Cell Regulation by Phosphotyrosine-Targeted Ubiquitin Ligases

Abstract

Three classes of E3 ubiquitin ligases, members of the Cbl, Hakai, and SOCS-Cul5-RING ligase families, stimulate the ubiquitination of phosphotyrosine-containing proteins, including receptor and nonreceptor tyrosine kinases and their phosphorylated substrates. Because ubiquitination frequently routes proteins for degradation by the lysosome or proteasome, these E3 ligases are able to potently inhibit tyrosine kinase signaling. Their loss or mutational inactivation can contribute to cancer, autoimmunity, or endocrine disorders, such as diabetes. However, these ligases also have biological functions that are independent of their ubiquitination activity. Here we review relevant literature and then focus on more-recent developments in understanding the structures, substrates, and pathways through which the phosphotyrosine-specific ubiquitin ligases regulate diverse aspects of cell biology.

INTRODUCTION

Phosphorylation and ubiquitination are among the commonest and best-studied posttranslational modifications of proteins. A phosphate group or ubiquitin molecule can trigger or obstruct protein-protein interactions, alter subcellular localization, stabilize a particular protein conformation, or have myriad other effects. Phosphorylation is directly catalyzed by protein kinases, but ubiquitination is more complex, requiring sequential activity of E1, E2, and E3 ubiquitin-activating, -conjugating, and -ligating enzymes (1–5). E3 ubiquitin ligases fall into two major groups: HECT domain ligases receive ubiquitin from an E2 enzyme and transfer it to a bound substrate, while RING-type ligases position an E2-ubiquitin conjugate near a substrate protein to facilitate ubiquitin transfer. Both phosphorylation and ubiquitination are reversible; protein phosphorylation is reversed by protein phosphatases and ubiquitination by deubiquitinating enzymes (DUBs) (6–8). Therefore, both phosphorylation/dephosphorylation and ubiquitination/deubiquitination can allow repeated cycles of protein modification. Reversible ubiquitination is particularly important in DNA repair and NF-κB signaling. However, many ubiquitination events lead irreversibly to protein destruction, allowing regulation of protein turnover. For example, K48 polyubiquitin chains primarily route cytosolic proteins to the proteasome, while modification of many Lys residues with single ubiquitin molecules (multimonoubiquitination) has several functions, including routing membrane proteins for destruction in the lysosome (9–11). The irreversibility of proteolysis means that the ubiquitin-proteasome and ubiquitin-lysosome pathways directly control protein life spans.

Protein phosphorylation and ubiquitination cross talk at many levels (12). In this review, we focus on situations where phosphorylation of a substrate creates a binding site for an E3 ligase, rendering ubiquitination dependent on prior phosphorylation of that substrate (13). Such phosphorylation-dependent substrate selection has particular importance because it can layer negative feedback onto an otherwise reversible phosphorylation event (Fig. 1). In principle, increasing the kinase activity in a simple kinase/phosphatase cycle simply increases the steady-state level of the phosphorylated substrate (Fig. 1A and C), but adding phospho-specific ubiquitination and proteolysis reactions alters the kinetics, causing the level of phosphorylated substrate to decay back to baseline as the total substrate pool is depleted (Fig. 1B and C). This constitutes a negative-feedback loop that is “hardwired”; no other regulatory inputs are required. In principle, the E3 ligase limits the duration of signaling by a substrate and may introduce a refractory period during which signaling cannot recur. Perhaps for this reason, phosphorylation-dependent ubiquitin ligases are key controllers of different steps in the cell cycle and in DNA repair, signal transduction, and other fundamental cellular events (12).

FIG 1.

Regulation by phosphorylation-dependent ubiquitin ligases. Phosphorylation-dependent ubiquitination can provide negative feedback. (A) A simple system with substrate protein X undergoing constitutive slow synthesis and equal slow degradation is acted on by a regulated protein kinase and a constitutive phosphatase. (B) A phosphorylation-dependent ubiquitin ligase can promote the ubiquitination of the phosphosubstrate and target it for lysosomal or proteasomal degradation. Ubiquitination may be reversed by deubiquitinases (DUBs). (C) Signal outputs over time for schemes A and B. In the simple kinase/phosphatase system (curve A), kinase activation causes a sustained increase in the level of phosphosubstrate. With the addition of a ubiquitin ligase and DUBs (curve B), kinase activation leads to a transient increase in the level of phosphosubstrate. Phosphorylation/dephosphorylation is assumed to be faster than ubiquitination/deubiquitination. Note that the ubiquitinated substrate can also be dephosphorylated and rephosphorylated, but this does not affect the negative feedback.

