Fbw7α and Fbw7γ Collaborate To Shuttle Cyclin E1 into the Nucleolus for Multiubiquitylation

Abstract

Cyclin E1, an activator of cyclin-dependent kinase 2 (Cdk2) that promotes replicative functions, is normally expressed periodically within the mammalian cell cycle, peaking at the G₁-S-phase transition. This periodicity is achieved by E2F-dependent transcription in late G₁ and early S phases and by ubiquitin-mediated proteolysis. The ubiquitin ligase that targets phosphorylated cyclin E is SCF^Fbw7 (also known as SCF^Cdc4), a member of the cullin ring ligase (CRL) family. Fbw7, a substrate adaptor subunit, is expressed as three splice-variant isoforms with different subcellular distributions: Fbw7α is nucleoplasmic but excluded from the nucleolus, Fbw7β is cytoplasmic, and Fbw7γ is nucleolar. Degradation of cyclin E in vivo requires SCF complexes containing Fbw7α and Fbw7γ, respectively. In vitro reconstitution showed that the role of SCF^Fbw7α in cyclin E degradation, rather than ubiquitylation, is to serve as a cofactor of the prolyl cis-trans isomerase Pin1 in the isomerization of a noncanonical proline-proline bond in the cyclin E phosphodegron. This isomerization is required for subsequent binding and ubiquitylation by SCF^Fbw7γ. Here we show that Pin1-mediated isomerization of the cyclin E phosphodegron and subsequent binding to Fbw7γ drive nucleolar localization of cyclin E, where it is ubiquitylated by SCF^Fbw7γ prior to its degradation by the proteasome. It is possible that this constitutes a mechanism for rapid inactivation of phosphorylated cyclin E by nucleolar sequestration prior to its multiubiquitylation and degradation.

INTRODUCTION

SCF ubiquitin ligases constitute a subset of the cullin ring ligase (CRL) family (1). SCF ligases contain the scaffold protein Cul-1, the linker protein Skp1, the ring finger protein Rbx1, and one of many adapter proteins known as F-box proteins. F-box proteins contain an F box that attaches them to the SCF core by binding to Skp1 and a substrate binding domain. Rbx1 binds the ubiquitin-conjugating enzyme hCdc34, which is responsible for transferring ubiquitin molecules to bound substrates. Fbw7 is an F-box protein that has been shown to form an active ubiquitin ligase when bound to an SCF core (2, 3). SCF^Fbw7 ubiquitin ligases regulate the turnover and thereby the activity of an increasing number of important cellular regulatory proteins, including cyclins E1 (2, 3) and E2 (4), c-Myc (5–7), c-Jun (8), Notch (9, 10), SREBP (11), PGC-1 (12), Mcl-1 (13, 14), and NF-κB2 (15–17). Fbw7, the targeting subunit of these ligases, recognizes substrate proteins only when they are phosphorylated on a motif known as the CPD (Cdc4 phosphodegron) (18, 19) (Saccharomyces cerevisiae Cdc4 being the founding member of the Fbw7 family), thereby linking turnover to protein kinase-based signaling pathways. Although the phosphodegron sequence is somewhat degenerate, most such sequences contain a phosphothreonine at position 0 followed by a proline at position +1 and a phosphoserine or phosphothreonine at position +4, although acidic amino acids can replace the second phosphoamino acid (20). The phosphodegron-binding component of Fbw7 is an eight-bladed β-propeller consisting of WD-40 repeats (19, 20). The crystallographically determined structure of a peptide corresponding to the cyclin E1 phosphodegron and surrounding amino acids bound to Fbw7 revealed a binding pocket on the β-propeller surface that can accommodate both phosphorylated residues with corresponding basic charges (20). Binding of substrates containing activated phosphodegrons presumably places them in appropriate proximity and with correct geometry for transfer of ubiquitins from a ubiquitin-conjugating enzyme, hCdc34, associated with the SCF ubiquitin ligase core.

