptotic gene, bcl-X L , and thereby enhance survival of erythroid cells (6,7). Activation of PI 3-kinase is required to override checkpoints in cell cycle progression, and, as such, it promotes proliferation of cells suffering various forms of stress (8,9). Despite these identified functions, however, EpoR mutants lacking domains required for Stat5, Ras, and PI 3-kinase activation continue to promote survival and proliferation of factor-dependent hematopoietic cells (10,11). Similar correlations have been derived from analysis of other cytokine receptors, including the receptor for IL-3 (12)(13)(14). Thus, Jak activation represents the only identified cytokine receptor-induced activity that is absolutely required to promote survival and proliferation.
Although the signaling intermediates downstream of Jak2 in Epo-induced growth and survival signals remain to be identified, several immediate-early gene targets have been linked to this signaling pathway. For instance, the membrane-proximal cytoplasmic domain of EpoR required for Jak2 activation and cell growth also mediates induced transcription of the immediate-early genes, c-myc and pim-1 (10). This is consistent with the previous observation that forced expression of both c-myc and pim-1 cooperate to render cells factor-independent for growth (15). Thus, transcriptional activation of c-myc and pim-1 are relevant targets of a growth signal mediated by EpoR. In keeping with its ability to inhibit p53-dependent apoptosis, Epo induces expression of anti-apoptotic genes, bcl-2 and bcl-X L , through the membrane-proximal domain of EpoR (16,17). Thus, bcl-2 and bcl-X L are likely transcriptional tar-gets of the survival signal initiated by a minimally active EpoR.
Signaling pathways activated by Type I cytokine receptors are subject to negative regulation by several mechanisms, including inhibition of Jak activity by phosphatases, and inhibited activation of Stats by PIAS and SOCS proteins (18). Cytokine signaling may also be influenced by protein turnover mediated by proteasome-dependent degradation of receptors, Jaks, and Stats (19 -22). Targeted degradation of these proteins can, in part, be mediated through recruitment of ubiquitin conjugating complexes associated with the SOCS-box of SOCS proteins (23).
Ubiquitination of cellular proteins is mediated by a multienzyme system (24,25). The first step is the formation of a thiol ester bond between the carboxyl terminus of ubiquitin and a ubiquitin-activating enzyme (E1). This is an ATP-dependent reaction resulting in release of AMP and PP i . The activated ubiquitin is then transferred to a ubiquitin-conjugating enzyme (E2). Ubiquitin ligase (E3) then mediates transfer of the activated ubiquitin from the E2 to lysine residues on the target protein. In this process, the ubiquitin ligase is principally responsible for selecting the protein target for ubiquitination. In mammalian cells, the identified components of ubiquitination cascades include a single E1, several E2s, and a large and growing number of E3s. Enzymes contributing ubiquitin ligase activity now comprise several protein families defined by retention of conserved domains, including HECT (homologous to EG-AP carboxyl terminus) domains, RING fingers, and the recently identified U-box (26).
The conjugation of ubiquitin to cellular proteins regulates a broad range of cellular functions (25, 27). Among the best characterized of these is the proteasome-dependent degradation of polyubiquitinated proteins. This result is associated with the formation of polyubiquitin chains linked through lysine 48 of the ubiquitin subunits. For example, EGF receptor internalization and degradation is regulated by polyubiquitination mediated by the RING finger protein, c-Cbl (28,29). Alternatively, polyubiquitin chains linked through lysine 63 of the ubiquitin subunits have been associated with modification of protein function rather than degradation. For example, NF-B activation by tumor necrosis factor-␣ receptors is mediated through the RING finger protein, Traf6. This is initiated through a cascade in which Traf6 is polyubiquitinated with chains linked through lysine 63 (30). Polyubiquitinated Traf6 then mediates activation of the serine kinase, Tak1, which then phosphorylates and activates the IB kinase (31). Phosphorylated IB is then targeted for lysine 48-linked polyubiquitination and subsequent proteolytic degradation, allowing the activation of the NF-B transcription factor. Thus, ubiquitination of receptors and signaling proteins can exert either positive or negative influence on the activities of receptor-activated signaling pathways.
In the present study, we identified a novel receptor-associated ubiquitin ligase, p33 RUL . We show that RUL interacts with EpoR, mediates ubiquitination of EpoR, and is inducibly phosphorylated on serine in response to Epo treatment. Expression of RUL mutants that lack ubiquitin ligase activity specifically inhibits Epo-induced expression of c-myc and bcl-2 and inhibits Epo-dependent proliferation and survival of factordependent hematopoietic cells. We also show that RUL is phosphorylated in response to growth factors acting through noncytokine receptors. Together, these observations suggest that RUL is a required component of Epo-dependent signaling and may function as a common regulator of mitogenic signaling pathways.

EXPERIMENTAL PROCEDURES
Cloning of p33 RUL -A bait construct for a yeast-3-hybrid screen was prepared consisting of the cytoplasmic domain of wild-type EpoR (32) fused to the DNA binding domain of Gal4 in the pBridge expression vector (Clontech). A wild-type Jak2 cDNA (33) was cloned into the second expression cassette of this vector, which is under control of the P Met25 promoter that is repressed in the presence of methionine. Yeast cells (strain Y190) harboring the EpoR⅐Jak2 bait plasmid were used to screen 1.2 ϫ 10 7 clones from a MATCHMAKER human bone marrow cDNA library (Clontech) for growth on media lacking leucine, tryptophan, histidine, and methionine. Prey plasmids from positive colonies were isolated and screened for growth on deficient media when retransformed into fresh yeast with the EpoR⅐Jak2 bait plasmid but not without the bait plasmid. To assess the effect of Jak2 expression on interactions, positive colonies were grown on media lacking leucine and tryptophan with or without methionine. Colonies were transferred to paper filters, permeabilized by freezing, and assayed for expression of the ␤-galactosidase reporter gene product by formation of blue color when incubated with 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal). A positive clone (p33 RUL ) identified through this procedure was sequenced and found to contain a complete open reading frame identical to human NOSIP (GenBank TM accession number 14424549).