This review focuses on the subset of phosphorylation-dependent ubiquitin E3 ligases that require phosphotyrosine (pY) in their substrates. These ligases fall into two well-studied groups, Cbl family proteins and cullin 5 (Cul5)-RING ligase complexes (CRL5s) bound to suppressor of cytokine signaling (SOCS) protein adaptors, as well as a less-studied ligase named Hakai. Here we introduce these ligases and discuss their strategies for binding pY, their substrates, and their biological functions. Because tyrosine protein kinases are master regulators of signal transduction cascades, pY-dependent ubiquitination might regulate cell growth, proliferation, motility, survival, and differentiation. Indeed, genetic evidence reviewed here shows the importance of Cbl and SOCS proteins in cancer, autoimmunity, and endocrine disorders, including diabetes. However, some Cbl and SOCS proteins have additional binding, scaffold, or adaptor functions independent of their ubiquitin ligase activity. This greatly complicates the identification of critical pY substrates and our understanding of how pY-dependent E3 ligases regulate cell biology.

A BRIEF HISTORY OF Cbl, Hakai, AND SOCS FAMILY PROTEINS

The c-Cbl gene was identified as the cellular homolog of the v-Cbl oncogene in Cas-Br-M mouse retrovirus (14). c-Cbl and its relatives Cbl-b and Cbl-3 contain an N-terminal pY-binding tyrosine kinase-binding (TKB) domain and a central zinc-binding C₃HC₄ RING motif (15–17). c-Cbl suppresses transformation in part by binding and downregulating receptor tyrosine kinases (RTKs), but the mechanism was unclear until Cbl's RING domain was found to stimulate ubiquitin transfer from an E2-ubiquitin donor to nearby proteins (18). pY substrates for Cbl family members bind to the TKB domain and receive ubiquitin from an E2 bound to the RING domain. The C termini of c-Cbl and Cbl-b contain proline-rich regions and phosphorylation sites that allow them to also act as signaling scaffolds (17).

Hakai (also called Cbl-like-1 and Cbll1) was first identified as a RING-containing, E-cadherin-binding protein that resembled Cbl (19). However, we now know that Hakai and Cbl have different mechanisms of pY substrate binding. Hakai dimerization creates an unusual HYB domain that binds pY substrates (20). The RING domains are directly adjacent to the HYB domain, allowing positioning of E2 and ubiquitin close to the pY residues in the substrate.

The SOCS proteins SOCS1-7 and CIS (cytokine-inducible SH2 protein) were discovered as cytokine-inducible inhibitors of signaling, STAT-induced STAT inhibitors, and cytokine-inducible SH2 proteins (21–25). They were initially shown to compete with STATs for binding to pY sites on cytokine receptors and to bind to and inhibit cytokine-activated kinases called JAKs. However, CIS bound to erythropoietin receptors through a pY site that was dispensable for signaling, calling into question the simple competition model. Moreover, erythropoietin induced CIS ubiquitination and proteasome inhibitors prolonged erythropoietin signaling (26). Later, SOCS1 was found to be induced by RTKs and to inhibit signaling at the level of the RTK and intracellular signaling proteins, such as Vav1 (27). Subsequently, SOCS1 was found to stimulate ubiquitination and degradation of tyrosine-phosphorylated Vav1 (28). We now know that all eight SOCS proteins contain a consensus sequence (a SOCS-BC box) that mediates their association with a multisubunit CRL5 complex (29–31). The other subunits of a SOCS-CRL5 complex which comprises elongin B, elongin C, cullin 5, and Rbx2 (RNF7) are invariant (32). (Under some circumstances, cullin 5 can form a complex with Rbx1 [33–35], but Rbx2 is more common). Rbx2 contains a RING domain and binds E2-ubiquitin conjugates. SOCS-CRL5 thus combines a substrate recognition domain (SH2) and a ubiquitin transfer domain (RING) in different subunits of a multiprotein complex. The RING protein is quite distant from the SH2 domain, potentially facilitating polyubiquitination.

