Nucleolar Targeting of the Fbw7 Ubiquitin Ligase by a Pseudosubstrate and Glycogen Synthase Kinase 3

Abstract

E3 ubiquitin ligases catalyze protein degradation by the ubiquitin-proteasome system, and their activity is tightly controlled. One level of regulation involves subcellular localization, and the Fbw7 tumor suppressor exemplifies this type of control. Fbw7 is the substrate-binding component of an SCF ubiquitin ligase that degrades critical oncoproteins. Alternative splicing produces three Fbw7 protein isoforms that occupy distinct compartments: Fbw7α is nucleoplasmic, Fbw7β is cytoplasmic, and Fbw7γ is nucleolar. We found that cancer-associated Fbw7 mutations that disrupt substrate binding prevent Fbw7γ nucleolar localization, implicating a substrate-like interaction in nucleolar targeting. We identified EBNA1-binding protein 2 (Ebp2) as the critical nucleolar factor that directly mediates Fbw7 nucleolar targeting. Ebp2 binds to Fbw7 like a substrate, and this is mediated by an Ebp2 degron that is phosphorylated by glycogen synthase kinase 3. However, despite these canonical substrate-like interactions, Fbw7 binding is largely uncoupled from Ebp2 turnover in vivo. Ebp2 thus acts like a pseudosubstrate that directly recruits Fbw7 to nucleoli.

The SCF^Fbw7 ubiquitin ligase controls the activities of critical cellular regulatory proteins, including cyclin E, c-Myc, c-Jun, and Notch (8, 18, 22, 23, 25, 27, 41, 46, 47, 50, 52, 54). The substrate recognition component of this SCF complex is the Fbw7 F-box protein, which binds to phosphorylated substrates and brings them into proximity with ubiquitin-conjugating enzymes (4, 30, 31, 39). Polyubiquitylated substrates are then rapidly eliminated by the 26S proteasome.

Fbw7 binding is mediated by a phosphorylated substrate motif, termed the Cdc4 phosphodegron (CPD) (24). In addition to hydrophobic residues, CPDs usually contain two critical phosphates (in the 0 and +4 positions) that are recognized by specific binding pockets within the Fbw7 WD40 repeats (10, 29, 42, 51, 55). In most cases, the central CPD threonine is phosphorylated by glycogen synthase kinase 3 (GSK3), thereby connecting Fbw7 activity to mitogenic signaling pathways (47). Mutations in tumors abrogate these binding interactions and allow oncogenic Fbw7 substrates to accumulate and contribute to carcinogenesis. In some tumors, substrate CPDs themselves are mutated, allowing oncoproteins to evade destruction by Fbw7 (47). More commonly, however, mutations target the Fbw7 substrate-binding domain, and Fbw7 is a tumor suppressor that is mutated in many cancers (1, 16, 34, 47). Three arginine residues within the WD40 repeats form a CPD-binding pocket and are subject to frequent missense mutations (here termed Fbw7^ARG). Fbw7^ARG account for approximately half of the Fbw7 mutations found in tumors and nearly all Fbw7 mutations in some cancers, such as T-cell acute lymphocytic leukemia (19, 20, 28, 43).

The Fbw7 gene encodes three mRNAs with unique 5′ exons and 10 common exons (40). The three Fbw7 protein isoforms share all known functional domains: (i) the WD40 repeats, which bind to substrate CPDs, (ii) the F-box, which recruits the remainder of the SCF complex for ubiquitin ligase activity, and (iii) the D-domain, which mediates dimerization (Fig. 1 B). However, the three Fbw7 isoforms differ in one crucial aspect: their localization. Fbw7α is nuclear, Fbw7β is cytoplasmic, and Fbw7γ is predominantly nucleolar (17, 49, 55). We previously identified cis-acting signals within the Fbw7α- and Fbw7β-specific 5′ exons that target these proteins to the nucleus and cytoplasm, respectively. The Fbw7γ-specific exon, however, is dispensable for nucleolar targeting, and the common exons specify nucleolar localization that can be overridden by dominant motifs in the Fbw7α- and Fbw7β-specific exons (49).

FIG. 1.

Regulation of Fbw7γ localization. (A) Mislocalization of common cancer-associated Fbw7γ arginine mutations. U2OS cells were transiently transfected with pFLAG-Fbw7γ or the indicated Fbw7γ^ARG mutant and stained with Flag antibody. (B) Schematic of the Fbw7γ isoform, the known functional domains, and the three arginine residues that form the substrate CPD-binding pocket. N, nuclear localization signal; D, dimerization domain; FB, F-box; WD40, WD40 repeats for substrate binding. (C) Ebp2 recruited Fbw7 to nucleoli in a trans-localization assay. Cytoplasmic Flag-Fbw7γ ΔNLS was coexpressed with nucleolar Fbw7-binding candidates and stained with Flag antibody to detect Fbw7γ ΔNLS and with Myc tag (MT) antibody to detect the coexpressed nucleolar candidate protein. (D) Alignment of Fbw7 phospho-degrons. The central threonine and the serine in the +4 position are required phosphorylation sites for Fbw7 binding (shown in bold).