cDNA Expression Constructs-Wild-type and mutant cDNA constructs were prepared by polymerase chain reaction, as previously described (10). EpoR⅐HA was prepared by inserting an HA tag within the carboxyl-terminal end of the EpoR cDNA (32) open reading frame in the expression vector, pXM. Mutant forms of the RUL cDNA were prepared that converted codon 49 from serine to alanine (RUL-S49A), or codon 186 from proline to alanine (RUL-P186A). An amino-terminal truncation of RUL (RUL-NT) was prepared containing a translation start site at codon 122 of the RUL open reading frame. Wild-type and mutant RUL cDNAs were cloned in-frame with the GST tag in the plasmid pACGHLT (BD Pharmingen) for preparation of baculovirus expression vectors, as previously described (12). For expression in mammalian cells, RUL cDNAs were cloned into pXM or the ponasteroneinducible expression vector, pIND (Invitrogen). An epitope tag was incorporated into RUL-pIND constructs to allow expressed proteins to be distinguished from endogenous RUL. A murine ubiquitin cDNA was obtained by reverse transcription-PCR from myeloid DA3 cells. This ubiquitin cDNA was prepared with a myc epitope tag at the aminoterminal end of the open reading frame and was cloned into the pXM expression vector.
Cell Lines and Culture Conditions-32D cells expressing EpoR (8) and NFS-60 cells (34) have been described, and both were maintained in RPMI 1640 media plus 10% fetal bovine serum (FBS) and recombinant murine IL-3 (70 pg/ml). Human TF-1 cells (ATCC) were maintained in RPMI medium plus 10% FBS and human granulocyte-macrophage/colony-stimulating factor (5 ng/ml). COS7 cells were maintained in Dulbecco's modified Eagle's medium plus 10% FBS. Transient transfection of COS7 cells was achieved using LipofectAMINE 2000 (Invitrogen) as directed by the manufacturer. Stable clones expressing EpoR⅐HA or RUL constructs were prepared by electroporation of plasmids into 32D cells. 32D-EpoR⅐HA cells were selected for their ability to grow in 5 units/ml Epo and subcloned by dilution. pIND-RUL plasmids were co-transfected with pVgRXR (Invitrogen). Transfected cells were selected in G418 (1 mg/ml) plus Zeocin (1 mg/ml) and were subcloned by dilution. Clonal cell lines were screened by Western blotting for expression of epitope-tagged RUL proteins when cultured in 6 M ponasterone for 24 h. Growth analyses were performed on at least three different clones expressing RUL-wt or RUL-NT and one RUL-S49A-expressing clone. It is noteworthy that establishing clones expressing RUL-S49A has been extremely difficult, presumably due to the potent growth inhibitory effect of this construct in combination with low level basal expression from the ponasterone-inducible system.
Growth Factor Stimulations-For stimulation experiments, cells were washed free of cytokine and were cultured in media plus 10% FBS for 4 h prior to stimulation with Epo (10 units/ml), IL-3 (2 ng/ml), G-CSF (10 ng/ml), or HGF (200 ng/ml). In experiments assaying RUL phosphorylation, cells were cultured in phosphate-free RPMI 1640 plus 10% FBS and [ 32 P]phosphorous (0.5 mCi/ml) for 2 h prior to growth factor treatment. Cell lines harboring inducible RUL expression constructs were treated with or without 6 M ponasterone for 20 h prior to stimulation with growth factors.
In Vitro Ubiquitination-To prepare proteins for analysis in ubiquitination reactions, insect Sf9 cells were infected with baculovirus encoding GST-RUL fusion proteins containing a thrombin-cleavable linker. Infected cells were lysed in NG buffer (0.5% Nonidet P-40, 10% p33 RUL , a Novel Receptor-associated Ubiquitin Ligase glycerol, 50 mM Tris, pH 8.0, 0.1 mM EDTA, 150 mM NaCl, 1 mM dithiothreitol, phosphatase inhibitors, and protease inhibitors), and GST fusion proteins were precipitated with glutathione-Sepharose (Amersham Biosciences). Precipitated proteins were washed and eluted by digestion with thrombin (50 units/ml; Sigma) in phosphate-buffered saline for 16 h at 4°C. Eluted fractions were assayed for protein content using the BCA protein assay (Pierce), and the presence of equivalent amounts of full-length wild-type or mutant RUL in purified preparations was confirmed by SDS-PAGE and Western blotting.
Ubiquitination reactions were performed with 15 g of purified wildtype or mutant RUL in a buffer containing 40 mM HEPES (pH 7.5), 5 mM MgCl 2 , 1 mM dithiothreitol, 10 mM ATP, protease inhibitors, and 0.5 g/l GST-ubiquitin (Calbiochem). The reactions were initiated by adding 5 l of fraction II rabbit reticulocyte lysate (Calbiochem) that had been pre-treated with 5 M ubiquitin-aldehyde. Reactions were incubated for 2 h at 37°C, and products were precipitated with glutathione-Sepharose. Precipitated proteins were separated by SDS-PAGE and Western blotted with anti-RUL antibodies.
Immunoprecipitation and Western Blotting-After growth factor stimulation, cells were lysed in NG buffer plus 60 mM octyl-␤-D-glucopyranoside and N-ethylmaleimide. Cell lysates (3 ϫ 10 7 cell equivalents) were immunoprecipitated with specific antibodies plus Protein A-Sepharose (Amersham Biosciences). Where indicated, the p33 RULimmunizing peptide or irrelevant peptide was added to cell lysates (100 g/ml) during immunoprecipitations. Precipitated proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Osmonics). Alternatively, cell lysates (5 ϫ 10 5 cell equivalents) were directly separated by SDS-PAGE. Membranes were probed with specific antibodies and visualized by chemiluminescence (Lumi-Light, Roche Applied Science) as directed by the manufacturer.