Fbw7 is frequently mutated in a broad spectrum of human malignancies (21). This is easily explained by the fact that many of the targets of SCF^Fbw7 are oncoproteins that are expected to accumulate to abnormal levels when Fbw7 is lost or reduced in expression. These include cyclin E (2–4, 21, 22), c-Myc (5–7), c-Jun (8), and Notch (9, 10). In addition, loss of Fbw7 elevates the levels of the antiapoptotic protein Mcl1 in tumor cells, rendering them resistant to chemotherapy (13, 14). The most prevalent malignancy-associated mutations correspond to missense mutations that replace arginines required for phosphodegron binding, but many other types of mutation have also been detected, including nonsense, frameshift, and deletion mutations (21). Although the vast majority of mutations are in the exons encoding the β-propeller structure and are therefore expected to directly affect substrate binding, mutations have been found in other functional domains of Fbw7, including the F box (which is required for tethering Fbw7 to the SCF core) (21), cellular localization signals (21), and cofactor association sites (23).

Analysis of Fbw7 function is complicated by the fact that the gene encodes three distinct splice-variant isoforms, designated α, β, and γ (24). All isoforms consist of a common structure encoded by 10 3′ exons and a unique amino-terminal domain encoded by one of three 5′ exons that are subject to differential transcriptional regulation. Fbw7α and Fbw7γ have been shown to be nuclear, but with Fbw7γ being primarily nucleolar and Fbw7α being excluded from the nucleolus (5). Fbw7β is cytosolic and localizes to membrane compartments, most likely via a predicted transmembrane domain near its amino terminus (25). Since most known targets of Fbw7 are nucleoplasmic, it would be logical that they be targeted by SCF^Fbw7α. Likewise, SCF^Fbw7γ would be expected to target nucleolar substrates. However, we made the surprising observation that ubiquitylation of cyclin E1 requires both Fbw7α and Fbw7γ (23, 26). We found that whereas SCF^Fbw7α binds phosphorylated cyclin E, it does not multiubiquitylate it (23). Instead, SCF^Fbw7α, in conjunction with the prolyl cis-trans isomerase Pin1, carries out a noncanonical isomerization of a proline-proline bond in the cyclin E phosphodegron to create a high-affinity interaction with SCF^Fbw7γ, which then carries out the ubiquitylation reaction (23). The experiments described above, which were carried out in HEK293A cells using RNA interference (RNAi)-mediated silencing, were reproduced in the same cell line in another study (27). However, in that same study, which used somatic cell deletion strategies, only Fbw7α was required for cyclin E turnover in HCT116 colon carcinoma cells based on deletion of the Fbw7 isoform-specific coding regions (27). This apparent discrepancy was resolved by showing that cells expressing low levels of cyclin E, characteristic of normal cells and many tumor-derived cells, require both Fbw7α and Fbw7γ for degradation of cyclin E, whereas cells overexpressing cyclin E, characteristic of some tumor-derived cells, including HCT116, require only Fbw7α (26). The possible reasons for this are discussed below. Based on these observations, we sought to determine the in vivo significance of the sequential role of Fbw7 isoforms in cyclin E degradation. In the present study, we report that cyclin E phosphodegron isomerization by SCF^Fbw7α/Pin1 potentiates binding to SCF^Fbw7γ, which then causes cyclin E to translocate or localize into the nucleolus, where it is multiubiquitylated prior to degradation by the 26S proteasome.

MATERIALS AND METHODS

Cell culture and treatments.

HEK293A cells were maintained in high-glucose Dulbecco's modified Eagle medium (DMEM; Invitrogen), supplemented with 2 mM l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen), and 10% fetal bovine serum (Invitrogen). hTERT-immortalized mammary epithelial (IME) cells were maintained in MCDB 131 medium (Invitrogen) with supplements as previously described (23). Tet-ON Saos-2 cells were maintained in Iscove's modified Dulbecco medium (IMDM) (Invitrogen), supplemented with 2 mM l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen), 10% tetracycline-free fetal bovine serum (Clontech), 100 μg/ml G418 (Invitrogen), and 0.1 μg/ml puromycin (Sigma). All cell lines were grown at 37°C in a humidified incubator with 5% CO₂. Fbw7 expression in stable Tet-ON Saos-2 cell lines was induced by treating cells with doxycycline (MP Biomedicals, LLC) for 24 h at a final concentration of 0.25 μg/ml. To inhibit proteasomal activity, cells were treated with 40 μM MG-132 (Cayman Chemical Company) or 1 μM epoxomicin (Calbiochem) for 5 h, if not otherwise indicated.

Plasmids, mutagenesis, and transfections.