pY SUBSTRATE BINDING AND UBIQUITINATION BY Cbl PROTEINS

Cbl family proteins contain a tyrosine kinase-binding or TKB domain and a RING domain, linked by a well-conserved linker helix region (LHR). c-Cbl and Cbl-b also share a proline-rich region and a C-terminal ubiquitin-associated (UBA) domain (16) (Fig. 2A). The TKB domain harbors a 4-helix bundle (4H), a Ca-binding EF hand, and an atypical SH2 domain (15) (Fig. 2B). The sequence of the SH2 domain is highly divergent, yet it forms a classic SH2 fold and binds pY peptides in a conventional way (15). The TKB domain allows Cbl to bind to pY proteins, such as active RTKs. Cbl may then become phosphorylated and bind other proteins, serving as an adaptor for protein complex assembly. In addition, pY proteins bound to the TKB domain can be ubiquitinated by an E2 enzyme bound to the RING domain. However, the structure of the TKB-LHR-RING region of c-Cbl revealed that an E2 enzyme bound to the RING domain would be unable to access the pY substrate by the TKB domain (36) (Fig. 2C, left). The mystery was solved when it was found that phosphorylation of Tyr371 in c-Cbl (Tyr363 in Cbl-b) within the LHR linker region induces a drastic conformational change that, in effect, reorients the substrate so that it is now close to the RING and E2 domains (37–40) (Fig. 2C, right). This phosphorylation switch may help explain how Cbl can alternatively promote signaling as an adaptor or inhibit signaling as an E3 ligase, as described below.

FIG 2.

Structures of three families of E3 ligases that recognize tyrosine-phosphorylated ligands. (A) Domain structure of c-Cbl and Cbl-b. Cbl-c is similar but terminates after the RING domain. (B) Crystal structure of the c-Cbl TKB domain in complex with a tyrosine-phosphorylated peptide derived from ZAP-70 (PDB code 2CBL) (15). In this panel and in panels D and G, α helices and β strands are shown in columns and ribbons, respectively. The peptide is shown as orange sticks, except for the pTyr residue, which is red. The regions in the TKB domain are colored according to the colors in panel A. The arginine residue of the SH2 domain (Arg294), essential for binding to the phosphate group of pTyr, is blue. (C) Activation of Cbl via a conformational change triggered by tyrosine phosphorylation of the LHR linker region. Shown here is the structure, in surface representation, of a c-Cbl–E2–substrate peptide complex in an inactive conformation (left, PDB code 1FBV [36]) or when activated through phosphorylation of Tyr371 in the linker region (right, PDB code 4A4C [37]). The E2 and RING domains are blue and magenta, respectively. The SH2 domain is cyan. A ZAP-70-derived phosphopeptide is orange. The LHR linker between the SH2 and the RING domains is represented by a green noodle, with Tyr371 highlighted in red. (D) Homodimeric structure of the Hakai HYB domain (PDB code 3VK6 [20]). The dimer coordinates six zinc ions (gray balls), two of which are located behind α1 and α1′ helices and are thus invisible in this figure. Four residues (i.e., His127, Tyr176, His185, and Arg189) from each monomer that contribute to phosphopeptide recognition are shown as blue sticks. (E) Diagram of a SOCS-CRL5 complex. (F) Structural model of a SOCS-CRL5 complex. The model was constructed according to reference 149, based on the following crystal structures: the cullin 5 C-terminal domain in complex with Rbx1 (PDB code 3DPL [150]), the cullin 5 N-terminal domain in complex with elongin B, elongin C, and Socs2 (PDB code 4JGH [151]), the structure of cullin 1, used here as a template for joining the N-terminal and C-terminal fragments of cullin 5 structures (PDB code 1LDJ [148]), and the Socs6 SH2 domain bound to a tyrosine-phosphorylated peptide (PDB code 2VIF [42]). The magenta molecule is Rbx1 (instead of Rbx2), since there are no available Rbx2-cullin complex structures. (G) SOCS6 SH2 domain in complex with a phosphorylated peptide derived from the juxtamembrane region of the receptor tyrosine kinase c-KIT (PDB code 2VIF [42]). The bound peptide is shown in orange sticks except for the pTyr residue, which is highlighted in red. The Arg409 residue of the SH2 domain, essential for pTyr recognition, is shown as blue sticks. The peptide is bound between the BG and EF loops.

pY SUBSTRATE SELECTION BY Hakai

Hakai was discovered as a Cbl-like protein that binds to the membrane protein E-cadherin (19). Hakai targets a cytoplasmic NVYYY motif in E-cadherin when it is phosphorylated on the second tyrosine (19, 20). Interestingly, the same motif is also the binding site for the Numb PTB domain, although Numb binds when the motif is nonphosphorylated (41). A NVYpYY peptide binds to an unusual structure formed by Hakai dimerization. The Hakai sequence contains a classic C₃HC₄ RING followed by a zinc-binding C₂H₂ motif known as a pTyr-binding (pTyrB) domain. In the dimer, two pTyrB regions come together in an antiparallel arrangement, mediated by strand exchange and stabilized by Zn ions and the RING domains, to create a basic channel that binds pY (20) (Fig. 2D). This dimeric structure is known as an HYB (Hakai pY-binding) domain. The HYB domain recognizes, in addition to the NVYpYY target sequence, the flanking acidic residues. In the same vein, Hakai has been shown to bind other acidic pY-containing sequences found in Cortactin and Dok1, but the physiological significance of these interactions is not defined (20). The Hakai-related proteins ZNF645, LNX1, and LNX2 also have RING-pTyrB homology, but the functions and structures of these domains are unknown (20).