RNA Isolation and Northern Blotting-Total RNA was prepared by TRIzol (Invitrogen) extraction from cells stimulated with or without Epo (10 units/ml) for 2 h. Total RNA (20 g) was separated on agarose gels and transferred to Hybond Nϩ membranes (Amersham Biosciences). Membranes were screened with 32 P-labeled cDNA probes for c-myc, pim-1, bcl-2, bcl-X L , ␤-actin, and GAPDH, as previously described (10,17). For each transcript, mRNA levels were measured on a PhosphorImager (Amersham Biosciences) and were normalized to GAPDH or ␤-actin levels as a control for loading. The percent change in Epo-induced levels between cells treated with or without ponasterone was calculated. The calculated change for cells expressing RUL-wt or RUL-NT were averaged from four experiments and compared using a two-tailed Student's t test. Significant differences were indicated by p values Յ 0.05.
siRNA Reagents-siRNA targeting constructs were prepared using the Silencer siRNA construction kit (Ambion) according to the directions of the manufacturer. The RUL siRNA used in Fig. 4B is derived from nucleotides 94 -114 of the murine RUL cDNA and resulted in the most efficient reduction of endogenous RUL proteins in COS7 cells among four candidate siRNA constructs that were tested. A GAPDH siRNA was supplied with the Silencer siRNA kit and was used as a control for suppression of a gene unrelated to p33 RUL .

A Novel EpoR Interacting Protein Is a Member of the U-box
Family of Ubiquitin Ligases-To identify novel signaling intermediates interacting with EpoR, a yeast-3-hybrid experiment was performed with a bait construct containing the cytoplasmic domain of EpoR and a separate expression cassette encoding wild-type Jak2. Yeast cells harboring the EpoR⅐Jak2 bait plasmid were used to screen a human bone marrow cDNA library, and a single clone survived re-screening when transformed into fresh yeast harboring the bait construct. In yeast containing a ␤-galactosidase reporter, this clone mediated formation of blue colonies when co-expressed with the EpoR⅐Jak2 bait plasmid (not shown). Suppression of Jak2 expression in these colonies inhibited, but did not abolish, expression of the ␤-galactosidase reporter and formation of blue colonies. Thus, the protein product of the clone appeared to preferentially associate with the EpoR cytoplasmic domain in the presence of Jak2.
When sequenced, the cloned cDNA was found to contain a complete open reading frame predicted to encode a 301-amino acid protein with an expected size of 33 kDa (p33). The sequence was identical to a protein recently identified in a yeast-2-hybrid screen for eNOS-interacting proteins, NOSIP (35).
Although not previously identified, we found that the amino acid sequence of this protein contained significant homology to the U-box family of ubiquitin ligases (Fig. 1A). The U-box is a modified form of the RING Finger domain that lacks metal-

FIG. 1. p33 RUL is homologous to the U-box family of ubiquitin ligases.
A, the human p33 RUL U-box is aligned with the U-box consensus from the SMART data base and nine nearby neighbors assigned to the U-box family. The 104 amino acids separating the two halves of the human p33 U-box are indicated (ٙ). Identical (black background) and similar (shaded background) amino acids are indicated. Conserved amino acids reported to inactivate U-box domains when mutated (38,39) are indicated by an arrowhead. The locations of predicted ␤-sheets (arrows) and an ␣-helix (bar) hypothesized to form the U-box core (26) are shown below the alignment. B, the location of the split U-box and helical domain of p33 RUL are shown. Serine residues conserved between mouse and human p33 RUL are shown (asterisks).

p33 RUL , a Novel Receptor-associated Ubiquitin Ligase
chelating cysteines and histidines (26). Like the RING Finger, the U-box is thought to form a structure consisting of three ␤-sheets and a single ␣-helix, which would be stabilized by salt bridges instead of chelated metal ions. The p33 U-box appears to be an atypical member of this family, because it is split into two halves separated by 104 amino acids. These intervening amino acids reside within the predicted loop between the helix and the third sheet. Interestingly, secondary structure modeling programs (36) predict that the helix beginning within the p33 U-box extends through the first 40 of the 104 intervening residues (Fig. 1B). Otherwise, the p33 U-box shows very high homology to the consensus U-box sequence, with a significant expectation (E) value of 6 ϫ 10 Ϫ11 obtained from the RPS-BLAST search (37).
Because the U-box family has only recently been identified, the structural requirements of this functional domain are not fully defined. Thus, it is not clear how the 104-amino acid intervening domain of the p33 U-box may affect its potential function as a ubiquitin ligase. Importantly, conserved amino acids that were reported to inactivate U-box proteins when mutated (38,39) are retained in the p33 U-box (Fig. 1A). To confirm the identity of p33 as a ubiquitin ligase, the ability of p33 to ubiquitinate itself in an in vitro reaction was assessed. As specificity controls, p33 point mutations were prepared converting essential U-box residues serine 49 or proline 186 to alanine. Wild-type and mutant cDNAs were expressed as GST fusion proteins in insect Sf9 cells and purified by glutathione-Sepharose precipitation and thrombin cleavage of the GST linker. Equivalent amounts of purified proteins were subjected to ubiquitination reactions in which a rabbit reticulocyte lysate provided ubiquitin activating (E1) and conjugating (E2) activities, and a GST-ubiquitin fusion protein provided the only ubiquitin source. As shown in Fig. 2, ubiquitination of wildtype p33 was demonstrated by the appearance of high molecular weight p33 in the glutathione-Sepharose-precipitated re-action product. Moreover, the presence of p33 conjugates larger than the predicted 67-kDa size resulting from addition of GSTubiquitin (34 kDa) indicated that p33 became polyubiquitinated, either at multiple sites or through the formation of polyubiquitin chains. As would be expected for a ubiquitin ligase, p33 ubiquitination was ATP-dependent and required activities present in the reticulocyte lysate. In contrast to wildtype p33, neither p33-P186A nor p33-S49A were efficiently ubiquitinated. The dependence of p33 ubiquitination on the retention of a functional U-box demonstrates that p33 itself is responsible for the ubiquitinations observed in these reactions. Thus, these observations support the assignment of p33 as a functional ubiquitin ligase.