Cyclin E and Fbw7 expression plasmids have been described previously (23). Full-length (FL) FBW7α and FBW7γ and their F-box deletion versions (ΔFα and ΔFγ) were amplified using Advantage-HF2 polymerase (Clontech) and introduced into the pSC-A PCR cloning vector (StrataClone). Tet-responsive Fbw7 expression plasmids were then created by subcloning (full length and ΔF) into the pTRE2-pur plasmid (Clontech). Tet-ON Saos-2 cells stably expressing pTRE2pur-Fbw7 were generated by selection in 100 μg/ml G418 (Gibco) and 1 μg/ml puromycin (Sigma). To generate HEK293A cells stably expressing Fbw7 isoforms (full length and ΔF), cells were infected with retroviruses expressing FLAG-Fbw7 (pBabe-puro) and subsequently selected in 1 μg/ml puromycin. Plasmids and viruses expressing empty vector were used as controls. Generation of retroviral particles has been described (23). To construct N-terminal mCherry-cyclin E expression vectors, cyclin E was subcloned into the pmCherry-C1 vector (Clontech). mCherry-cyclin E^P382I and mCherry-cyclin E^T380A were generated by using the QuikChange XL system (Stratagene) with the following primers; P382I-Forward, 5′-GGG CTC CTC ACC CCG ATA CAG AGC GGT AAG AAG-3′; P382I-Reverse, 5′-CTT CTT ACC GCT CTG TAT CGG GGT GAG GAG CCC-3′; T380A-Forward, 5′-CCA GTG GGC TCC TCG CCC CGC CAC AGA G-3′; and T380A-Reverse, 5′-CTC TGT GGC GGG GCG AGG AGC CCA CTG G-3′. Plasmid transfections were performed by calcium phosphate precipitation or with the TransIT-LT1 transfection reagent (Mirus, Madison, WI). Transient small interfering RNA (siRNA) transfections were performed using the HiPerFect transfection reagent, according to the manufacturer's protocol (Qiagen), or Lipofectamine RNAiMAX (Invitrogen), using siRNA oligonucleotide sequences for Fbw7α, Fbw7γ, and Pin1 or nontargeting control siRNAs (scrambled siRNAs or enhanced green fluorescent protein (EGFP)-labeled siRNA). siRNAs were purchased from either Dharmacon or Microsynth.

Antibodies.

The following primary antibodies were used in this study: mouse monoclonal anti-cyclin E (HE12 and H172; Santa Cruz), anti-UBF (F-9; Santa Cruz), anti-RPA194 (Santa Cruz), anti-β-actin (A5441; Sigma), anti-FLAG (M2; Sigma), anti-γH2AX (Ab-2893; Abcam), anti-NPM (ab-37695; Abcam), rabbit polyclonal antihemagglutinin (anti-HA) (Y11; Santa Cruz), anti-Fbw7 (3), anti-Pin1 (ab-12107; Abcam), antifibrillarin (a gift from Eng Tan, The Scripps Research Institute), antinucleolin (C-23; Santa Cruz), and anti-PCNA (PC-10; Santa Cruz). Secondary antibodies included horseradish peroxidase (HRP)-conjugated anti-mouse and anti-rabbit IgG (Rockland), fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse IgG (Dako), and Cy3-conjugated goat anti-mouse and FITC-conjugated goat anti-human IgG (Invitrogen).

Quantitative RT-PCR.

Total RNA was extracted using TRIzol reagent (Invitrogen), and 1 μg was reverse transcribed into cDNA by using Superscript reverse transcriptase (Invitrogen) with random primers. Quantitative real-time PCR (RT-PCR) was performed using Fbw7 and GAPDH TaqMan primers and probes (Applied Biosystems), as previously described (23). Gene expression levels were analyzed on an ABI 7300 real-time PCR system (SDS software, version 1.3.1) and normalized based on GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA expression.

Immunoblotting and ubiquitylation assay.