The unique compartmentalization and transcriptional controls of each Fbw7 isoform suggest that they perform nonredundant functions, and it remains unclear which isoform(s) has tumor-suppressor activity. Proliferating cells express all isoforms. Adult mouse tissues ubiquitously express Fbw7α, whereas Fbw7β is predominant in neural tissues and Fbw7γ is predominant in muscle (7, 17, 21, 40, 41). Fbw7α is the most abundant isoform and carries out the bulk of known Fbw7 functions (most known substrates are nucleoplasmic) (6), whereas no clear cytoplasmic Fbw7β substrates are known. Fbw7γ regulates Myc in the nucleolus (3, 49) and may be required for cyclin E degradation in some contexts (45). Importantly, Fbw7γ nucleolar localization is disrupted in some cancers. The simian virus 40 (SV40) large T antigen prevents nucleolar Fbw7γ localization by acting as a decoy substrate (48). Nucleophosmin 1 (NPM1) has also been implicated in Fbw7γ nucleolar localization, and leukemia-associated NPM1 mutations impair Fbw7γ localization (3).

We found that tumor-associated Fbw7γ ^ARG mutants cannot localize to nucleoli. Because Fbw7^ARG mutations prevent binding to substrate CPDs, we hypothesized that Fbw7 nucleolar localization requires a substrate-like interaction. Although we independently found that NPM1 is required for Fbw7γ nucleolar targeting, NPM1 does not contain a consensus CPD, and we sought to identify nucleolar proteins that bind to Fbw7 like a substrate. We report that EBNA1-binding protein 2 (Ebp2) is the critical nucleolar factor that directly mediates Fbw7 nucleolar localization. We show that Ebp2 contains a canonical CPD that is phosphorylated by GSK3 on threonine 3 and mediates Fbw7 binding. Importantly, Ebp2 expression, its phosphorylation on threonine 3, and GSK3 activity are all required for Fbw7 nucleolar localization. Enforced Ebp2 expression also translocates Fbw7α to nucleoli, and we identified cancer cell lines with predominant nucleolar localization of Fbw7α. Finally, we found that although Ebp2 is an Fbw7 substrate in some contexts, Fbw7 binding is largely uncoupled from Ebp2 turnover in vivo, and this may result from the selective inability of the SCF^Fbw7-associated ubiquitin-conjugating enzyme hCdc34 to ubiquitylate Ebp2. We conclude that Ebp2 is a pseudosubstrate that directly targets Fbw7 to nucleoli.

MATERIALS AND METHODS

Cell culture, transfections, immunofluorescence, and lysate preparation.

U2OS (human osteosarcoma), HEK-293A (human embryonic kidney), human primary foreskin fibroblast (HFF), and HeLa and C33A (both cervix) cell lines were cultured under standard conditions in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. For transient transfections, cells were seeded into 6-cm dishes and transfected the next day by calcium precipitation overnight at 30 to 40% confluence. Twenty-four hours after washing and replacing the medium, cells were harvested. For immunofluorescence, cells were seeded and transfected on glass coverslips. Slips were fixed with ice-cold methanol/acetone (1:1) for 5 min, air dried, and immunostained with primary antibody. After washing with phosphate-buffered saline, slips were incubated with secondary fluorescein isothiocyanate-coupled antibody, washed, dried, and mounted. Images were captured using a Nikon Eclipse E800 microscope, Nikon Plan Flour 40×/1.30 numerical aperture oil objective, APOT RT Slider camera, and SPOT RT software v3.0.

All protein extracts were made with NP-40- or Tween 20-containing lysis buffer. Protein expression and protein complexes were analyzed by standard procedures (SDS-PAGE, Western blotting, immunoblotting, and immunoprecipitation [IP]). As observed by others (36), direct immunoprecipitation of Ebp2, even of tagged overexpressed Ebp2, has proved to be extremely inefficient despite its apparent solubility in cleared cell lysates. This is independent of cell lysis conditions (Tween 20, NP-40, radioimmunoprecipitation assay buffer, TX-100), epitope (Flag, hemagglutinin [HA], or Myc tag) or position (N versus C terminal) of the antibody tag, or with antibody targeting Ebp2 directly, indicating that most Ebp2 is inaccessibly sequestered within native protein lysates. C-terminal truncation of Ebp2 resolved this problem (data not shown), and a C-terminally truncated version of Ebp2 (amino acids 1 to 211) was therefore used for the IP shown below in Fig. 5A.

Antibodies, plasmids, and cDNAs.

The following primary antibodies were used in this study: anti-Flag (M2; Sigma), anti-HA (12CA-5), mouse anti-myc tag (9E-10), rabbit anti-myc tag (catalog number 2278; Cell Signaling), anti-Ebp2 (rabbit polyclonal antibody) (38), phospho-T58 Myc (primarily recognizes pT3-Ebp2; catalog number 9401S; Cell Signaling), NPM1 (catalog number 32-5200; Zymed), GSK-3β (catalog number G22320; Transduction Laboratories). GSK inhibitor VIII was purchased from Calbiochem.

pFLAG-Fbw7γ and its F-box deletion have been described elsewhere (50). All point mutants and deletions were introduced by using the QuikChange method (Stratagene) and confirmed by sequencing. Ebp2 and the other nucleolar candidates were cloned from U2OS, 293A, or HeLa cDNA and subcloned into a Myc-tagged version of the pCS2+ vector and sequenced. The other candidate genes not shown were nucleostemin, DDX21, PES1, NUSAP, PP1γ, p120, NOPP34, GNL3L, and several ribosomal subunits. Plasmids expressing the GSK3 interaction domain (GID) and GID*, a nonfunctional point mutant of this peptide, have been described elsewhere (11). Small interfering RNAs (siRNAs; Integrated DNA Technologies) were directed at the following Ebp2 target sites: si-1 (273-GGACCAGAAAGCTGTTGATCCAGAA), si-2 (640-GAGGGAGATCAGAAACCTCTGGCAC). The NPM1 shRNA target sequence was 161-AGGATGAGTTGCACATTGTTGAAGC. The GSK3 shRNA has been described elsewhere (51).