p33 Interacts with EpoR in Vivo-To verify the interaction between p33 and EpoR, co-precipitation experiments were performed from lysates of COS7 cells transiently transfected with wild-type p33 and an HA-tagged EpoR (EpoR⅐HA). As shown in Fig. 3A, EpoR was efficiently precipitated by p33-specific antibodies but not when competed by the p33 peptide to which the antiserum was raised. As expected, an irrelevant peptide failed to disrupt immunoprecipitation of the EpoR⅐p33 complex. To assess whether endogenous p33 interacts with receptors in factor-dependent hematopoietic cells, murine myeloid 32D cells stably expressing EpoR⅐HA were evaluated for co-precipitation with endogenously expressed p33. EpoR⅐HA was detected in p33 immunoprecipitations and co-migrated with the EpoR⅐HA complex that was immunoprecipitated by HA antibodies (Fig.  3B). The p33-precipitated EpoR⅐HA was also competed by the p33-immunizing peptide. Similarly, anti-HA immunoprecipitates from 32D-EpoR⅐HA cells contained significant amounts of p33 protein compared with HA immunoprecipitates from parental 32D cells. Interaction between p33 and EpoR was also assessed in human TF-1 cells, which express both proteins endogenously. As shown in Fig. 3C, EpoR was precipitated by p33-specific antiserum from TF-1 cell lysates but not in the presence of the p33-immunizing peptide. Interaction between p33 and EpoR appeared to be constitutive and did not require receptor activation, because co-precipitation was observed from lysates of unstimulated cells. However, in transfected COS7 cells, 32D-EpoR⅐HA cells, and TF-1 cells, Epo treatment reproducibly increased co-precipitation of EpoR and p33, suggesting that this interaction is enhanced by receptor activation. Based on the demonstrated receptor-association of this ubiquitin ligase, we refer to the cloned protein as p33 RUL .
p33 RUL Mediates Ubiquitination of EpoR-The association between EpoR and RUL suggests the possibility that EpoR may be targeted for ubiquitin ligation. To assess the potential for ubiquitination of EpoR, COS7 cells were transiently transfected with EpoR⅐HA and an myc-tagged ubiquitin. As shown in Fig. 4A, significant ubiquitination was detected in HA immunoprecipitates from cells co-expressing EpoR⅐HA and myc-ubiquitin, but not from cells expressing either construct alone. The high molecular weight of the ubiquitinated species in EpoR⅐HA precipitates suggests the presence of multiple ubiquitinations on each target. In repeated experiments, Epo treatment had no reproducible effect on the degree of EpoR⅐HA ubiquitination. Similarly, overexpression of RUL did not alter ubiquitination of EpoR⅐HA. The constitutive ubiquitination of transfected EpoR could be mediated by endogenous RUL expressed in COS7 cells. To test this possibility, COS7 cells were transfected with EpoR⅐HA plus RUL siRNA (Fig. 4B). Transfection with RUL siRNA resulted in reduced expression of endogenous RUL protein and significantly reduced the ubiquitination of transfected EpoR⅐HA. By comparison, transfection with GAPDH siRNA had little effect on ubiquitination of EpoR⅐HA. To further assess the role of RUL in EpoR ubiquitination, COS7 cells were FIG. 2. Ubiquitin ligase activity of p33 RUL . A, in vitro ubiquitination reactions were performed as described under "Experimental Procedures" with purified wild-type p33 RUL or point mutants predicted to inactivate the U-box domain (P186A and S49A). Reactions were performed in the presence (ϩ) or absence (Ϫ) of ATP, or rabbit reticulocyte lysate (Fr.II) as the source of E1 and E2 activity. Reaction products specific to p33 RUL activity are indicated (Ub). A nonspecific product is also indicated (NS). B, to confirm equivalent wild-type and mutant protein in each reaction, 300 ng of each purified protein was separated by SDS-PAGE and Western blotted with antibodies specific for p33 RUL . p33 RUL , a Novel Receptor-associated Ubiquitin Ligase also transfected with an amino-terminally truncated form of RUL (RUL-NT) lacking 122 amino acids, including the aminoterminal portion of the split U-box and the predicted helix (see Fig. 1). Overexpression of RUL-NT in COS7 cells significantly reduced the ubiquitination of transfected EpoR⅐HA as compared with cells expressing only endogenous RUL (Fig. 4B). Thus, RUL-NT appears to have dominant negative activity, consistent with the observation that RUL mutants lacking a functional U-box also lack ubiquitin ligase activity. Notably, the reduced degree of EpoR ubiquitination resulting from RUL-NT or RUL siRNA transfection did not result in any detectable stabilization of EpoR⅐HA protein. Similarly, treatment of transfected COS7 cells with proteasome inhibitors did not enhance EpoR⅐HA expression (not shown). Therefore, at least in COS7 cells, RUL-mediated ubiquitination of EpoR did not appear to result in degradation of this receptor.
The state of RUL ubiquitination was also assessed in transfected COS7 cells (Fig. 4C). Significant ubiquitinated protein was detected in RUL immunoprecipitates from COS7 cells overexpressing wild-type RUL. Ubiquitination of overexpressed RUL was not significantly enhanced by co-expression with EpoR or by Epo treatment of transfected cells. However, Epo treatment induced detectable ubiquitination in RUL immunoprecipitates from cells expressing only endogenous RUL plus transfected EpoR⅐HA. Thus, overexpressed RUL is constitutively ubiquitinated in COS7 cells and may mask inducible ubiquitinations that are observable with physiological levels of RUL expression. Whether more physiological levels of EpoR expression will similarly unmask Epo-induced ubiquitination has not been determined.