pY SUBSTRATE RECOGNITION BY SOCS PROTEIN ADAPTORS FOR CRL5

The C termini of all SOCS proteins contain an SH2 domain that binds pY, followed closely by a SOCS-BC box for binding to Cul5 and the elongin BC subunits of CRL5 (Fig. 2E). The SH2 domain positions a pY substrate on the same side of Cul5 as Rbx2, which contains a RING domain and positions an E2-ubiquitin conjugate for ligation (Fig. 2F). SOCS SH2 domains bind pY peptides through modified SH2-pY interactions, as shown for the SOCS6 SH2 domain in complex with a juxtamembrane phosphorylation site, Tyr568, in c-Kit (42) (Fig. 2G). Although the SOCS SH2 domain adopts a typical SH2 fold, it features an unusually long BG loop, and together with the EF loop, they form an elongated channel to allow interactions with residues between 3 and 6 residues C terminal to the pY in the c-Kit peptide. This unique binding mode dictates that only certain phosphopeptides (and the corresponding proteins) would be targeted by SOCS6.

SOCS proteins may also recruit non-pY substrates to CRL5. SOCS7 contains a proline-rich region that binds the SH3 domains of Nck, Grb2, PLCγ, and vinexin (43, 44). SOCS7 shuttles between the nucleus and cytoplasm and acts as a bridge between septins and Nck (45). However, there is no evidence that tyrosine phosphorylation or ubiquitination is involved. Tyrosine phosphorylation of SOCS3 creates a binding site for the SH2 domains of Nck and Crk/CrkL (46). It is again unclear if Nck or Crk/CrkL is a substrate for ubiquitination. Finally, using overexpressed proteins, the N-terminal region of SOCS6 was shown to bind to the SH2 kinase region of active (but not inactive) Lck and less well to other Src-related kinases (47).

SUBSTRATES AND FUNCTIONS OF Cbl

The first insights into Cbl function came from Caenorhabditis elegans genetics (48). Mutation of Cbl restored normal development to worms containing a weakly active mutant epidermal growth factor receptor (EGFR), implying that c-Cbl inhibits the EGFR. Subsequent animal cell studies showed that EGF stimulates the binding of c-Cbl to EGFRs and Cbl tyrosine phosphorylation (49). While EGF-activated EGFRs are targeted to the lysosome, an EGFR relative, ErbB4, which does not bind Cbl, simply recycles (50). This suggested a role for c-Cbl in trafficking between endosomes and lysosomes. Indeed, overexpression of c-Cbl stimulated EGF-dependent EGFR ubiquitination, internalization, and degradation, which was dependent on the Cbl RING domain (39, 51). Following the discovery that the c-Cbl RING domain has intrinsic E3 ligase activity (18), the results suggested that EGF stimulation induces c-Cbl binding and EGFR ubiquitination and that ubiquitination then signals lysosomal trafficking. Subsequent studies have filled in critical details about how c-Cbl binds and ubiquitinates the EGFR and about the importance of ubiquitination for lysosomal degradation (52).

The basic steps in EGFR downregulation by c-Cbl are as follows (52). EGF binding activates the EGFR kinase domain and leads to autophosphorylation on Tyr1068 and Tyr1086. These sites then bind an adaptor protein, Grb2, which binds in turn to a proline-rich region in c-Cbl. The complex is further stabilized by a second, direct interaction of the c-Cbl TKB domain with the EGFR pY1045 site. c-Cbl then transfers ubiquitin onto multiple lysine residues within the EGFR (53). These lysines are mono-, not poly-, ubiquitinated (54, 55). While multimonoubiquitination can trigger receptor endocytosis, it is not required (52, 55). Rather, multimonoubiquitination stimulates sorting to multivesicular bodies and degradation in the lysosome. Sorting is presumably mediated by ubiquitin-binding domains contained in Hrs and other components of the ESCRT0 complex (52). Thus, a mutant EGFR lacking all ubiquitination sites is internalized but not degraded (53), and wild-type EGFRs are internalized but not degraded in c-Cbl knockout fibroblasts (56).

c-Cbl-dependent multimonoubiquitination also downregulates other RTKs and the gp130 cytokine receptor, thereby inhibiting responses to a range of growth factors (57, 58). Removal of Cbl might thus be expected to increase mitogenic signaling. Indeed, c-Cbl is frequently mutated in lung cancer (59). EGFR is also frequently mutated in lung cancer, and some mutations in EGFR may decrease its downregulation by c-Cbl, leading to increased EGFR activity (60). EGFR ubiquitination is also implicated in breast cancer; an EGFR-specific DUB named Cezanne-1 is upregulated in many cases of breast cancer (61). c-Cbl and Cbl-b are also required for routing activated, tyrosine-phosphorylated T cell receptors to the lysosome for degradation, thus downregulating the immunological response and suppressing autoimmunity (62).