Kinase and ubiquitylation assays.

For the GSK3 kinase reaction, truncated forms of Ebp2 (amino acids 1 to 211) and its T3A mutant (serving as a control) were immunoprecipitated from transfected cell lysates and subjected to standard reactions using commercial GSK3 enzyme (catalog number P6040S; buffer provided; BioLabs). In vitro ubiquitylation of Ebp2 was carried out on soluble, in vitro-transcribed/translated (with 20 μl of rabbit reticulocyte lysate) Ebp2, or its CPD mutant, with the following recombinant proteins: Fbw7/Skp1 complex (825 nM), neddylated Cul1/Rbx1 complex (1 μM), ubiquitin (75 μM), UbcH5 or hCdc34 (5 μM), and UBE1 (560 nM). Reaction mixtures also contained ATP (2 mM), MgCl₂ (10 mM), and 50 μM bortezomib.

RESULTS

Ebp2 recruits Fbw7 to nucleoli.

Fbw7γ is normally a nucleolar protein in proliferating cells. In contrast, we noted that each of the three most common Fbw7γ ^ARG mutants failed to localize to nucleoli but were instead nucleoplasmic (Fig. 1A). Fbw7^ARG thus represents another type of cancer-associated mutation that disrupts Fbw7 nucleolar localization. Because these arginine residues make critical contacts with phosphorylated substrate CPDs (Fig. 1B), we speculated that Fbw7γ nucleolar targeting requires a substrate-like interaction that is disrupted by Fbw7^ARG mutations. We searched databases for nucleolar proteins that contain potential consensus CPD motifs, cloned 10 candidate genes (see Materials and Methods), and tested them in an immunofluorescence-based binding assay (48). Because wild-type Fbw7γ is already nucleolar, this assay utilized an Fbw7γ mutant that lacks the nuclear localization signal and mislocalizes to the cytoplasm (Fbw7γ ΔNLS). We thus screened for proteins that could relocalize Fbw7γ ΔNLS to nucleoli.

We coexpressed Fbw7γ ΔNLS and nucleolar candidate proteins in U2OS cells and examined each protein by immunofluorescence. As expected, Fbw7γ ΔNLS was mislocalized to the cytoplasm (Fig. 1C). One candidate, Ebp2, efficiently recruited the cytoplasmic Fbw7γ mutant to the nucleolus, suggesting that both proteins interacted. No other candidates relocalized Fbw7γ ΔNLS to nucleoli, including NPM1 and Pbk1, the human homolog of yeast cic1p, which binds to the Fbw7 homolog in Saccharomyces cerevisiae (13) (Fig. 1C and data not shown).

Ebp2 binding to Fbw7 is regulated by Ebp2 phosphorylation.

Ebp2 is a 40-kDa nucleolar protein that is essential for mammalian cell proliferation and ribosome biogenesis (12, 14, 44). Ebp2 expression is mitogen regulated in some contexts and may impact mitogenic signaling, although it is not known whether this requires its ribosomal functions (35). Ebp2 also regulates mitotic segregation of Epstein-Barr virus episomal genomes via its interaction with Ebna1 (15, 26, 53).

Ebp2 contains a consensus CPD at its extreme N terminus that is centered on threonine 3 and highly related to the c-Myc T58 degron (Fig. 1D). Remarkably, we had previously identified Ebp2 as the major cellular protein detected by an anti-phospho-T58 Myc antibody in immunoblot assays, and this required Ebp2 T3 phosphorylation (9). Moreover, Ebp2 knockdown completely eliminates the nucleolar immunostaining detected by this antibody in immunocytochemical assays (data not shown). This cross-reactivity further supports the idea that the Ebp2 and Myc CPDs are structurally similar and enables the immunodetection of Ebp2 T3 phosphorylation. Because Ebp2 is T3 phosphorylated in vivo, we hypothesized that the nucleolar targeting of Fbw7γ ΔNLS in Fig. 1C is mediated via a direct interaction with phosphorylated Ebp2. We confirmed that Ebp2 and Fbw7 bind each other by using reciprocal coimmunoprecipitation (Fig. 2 A). This binding was dependent on the Ebp2 CPD and, like other Fbw7 substrates, mutation of either the central phospho-threonine (T3) or the serine in the +4 position (Ser7) to alanine residues prevented binding (Fig. 2A and data not shown). As expected, the interaction was also disrupted by Fbw7^ARG mutations (Fig. 2B; see also Fig. 6C, below).

FIG. 2.