Expression of p33 RUL U-box Mutants Inhibits Epo-and IL-3-dependent Survival, Proliferation, and Expression of Immediate-early Genes-To test the hypothesis that the ubiquitin ligase activity of RUL contributes to growth of factor-dependent hematopoietic cells, we examined the effect of ubiquitin ligase-deficient RUL mutants (RUL-S49A and RUL-NT) on 32D cell growth. Mutant and wild-type RUL cDNAs were prepared with epitope tags and cloned into a ponasterone-inducible vector for stable expression in 32D-EpoR cells. Clonal cell lines expressing each of these constructs were obtained while being cultured in medium containing IL-3, and ponasteroneinducible expression of different RUL forms was confirmed by Western blotting (Fig. 5A). In 32D-EpoR cells expressing wildtype RUL, ponasterone treatment had no significant effect on either Epo-or IL-3-induced proliferation (Fig. 5B). In contrast, cells expressing RUL-NT or RUL-S49A grew significantly slower when cultured in ponasterone as compared with cells cultured in its absence. Ponasterone treatment also resulted in reduced viability of cells expressing RUL-NT (less than 50% viability), whereas RUL-S49A cultures lacked viable cells by 48 h after addition of ponasterone. By comparison, parental 32D-EpoR and RUL-wt cultures remained greater than 90% viable throughout the course of the growth assay. Even in the absence of ponasterone, RUL-NT and RUL-S49A clones grew slower than parental controls or other inducible clones. In fact, in the absence of selection (G418 and Zeocin) cultures with mutant RUL proteins rapidly lost expression of the inducible construct, suggesting that even low levels of uninduced expression inhibit growth (not shown).
To determine whether reduced viability in cells expressing mutant RUL was associated with apoptosis or necrosis, cells were stained with annexin V and propidium iodide. Ponasterone-induced expression of RUL-NT or RUL-S49A increased the apoptotic population in these cultures, as indicated by an increase in the propidium iodide-low, annexin V-high population (lower right quadrant, Fig. 5C). No significant apoptosis was observed in RUL-wt cultures in the presence or absence of ponasterone. The lack of significant effects associated with expression of wild-type RUL, as compared with the considerable inhibition of growth and viability mediated by RUL mutants lacking a functional U-box domain, are consistent with a requirement for RUL U-box function in mediating cytokine-dependent cell growth and survival.
The possibility that expression of RUL mutants might alter the stability of EpoR or associated proteins was assessed in 32D-EpoR cells expressing inducible RUL-NT (Fig. 6). Epo treatment of cells for up to 90 min had no observable effect on expression levels of EpoR, Jak2, or endogenous RUL. Similarly, ponasterone-induced expression of RUL-NT did not alter expression of these proteins prior to or following Epo treatment. This observation is consistent with the lack of effects on EpoR FIG. 3. p33 RUL co-precipitates with EpoR. A, COS7 cells were transfected with equal amounts of EpoR⅐HA-and p33 RUL -expressing plasmids. 36 h after transfection, cells were treated for 12 min with 10 units/ml Epo (ϩ) or left untreated (Ϫ). Cells were lysed and immunoprecipitated (IP) with p33 RUL antibodies in the presence (ϩ) or absence (Ϫ) of the p33 RUL -immunizing peptide or an irrelevant peptide. Precipitated proteins were Western blotted with antibodies specific for the HA-tagged EpoR⅐HA or p33 RUL . B and C, parental 32D cells and a clone stably expressing EpoR⅐HA (B) or TF-1 cells endogenously expressing EpoR and p33 RUL (C) were treated for 12 min with 10 units/ml Epo (ϩ) or left untreated (Ϫ). Cells were lysed, and parallel aliquots were immunoprecipitated (IP) with antibodies specific for p33 RUL in the presence (ϩ) or absence (Ϫ) of the p33 RUL -immunizing peptide. EpoR⅐HA was precipitated from 32D cell lysates with anti-HA antibodies. Precipitated proteins were Western blotted with antibodies specific for HA (EpoR⅐HA), wild-type human EpoR, or p33 RUL . p33 RUL , a Novel Receptor-associated Ubiquitin Ligase stability by RUL-mediated ubiquitination in COS7 cells (Fig.  4).
The effect of RUL activity on Epo-induced expression of immediate-early genes was also assayed in 32D-EpoR cells expressing inducible RUL-NT or RUL-wt. Epo-induced expressions of c-myc, pim-1, bcl-2, and bcl-X L were each modestly enhanced by ponasterone treatment in RUL-wt cells (Fig. 7, A  and B), although these differences were not statistically significant (p Ͼ 0.1). By comparison, expression of RUL-NT significantly inhibited Epo-induced expression of c-myc and bcl-2. These findings are consistent with the reduced growth and viability of RUL-NT cultures compared with parental and RUL-wt controls. Epo-induced bcl-X L expression was also reproducibly lower in RUL-NT cells as compared with RUL-wt cells; however, bcl-X L levels were not further reduced by ponasterone-dependent expression of RUL-NT. Thus, in contrast to specific inhibition of c-myc and bcl-2, induced expression of RUL-NT did not significantly inhibit Epo-dependent expression of bcl-X L . Likewise, Epo-induced pim-1 expression did not significantly differ between RUL-NT and RUL-wt cells. Thus, conditional expression of a dominant-negative RUL mutant selectively inhibited Epo-induced expression of some, but not all, immediate-early gene targets associated with survival and proliferation of factor-dependent hematopoietic cells.

