Tyrosine Phosphorylation Disrupts Elongin Interaction and Accelerates SOCS3 Degradation*

The suppressors of cytokine signaling (SOCS) are negative feedback inhibitors of cytokine and growth factor-induced signal transduction. The C-terminal SOCS box region is thought to regulate SOCS protein stability most likely via an elongin C interaction. In the present study, we have found that phosphorylation of SOCS3 at two tyrosine residues in the conserved SOCS box, Tyr 204 and Tyr 221 , can inhibit the SOCS3-elongin C interaction and activate proteasome-mediated SOCS3 degradation. Jak-mediated phosphorylation of SOCS3 decreased SOCS3 protein half-life, and phosphorylation of both Tyr 204 and Tyr 221 was required to fully destabilize SOCS3. In contrast, a phosphorylation-deficient mutant of SOCS3, Y204F,Y221F, remained stable in the presence of activated Jak2 and receptor tyrosine kinases. SOCS3 stability correlated with the relative amount that bound elongin C, because

The suppressors of cytokine signaling (SOCS) 1 family of proteins can be transcriptionally activated by a broad range of extracellular ligands and functions in a classical feedback loop to regulate signal transduction through multiple cytokine and growth factor receptors (1)(2)(3)(4)(5)(6). The prototype members of this family, CIS and SOCS1, were initially cloned as cytokine-inducible immediate-early genes that could inhibit activation of signal transducers and activators of transcription (STAT) and biological responses to several cytokines (7)(8)(9). CIS regulates signaling through the erythropoietin and IL-3 receptors, whereas SOCS1 has been shown to be essential for proper regulation of IL-6 and interferon ␥ responses (10,11). These two molecules represent a subfamily of SOCS proteins that contain a central SH2 domain and inhibit cytokine-induced STAT activation by binding phosphotyrosine STAT docking sites on activated receptors or by interacting with phosphorylated tyrosine residues in the catalytic loop of receptor-associated Jaks to occlude the catalytic cleft and inhibit kinase activity and downstream signal transduction (12)(13)(14). Another member of the family, SOCS-3, was initially reported to inhibit signal transduction by binding to the activation loop of the Janus kinases. But SOCS-3 also exerts at least part of its effect by directly binding to specific phosphotyrosine motifs of activated cytokine receptor subunits (12,15,16).
Overexpression studies have implicated SOCS3 in the negative regulation of a number of cytokine signaling pathways. Mice nullizygous for SOCS3 demonstrated embryonic lethality caused by impaired placental development (17). Analysis of these mice indicated a role for SOCS3 in modulating leukemia inhibitory factor signaling during trophoblast giant cell differentiation. Furthermore, tetraploid rescued SOCS3Ϫ/Ϫ mice died within 3 weeks of birth because of heart failure. Phenotypically these mice displayed growth retardation and lethargy. Post-mortem analysis revealed cardiac monocyte hypertrophy presumably caused by a loss of SOCS3-mediated inhibition of leukemia inhibitory factor and CT-1 signaling.
The presence of a C-terminal SOCS box homology region in all SOCS proteins suggests that this domain can play an essential role in SOCS function. More than 50 proteins have been identified in mammals, Drosophila, and Caenorhabditis elegans that possess diverse protein-protein interaction motifs such as SH2 domains, SPRY domains, WD40 motifs, and ankyrin repeats, linked to this common homologous region known as the SOCS box, a C-terminal domain of ϳ40 amino acids (5,6). It has recently been shown that the SOCS box is a protein-protein interaction domain that mediates complex formation with elongin C (18,19). Elongin C is a component of ubiquitin ligases that include elongin B, the ring finger protein Roc1, and one of the scaffold proteins Cul2 or Cul5 (5,6,19,20). The function of the SOCS-elongin C interaction, however, is not well understood. Some recent data suggest that elongin C links SOCS proteins to the proteasome and targets them for degradation. Consistent with this model are the findings that SOCS1 induces the degradation of Jak2, the TEL-Jak2 oncogenic fusion proteins, and IRS-1/2 (21)(22)(23)(24)(25). In addition, mutations or post-translational modifications of SOCS1 that disrupt the elongin C interaction stabilize the protein and increase its half-life (18,26). However, there is also evidence that elongin C can stabilize SOCS protein expression and that disruption of this interaction leads to proteasome-mediated SOCS destruction (19,27). A well characterized elongin C-binding partner, the Von Hippel-Lindau (VHL) tumor suppressor, regulates the stability of a hypoxia-induced transcription factor HIF-1␣. Mutations in the VHL SOCS box that interfere with its interaction with elongin C destabilize the protein result in marked reduction in protein levels and are responsible for the loss of function phenotype in human VHL syndrome (28 -31). Thus, the binding of elongin C to SOCS proteins may regulate protein stability.