For immunoblotting, cells were lysed in M-RIPA (modified radioimmunoprecipitation assay) buffer (50 mM Tris-HCl, pH 8.0, with 150 mM sodium chloride, 1.0% Igepal CA-630 [NP-40], 0.5% sodium deoxycholate, and 0.05% sodium dodecyl sulfate) and sonicated on ice. Protein extracts were collected after centrifugation at 18,900 × g in 4°C, and protein concentration was quantified spectrophotometrically using the Bradford assay (Bio-Rad), according to the manufacturer's instructions. Whole-cell protein extracts were boiled in SDS loading buffer (Invitrogen), separated using 7.5 or 10% reducing SDS-PAGE gels (Invitrogen), transferred to Immobilon polyvinylidene difluoride (PVDF) membranes (Millipore), and probed with the appropriate primary and secondary antibodies. Visualization was performed with an enhanced chemiluminescence detection system (Pierce). For immunoprecipitation and in vivo ubiquitylation assays, HEK293A cells were transfected with a plasmid expressing HA-ubiquitin for 24 h and treated with 40 μM MG-132 for 5 h before harvesting. Cells were collected, washed with phosphate-buffered saline (PBS), and separated into enriched nucleolar and nucleolus-depleted protein fractions, as described below. Equal amounts of total cyclin E protein from each fraction were subsequently denatured by incubation in 1% SDS and 10 mM N-ethylmaleimide (Sigma). Extracts were sonicated, boiled for 10 min, and diluted 10 times in M-RIPA buffer (without SDS), followed by incubation with anti-cyclin E (HE172) or control IgG antibodies for 2 h. Immune complexes were adsorbed onto GammaBind Sepharose beads (GE health care) for 1 h at 4°C. Beads were washed four times with lysis buffer and boiled in SDS loading buffer, followed by separation of proteins on 7% SDS-PAGE gels. Proteins were transferred to PVDF membranes, and ubiquitylated cyclin E was visualized by immunoblot analysis using anti-HA antibodies, as described above.

Immunofluorescence microscopy.

Cells were cultured on coverslips in complete medium, removed, and washed in PBS, followed by fixation in 2% paraformaldehyde (wt/vol) in PBS (pH 7.6) for 10 min at room temperature. Coverslips were transferred to −20°C methanol for 10 min, washed three times with PBS, and blocked with 2% bovine serum albumin (BSA), 0.2% Tween 20, 5% glycerol in PBS. Cells on coverslips were subsequently incubated with primary antibodies (in 1% BSA-PBS) overnight at 4°C, washed three times in Tris-buffered saline (TBS), 0.05% Tween 20 (TBS-T), followed by incubation with secondary antibodies (diluted in TBS) for 1 h at room temperature. DNA was visualized by the addition of DAPI (4′,6-diamidino-2-phenylindole) or Hoechst stain (Sigma), and samples were mounted with Mowiol (Calbiochem). Cells were analyzed at a magnification of ×63 or ×100, and images were captured using a Zeiss META LSM510 laser scanning confocal microscope or a Zeiss Axioplan 2 microscope (Carl Zeiss, Inc.). Images were recorded with an empirically chosen, constant exposure time, and fluorescence intensity was quantified using Axiovision 4.5 software (Carl Zeiss, Inc.). The ratio of the average pixel intensities between nucleolar and nucleoplasmic cyclin E staining was determined using Image J software (National Institutes of Health). From each experiment, the mean cyclin E fluorescence intensity (after background correction) was calculated from analysis of at least 20 individual cells. In addition, where indicated, the number of cells with cyclin E-positive nucleoli was calculated from analysis of at least 100 cells. Fields for analysis were chosen at random, and all cells within a field were scored. All data represent averages from at least three independent experiments. Significance was calculated with Student's t test using excel Software. Differences were considered significant when P was <0.05.

Nucleolar fractionation.