Despite the importance of c-Cbl in receptor trafficking, cancer, and immunity, mutant mice completely lacking c-Cbl are viable and fertile (63). The mild phenotype may be due to redundancy with Cbl-b or Cbl-c. Indeed, close examination revealed alterations in hematopoiesis, T cell response, mammary epithelial growth, and fat and energy metabolism, consistent with c-Cbl stimulating the downregulation of various receptors (63, 64). Curiously, point mutation of the c-Cbl RING domain caused a much stronger phenotype—death in utero—suggesting that c-Cbl has significant developmental roles (65). The c-Cbl RING mutant may compete with Cbl-b and Cbl-c, allowing the true scope of c-Cbl functions to be exposed. However, the interpretation is challenging because evidence suggests that the c-Cbl RING mutant has gain-of-function effects. First, mice expressing half the normal amount of the RING mutant were viable and showed the same fat and energy defects as the knockout mice, indicating that these phenotypes likely require ubiquitin transfer (65). Second, the RING mutant mice suffer complete thymic ablation due to hyperactivation of phosphatidylinositol 3′ (PI3′) kinase, not seen in the knockout mice (66). The results suggest that both signal inhibition by c-Cbl's RING and signal activation by c-Cbl's adaptor function occur in vivo.

Human cancer mutations also indicate that c-Cbl has both positive and negative effects in vivo. Many c-Cbl mutations found in myeloid neoplasms are heterozygous missense and frameshift mutations (67, 68). These mutations generally leave the TKB domain and LHR intact, so the mutant proteins may act as dominant-interfering alleles, competing with endogenous Cbl proteins for binding sites on the receptor. However, they may have alternatively gained mitogenic activity. Tyr371 substitutions, frequently found in spontaneous neoplasms and in a cancer predisposition syndrome called Noonan syndrome, present a curious case (67, 69). Since phosphorylation of Tyr371 stimulates ubiquitination of bound pY substrates (Fig. 2C), Tyr371 mutation is likely to inhibit ubiquitination. However, these alleles transform cells only when endogenous c-Cbl is absent (68, 69). This is difficult to explain as a dominant interfering effect and suggests that these c-Cbl mutants may have gained oncogenic activity. c-Cbl mutants that have decreased ubiquitination activity may be more stable, and increased c-Cbl protein levels may stimulate signaling through an adaptor function.

Some c-Cbl functions may be due to ubiquitination of cytoplasmic tyrosine kinases and pY proteins, but ubiquitination-independent mechanisms may also play a role. ZAP70 is a case in point. A major phenotype of c-Cbl knockout mice is lymphoid hyperplasia and increased T cell signaling via the T cell receptor-associated kinase ZAP70 (63, 70, 71). However, ZAP70 degradation is not altered. Similarly, when Cbl-b is overexpressed in T cells, the p85 subunit of PI3′ kinase is ubiquitinated, but its steady-state abundance and half-life are not affected (72). Nevertheless, ubiquitination inhibits the recruitment of p85 to the coreceptor CD28. Decreased p85 ubiquitination and increased PI3′ kinase activity may explain the hyper-activation of Cbl-b knockout T cells and consequent autoimmunity, but p85 degradation may not be involved (73, 74). Cbl-b also regulates the phosphatase PTEN, independently of its ubiquitin ligase activity (75). These results suggest ubiquitin-independent and degradation-independent effects of both c-Cbl and Cbl-b.

c-Cbl has also been implicated in the degradation of the Src kinase. Activation induces Src downregulation by ubiquitination and proteasomal degradation (76, 77). Overexpressed Cbl was found to stimulate ubiquitination of Src and Fyn (a Src-related kinase), implicating endogenous Cbl in normal cells (78, 79). Indeed, Fyn protein levels are increased in c-Cbl knockout cells (79), and overexpressed truncated Cbl TKB inhibits Src turnover, as expected if the TKB has a dominant negative effect (80). However, overexpressed Cbl TKB may inhibit the binding of other Src-specific E3 ligases. Indeed, other evidence, described below, suggests that SOCS-CRL5 complexes may be responsible for downregulating endogenous Src. Moreover, genetic experiments in osteoclasts are not consistent with Cbl inhibiting Src but are consistent with Src requiring Cbl for signaling (81–83). Src and other tyrosine kinases may phosphorylate Cbl at sites that bind PI3′ kinase and Crk, thus stimulating signaling (55). Phosphorylation of Cbl by Src also stimulates Cbl ubiquitination and degradation. Decreased Cbl means that there is less downregulation of RTKs and more RTK activity (78, 84). This may contribute to the synergy between Src and RTKs for transformation (85).