We have previously demonstrated that SOCS3 is phosphorylated by Jaks and receptor tyrosine kinases at two tyrosine residues, Tyr 204 and Tyr 221 , within the conserved SOCS box region and that phosphorylated Tyr 221 interacts with the SH2 domains of the Ras inhibitor p120 RasGAP (32). Here, we demonstrate that SOCS3 tyrosine phosphorylation can regulate protein stability and elongin C interaction. Tyrosine phosphorylation decreased SOCS3 protein half-life by disrupting the interaction between SOCS3 and elongin C. Furthermore, the data demonstrated that a SOCS3/SOCS1 chimera (3/1/3), which bound much more strongly to elongin C, was significantly more stable than wild-type SOCS3.
Plasmids/Antibodies/Fusion Proteins-C-terminally FLAG-tagged wild-type and mutant SOCS3 cDNAs were cloned into the mammalian expression vector pME18S have been described previously (32,35). C-terminally HA-tagged SOCS3 and 3/1/3 chimera were constructed by engineering a ClaI site at the 3Ј end of the SOCS3 coding sequence by PCR-based mutagenesis, followed by cloning into pME18S encoding a 5Ј in-frame ClaI site followed by the sequence for a hemagglutinin epitope tag. SOCS 3/1/3 and 3/V/3 chimeras were generated using a fusion PCR approach with overlapping (sense and antisense) mutagenic oligonucleotide primers encoding the SOCS1 or VHL BC boxes. The 5Ј PCR primer spanned the unique SacII site in the SOCS3 cDNA and the 3Ј primer covered the 3Ј coding sequence of SOCS3 followed by a NotI restriction site. The PCR product containing the chimeric sequence was swapped with the SacII/NotI fragment from the WT SOCS3 expression vector. All of the constructs were confirmed by sequencing (Seqwright, Houston, TX). FLAG-tagged SOCS3 and 3/1/3 chimera were cloned into the retroviral vector pMX-IRES-GFP, which has been described elsewhere (36,37). GST fusion constructs were produced by cloning the SOCS3, 3/1/3 or 3/V/3 chimeras into the bacterial expression vector pGEX4T (Clontech, Palo Alto, CA). Transformation of DH5␣ competent Escherichia coli (Invitrogen), isopropyl-␤-D-thiogalactopyranoside induction, and purification of GST fusion proteins were performed according to standard methods. N-terminally FLAG-tagged elongin B and N-terminally Myc-tagged elongin C constructs were generous gifts from Drs. Doug Hilton and Tracy Willson (Walter and Eliza Hall Institute, Parkville, Australia). The GST-JAK2-JH1 constructs were kindly provided by Dr. Akihiko Yoshimura (Fukuoka, Japan). All of the restriction endonucleases and DNA-modifying enzymes were purchased from New England Biolabs (Beverly, MA). FLAG antibody for FLAG-SOCS3 immunoprecipitation was purchased from Sigma.
Transfections/Infections-293T, COS-7, and PlatE cells were transfected with Effectene lipid-based transfection reagent (Qiagen, Crawley, UK) according to the manufacturer's instructions. For retroviral infections, 2 ml of supernatant from transfected Phoenix A or PlatE cells was added to 2 ml of A431 or 1881 cells at a concentration of 2 ϫ 10 5 cells/ml and incubated for 18 h at 37°C. The cells were washed once and resuspended in fresh culture medium. Cells stably expressing the SOCS3-IRES-GFP retrovirus were sorted by flow cytometry for green fluorescent protein expression.