Endogenous p33 RUL Is Transiently Phosphorylated following Cytokine Stimulation of Factor-dependent Hematopoietic Cells-Because RUL was identified based on interaction with
EpoR⅐Jak2 complexes, we attempted to demonstrate cytokinedependent tyrosine phosphorylation of endogenous RUL by immunoprecipitation and Western blotting with anti-phosphotyrosine antibodies. However, we found no circumstances under which endogenous RUL was detectably phosphorylated on tyrosine (not shown). Thus, we have found no evidence to support a role for direct Jak2 phosphorylation of RUL in hematopoietic cells. However, RUL immunoprecipitations performed from 32 P-labeled 32D cells stably expressing wild-type EpoR (32D-EpoR) demonstrated that 32 P was incorporated into endogenous RUL following stimulation with either Epo or IL-3 (Fig. 8A). Specifically, a major 32 P-labeled protein present in RUL immunoprecipitates from Epo-or IL-3-treated cells completely co-migrated with RUL protein visualized by Western blotting. Moreover, immunoprecipitation of the cytokine-induced phosphoprotein was efficiently competed by the RUL peptide to which the antiserum was raised. Little or no phosphorylated RUL was detectable in immunoprecipitates from untreated cells. To determine the site of cytokine-induced RUL phosphorylation, the Epo-induced 32 P-RUL band was excised and subjected to two-dimensional phosphoamino acid analysis. As shown in Fig. 8B, 32 P-RUL was phosphorylated on serine residues without any detectable phosphorylation at threonine or tyrosine. Identical results were obtained from a similar analysis of IL-3-induced 32 P-RUL (not shown).
Most receptor-activated phosphorylation events occur rapidly and are transient following continuous receptor activation. nM siRNA derived from p33 RUL or GAPDH. 24 h after transfection, cells were treated with or without Epo, lysed, and assayed for ubiquitination in anti-HA immunoprecipitates, as described for panel A. Additionally, total cellular lysates were directly separated by SDS-PAGE and Western blotted with antibodies specific for the HA-epitope, p33 RUL , or GAPDH. C, COS7 cells were transfected with myc-ubiquitin, plus EpoR⅐HA-and/or RUL-expressing plasmids. After treatment with or without Epo, cells were lysed and immunoprecipitated with p33 RULspecific antibodies. Ubiquitination of precipitated proteins was detected by Western blotting with anti-myc antibodies. Detection of the precipitating antibody (Ig) in the Western blot is indicated. Precipitated RUL protein was also detected by Western blotting with anti-p33 RUL antibodies.

FIG. 4. RUL mediates EpoR ubiquitination in COS7 cells and is
inducibly ubiquitinated after Epo treatment. COS7 cells were transfected with equal amounts of myc-tagged ubiquitin, EpoR⅐HA, or p33 RUL expressing plasmids, as indicated. 24 h after transfection, parallel cultures were treated with 10 units/ml Epo for 12 min (ϩ) or left untreated (Ϫ). Cells were lysed, and EpoR⅐HA was immunoprecipitated using agarose-coupled anti-HA antibodies. Ubiquitinated proteins present in immunoprecipitates were detected by Western blotting with anti-myc antibodies. Blots were also stripped and reprobed with anti-HA antibodies for detection of EpoR⅐HA. B, COS7 cells were transfected with myc-ubiquitin plus EpoR⅐HA alone (control) or were cotransfected with RUL-wt or RUL-NT expression plasmids, or with 10

p33 RUL , a Novel Receptor-associated Ubiquitin Ligase
The time course of Epo-induced RUL phosphorylation was assessed in 32 P-labeled 32D-EpoR cells (Fig. 9). Epo-induced phosphorylation of RUL was maximal 10 min after addition of cytokine. By 30 min of Epo stimulation, all detectable RUL phosphorylation was lost. A similar time course was observed for Epo-stimulated Stat5 tyrosine phosphorylation (lower panels, Fig. 9). Although anti-phosphotyrosine Western blotting detected Stat5 tyrosine phosphorylation 5 min after Epo treatment, maximal phosphorylation of Stat5 occurred between 10 and 15 min of stimulation, identical to the RUL time course. Thus, phosphorylation of RUL and Stat5 phosphorylation follow a nearly identical time course of transient induction after cytokine treatment.
To assess whether inhibited proliferation of 32D cells expressing mutant RUL is associated with defects in the phosphorylation of endogenous RUL, the effect of Epo stimulation was evaluated in these cells (Fig. 10). Epo treatment induced significant RUL phosphorylation in cells expressing RUL-wt, whereas inducible phosphorylation of endogenous RUL was barely detectable in RUL-NT-expressing clones. RUL-NT appeared to be weakly phosphorylated, but its phosphorylation was not induced by Epo treatment. These data suggest that RUL-NT has dominant-negative activity with regard to its ability to inhibit phosphorylation of endogenous RUL and are consistent with the ability of RUL-NT to inhibit endogenous RUL ubiquitin ligase activity in COS7 cells (see Fig. 4B). We have been unable to assess RUL phosphorylation in p33-S49Aexpressing cells, because inducible expression of this mutant is repeatedly lost during expansion of cultures to the quantities needed for phosphorylation assays (not shown).