The original concept that Cbl family proteins target cytoplasmic proteins like p85, ZAP70, and Src for degradation is difficult to reconcile with the findings that c-Cbl catalyzes monoubiquitination of EGFR and gp130 (54, 55, 58). Monoubiquitination triggers lysosomal but not proteasomal degradation, and cytoplasmic proteins are not normally degraded by the lysosome. However, monoubiquitination can affect the binding or enzymatic properties of substrates independently of degradation. Indeed, monoubiquitination of CARMA1 by Cbl-b inhibits binding of CARMA1 to Bcl10 without inducing degradation, thereby inducing NKT cell anergy (86). In general, Cbl family proteins may stimulate degradation of transmembrane proteins by the lysosome but have other effects, not related to degradation, on cytoplasmic proteins.

CADHERIN REGULATION BY Hakai

E-cadherin is the best documented substrate for Hakai (19). Hakai overexpression increases E-cadherin ubiquitination, endocytosis, and lysosomal degradation, dependent on tyrosine phosphorylation and the NVYYY sequence. This sequence is part of a longer motif that binds p120Ctn, a protein that stabilizes E-cadherin on the cell surface (87). p120 binds only the nonphosphorylated sequence; phosphorylation thus switches E-cadherin from binding to p120 to binding to Hakai (19, 88–90). Competition between Hakai, p120Ctn, and Src for the same site makes it difficult to interpret the effects of E-cadherin mutations and knockdown or overexpression of Hakai or p120Ctn on E-cadherin stability. Nevertheless, there is good evidence that Hakai promotes E-cadherin downregulation from adherens junctions and induces the epithelial-mesenchymal transition (EMT) (91). Thus, removing calcium induces downregulation of E-cadherin by a series of steps, including activation of Cdc42, EGF receptor, and Src; phosphorylation of E-cadherin; Hakai-dependent ubiquitination; and lysosomal degradation (92). Overexpressing Hakai, or stimulating E-cadherin phosphorylation by activating Src or RTKs, also drives E-cadherin downregulation and junction disassembly (19, 93). Transforming growth factor β (TGFβ), in combination with Ras pathway activation, stimulates Hakai gene expression and induces E-cadherin downregulation and EMT (94). Despite this evidence, it remains unclear whether endogenous Hakai regulates EMT in vivo. A Hakai knockout mouse has not been described. Knockdown of Hakai in cancer cell cultures does not restore cell-cell junctions but does inhibit cell proliferation, independently of E-cadherin (95). Hakai expression inversely correlates with E-cadherin in some cancers but not others (96). It will be important to determine whether Hakai is required for EMT during normal embryonic development and cancer progression.

CELL REGULATION BY SOCS-CRL5 COMPLEXES AND BY SOCS PROTEINS ACTING ALONE

SOCS proteins are best understood as inhibitors of cytokine receptors and RTKs. Cytokines induce the transcription of SOCS1, -2, and -3 and CISH genes by a cascade of cytokine receptor phosphorylation, JAK kinase binding and activation, and phosphorylation of STAT transcription factors (97–99). The phosphorylated STATs then move to the nucleus and activate expression of many genes, including SOCS genes. SOCS proteins then provide feedback inhibition by binding to cytokine receptors and JAK kinases. Accordingly, mutations of SOCS1 (100–102) and SOCS3 (103–105) have major effects on erythropoiesis, the immune system, and the placenta, while CISH deletion also affects immunity but in a more subtle way (106).

While SOCS proteins may inhibit signaling by CRL5-dependent downregulation of tyrosine-phosphorylated cytokine receptors or JAKs, several SOCS proteins have major effects that are independent of CRL5. Specifically, SOCS1 and -3 bind poorly to the remainder of the CRL5 complex (107) and are able to inhibit signaling through two ubiquitin-independent mechanisms. On the one hand, they bind through their SH2 domains to phosphorylated cytokine receptors, thereby competing for STAT binding, and on the other, they bind to and inhibit JAKs via the “kinase inhibitor regions” (KIRs) immediately N terminal to their SH2 domains (98). Indeed, germ line mutation of the SOCS1 BC-SOCS box has a much reduced effect on gamma interferon signaling in vivo than a complete SOCS1 gene ablation, suggesting that CRL5 binding is largely dispensable for in vivo function (108). Even so, under some conditions, SOCS1 and -3 can stimulate ubiquitination; SOCS1 stimulates polyubiquitination and proteasomal degradation of pY-JAK2 when overexpressed (109), and SOCS3 can stimulate ubiquitination of JAK2 and gp130 in vitro (110).