Immunoprecipitations/GST Pull-down Experiments/Western Blotting-The cells were lysed in buffer containing 150 mM NaCl, 50 mM Tris-HCl, 2 mM EDTA, 0.875% Brij 97 (Sigma), 0.125% Nonidet P-40 (British Drug Houses, Poole, UK), 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , and 1 mM NaF. The lysates were centrifuged at 12,000 ϫ g for 5 min at 4°C to remove nuclei. The lysates were immunoprecipitated with appropriate antibodies as described in the figure legends. Rabbit polyclonal anti FLAG antibodies were purchased from Zymed Laboratories Inc. (South San Francisco, CA). Sepharose-conjugated M2 monoclonal was from Sigma. Monoclonal anti-Myc 9E10, rabbit anti-hemagglutinin tag, rabbit anti-Myc, and rabbit anti-GST antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-RasGAP antibodies were from New England Biolabs (Beverly, MA). Anti-phosphotyrosine 4G10 was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). For GST pull-down experiments, the cells were lysed as described above, and 5 g of GST, GST-wild-type SOCS3-GST, or GST-3/1/3 were added to the lysates in combination with 30 l of a 50% slurry of glutathione-Sepharose beads (Sigma). The lysates were incubated with the fusion proteins for 4 h at 4°C. The precipitates were washed in lysis buffer, boiled, and resolved by SDS-PAGE. The proteins were electroblotted onto nylon membranes (Immobilon-P, Millipore, Bradford, MA). The membranes were probed with appropriate antibodies followed by incubation with horseradish peroxidase-labeled anti-mouse or anti-rabbit secondary antibodies (Bio-Rad). The proteins were detected with chemiluminescent substrate (Pierce). Pulse

RESULTS
We have previously reported that SOCS3 is tyrosine-phosphorylated by several families of kinases, including Jaks and receptor tyrosine kinases (32). In mutagenesis studies, we have identified Tyr 204 and Tyr 221 within the conserved SOCS box motif as the major targets of phosphorylation by Jak1, Jak2, the platelet-derived growth factor receptor, and epidermal growth factor receptor (EGFR) and demonstrated that phosphorylated Tyr 221 binds the Ras inhibitor p120 RasGAP, thus modulating the ERK/MAP kinase pathway (32). The previous observation that the SOCS box regulates protein stability (18 -27) suggested that Tyr 204 and Tyr 221 phosphorylation might regulate this property of SOCS3.
LPS is known to induce SOCS3 expression. The murine monocytic cell line RAW264 was treated with LPS to assess whether endogenous SOCS3 protein induced by LPS would become tyrosine-phosphorylated. LPS-induced SOCS3 was not tyrosine-phosphorylated (Fig. 1A, top panel), presumably because LPS does not trigger tyrosine kinase activity. In these cells treated with LPS we have observed SOCS3 expression for periods up to 24 h without turnover of SOCS3 (results not shown). However, SOCS3 tyrosine phosphorylation can be induced by treating cells with LPS in the presence of the proteintyrosine phosphatase inhibitor sodium pervanadate. RAW264 cells were stimulated with LPS for up to 4 h. The cells were also incubated with or without pervanadate during the final 15-60 min of the experiment. SOCS3 protein levels were markedly increased in cells that had been activated by LPS for 3 h (Fig.  1A, bottom panel, lanes 1 and 3), but there was no detectable tyrosine phosphorylation of SOCS3 under these conditions (Fig. 1A, top panel, lane 3). In contrast, pervanadate treatment combined with LPS stimulation resulted in SOCS3 tyrosine phosphorylation (Fig. 1A, top panel, compare lanes 3 and 4). Interestingly, SOCS3 protein levels were significantly reduced upon extended treatment with pervanadate (Fig. 1A, bottom panel, lanes 3-7). These results suggested that tyrosine phosphorylation can markedly enhance SOCS3 degradation.
To demonstrate whether phosphorylation-induced SOCS3 destabilization occurred in a ligand-dependent manner, we used the human tumor cell line A431, which is transformed by the overexpression of EGFR (38). Stimulation of A431 cells with EGF after serum starvation results in phosphorylation of known EGFR substrates, such as STAT3, Shc, Grb2, and the EGFR itself (38). Using retroviral-mediated gene transfer, we stably expressed SOCS3 in A431 cells. To demonstrate that direct phosphorylation of SOCS3 was required to destabilize the protein, we also expressed a phosphorylation-deficient SOCS3 mutant, Y204F,Y221F, which we have previously shown not to be phosphorylated by Jak1, Jak2, and plateletderived growth factor receptor (32). A431 cells were serumstarved for 8 h and incubated with 50 ng/ml EGF for 30 or 60 min, and SOCS3 protein levels as well as tyrosine phosphorylation were determined by immunoblotting. As shown in Fig.  1B, EGF stimulation of A431 cells resulted in strong tyrosine phosphorylation of wild-type SOCS3 (top panel, lanes 4 -6), whereas EGF-induced phosphorylation of SOCS3 Y204F, Y221F was barely detectable (lanes 7-9). Interestingly, phosphorylation of wild-type SOCS3 resulted in a marked decrease in SOCS3 protein levels, but protein levels of the phosphorylation mutant remained unchanged after EGF stimulation (Fig.  1B, bottom panel, lanes 4 -6 and 7-9).