Phosphorylation-dependent interaction between Ebp2 and Fbw7γ. (A) Coimmunoprecipitation of Fbw7 and Ebp2. 293A cells were transfected with Flag-tagged Fbw7 and Myc-tagged Ebp2 (or their mutants) as indicated, and lysates were immunoprecipitated in reciprocal fashion and analyzed by Western blotting using tag antibodies. The asterisk marks the heavy chain of the IP antibody. The Fbw7γ ΔF mutant has a deletion of the F-box, and an alanine replaces Thr3 in the Ebp2-T3A mutant. (B) Cancer-associated Fbw7^ARG mutants fail to bind Ebp2. The assay was similar to that shown in panel A, except that Ebp2 was detected with an Ebp2-specific primary antibody instead of via its Myc tag, which demonstrated the amount of relative overexpression compared with endogenous Ebp2 (typically within 1 order of magnitude). The lysate Ebp2 blot was subsequently reprobed for Fbw7 (Flag antibody) without prior stripping. (C) Nucleolar localization of Fbw7γ ΔNLS is dependent on Ebp2-T3 phosphorylation. U2OS cells were cotransfected with Flag-Fbw7γ ΔNLS and Myc-tagged Ebp2 or its T3A mutant, and cells were stained for both proteins with tag antibodies. (D) Dominant negative effect of the Ebp2/T3A mutant. U2OS cells were cotransfected with wild-type Flag-Fbw7γ and Myc-tagged Ebp2 or its T3A mutant and stained as described above. (E) Ebp2/T3A reduces the T3 phosphorylation of endogenous Ebp2 in nucleoli in a dose-dependent manner. Cells were transfected with Myc-tagged Ebp2/T3A and costained with 9E-10 and phospho-T3 Ebp2 antibody, which only recognizes the endogenous Ebp2 in this case.

Coexpression of Fbw7γ modestly reduced Ebp2 abundance compared with inactive Fbw7 mutants that are crippled by an F-box deletion (Fbw7γ ΔF) (Fig. 2A) or Fbw7^ARG mutation (Fig. 2B), and reduced amounts of Ebp2 precipitated with Fbw7γ than with inactive Fbw7γ ΔF (Fig. 2A). These findings are typical of other Fbw7 substrates and suggest that binding to active Fbw7 facilitates Ebp2 degradation (see below). We also noted high-molecular-weight (HMW) species of Fbw7-bound Ebp2 that are reminiscent of polyubiquitylated proteins but that we were unable to further define (Fig. 2A; see also Fig. 6B). Because these HMW Ebp2 forms were also associated with inactive Fbw7γ ΔF, they cannot result directly from SCF^Fbw7 activity and may instead reflect another type of modification. Indeed, yeast Ebp2 is sumoylated (37).

We next examined whether Ebp2 phosphorylation is required to direct Fbw7γ to nucleoli. Figure 2C shows that the nonphosphorylatable Ebp2/T3A mutant could not recruit Fbw7γ ΔNLS to nucleoli, although Ebp2/T3A itself is exclusively nucleolar. Remarkably, Ebp2/T3A expression displaced wild-type Fbw7γ from nucleoli and dominantly prevented endogenous Ebp2 from targeting Fbw7γ to nucleoli (Fig. 2D). Indeed, we observed loss of endogenous Ebp2 T3 phosphorylation in nucleoli of cells expressing Ebp2/T3A in a dose-dependent manner (Fig. 2E), thereby explaining the exclusion of Fbw7γ from nucleoli. This dominant inhibition likely results from either saturation of nucleolar Ebp2-binding sites by exogenous Ebp2/T3A or by preventing endogenous Ebp2 phosphorylation, perhaps by competing for binding to its kinase(s).

Endogenous Ebp2 is required for Fbw7 nucleolar localization.

The inhibition of Fbw7γ nucleolar targeting by exogenous Ebp2/T3A suggested that endogenous Ebp2 localizes wild-type Fbw7γ to nucleoli. We directly tested this possibility by using siRNA to knock down endogenous Ebp2 expression. Two independent siRNAs that greatly reduced endogenous Ebp2 protein abundance (Fig. 3 A) efficiently inhibited Fbw7γ nucleolar localization, whereas a control siRNA did not (Fig. 3B). Potential siRNA-mediated off-target effects are unlikely when using multiple siRNAs. However, to eliminate the possibility that the mere lack of Ebp2 in siRNA-treated cells leads indirectly to Fbw7γ mislocalization, we reintroduced siRNA-insensitive Ebp2 containing silent point mutations within the siRNA seed targeting region, either as the wild type or the T3A mutant. Figure 3C demonstrates that these constructs were completely resistant to siRNA-mediated knockdown. Importantly, when coexpressed with Ebp2 siRNA, only wild-type siRNA-resistant Ebp2, but not the T3A mutant, rescued nucleolar targeting of Fbw7γ (Fig. 3D). Thus, endogenous Ebp2 directly regulates Fbw7γ nucleolar targeting, and this requires the Ebp2 T3 CPD.

FIG. 3.

Endogenous Ebp2 localizes Fbw7 to nucleoli. (A) Knockdown of endogenous Ebp2 (∼40 kDa) with different siRNAs and analysis by Western blotting for Ebp2 and phospho-T3 Ebp2. The asterisk marks a cross-reaction (∼60 kDa) by this Ebp2 antibody, which served as a loading control. (B) Dissociation of Fbw7γ from nucleoli in the absence of Ebp2. U2OS cells were cotransfected with Flag-Fbw7γ and siRNAs as indicated and were stained with Flag antibody to detect Fbw7γ and with an anti-Ebp2 antibody (si-c, a nonspecific control siRNA). (C and D) Verification of Ebp2 knockdown specificity. The absence of Ebp2 in knockdown cells was replaced by cotransfection of siRNA-resistant Myc-tagged constructs encoding either wild-type or T3A mutant Ebp2 along with si-Ebp2 and analyzed as described above. Note that only wild-type Ebp2, but not T3A mutant Ebp2, can rescue Ebp2-mediated Fbw7 nucleolar localization. The asterisk in panel C marks a cross-reaction of the Myc-tag antibody, and this served as a loading control.