Nucleolar fractionation was based on the method originally described (28). To prepare nucleolus-enriched and nucleolus-depleted protein fractions, HEK293A cells were grown on 6- by 15-cm dishes in complete medium. Briefly, cells were collected by trypsinization, washed three times in PBS, resuspended in 5 ml of buffer A (10 mM HEPES [pH 7.9], 10 mM KCl, 1.5 mM MgCl₂, 0.5 mM dithiothreitol [DTT]), and supplemented with complete protease inhibitor cocktail (Roche) and Halt phosphatase inhibitor cocktail (Pierce). The cell lysate was homogenized on ice by 15 strokes of a Dounce homogenizer with a tight pestle and centrifuged at 218 × g for 5 min at 4°C to separate the cytoplasmic fraction (supernatant) from the nuclear fraction (pellet). The nuclear fraction was resuspended in 3 ml of buffer S1 (0.25 M sucrose, 10 mM MgCl₂) and then layered on 3 ml of buffer S2 (0.35 M sucrose, 0.5 mM MgCl₂) followed by centrifugation at 1,430 × g for 5 min at 4°C. The pelleted nuclei were resuspended in 3 ml buffer S2 and sonicated six times for 10 s each time on ice. Sonication efficiency was analyzed using a phase-contrast microscope before layering on 3 ml of buffer S3 (0.88 M sucrose, 0.5 mM MgCl₂) and centrifugation at 2,800 × g for 10 min at 4°C. The sedimented pellet contained the purified nucleoli and the supernatant the nucleoplasmic fraction. Nucleoli were washed twice by resuspension in 500 μl of S2 buffer followed by centrifugation at 2,000 × g at 4°C. The cytoplasmic and nucleoplasmic fractions were pooled and further clarified by centrifuging at 2,800 × g for 10 min at 4°C. This fraction was referred to as the nucleolus-depleted fraction. Nucleolar enrichment was verified by immunoblotting with antibodies specific for the nucleolar proteins RPA194 and UBF.

EdU incorporation and flow cytometry analysis.

Saos2 cells (2 × 10⁵/10-cm dish) were synchronized using aphidicolin (2 μM). After 16 h in aphidicolin, cells were released for 2 h, washed with PBS, and pulsed for 15 min with 20 μM EdU (Invitrogen). Cells were then resuspended in 100 μl PBS–1% BSA, fixed, and analyzed according to the manufacturer's instructions. Before harvesting, asynchronous cells were pulse-labeled for 15 min with 20 μM EdU. Cells were fixed with 100 μl of Click-iT fixative while being vortexed and were incubated at room temperature for 15 min. The cells were then washed with 1 ml PBS–1% BSA, and the pellet was resuspended in 1 ml 1× saponin permeabilization buffer (1× saponin in PBS–1% BSA) for 10 min at room temperature. Cells were then washed with 1 ml PBS–1% BSA and stained with Click-iT reaction cocktail (containing azide-modified Alexa Fluor 647, CuSO₄, and Click-iT EdU buffer additive) and propidium iodide (2 μg/ml) for 30 min at room temperature. Cells were then analyzed using FACS LSRII and CellQuest software (Becton, Dickinson).

RESULTS

Cyclin E accumulates in the nucleolus when the proteasome is inhibited.

As part of this investigation, we determined the subcellular localization of cyclin E under conditions of proteosomal inhibition. Saos2 cells were treated with the proteasome inhibitor epoxomicin and analyzed for cyclin E localization by immunofluorescence microscopy (Fig. 1A to C). In the majority of cells, cyclin E accumulated in large nuclear bodies that were determined to be nucleoli because they also stained positive when antibodies specific for the nucleolar marker nucleophosmin (NPM) were used (Fig. 1A and B). A significant fraction of non-epoxomicin-treated cells also exhibited nucleolar cyclin E staining (Fig. 1B). This is consistent with a previous study (29). However, the intensity of this staining was significantly higher in epoxomicin-treated cells (Fig. 1C). Similar results were obtained using HEK293A cells treated with epoxomicin and with hTERT-immortalized human mammary epithelial cells treated with the alternative proteasome inhibitor MG-132 (see Fig. S1 in the supplemental material). Accumulation of cyclin E in the nucleoli of epoxomicin-treated cells was confirmed by Western blotting of purified nucleoli (Fig. 1D).

Fig 1.

Cyclin E accumulates in the nucleolus upon proteasomal inhibition. (A) Saos2 cells were treated with aphidicolin (2 μM) to arrest cells in early S phase and epoxomicin (bottom) to inhibit proteasomal activity and stained with cyclin E antibody (red). Nucleoli were visualized by staining with NPM antibody (green) and DNA with DAPI (blue). As a control, cells were treated with the solvent dimethyl sulfoxide (DMSO) (top). Bar, 10 μm. (B) Quantification of the percentage of cells with cyclin E-positive nucleoli upon treatment with epoxomicin, compared to the DMSO control. (C) Average intensities of nucleolar relative to nucleoplasmic cyclin E in Saos2 cells after treatment with epoxomicin as described for panel A. The nucleolar-to-nucleoplasmic ratio in DMSO-treated cells was set to 1. (D) Saos2 cells were treated with proteasome inhibitor epoxomicin or DMSO as described for panel A, and a nucleolus-enriched fraction was extracted (see Materials and Methods). Proteins were separated by SDS-PAGE followed by immunoblotting with cyclin E antibody or UBF antibody as a marker for nucleolar purity.