The ubiquitination activities of SOCS1 and -3 may be more important in suppressing signaling through adhesion receptors and RTKs (111, 112). Overexpressed SOCS3 binds focal adhesion kinase (FAK) through a pY in the activation loop and stimulates FAK turnover (111). This may be important in the adhesive response of B lymphocytes to the chemokine CXCL12 (113). As B cells mature, SOCS3 expression increases and CXCL12-induced FAK activation and adhesion decrease. SOCS2 has a similar function in regulating the FAK-like kinase Pyk2 in NK cells (114). SOCS2 expression is induced in NK cells by interleukin 15 (IL-15) and binds Pyk2 through the activation loop pY. SOCS2 stimulates Pyk2 ubiquitination and degradation. Accordingly, the knockdown of SOCS2 caused the accumulation of pY-Pyk2 and blocked NK cell effector functions.

Perhaps the best-understood target for SOCS2 is the growth hormone (GH) receptor, which signals through a classical JAK-STAT mechanism. SOCS2 knockout causes mice to grow excessively, to 40% larger than their littermates, with symptoms of increased GH and insulin-like growth factor 1 (IGF1) signaling (115). Mutant cells have increased responses to GH but not IGF1, which is likely activated indirectly (116). Genetic experiments revealed that the SOCS2 phenotype requires increased STAT5b signaling from pY595 of the GHR (116). SOCS2 inhibition of GH signaling requires the SOCS box and SH2 domain, pointing to a role for CRL5 and suggesting that SOCS2 brings CRL5 to pY-GHR and stimulates ubiquitination and degradation (117). Curiously, overexpression of SOCS2 in a transgenic mouse also stimulated growth (118), suggesting that SOCS2 both stimulates and inhibits GH responses. One potential explanation is that overexpressed SOCS2 (and perhaps SOCS6 and SOCS7) negatively regulates other SOCS proteins (119). However, it is not clear whether this mechanism explains the increased growth of SOCS2 transgenic mice.

SOCS4, -5, -6, and -7 are more widely expressed than SOCS1, -2, and -3 and CISH (112). They have functions beyond cytokine signaling. For example, both SOCS4 and -5 are induced by EGF (120). Overexpressed SOCS5 inhibits EGFR signaling through its SOCS box and SH2 domain (120, 121). EGFR degradation is increased, but curiously, both inactive and active receptors are degraded (120). This suggests that SOCS5 might downregulate the EGFR independently of pY, but the recognition mechanism and biological significance are unknown. SOCS5 knockout mice have no apparent phenotype (122), and the SOCS4 mutant phenotype has not been reported. SOCS4 and -5 are closely related and potentially overlap in function, so double mutation may be necessary to detect phenotypes.

SOCS6 and -7 may also have overlapping functions. They are closely related in their SH2 domains, although they diverge considerably in their N-terminal regions. Both SOCS6 and SOCS7 bind to pYVϕϕ sequences, where ϕ is a hydrophobic residue (123). Both can bind to insulin receptor substrates 2 and 4 (IRS2 and IRS4, respectively) and to the p85 subunit of PI3′ kinase. However, deletion of SOCS6 has little effect on insulin sensitivity or other reported phenotypes, although the mice were a little smaller than controls (123). Potential overlapping functions with SOCS7 may conceal SOCS6 mutant phenotypes. SOCS6 also binds the juxtamembrane region of the RTK c-Kit, and overexpressed SOCS6 inhibits c-Kit activation and shortens its half-life (42, 124). However, it is unclear whether SOCS6 triggers c-Kit lysosomal or proteasomal degradation or whether endogenous SOCS6 regulates c-Kit.

Unlike with SOCS6, deletion of SOCS7 causes many phenotypes, depending on the particular allele and the genetic background. Systemic SOCS7 deletion was reported to hyper-activate mast cells, with increased proinflammatory cytokine production and severe skin disease (125). SOCS7 deletion caused hydrocephalus in other mouse strains (126). Systemic SOCS7 deletion can also increase insulin sensitivity by stabilizing IRS proteins, which mediate insulin signaling, in pancreatic β cells (127). Conditional knockout of SOCS7 in neural progenitors during brain development altered the migration of specific classes of neurons, consistent with hyper-responsiveness to reelin, a migration-stimulatory factor (128). SOCS7 binds to Dab1, a protein that is phosphorylated at tyrosine in reelin-stimulated neurons, and stimulates ubiquitination of pY-Dab1 in vitro. Dab1 protein accumulates in misplaced neurons in SOCS7 and Rbx2 mutant brains, consistent with a requirement for SOCS7-CRL5-dependent turnover of pY-Dab1 for normal neuron migration.