Because NPM1 knockdown also mislocalizes Fbw7γ (3) (Fig. 4 A), we examined if NPM1 regulates Ebp2, or vice versa. NPM1 silencing did not affect Ebp2 or pT3-Ebp2 abundance (Fig. 4B), nor did NPM1 knockdown prevent Ebp2 from localizing to nucleoli (Fig. 4C). NPM1 knockdown did cause subtle alterations in pT3 Ebp2 immunostaining (as detected with the phospho-antibody), but this was in the setting of grossly aberrant nucleolar morphology that was visible by bright-field microscopy (Fig. 4C and D). Indeed, others have also found that NPM1 silencing distorts nucleolar structure (2). We also examined how Ebp2 may impact NPM1 and found that Ebp2 knockdown did not alter NPM1 expression or localization (Fig. 4E and F). Thus, although both Ebp2 and NPM1 are needed to target Fbw7 to nucleoli, their mechanistic relationship remains uncertain. In light of the grossly disrupted nucleolar structure following NPM1 knockdown, the failure to localize Fbw7 to nucleoli may be indirect and largely reflect nucleolar disorder caused by NPM1 silencing.

FIG. 4.

Nucleophosmin (NPM1) is essential for Fbw7 nucleolar localization. (A) NPM1 is required for proper Fbw7 localization to nucleoli. U2OS cells were infected with a retrovirus encoding a hairpin against NPM1 or a nonspecific control hairpin. After puromycin selection, cells were transfected with Flag-Fbw7γ and subsequently stained for NPM1 and Flag. U2OS cells tolerated NPM1 knockdown. (B) Verification of NPM1 knockdown. Cells from the same experiment as shown in panel A were analyzed by Western blot before (whole-cell extract [WCE]) and after (supernatant [SN]) lysate clearing. Ebp2 and its T3 phosphorylation appeared unaffected by NPM1 knockdown. (C) Altered staining patterns of Ebp2 and T3-phospho-Ebp2 upon NPM1 knockdown. Cells identical to those shown in panel A were stained with antibodies to Ebp2 and pT3-Ebp2. Note the altered nucleolar pT3 staining in shNPM1 cells. The single bright cell in the pT3 staining resembles a cell in mitosis, where this antibody is known to recognize another protein (9). (D) NPM1 knockdown affects nucleolar appearance. Identical cells as those in panel A were analyzed by bright-field microscopy. Note the loss of dense-appearing nucleoli in U2OS cells after NPM1 knockdown. (E and F) NPM1 is unaffected by Ebp2 knockdown. U2OS cells were transfected with the indicated siRNAs and analyzed by Western blotting (E) and immunofluorescence (F).

GSK3 phosphorylates Ebp2 T3 and is required for Fbw7γ nucleolar localization.

Most Fbw7 substrate CPDs are phosphorylated by GSK3. The presence of an Ebp2 serine residue in the +4 position relative to T3 suggests classic priming-dependent phosphorylation by GSK3, and we previously established that S7 is needed for T3 phosphorylation (9). We found that purified GSK3 readily phosphorylated Ebp2 on T3 in vitro (Fig. 5 A). Conversely, treating cells with a pharmacologic GSK3 inhibitor (GSK inhibitor VIII) reduced Ebp2 T3 phosphorylation in 293A cells (Fig. 5B) and U2OS cells (data not shown). We also used a short GSK3 inhibitory peptide consisting of the axin GSK3 interaction domain (GID) to inactivate GSK3 in vivo (11). The GID peptide greatly reduced T3 phosphorylation of exogenous Ebp2 in 293A cells (Fig. 5C) and eliminated nucleolar pT3 staining of endogenous Ebp2 in U2OS cells (Fig. 5D), whereas a nonfunctional point mutant of this peptide (GID*) had no effect. Finally, we demonstrated loss of endogenous Ebp2 T3 phosphorylation in primary human fibroblasts upon knockdown of endogenous GSK3 (Fig. 5E). In sum, these findings indicate that GSK3 is the primary kinase that phosphorylates Ebp2 on T3.

FIG. 5.

GSK3 is the major physiologic Ebp2-T3 kinase. (A) In vitro kinase reaction with GSK3. Myc-tagged Ebp2 or the T3A mutant were transfected into U2OS cells, purified by immunoprecipitation, subjected to a GSK3 kinase assay, and analyzed by Western blotting. The asterisk marks a cross-reaction of the GSK3 antibody. The Ebp2 constructs used in this experiment were C-terminally truncated to facilitate immunoprecipitation (see Materials and Methods). (B) Inhibition of endogenous GSK3 blocks Ebp2-T3 phosphorylation. Cells were transfected with Myc-tagged Ebp2 or the T3A mutant and exposed overnight to GSK3 inhibitor VIII (25 μM). (C) A small peptide derived from the axin GID quenches Ebp2-T3 phosphorylation. Cells were transfected with Myc-tagged Ebp2, and plasmids encoding GID or an inactive point mutant of GID (GID*) and lysates were analyzed for exogenous Ebp2 and T3 phosphorylation by Western blotting. The first lane is an empty vector control. (D) GID eliminates nucleolar phospho-T3 Ebp2. Cells were transfected with Myc-tagged GID or GID* and costained for the Myc-tagged GID peptide and endogenous phospho-T3 Ebp2. (E) Knockdown of endogenous GSK3 (both α and β) eliminates endogenous Ebp2 T3 phosphorylation in primary human foreskin fibroblasts. HFF cells were transduced with retrovirus encoding hairpin RNAs against Ebp2 or GSK3 (α and β), or a control hairpin (sh-c), and Western blotted as indicated. The GSK3 hairpin was verified previously to target both GSK3 paralogs (51). (F) GSK3 is essential for nucleolar Fbw7γ localization. Flag-Fbw7γ-expressing cells were cotransfected with GID/GID* or exposed to LiCl (30 mM overnight) or GSK inhibitor VIII (25 μM overnight) and stained for Fbw7γ using Flag antibody.