To determine the half-life of SOCS3 in the phosphorylated and unphosphorylated states, we stably expressed SOCS3 in a murine B cell line, 1881. These cells were treated with cycloheximide in the absence or presence of pervanadate to induce SOCS3 tyrosine phosphorylation. As shown in Fig. 1C, the half-life of unphosphorylated SOCS3 in this experiment was 8 h, whereas the half-life of phosphorylated SOCS3 was ϳ4 h. Significant amounts of unphosphorylated SOCS3 were detectable at 6 and 8 h following cycloheximide treatment. In pervanadate-treated cells, however, SOCS3 protein was barely detectable 6 h following cycloheximide treatment, and no SOCS3 was present at the 8-h time point. These data suggest that tyrosine phosphorylation of SOCS3 in the conserved SOCS box region results in protein destabilization and accelerated degradation.
We next examined the half-life of phosphorylated and unphosphorylated SOCS3 in pulse-chase experiments. COS7 cells were transfected with FLAG-tagged wild-type or Y204F,Y221F SOCS3 in combination with wild-type or kinase-inactive (FF) Jak2 catalytic domain (JH1) constructs. After [ 35 S]cysteinemethionine labeling, the cells were lysed at time points from 0 -2 h post-chase, and SOCS3 was immunoprecipitated with anti-FLAG antibodies. As shown in Fig. 2, the half-life of SOCS3, phosphorylated by the active Jak2 JH1 domain, was ϳ15-30 min ( Fig. 2A). In contrast, wild-type SOCS3 co-expressed with the catalytically inactive Jak2 mutant was significantly more stable than phospho-SOCS3 (half-life of 60 min; Fig. 2B). The phosphorylation-deficient Y204F,Y221F SOCS3 mutant had a half-life of Ͼ90 min (Fig. 2C), even in the presence of the kinase-active Jak2 JH1 domain. Likewise, Y204F and Y221F SOCS3 mutants were significantly more stable than WT SOCS3 when expressed with kinase-active Jak2 (Fig.  2, D and E), again suggesting that tyrosine phosphorylation may contribute to the destabilization of SOCS3 protein. The difference in the observed half-life of SOCS3 in 1881 and COS7 cells was observed in many experiments and may reflect the relative abundance of an E3 ligase or a difference between 35 S labeling and cycloheximide treatment to determine protein half-life. We further explored the contribution of each tyrosine to phosphorylation-dependent SOCS3 degradation by comparing protein levels of wild-type SOCS3, SOCS3 single mutants Y204F and Y221F, and the Y204F,Y221F double mutant. COS-7 cells were transfected with a wild-type SOCS3 expression plasmid or one of the mutant SOCS3 constructs in combination with wild-type or kinase-inactive Jak2 JH1 domain. Cell lysates were immunoprecipitated with anti-FLAG, and the precipitates were probed for FLAG-epitope or phosphotyrosine. As shown in Fig. 3A, wild-type SOCS3 became tyrosine-phosphorylated in the presence of wild-type Jak2 JH1 domain (top panel, lane 2) and resulted in low levels of SOCS3 protein expression (bottom panel, lane 2). As expected, the single tyrosine mutants were phosphorylated to a lesser degree than wild-type SOCS3 (top panel, lanes 3 and 4). Although the Y221F mutant appeared to be phosphorylated to a similar extent as wild-type SOCS3, when normalized for relative protein expression, the phosphorylation signal was reduced compared with wild type. In agreement with our pulse-chase analysis, levels of the SOCS3 Y204F,Y221F mutant with the active Jak2 kinase as well as wild-type SOCS3 expressed with the kinase-inactive Jak2 construct are both higher than phosphorylated wild-type or single mutant SOCS3 proteins. These results suggest that phosphorylation of either Tyr 204 or Tyr 221 resulted in a partial SOCS3 destabilization, but phosphorylation of both tyrosines produced more complete protein degradation.
To determine whether the phosphorylated SOCS3 was degraded by a proteasome-mediated pathway, 293T cells were transfected with a wild-type SOCS3 with or without Jak1, and treated with the proteasome inhibitor MG132. The cell lysates were probed for SOCS3 or Jak1 and as shown (Fig. 3B) low levels of SOCS3 expression were observed when co-expressed with Jak1 and SOCS3 levels were restored by pretreatment with MG132.
To estimate the relative half-lives of both wild-type SOCS3 and the SOCS3 Y204F,Y221F phosphorylation-deficient COS-7 cells, the transfectants were subsequently treated with cycloheximide for up to 8 h, and the lysates were immunoblotted with anti-FLAG antibody. As shown in Fig. 3C (lanes 2-6), the half-life of the wild-type SOCS3 protein was ϳ2 h when coexpressed with the Jak1 kinase, whereas the SOCS3 Y204/ 221F mutant, which lacks tyrosine phosphorylation sites, appeared to be much more stable (4 -8 h half-life) when coexpressed with the active kinase (Fig. 3C, lanes 8 -12). These results further demonstrate that phosphorylation-dependent degradation of SOCS3 is controlled by the SOCS box tyrosine residues Tyr 204 and Tyr 221 as shown by the relative stability of the Y204F,Y221F mutant.