Cyclin E requires both Fbw7α and Fbw7γ for nucleolar accumulation.

We showed previously that SCF complexes containing Fbw7α and Fbw7γ, respectively, are required for ubiquitylation and degradation of cyclin E under conditions where cyclin E specifically is not overexpressed (23, 26). The localization of cyclin E to the nucleolus in response to proteasomal inhibition suggests that a step in the cyclin E degradation pathway may be located in the nucleolus. It has been shown previously that one of the splice variant isoforms, Fbw7γ, is concentrated in the nucleolus (5) and that SCF^Fbw7γ is the isoform-specific ligase responsible for ubiquitylating cyclin E (23). On the other hand, SCF complexes containing Fbw7α, which serve as a cofactor for Pin1-mediated isomerization of a proline-proline bond in the cyclin E phosphodegron, are also indirectly required for ubiquitylation (23). Fbw7α resides in the nucleoplasm but is excluded from the nucleolus (5) (see Fig. 3A). To establish the roles of Fbw7 isoforms in cyclin E localization, Fbw7α and Fbw7γ were silenced using RNAi in Saos2 cells, and immunofluorescence microscopy was carried out (Fig. 2A; also, see Fig. S2 in the supplemental material for qPCR confirmation of silencing). In addition, confirmation of isoform-specific silencing using these siRNAs has been published elsewhere (16). Lowering the levels of either isoform sharply reduced both the intensity of cyclin E nucleolar staining as well the percentage of cells exhibiting localization of cyclin E to the nucleolus (Fig. 2B and C). Similar results were obtained using HEK293A cells (Fig. 2D and E). These data indicate that both Fbw7α and Fbw7γ have roles in cyclin E nucleolar localization in addition to their role in cyclin E ubiquitylation and proteolysis.

Fig 3.

F-box-deleted Fbw7 alleles differentially affect cyclin E localization. (A) Saos2 cells were treated with doxycycline to induce expression of wild-type or F-box-deleted alleles (ΔF) of Fbw7α and Fbw7γ as indicated. After 8 h, cells were treated with aphidicolin (2 μM) and cultured for an additional 16 h. Cells were washed, fixed, and analyzed by immunofluorescence microscopy. Examples of cellular localization of Fbw7 after staining with Fbw7 antibodies (green) and cyclin E antibodies (red). DAPI (blue) was used to visualize nuclei. The rightmost panel in each row shows the merged image. The bottom row shows empty vector control cells; nucleoli are shown by staining with NPM antibodies (green). Bar, 10 μm. (B) Quantification of average intensities of nucleolar relative to nucleoplasmic endogenous cyclin E in Saos2 cells after doxycycline-induced expression of ΔF-Fbw7α or ΔF-Fbw7γ, compared to empty vector cells. The nucleolar-to-nucleoplasmic ratio in empty vector control cells was set at 1. (C) Percentage of cells with cyclin E-positive nucleoli after doxycycline-induced expression of ΔF-Fbw7α or ΔF-Fbw7γ, compared to empty vector cells. Error bars represent standard deviations (n = 3). Representative stainings are shown in panel A. (D) A nucleolus-enriched fraction was extracted from doxycycline-treated ΔF-Fbw7α, ΔF-Fbw7γ, or empty vector Saos2 cells and analyzed by immunoblotting with cyclin E antibody. The nucleolus-enriched fraction was also immunoblotted with antibody against the nucleolar protein UBF.

Fig 2.