p33 RUL Phosphorylation Is Induced by Growth Factors Acting through Cytokine and Non-cytokine Receptors-To begin
assessing the relevance of RUL to mitogenic signaling pathways regulated by other classes of growth factor receptors, RUL phosphorylation status was evaluated in factor-dependent NFS-60 cells. NFS-60 cells are unique in that they can be maintained in hepatocyte growth factor (HGF), G-CSF, or IL-3 (34). HGF functions through activation of the receptor tyrosine kinase, c-Met (40), whereas receptors for G-CSF and IL-3 are members of the same family of cytokine receptors as EpoR. Notably, RUL immunoprecipitated from 32 P-labeled NFS-60 cells was equivalently phosphorylated after treatment with IL-3, G-CSF, or HGF (Fig. 11). Specificity in the signaling pathways activated by these different growth factors is indicated by the induction of Stat5 phosphorylation only in response to IL-3 and G-CSF. By contrast, each of these growth factors induced robust phosphorylation of Erk. Thus, RUL phosphorylation may be a common target of growth factor receptors of differing families. DISCUSSION In this study, we identified a novel member of the U-box family of ubiquitin ligases as an EpoR-interacting protein. This receptor-associated ubiquitin ligase (RUL) interacts with EpoR in factor-dependent hematopoietic cells, mediates ubiquitinauninduced. At regular intervals, cultures were sampled, and viable cells were enumerated after staining with Trypan Blue. Cultures approaching saturation density (ϳ1 ϫ 10 6 cells/ml) were diluted 1/10 with the appropriate medium. C, the apoptotic state of cultures described in panel B was assessed from samples of Epo-treated RUL-wt and RUL-NT cells obtained 48 h after initiation of ponasterone treatment, whereas RUL-S49A cultures were sampled at 20 h. Collected cells were stained with propidium iodide plus annexin V and were analyzed by flow cytometry as previously described (55). The percentage of apoptotic cells (annexin V-high, propidium iodide-low) relative to the total population is indicated in the lower right quadrant. Each density plot represents at least 10,000 recorded events.  Recently, the human RUL cDNA (also called NOSIP) was isolated in a yeast-2-hybrid screen for proteins interacting with endothelial nitric-oxide synthase (eNOS) (35,41). Transient overexpression in Chinese hamster ovary cells was found to relocalize overexpressed eNOS from plasma membrane caveolae to the cytoplasm, and endogenous RUL could be co-precipitated with eNOS. However, it remains to be determined if there is a functional consequence to this interaction for endogenously expressed eNOS. RUL did not associate with other forms of NOS, suggesting that this potential interaction may not be relevant to RUL function in hematopoietic cells that do not express eNOS (42). The functional identification of RUL as a ubiquitin ligase was not reported previously. Our data suggest a role for the ubiquitin ligase activity of RUL in the regulated proliferation of hematopoietic cells based on the ability of RUL U-box mutants to inhibit factor-dependent growth of 32D cells.
Assignment of RUL to the ubiquitin ligase family is based on the presence of an atypical form of the U-box homology, which is split by 104 intervening amino acids. The intervening sequence in the RUL U-box occurs within a predicted loop between the ␣-helix and the third ␤-sheet and thus may not disrupt the hypothesized core structure of this domain (26) and resulting activity. The in vitro ubiquitination experiment shown in Fig. 2 and the dependence on endogenous RUL for EpoR ubiquitination in COS7 cells support the conclusion that RUL is a functional ubiquitin ligase that is dependent on essential residues conserved within the U-box domain. It is presently unclear whether this activity is dependent on, or influenced by, the 104 residues splitting the RUL U-box. This intervening sequence represents the most significant divergence between the human and murine proteins (not shown). However, this does not preclude a functional role in regulating RUL ubiquitin ligase activity, including specificity for ubiquitin conjugating enzymes and targets of ubiquitination.
It remains to be determined if RUL ubiquitin ligase activity is directly influenced by activation of cytokine receptors. Although overexpressed RUL was constitutively ubiquitinated in COS7 cells, ubiquitination of endogenous RUL was enhanced after EpoR activation. Similarly, we have repeatedly observed Epo-induced increases in high molecular weight forms of RUL in Epo-treated 32D-EpoR cells (not shown). However, we have not been able to detect high molecular weight RUL species from either COS7 or 32D cells with several anti-ubiquitin antibodies. Thus, additional experimentation will be required to determine the effects of cytokine receptor activation on endogenous RUL ubiquitination in factor-dependent cells.
The consequences of protein ubiquitination vary depending on the structure of the ubiquitin conjugate (27). Polyubiquitin chains conjugated through lysine 48 generally target proteins for proteasome-dependent degradation, whereas monoubiquitination or polyubiquitin chains conjugated through other lysines generally result in modified function or altered subcellular localization of targeted proteins. Several reports demonstrating ubiquitin ligase activity of U-box proteins have shown that their activities include formation of polyubiquitin chains linked through multiple lysine residues of ubiquitin, in addition to lysine 48 and lysine 63 (38,39,43). We observed that RUL autoubiquitination and ubiquitination of EpoR consist of either multiple monoubiquitinations or polyubiquitination. It is presently unclear which lysine linkages might be involved in polyubiquitination mediated by RUL. However, we have not observed any detectable changes in expression levels of EpoR or associated proteins, including Jak2, as a function of expression of wild-type or mutant RUL (see Figs. 4 and 6).
Several studies have implicated ubiquitination in regulating the stability of proteins associated with Epo and IL-3 signaling pathways. EpoR, Jak2, and Stat5 have been reported to be targets of proteasome-dependent degradation following cytokine stimulation (19 -22). Recently, activated Jak2 was shown to be ubiquitinated by a complex containing SOCS1, resulting in proteasome-dependent degradation of the activated Jak2 (23). These observations imply a role for ubiquitination in the negative regulation of cytokine signaling by targeting elements of the signaling pathways for proteolytic destruction. By contrast, the growth inhibitory activity of RUL mutants lacking ubiquitin ligase activity implies a role for RUL-mediated ubiquitination in promoting survival and proliferation. The possibility that RUL acts as a positive regulator of signaling pathways may be more analogous to the role of TRAF6 in tumor necrosis factor-␣ signaling, where TRAF6-mediated ubiquiti- . Cells were lysed and immunoprecipitated with RUL antibodies. Precipitated proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes. 32 P-Labeled proteins were visualized on a PhosphorImager. Membranes were probed with RUL antibodies and then stripped and reprobed with horseradish peroxidase-coupled antibodies against the V5 epitope contained in the RUL-NT expression construct.