It has been established that the SOCS box mediates a direct physical interaction with elongin C (18,19). It is believed that interaction with elongin C may link SOCS proteins to the proteasome and target them for degradation. However, several studies have shown that VHL mutants deficient in elongin C binding are unstable and are expressed at very low levels compared with wild-type VHL, suggesting that elongin C can stabilize SOCS protein expression under certain conditions  3. Jak kinases accelerate the SOCS3 degradation. A, COS7 cells were transfected with GST-Jak2 JH1 alone or in combination with wild-type (WT) SOCS3, the Y204F SOCS3, the Y204F SOCS3, the Y204,221F SOCS3, or mutant wild-type SOCS3 with a kinase-inactive mutant of Jak2 (Jak2-JH1-FF) as indicated. The lysates were immunoprecipitated (IP) with FLAG antibody and blotted for phosphotyrosine (upper panel or FLAG) (all SOCS3 constructs were FLAGtagged). B, 293T cells were transfected with a wild-type SOCS3 with or without Jak1, and treated with the proteasome inhibitor MG132 as indicated and lysates were probed for SOCS3 or Jak1. C, COS-7 cells were transfected with Jak1 and wild-type SOCS3 or SOCS3 Y204/221F, and either untreated (control, con) or treated with cycloheximide (CHX) for up to 8 h. The lysates were blotted for SOCS3 (upper panel) or Jak1 (lower panel). (28). Thus, in some contexts, disruption of the SOCS-elongin C interaction results in rapid proteasome-mediated protein turnover. We were therefore interested in determining the effect of tyrosine phosphorylation on the SOCS3-elongin C interaction.
To better evaluate the effect of tyrosine phosphorylation, we were interested in determining the position of the two tyrosines within the SOCS box. The three-dimensional structure of a VHL-elongin C complex has been solved by x-ray crystallography (30). Based on this structural data and BC box sequence homology, we have generated a model for the structure of the SOCS3 SOCS box (Fig. 4B). The SOCS box contains three ␣ helices (␣1, ␣2, and ␣3), joined by short segments of random coil sequences. The interaction with elongin C leads to the formation of a four-helix cluster involving ␣1, ␣2, and ␣3 of SOCS3 and the C-terminal helix of elongin C. It has been demonstrated that the ␣1 helix of the VHL SOCS box contains the major elongin C contact region (BC box), whereas the ␣2 and ␣3 helices function to stabilize ␣1, increasing its binding to the elongin BC complex (28,30). Based on the position of Tyr 204 and Tyr 221 within the ␣2 and ␣3 helices, respectively, we hypothesized that phosphorylation of these residues could disrupt elongin C binding and trigger SOCS3 degradation.
To test the theory that SOCS3 protein stability correlates with its binding to elongin C, we analyzed the effect of SOCS3 tyrosine phosphorylation on elongin C binding in GST pulldown experiments. As shown in Fig. 5A, a SOCS3-GST fusion protein bound Myc-tagged elongin C in lysates of transfected 293T cells (top panel, lane 4). In contrast, when the SOCS3-GST fusion protein was first tyrosine-phosphorylated in vitro with recombinant Lck, it failed to interact with elongin C (top panel, lane 5). Control blots showed that the SOCS3-GST fusion protein was strongly phosphorylated by recombinant Lck, and the levels of all the GST proteins added to the cell lysates were equal (Fig. 5A, bottom panel).
Because it has been demonstrated that SOCS1 binds elongin C more strongly than SOCS3 (18,19), we reasoned that introduction of the SOCS1 or VHL ␣1 helix (BC box) would generate a SOCS3 mutant with a stronger interaction with elongin C and thus greater protein stability than wild-type SOCS3. To test this hypothesis we generated SOCS3/SOCS1/SOCS3 and SOCS3/VHL/SOCS3 chimeras (3/1/3 and 3/V/3) in which the ␣1 helix of SOCS3 was replaced by the ␣1 helix of SOCS1 or VHL, respectively (Fig. 4A). We compared elongin C binding by SOCS3, 3/1/3, and 3/V/3 in GST pull-down and co-immunoprecipitation experiments. As shown in Fig. 5B, the GST-3/1/3 and GST-3/V/3 chimera fusion proteins bound significantly higher levels of elongin C than GST-wild-type SOCS3 protein, suggesting that the chimeras bind elongin C more strongly than wild-type SOCS3. In addition, we transiently expressed elongins B and C in combination with HA-tagged wild-type SOCS3 or the 3/1/3 chimera. Following immunoprecipitation of elongin C with anti-Myc antibodies, the precipitates were immunoblotted with anti-HA for the presence of SOCS3. As shown in Fig. 5C, and in agreement with our GST pull-down results, the 3/1/3 chimera interacted with elongin C better than wild-type SOCS3. Similar results were obtained with the 3/V/3 chimera (not shown).