CELL REGULATION BY CRL5: STUDIES WITH Cul5 AND Rbx2

Removal of one of the CRL5 invariant subunits, Cul5 or Rbx2, might be expected to inhibit all SOCS gene functions, unveiling phenotypes that might be concealed by redundancy when a single SOCS gene is deleted. However, SOCS proteins are not the only substrate adaptors for CRL5: approximately 30 other proteins that lack SH2 domains may also recruit non-pY substrates to CRL5 (129). Removing Cul5 or Rbx2 may affect these nonphosphorylated substrates as well as SOCS protein ligands. One way to determine which Cul5 or Rbx2 phenotypes might involve pY-dependent ubiquitination is to focus on phenotypes and substrates that also require tyrosine kinases.

Along these lines, the removal of Cul5 and/or Rbx2 from fibroblasts, epithelial cells, and developing brain increases the level of active but not inactive Src and Fyn proteins but not mRNA (128, 130–133). This suggests that CRL5 complexes regulate Src/Fyn stability, but changes in half-life have not been reported. Knockdown or knockout of Cul5 or Rbx2 also increases fibroblast transformation and epithelial cell proliferation and migration, depending on Src, suggesting that the pY substrates are involved (130–133). The specific pY substrates suppressed by Cul5 in fibroblasts are not known, but the Src substrate p130Cas is required for the increased proliferation and migration of Cul5-deficient epithelial cells (130). p130Cas turnover requires Src, Cul5, and SOCS6. Phosphorylated p130Cas binds SOCS6. This suggests that SOCS6-CRL5 inhibits proliferation and migration by targeting p130Cas for ubiquitination and degradation after it has been phosphorylated by Src. The full picture is more complicated, however. Combined knockdown of several SOCS proteins is required to stimulate cell migration to the same extent as Cul5 (130). SOCS7 knockdown stimulates migration of some cancer cell lines (134). In addition, SOCS3 overproduction inhibits keratinocyte proliferation and migration and wound healing in vivo (135). Therefore, several SOCS proteins probably target multiple pY proteins for ubiquitination, and together they inhibit cell proliferation and migration.

Such inhibitory effects suggest that SOCS-CRL5s may function as tumor suppressors. Indeed, SOCS6 is strongly downregulated in the most-severe cases of prostate cancer (136), hepatocellular carcinoma (137), lung cancer (138), and gastric cancer (139). SOCS3 is downregulated in head and neck cancer (140) and Barrett's adenocarcinoma (141). High expression of SOCS4 and -7 is associated with improved clinical outcome in breast cancer (142), loss of SOCS5 correlates with poor outcome in liver cancer (143), and decreased SOCS2 expression correlates with shorter recurrence-free survival in prostate cancer (136). In addition, SOCS1 and -2 are hypermethylated and underexpressed in more-severe cases of ovarian and breast cancer (144, 145). These findings are consistent with tumor suppression by SOCS proteins.

However, it should be noted that some evidence suggests that the Rbx2 gene may also function as an oncogene in certain contexts. Rbx2 levels increase in some tumors (35). Tan et al. reported that Rbx2 is required for an Rbx2-Cul1-Fbxw7 complex to stimulate ubiquitination and degradation of the tumor suppressor NF1, consistent with the Rbx2 gene being an oncogene (146). On the other hand, other research indicates that NF1 is degraded by a Cul3 mechanism (147) and that Fbxw7 normally binds to Cul1-Rbx1 via Skp1 as opposed to elongin BC (148). It is an open question whether Rbx2 may function separately from Cul5.

SUMMARY AND FUTURE PROSPECTS

Despite having different mechanisms for binding to their substrates and associating with E2-ubiquitin conjugates, Cbl, Hakai, and SOCS proteins all have the potential to promote pY-dependent ubiquitination and provide negative regulation of tyrosine kinase signaling pathways. This potential has been brought out by tissue culture and overexpression experiments. However, Cbl and SOCS proteins, and quite probably Hakai, also have ubiquitin-independent functions, making it difficult to tease out the biological importance of pY-dependent ubiquitination in vivo. Indeed, for Cbl and SOCS proteins, where knockout data are available, some of the phenotypes are likely due to loss of ubiquitin-independent functions. A continuing challenge for the future is the identification of bona fide in vivo pY proteins whose ubiquitination and degradation by pY-dependent ubiquitin ligases is critical for normal development or homeostasis. Given the low stoichiometry of many tyrosine phosphorylation events, pY-dependent turnover may not detectably change the bulk abundance or half-life of a substrate. New approaches for substrate identification will be key.