We also found that Fbw7γ nucleolar localization requires GSK3 activity. Expression of GID, but not GID*, as well as two pharmacologic GSK3 inhibitors (lithium chloride and GSK3 inhibitor VIII), each prevented Fbw7γ nucleolar localization, although the various inhibitors led to subtly different staining patterns, perhaps due to varied inhibition efficiencies (Fig. 5F). Collectively, these data indicate that GSK3 is the dominant physiologic kinase that phosphorylates Ebp2's CPD and regulates Fbw7γ nucleolar localization. Because Ebp2 T3 phosphorylation by GSK3 requires a priming phosphorylation on S7, the Ebp2 S7 kinase may also regulate Fbw7 localization.

Nucleolar targeting of Fbw7α by Ebp2.

In most cell types Fbw7α localizes predominantly to the nucleoplasm. Because Fbw7γ nucleolar localization is mediated via the substrate-binding domain present in all Fbw7 isoforms, we tested whether Ebp2 can also target Fbw7α to nucleoli. Enforced expression of Ebp2, but not the Ebp2/T3A mutant, also recruited Fbw7α to nucleoli (Fig. 6 A, upper panels), which was highlighted by using a cytoplasmic Fbw7α ΔNLS mutant (Fig. 6A, lower panels). Fbw7α bound to the Ebp2 phospho-degron (Fig. 6B) and required an intact substrate-binding domain (Fig. 6C). In a limited screen, we identified two transformed cell lines that preferentially localized exogenous Fbw7α to the nucleolus: HeLa and C33A cells (Fig. 6D, upper panels). Both lines are of cervical origin, although it is unclear whether nucleolar Fbw7α localization is a common feature of cervical carcinoma cell lines. Fbw7α^ARG mutations (Fig. 6D, lower panels) or knockdown of endogenous Ebp2 (data not shown) displaced Fbw7α from nucleoli in these cells. HeLa and C33A cells contain modestly elevated levels of T3-phosphorylated Ebp2, which could account for the nucleolar distribution of Fbw7α in these cells (Fig. 6E; note that VA-13 expresses the SV40 T-antigen and therefore fails to localize Fbw7 to nucleoli [48]). Thus, both of the nuclear Fbw7 isoforms can localize to nucleoli via Ebp2, whereas Fbw7β remains anchored within the cytoplasm via its transmembrane domain and cannot be targeted to nucleoli by Ebp2 (data not shown). The factors that favor Fbw7α partitioning to the nucleoplasm versus the nucleolus remain incompletely understood but may involve dominant signals in the Fbw7α-specific N terminus and pT3-Ebp2 abundance in different cellular contexts.

FIG. 6.

Interaction between Ebp2 and Fbw7α. (A) Ebp2 overexpression recruits Fbw7α to the nucleolus. U2OS cells were transfected as indicated and stained for Fbw7α using Flag antibody. In the Fbw7α ΔNLS mutant both of the redundant NLS sequences have been deleted. (B) Phosphorylation-dependent binding of Fbw7α and Ebp2. Reciprocal coimmunoprecipitation from transfected 293A cell lysates, as in the assay shown in Fig. 2A, used Flag-tagged Fbw7α and Myc-tagged Ebp2. Fbw7α-ΔF is an F-box deletion mutant; in Ebp2/T3A Thr3 was replaced with alanine. The asterisk marks the heavy chains of the IP antibodies. (C) Cancer-associated Fbw7α mutants fail to bind to Ebp2. 293A cells were transfected as indicated and analyzed as described above using Flag and Myc tag antibodies. (D) Identification of cell lines with nucleolar association of Fbw7α. HeLa and C33A cells were transfected with Fbw7α or the Fbw7α^ARG (R505L) mutant and stained for Fbw7α using Flag antibody. (E) Comparison of endogenous Ebp2 versus phospho-T3 Ebp2 in a panel of cell lines. Equal amounts of cleared cell lysates were analyzed by Western blotting for their steady-state amounts of Ebp2 and its T3 phosphorylation. Grb2 was included as a loading control.

Ebp2 is a substrate for exogenous Fbw7 but acts like a pseudosubstrate in vivo.

The findings that (i) Ebp2 contains a canonical CPD that is phosphorylated by GSK3, (ii) the Ebp2 CPD is recognized by Fbw7 via its substrate-binding domain, and (iii) the Ebp2-Fbw7 interaction is facilitated by mutation of the F-box all suggest that Ebp2 could simply be an Fbw7 substrate. We thus determined whether Fbw7 regulates Ebp2 abundance and ubiquitylation. Ectopic Fbw7γ reduced the steady-state abundance of coexpressed Ebp2, but not that of Ebp2/T3A (Fig. 7 A to C; see also Fig. 2A and B), and this was more pronounced when specifically analyzing the T3-phosphorylated pool of Ebp2 (Fig. 7C). In contrast, two inactive Fbw7 mutants failed to eliminate Ebp2: the Fbw7^ARG mutant, which prevents substrate binding (Fig. 7B), and Fbw7 ΔF, which prevents Skp1 binding (Fig. 7C). These data suggest that overexpressed Fbw7γ can drive the turnover of ectopically expressed Ebp2 in cells. We further tested this model by developing a reconstituted in vitro ubiquitylation system. We found that highly purified SCF^Fbw7 catalyzed robust and rapid ubiquitylation of Ebp2 prepared by in vitro translation, and this depended upon the Ebp2 CPD (Fig. 7D). Surprisingly, Fbw7-mediated Ebp2 ubiquitylation completely depended upon the E2 ubiquitin-conjugating enzyme used in these reactions: UbcH5 actively ubiquitylated Ebp2, whereas hCdc34 did not but was active toward cyclin E (Fig. 7E) (see Discussion).