We next compared the half-lives of wild-type SOCS3 and the 3/1/3 chimera. We stably expressed the wild-type and chimeric cDNAs in 1881 cells, using retroviral-mediated gene transfer. Each cell line was treated with 20 M cycloheximide for an extended time course and analyzed for SOCS3 expression by immunoblotting. As shown in Fig. 5D, the half-life of wild-type SOCS3 in the 1881 cells was ϳ4 h (top panel). Treatment of the cells with pervanadate to induce SOCS3 tyrosine phosphorylation markedly reduced the half-life to ϳ1-2 h (middle panel). In contrast, the 3/1/3 chimera, which bound elongin C more strongly than WT SOCS3, was stable beyond 8 h of cycloheximide treatment (bottom panel). Expression of wild-type SOCS3 was rescued by treatment of the cells with the proteasome inhibitor MG132 (not shown). Thus, our findings indicate that SOCS3 protein expression can be stabilized through interaction with elongin C. Phosphorylation Tyr 204 and Tyr 221 resulted in loss of elongin C binding and accelerated SOCS3 turnover, whereas mutations that increase elongin C interaction significantly extended the half-life of SOCS3.
We have also demonstrated that SOCS3 binds to the Ras inhibitor p120 RasGAP in ligand-stimulated cells (12,32). It was therefore possible that in ligand-activated cells, a subpopulation of SOCS3 is phosphorylated, destabilized, and in a complex with SH2-containing proteins such as p120 RasGAP, although distinct pools of SOCS3 remain stable and bound to elongin C. To determine whether SOCS3 can form discrete complexes with different binding partners, we separated lysates of IL-2-stimulated CTLL-2 cells over a size exclusion column to determine the relative molecular masses of SOCS3containing complexes. By immunoblotting, we identified the fractions that contained SOCS3, phosphorylated SOCS3, and

FIG. 4. Schematics of SOCS box chimeras and model of threedimensional structure of SOCS3-elongin C.
A, schematic representation SOCS box chimeras, the SOCS3-elongin C interaction, the SOCS3 phosphorylation mutants, and the SOCS 3/1/3 chimeras used in these studies SOCS3. Shown are the kinase inhibitory region (KIR), the SH2 domain, and the C-terminal SOCS box. Below is an alignment of the SOCS1, VHL, and SOCS3 SOCS box sequences, showing the positions of the elongin C-binding ␣1 helix (BC box), as well as the ␣2 and ␣3 helices. ESS, extended SH2 domain. B, model of the predicted three-dimensional structure of the SOCS3-elongin C interaction (left image) and expanded model of the SOCS3 SOCS box (right image). As shown on the left, the major elongin C contact site is the ␣1 helix, and the ␣2 and ␣3 helices are believed to stabilize the structure of the ␣1 helix. The SOCS box model on the right shows the positions of Tyr 204 and Tyr 221 . Tyr 204 is located within the ␣2 helix, whereas Tyr 221 is just C-terminal to the ␣3 helix.
p120RasGAP. As shown in Fig. 6, SOCS3 eluted from the column in several fractions, representing a range of molecular weights (top panel, fractions [11][12][13][14][15][16][17][18]. Only a subset of the total cellular pool of SOCS3 co-eluted with p120 RasGAP (fractions [11][12][13], whereas the majority of SOCS3 eluted in fractions corresponding to complexes of smaller molecular weight or monomeric SOCS3 (fractions 14 -18). DISCUSSION The expression of SOCS proteins must be tightly controlled to maintain an equilibrium between activation and inhibition (26). Our findings suggest that SOCS3 protein stability can be controlled by phosphorylation-induced changes to regulate feedback inhibition of cytokine signal transduction. Clearly tyrosine phosphorylation can dissociate SOCS3 from the elon-  gin-BC complex and accelerate its degradation. The phosphorylated tyrosines are found at the C-terminal SOCS box and are phosphorylated in response to cytokines and growth factors (32), suggesting that this turnover must be an important cytokine mediated response. The importance of the 50-amino acid conserved SOCS box in the regulation of SOCS activity and elongin C binding has been demonstrated by several groups (18 -20, 26 -31). Recent studies have shown that the stability and function of the SOCS1 and VHL proteins are regulated through the SOCS box and its interaction with elongin C. However, the reports have been conflicting, because the SOCS box-elongin C interaction can both stabilize SOCS proteins and enhance SOCS protein turnover (18 -20, 26 -31). Several studies suggest that the interaction of elongin C with the SOCS box targets SOCS1 for proteasome-mediated degradation (19,25,26). In support of this model, N-terminal serine phosphorylation of SOCS1 or SOCS box mutations that disrupt the SOCS1-elongin C interaction stabilize the protein and abolish its ability to direct the degradation of signaling proteins (26). The suggestion is that elongin C can couple SOCS1 to the proteasome and enhance degradation of SOCS1-containing complexes.