Fbw7α and Fbw7γ are essential for nucleolar localization of endogenous cyclin E. (A) Saos2 cells were transfected with siRNAs targeting Fbw7α (middle row) and Fbw7γ (bottom row) or a scramble sequence (top row). After 32 h, cells were treated with aphidicolin (2 μM) and cultured for an additional 16 h. Cells were washed, fixed, and analyzed by immunofluorescence microscopy using specific antibodies recognizing cyclin E (red) and NPM (green). DAPI (blue) was used to visualize nuclei. The right column shows the merged images. Bar, 10 μm. (B) Relative average intensities of nucleolar to nucleoplasmic endogenous cyclin E in Saos2 cells after transfection with isoform-specific Fbw7 siRNAs or scrambled control as described for panel C. The nucleolar-to-nucleoplasmic ratio in scrambled control transfected cells was set at 1. (C) The percentage of cells with cyclin E positive nucleoli after transfection with isoform-specific Fbw7 siRNAs or scrambled control as described for panel A. Error bars represent standard deviations (n = 3). Representative stainings are shown in panel A. (D) HEK293A cells were transfected with siRNAs targeting Fbw7α (middle row) and Fbw7γ (bottom row) or a scramble sequence (top row). After 32 h, cells were treated with aphidicolin (2 μM) and cultured for an additional 16 h. Cells were washed, fixed, and analyzed by immunofluorescence microscopy using specific antibodies recognizing cyclin E (red) and fibrillarin (green). DAPI (blue) was used to visualize nuclei. The right column shows the merged images. Bar, 10 μm. (E) Average intensities of nucleolar relative to nucleoplasmic endogenous cyclin E in HEK293A cells after transfection with isoform-specific Fbw7 siRNAs or scrambled control as described for panel D.

Fbw7γ protein but not associated ligase activity is required for nucleolar translocation of cyclin E.

To further characterize the roles of Fbw7α and Fbw7γ in cyclin E localization, we expressed both wild-type and F-box deletion versions of these proteins in Saos2 cells using a conditional tetracycline-inducible promoter and carried out immunofluorescence microscopy (see Fig. S3 in the supplemental material for confirmation of Fbw7 isoform specificity in the images). F-box-deleted alleles (ΔF) of F-box protein substrate adapters for SCF ubiquitin ligases produce proteins that can bind substrates but not the SCF core and therefore have dominant negative properties (3). Overexpression of nonmutated Fbw7 isoforms generally reduced the level of cyclin E staining (Fig. 3A, rows I and II), consistent with accelerated proteolysis of cyclin E, as shown previously (3). On the other hand, overexpression of Fbw7αΔF or Fbw7γΔF promoted accumulation of cyclin E (Fig. 3A, rows III and IV), consistent with a dominant negative phenotype and with the previous observation that both isoforms are required to degrade cyclin E (23, 26). More importantly, whereas expression of Fbw7αΔF caused accumulation of cyclin E in the nucleoplasm and elimination of most nucleolar localization, expression of Fbw7γΔF caused hyperaccumulation of cyclin E in the nucleolus (Fig. 3A, rows III and IV). Quantitation of the percentages of cells showing nucleolar staining and the intensities of nucleolar staining confirmed these observations (Fig. 3B and C), as did biochemical purification of nucleoli followed by Western blotting (Fig. 3D). Similar results were obtained using HEK293A cells (see Fig. S4 in the supplemental material). Therefore, whereas expressing a non-SCF-associated mutant Fbw7α blocked nucleolar accumulation of cyclin E, a similar allele of Fbw7γ promoted nucleolar accumulation of cyclin E. Taken together with the results of isoform-specific Fbw7 silencing, these data suggest that SCF^Fbw7α-dependent cyclin E phosphodegron isomerization and subsequent binding of the isomerized protein to Fbw7γ are required for localization of cyclin E into the nucleolus. The latter function does not require association with the SCF core, but the hyperaccumulation of cyclin E in the presence of Fbw7γΔF confirms that association with the SCF core is required for cyclin E degradation.

Ubiquitylated cyclin E is detected mostly in the nucleolus.