FIG. 7.

Ebp2 is an Fbw7 substrate in vitro. (A) Fbw7γ degrades Ebp2 in a CPD-dependent fashion. 293A cells were transfected with Flag-tagged Fbw7 and Myc-tagged Ebp2 constructs as indicated, and lysates were analyzed by Western blotting using tag antibodies. (B) Fbw7γ^ARG is incompetent for Ebp2 turnover. 293A cells were transfected with Myc-tagged Ebp2 and increasing amounts (1, 3, and 6 μg of Flag-Fbw7 plasmid/6-cm dish) of either wild-type Fbw7γ or an Fbw7γ^ARG mutant (R387L). Lysates were analyzed by Western blotting using tag antibodies. (C) Ebp2 turnover requires Fbw7's F-box. 293A cells were transfected as indicated with tagged constructs and analyzed as described above. Note that the Fbw7-mediated Ebp2 turnover is more obvious when specifically analyzing the T3-phosphorylated pool of Ebp2. The Ebp2 blot was sequentially reprobed for Fbw7, and the asterisk on the Fbw7 blot marks the prior Ebp2 signal. (D) In vitro ubiquitylation of Ebp2 using UbcH5. In vitro-translated Myc-tagged Ebp2, or its CPD mutant, was subjected to ubiquitylation by purified Fbw7 for 5 or 60 min and analyzed by Western blotting using Myc tag antibody. (E) Selective substrate rejection by hCdc34. The assay was similar to that shown in panel D but was designed to compare two different ubiquitin-conjugating enzymes (UbcH5 and hCdc34) in cyclin E and Ebp2 ubiquitylation. Both substrates were assayed in parallel and blotted for their Myc tags on the same membrane.

Although these data support the idea that Fbw7 can promote phosphorylation-dependent Ebp2 degradation in some contexts, this was not the case for the endogenous proteins in vivo. We inactivated endogenous Fbw7 by two approaches: either acutely with siRNA or by using gene targeting to delete both Fbw7 alleles (6). Fbw7 inactivation by either approach did not increase either endogenous Ebp2 or T3-phosphorylated Ebp2 abundance or Ebp2 stability (data not shown). Other aspects of the Ebp2-Fbw7 relationship are also atypical for Fbw7 substrates. For example, if prolonged Ebp2 binding is required for Fbw7's transport to and persistence in the nucleolus, their interaction must be uncoupled from rapid turnover, at least in part. This idea was supported by coimmunostaining experiments, where we found that ectopic Fbw7γ readily costained with endogenous phospho-T3 Ebp2 and did not decrease its abundance in most cells (data not shown). Thus, Fbw7 binding to endogenous Ebp2 appears insufficient for rapid Ebp2 degradation in vivo. Despite our ability to force Ebp2 turnover and ubiquitylation in vitro and in transfections, endogenous Ebp2 behaves most like an Fbw7 pseudosubstrate in vivo.

DISCUSSION

Alternative splicing of the Fbw7 ubiquitin ligase produces three protein isoforms that occupy all major subcellular locations. Each isoform is under its own transcriptional control, and this genomic organization likely allows precise control of Fbw7 activity in different cellular compartments in response to distinct signals. In this study we identified the molecular pathway that leads to Fbw7 nucleolar localization. Fbw7 directly binds to the abundant nucleolar protein Ebp2 in a substrate-like fashion, and this mediates Fbw7 nucleolar targeting. Fbw7-Ebp2 binding appears to occur more efficiently in the nucleoplasm, since Fbw7 ΔNLS mutants localize mainly to the cytoplasm and endogenous Ebp2 inefficiently tethers NLS mutants to the nucleolus. Increased Ebp2 expression can force Fbw7 to the nucleolus, even though ectopic Ebp2 overexpression was only severalfold greater than endogenous Ebp2 in these experiments (data not shown).

Because of their very low abundances, the endogenous Fbw7β and Fbw7γ proteins cannot be detected by either immunoblotting or immunostaining, and endogenous Fbw7α, which is more abundant, can only be detected by immunoblotting after direct immunoprecipitation. We and others have thus had to rely upon ectopic Fbw7 for localization studies. However, the distinct localizations of the three ectopically expressed isoforms (even when expressed at similar and low abundances) and our identification of isoform-specific cis-acting localization signals make it very likely that the endogenous Fbw7 proteins exhibit similar localization patterns.

A surprising aspect of our study is that enforced Ebp2 expression also translocates Fbw7α to nucleoli, and we identified a few cell lines with aberrant nucleolar expression of ectopic Fbw7α that is dependent upon Ebp2. The low levels of endogenous Fbw7 compared with phosphorylated Ebp2 raise the possibility that some (or most) endogenous Fbw7α is also nucleolar. Although we cannot exclude this possibility, ectopic Fbw7α remains nucleoplasmic even when expressed at levels approaching the limit of detection (data not shown). Moreover, the Fbw7α-specific N terminus dominantly overrides nucleolar localization when fused to Fbw7γ, indicating that the Fbw7α isoform contains additional regulatory signals (49). However, because the partitioning of ectopically expressed Fbw7α between the nucleoplasm and nucleolus depends upon the relative stoichiometry of Fbw7α and Ebp2, endogenous Fbw7α localization may also be regulated by Ebp2 abundance.