In contrast, in some reports the SOCS box has been shown to stabilize SOCS1 through its interaction with elongin C, and disruption of the SOCS-elongin C interaction reduces the halflife of SOCS1. In one study, expression of a dominant-negative SOCS1 mutant (F59D) destabilized endogenous SOCS1 and SOCS3 protein in transfection experiments (27). In addition, gene-targeted mice have been engineered that express a SOCS1 SOCS box deletion mutant (39). The mutant SOCS1 protein was unstable and expressed at lower levels than wildtype SOCS1, strongly supporting a role for the SOCS box in maintaining protein stability.
We have shown that phosphorylation of tyrosine residues in the SOCS box can block interaction with elongin C and can activate rapid proteasome-mediated degradation. Because the phosphorylation-induced destabilization of SOCS3 correlated with its dissociation from elongin C, this demonstrated that SOCS3 can be stabilized by interaction with the elongin B/C complex. Similarly, the well characterized SOCS family member, the VHL tumor suppressor protein can form part of a multi-subunit complex containing elongins B and C in combination with Roc1 (Rbx1) and the scaffold protein Cul2 (CB-C VHL , (Cul2-elongin BC-VHL)) (31). The formation of CBC VHL complexes stabilized VHL protein and prevented its ubiquitination. In contrast, formation of a complex containing VHL, Cul-2, and the ring finger protein Rbx-1 in the absence of elongins B and C resulted in efficient ubiquitination of VHL.
In further support of this idea SOCS3 mutants that interact more strongly with elongin C are more stable than wild-type SOCS3. The chimeric SOCS molecules, in which the elongin C-binding ␣1 helix of SOCS3 was replaced by that of SOCS1 or VHL, were significantly more stable than wild-type SOCS3 in cycloheximide-treated cells. This suggested that stable SOCS3 expression depended on the formation of a complex with elongin C. Post-translational modification such as tyrosine phosphorylation, which disrupts the complex, destabilized SOCS3, whereas mutations that increase interaction with elongin C substantially extend the half-life of SOCS3 protein.
In our proposed structure of SOCS3, the major elongin C contact residues are located in a single ␣ helical subdomain (␣1) of the SOCS box, and Tyr 204 and Tyr 221 are located Cterminal to the elongin C binding site, within the ␣2 and ␣3 helices, respectively. VHL mutations within the ␣1 region are often responsible for the loss-of-function phenotype in human von Hippel-Lindau syndrome (29,30). However, a subset of VHL syndrome patients have mutations in the ␣2 or ␣3 helices, and it has been hypothesized that the structural integrity of these two subdomains may be required for high affinity binding to elongin C (30). Neither Tyr 204 nor Tyr 221 are located in the SOCS3/elongin C interface, so they cannot directly interfere with elongin C binding. Rather, the proximity of Tyr 204 and Tyr 221 to the SOCS3 BC box would suggest that phosphorylation of these residues can result in SOCS3 destabilization by displacing the ␣2 and ␣3 helices via steric hindrance or electrostatic repulsion of the two phosphate groups. The resulting changes in the structure of the SOCS box may destabilize the three helices (␣1, ␣2, and ␣3), resulting in inhibition of elongin C binding and activation of proteasome-mediated SOCS3 degradation.
In agreement with this hypothesis, we have found that phosphorylation of either tyrosine residue alone results in partial destabilization of SOCS3, but phosphorylation of both tyrosines is required for complete protein degradation (Figs. 2  and 3). Phosphorylation of Tyr 204 , adjacent to the BC box, may change the conformation of the BC ␣1 and ␣2 helices, but it is possible that phosphorylation of Tyr 221 may enhance this conformational change, and thus both tyrosines can play a role in SOCS3 turnover.