FIG. 11. RUL phosphorylation is mediated by cytokine and non-cytokine receptors. NFS-60 cells were labeled with [ 32 P]phosphate and treated for 12 min with IL-3, G-CSF, HGF, or left untreated (no factor). Cells were lysed and immunoprecipitated (IP) with RUL antibodies or a mixture of Stat5A and Stat5B antiserum. Precipitated proteins or total cell lysates were separated by SDS-PAGE and transferred to nitrocellulose membranes. 32 P-Labeled proteins in RUL IPs were visualized on a PhosphorImager. Membranes were also probed with antibodies against RUL, phosphotyrosine, or phospho-Erk. p33 RUL , a Novel Receptor-associated Ubiquitin Ligase nation directly activates a serine kinase, TAK1, and ultimately leads to activation of the transcription factor, NF-B (30,31). A relationship between RUL activity and transcription is suggested by the inhibited expression of c-myc and bcl-2 mRNA in cells expressing truncated RUL (RUL-NT), a mutant lacking a complete U-box domain. Notably, Epo-induced expression of pim-1 and bcl-X L were not substantially altered by RUL-NT expression, suggesting that RUL-NT did not result in a general inhibition of cytokine-induced gene expression. Such selective inhibition of certain targets of Epo-induced transcription suggests involvement of RUL in some, but not all, signaling pathways responsible for cytokine-induced expression of immediate-early genes. However, the specific reduction in Epo-induced expression of genes associated with growth and survival (c-myc and bcl-2) is consistent with the reduced proliferation and increased apoptotic death of RUL-NT-expressing cells.
The mechanism leading to cytokine-dependent expression of c-myc and bcl-2 has not been clearly identified. Epo and IL-3 have been reported to activate NF-B (44,45), a transcription factor that can be activated through the activity of ubiquitin ligases. Although we have not assessed the activity of NF-B in cells expressing wild-type or mutant RUL, NF-B is more usually associated with bcl-X L rather then bcl-2 expression (46). Epo-induced expression of these immediate-early genes is Jak2dependent, because this activity is lost from receptors that do not associate with Jak2, or in cells expressing dominant-negative Jak2 (17,(47)(48)(49). Notably, RUL-NT appears to function as a dominant negative, at least with regard to its ability to inhibit RUL-dependent ubiquitination of EpoR as well Epoinduced phosphorylation of endogenous RUL. Thus, along with Jak2, RUL may represent a required component of cytokineinduced signaling pathways regulating expression of c-myc and bcl-2.
The rapid and transient phosphorylation of RUL after cytokine treatment links this protein with the early events regulated by cytokine receptors. Indeed, the kinetics of RUL phosphorylation demonstrates that this event occurs at the same time as other events thought to occur within the receptor complex, including Jak2-mediated phosphorylation of Stat5. Whether RUL phosphorylation is related to or downstream of Jak2-dependent signaling pathways is presently unclear. However, the apparent requirement of both activities in the regulated transcription of c-myc and bcl-2 suggests some relationship between RUL and Jak2. We have found that activated Jak2 can phosphorylate RUL on tyrosine in vitro (not shown). However, we have found no evidence that this occurs in cytokine-treated hematopoietic cells. It remains possible that RUL phosphorylation occurs downstream of Jak2 through subsequent activation of a serine kinase. Importantly, RUL phosphorylation is not dependent on Jak activity, because RUL was inducibly phosphorylated following HGF treatment of NFS-60 cells. HGF acts through the receptor tyrosine kinase, c-Met (40), and regulates proliferation of NFS-60 cells without activating Jak kinases or Stat transcription factors (34). Thus, RUL phosphorylation may represent a common target of growth factor receptors from differing families. Consistent with a potential general role in proliferation regulated by a variety of growth factors, we observed ubiquitous expression of RUL mRNA in both human and mouse tissues (not shown).
The kinase that mediates serine phosphorylation of RUL has not been identified, and serine kinase activity has not generally been observed within cytokine receptor complexes. However, PKA has been reported to mediate IL-3-dependent serine phosphorylation of the IL-3 receptor (50). In this regard, serine 36 of RUL does conform to the consensus sequence for PKA phosphorylation ( 34 RLS 36 , fitting the consensus sequence (R/K)X(S/ T)). Other reports have suggested a role for protein kinase C in regulating signaling activity at EpoR, including possible involvement in induced expression of c-myc and bcl-2 (51)(52)(53)(54). Several protein kinase C consensus sites are present in the carboxyl-terminal half of RUL, although these sites are retained in the truncated RUL-NT that is not phosphorylated in response to cytokine treatment. Whether these or other serine kinases can phosphorylate RUL is currently being evaluated.
It is presently unclear how growth factor-induced phosphorylation may affect RUL function. By analogy, EGF receptor ubiquitination by the RING finger E3, Cbl, has been reported to require prior phosphorylation of Cbl by the EGF receptor tyrosine kinase (28). It is therefore possible that growth factorinduced phosphorylation of RUL modifies its activity either through altered ubiquitin ligase activity or altered protein associations. In that regard, one potential site of RUL phosphorylation is at serine 49, a residue required for RUL ubiquitin ligase activity in vitro.
We have yet to determine if this site is phosphorylated in growth factor-treated hematopoietic cells. Although EpoR activation in COS7 cells increased ubiquitination of endogenous RUL, we have not determined if this was associated with phosphorylation of RUL in these cells. Clearly, many questions remain concerning the involvement of RUL with mitogenic signaling pathways regulated by cytokine receptors particularly with regard to the specific role of its ubiquitin ligase activity and the status of ubiquitinated proteins in factor-dependent hematopoietic cells. However, the significance of this role is suggested by the specific inhibitory effects of RUL mutants lacking ubiquitin ligase activity and the association of endogenous RUL with early events activated by growth factor receptors.