Ebp2: substrate or regulator of Fbw7?

The idea that a substrate localizes its ubiquitin ligase to a cellular compartment is somewhat paradoxical, since binding would likely cause the substrate's destruction. However, although Ebp2 contains a consensus CPD that is phosphorylated by GSK3, Ebp2 is a poor Fbw7 substrate in transfection-based degradation assays compared with bona fide substrates, such as cyclin E or c-Myc. Moreover, we observed costaining of ectopic Fbw7 with endogenous T3-phosphorylated Ebp2 protein in most cells, even without the use of proteasome inhibitors to block Ebp2 degradation. Finally, endogenous Ebp2 abundance was unaffected by Fbw7 knockdown or knockout. We also considered the idea that Ebp2 might be a conditional Fbw7 substrate, but we failed to find any physiologic contexts that promoted endogenous Ebp2 degradation by endogenous Fbw7. Ebp2 degradation thus appears largely uncoupled from Fbw7 binding, and Ebp2 acts like an Fbw7 pseudosubstrate in vivo.

We investigated several possible mechanisms that might protect Ebp2 from Fbw7-mediated destruction, including ubiquitin hydrolase activities that remove ubiquitin from Ebp2, or posttranslational modifications, such as sumoylation, that could block the conjugation of ubiquitin to Ebp2 lysine residues. We did not find any evidence of these mechanisms preventing Ebp2 degradation. We also considered that the predicted low affinity of the Ebp2 degron itself, which is missing key upstream residues that normally contact the Fbw7-binding pockets (Fig. 1D), could lead to inefficient Ebp2 degradation in vivo. Indeed, kinetic studies indicate that the initial ubiquitin ligation is the rate-limiting step for substrate polyubiquitylation by SCFs and that substrates often dissociate from SCFs before they can become ubiquitylated (32). Thus, the low-affinity Ebp2 degron may result in a dynamic Fbw7 interaction that suffices to retain Fbw7 within the nucleolus but is too weak to initiate ubiquitylation. To test this idea we extended the Ebp2 CPD so it conforms to the Myc T58 degron (up to the −5 position). While this high-affinity Ebp2 CPD enhanced Fbw7 binding, it only marginally increased Ebp2 turnover (data not shown). Thus, while low CPD affinity may contribute to Ebp2 stability in vivo, it is unlikely to be the major mechanism that prevents Ebp2 turnover. Another consequence of low Ebp2 CPD affinity may be that it allows low-abundance, bona fide Fbw7 substrates to effectively compete with the more abundant Ebp2 for Fbw7 binding, since true substrates contain high-affinity CPDs. Moreover, some proteins that are not Fbw7 substrates may contain phosphorylated sequences that resemble CPDs and could possibly bind inappropriately to Fbw7. Another function of the Ebp2-Fbw7 interaction may thus be to protect these nonsubstrates with CPD-like motifs from unwanted destruction, and Ebp2 may set a sensitivity threshold that adds specificity to Fbw7's substrate selection. Indeed, this may be a general role for pseudosubstrates that contain suboptimal recognition motifs.

An unexpected mechanism that may protect Ebp2 from Fbw7 involves our finding that Ebp2 ubiquitylation is highly dependent upon the ubiquitin-conjugating enzymes included within the in vitro reaction: UbcH5 robustly ubiquitylates Ebp2, whereas UbcH3/hCdc34 does not catalyze Ebp2 ubiquitylation at all. In contrast, UbcH3/hCdc34 actively ubiquitylates cyclin E. The budding yeast Fbw7 homolog Cdc4 works in conjunction with Cdc34 to ubiquitylate its substrates, and hCdc34 is thought to be the most physiologic ubiquitin-conjugating enzyme for human SCF^Fbw7. In fact, a recent report found that SCF^Fbw7 utilizes hCdc34, but not UbcH5, to ubiquitylate c-Myc in vivo (33). We thus favor the idea that Ebp2 cannot be utilized as a substrate by hCdc34. Substrate selection by the E2 enzyme may thus provide another level of regulatory control for the Fbw7 pathway and could explain Fbw7's inability to degrade Ebp2 in vivo.

In summary, we have demonstrated that Fbw7's interaction with a pseudosubstrate facilitates its nucleolar localization and is regulated by the same CPD motif and signaling pathway used to degrade bona fide Fbw7 substrates. A similar mechanism involving a pseudosubstrate also localizes the β-TrCP F-box protein to the nucleus, and pseudosubstrates may thus play broad roles in regulating SCF ligases (5). Our attempts to study a possible regulatory role for Ebp2 on Fbw7 function have been severely hampered by the fact that Ebp2 silencing rapidly impairs cell proliferation and ribosomal function. We have thus been unable to unambiguously determine the consequences of Ebp2 knockdown on Fbw7 substrate stability in vivo. The extent to which Ebp2 may regulate Fbw7 function therefore remains an open question, although we believe that our ongoing efforts to create endogenous Ebp2 T3A knock-in models will address this issue. This work, and future studies of the relationships between Fbw7 binding and Ebp2 degradation, may lead to novel insights into Fbw7's physiologic roles and SCF function in general.