It is not surprising that SOCS3 stability may be controlled by phosphorylation because expression of a wide range of proteins can be modulated by phosphorylation. For example, the cell cycle regulator p27 is phosphorylated on threonine 187 by several families of kinases, including Akt, and cyclin-dependent kinases, which induces its interaction with Skp2, a component of an E3 ubiquitin ligase (40). Tyrosine phosphorylation has also been identified as a mechanism of activating proteasome-mediated degradation of kinases. Phosphorylated tyrosine residues on activated kinases such as EGFR, plateletderived growth factor receptor, and Zap-70 recruit the ring finger protein c-Cbl through its phosphotyrosine-binding domain. c-Cbl bridges the kinases to an E2 ubiquitin ligase and facilitates protein turnover and down-regulation of signal transduction (41). We have found that phosphorylated fulllength SOCS3 as well as SOCS3-derived phosphopeptides fail to interact with c-Cbl, 2 suggesting that c-Cbl may not be involved in SOCS3 degradation.
It has recently been shown that SOCS1 binds to a range of proteins including the TEL-Jak2 oncogene, the receptor tyrosine kinase c-Kit, and the Rac guanine nucleotide exchange factor Vav triggering degradation, suggesting that their halflife can be regulated by SOCS1 (21,22,42,43,51). SOCS3 can associate with the IL-6 receptor subunit gp130, as well as the receptors for IL-2, erythropoietin, leptin, insulin, G-CSF, EGF, and platelet-derived growth factor (12,15-17, 44 -50, 52). We have previously reported that phosphorylated SOCS3 tyrosine 221 binds the SH2 domains of the Ras inhibitor p120 RasGAP (32). Our data suggest that a potential mechanism for the negative regulation of RasGAP activity by SOCS3 may be via proteasome-mediated degradation of a SOCS3-RasGAP complex. We have found that a number of high molecular weight RasGAP species can be detected in SOCS3 immunoprecipitates from cytokine-activated cells, suggesting that these species were unstable and targeted for degradation.
An alternative function for the SOCS box-elongin C interaction has been suggested by the finding that the ring finger protein Rbx-1 (Roc-1) could bind and directly ubiquitinate VHL in the absence of the other components of the CBC complex, and expression of elongins B and C inhibited VHL ubiquitination (31). Therefore, the elongin BC complex may function to block a ring finger protein-binding site and protect SOCS proteins from degradation. Mutations that prevent the interaction of the VHL SOCS box and elongin C may expose an Rbx-1 binding site on VHL and result in rapid turnover and a decrease in steady-state levels of VHL protein. Interestingly, SOCS1 can directly interact with the novel ring finger protein TRIM8/GERP (40). Expression of TRIM8 decreased SOCS1 protein levels and greatly reduced the half-life of SOCS1, which suggests that the stability of both VHL and SOCS1 is regulated by a mechanism that involves the recruitment of a ring finger protein to initiate proteasome-mediated ubiquitination. The effect of elongin BC expression on TRIM8-mediated SOCS1 degradation has not been analyzed, so it is unclear whether elongin C can protect SOCS1 from degradation in a manner similar to VHL.
These recent studies as well as our present data are consistent with a model in which the elongin BC complex protects SOCS proteins from proteasome-mediated degradation while also acting as a molecular bridge between SOCS molecules and the proteasome. In the case of SOCS3, the function of an elongin BC interaction may be to localize it to a latent ubiquitination complex that is inactive until triggered by phosphorylation-induced dissociation of the SOCS3-elongin C interaction. Bringing tightly regulated ubiquitination machinery into physical contact with SOCS3 would facilitate rapid SOCS3 turnover in response to an appropriate stimulus that releases elongin C from the SOCS box. In support of this model, a naturally occurring isoform of SOCS3, which is generated by alternative translation initiation, has been identified that appears to encode a truncated SOCS3 protein, which is more stable than the full-length protein (23). Analysis of its biological activity demonstrated an enhanced ability to inhibit cytokine signaling, presumably though increased stability. The truncated SOCS3 isoform lacks the first 12 N-terminal amino acids including a lysine residue at position 6, which has been shown to be the main site for ubiquitination on SOCS3, further underlining the potential importance for E3 ligase mediated regulation of SOCS3 protein expression.
Sequence comparison reveals a high degree of conservation of SOCS box tyrosine residues. The amino acid corresponding to SOCS3 Tyr 221 is conserved in ϳ70% of all SOCS family members. In addition, Tyr 204 is conserved in SOCS proteins WSB-2, ASB-1, and ASB-2 (5, 6), suggesting that tyrosine phosphorylation may regulate the stability of these molecules as well. It will be important to determine whether other SOCS proteins can also become phosphorylated at the C terminus and thus undergo rapid degradation.