Based on the observations that SCF^Fbw7γ is the ubiquitin ligase that preferentially ubiquitylates cyclin E in vitro and is primarily nucleolar, we hypothesized that cyclin E activated by phosphodegron phosphorylation is specifically translocated into the nucleolus and ubiquitylated there by SCF^Fbw7γ. We therefore characterized the ubiquitylation state of cyclin E in the nucleolus. Nucleoli were purified with or without prior treatment with the proteasome inhibitor MG132 and analyzed by SDS-PAGE for cyclin E ubiquitylation (Fig. 4A). Compared to nucleolus-depleted extract and whole-cell extract, nucleolar extract was enriched for a ladder of lower-mobility cyclin E-cross-reacting species. The relative abundance of these species increased significantly after proteasomal inhibition, consistent with their being multiubiquitylated derivatives of cyclin E. To test this directly, two different protocols were used. In the first method, cells were transfected with a plasmid expressing HA-ubiquitin and treated with MG132 to inhibit proteasomal degradation. Nucleolar and nucleolus-depleted extracts were prepared and immunoprecipitated with cyclin E antibody, followed by SDS-PAGE and detection with either cyclin E (input) or anti-HA antibodies (immunoprecipitate) (Fig. 4B). To avoid potential artifacts associated with ubiquitin overexpression, nucleolar and nucleolus-depleted extracts were also prepared from nontransfected cells and cyclin E immunoprecipitates were analyzed by SDS-PAGE using antiubiquitin antibodies for detection (Fig. 4C). Consistent with the data in Fig. 4A, multiubiquitylated cyclin E was detected primarily in the nucleolar fraction. The data confirming the purity of nucleolar extracts are shown in Fig. S5A in the supplemental material. If translocation of cyclin E into the nucleolus depends on binding of its activated phosphodegrons to Fbw7γ, then phosphorylated cyclin E should accumulate in the nucleolus. Indeed, blotting of nucleolar and nucleolus-depleted extracts with antibodies specific for the two phosphorylated cyclin E phosphodegrons indicates that most cyclin E with activated phosphodegrons is nucleolar, much of it multiubiquitylated (Fig. 4D).

Fig 4.

Multiubiquitylated cyclin E exists primarily in the nucleolus. (A) hTERT-immortalized human mammary epithelial cells were treated with aphidicolin to arrest cells in early S phase and then treated with MG132 for 5 h before harvesting. Total cell extract and nucleolus-depleted and -enriched fractions were prepared, and proteins were then separated by SDS-PAGE followed by immunoblotting for endogenous cyclin E, nucleolin, and PCNA, as indicated. Note the ladder of lower-mobility cyclin E-cross-reacting species in the nucleolus-enriched fraction (middle lanes), compared to the nucleolus-depleted fraction (left lanes) and total cell extract (right lanes). The asterisk indicates a nonspecific band that cyclin E antibody cross-reacts with. (B) 293A cells were transfected with a plasmid expressing HA-ubiquitin for 24 h and treated with MG132 for 5 h before preparation of nucleolus-enriched and nucleolus-depleted fractions. Equal amounts of cyclin E protein (5% input is shown in the lower panel) was used for immunoprecipitation, and cyclin E ubiquitylation was analyzed in each fraction by immunoblotting with HA antibody. (C) 293A cells were treated with aphidicolin to arrest cells in early S phase and then treated with MG132 for 5 h before preparation of nucleolus-enriched and nucleolus-depleted fractions. Equal amounts of cyclin E protein (5% input is shown in the lower panel) were used for immunoprecipitation, and cyclin E ubiquitylation was analyzed in each fraction by immunoblotting with antiubiquitin antibody. (D) Total nucleolus-depleted and nucleolus-enriched fractions were prepared from 293A cells, and samples were analyzed by immunoblotting with antibodies against total cyclin E or cyclin E phosphorylated on threonine 62 (T62) or threonine 380 (T380), as indicated. Note the ladder of lower-mobility cyclin E species enriched in the nucleolar fractions.

Cyclin E phosphodegron prolyl isomerization is required for nucleolar localization.

We show above that reduction in levels of Fbw7α prevented nucleolar accumulation of cyclin E. We showed previously that SCF^Fbw7α is required for Pin1-mediated noncanonical isomerization of the proline-proline bond in the primary cyclin E phosphodegron but does not carry out multiubiquitylation of cyclin E (23). These data suggest that prolyl isomerization of the primary cyclin E phosphodegron is required for nucleolar localization of cyclin E. To assess the role of Pin1-dependent phosphodegron isomerization, we determined the localization of cyclin E by immunofluorescence microscopy after RNAi-mediated silencing of Pin1. A Western blot confirming the silencing of Pin1 by siRNA is shown Fig S5B in the supplemental material. Silencing of Pin1 prevented nucleolar localization of cyclin E (Fig. 5A and B), although cyclin E levels were generally elevated as has been previously shown (23, 30). These effects are also seen when nucleoli are purified from Pin1-silenced cells and analyzed by Western blotting (Fig. 5C). Pin1 isomerization affects the functions of many cellular proteins and therefore its role in nucleolar translocation of cyclin E could be indirect. To circumvent this possibility, we analyzed a cyclin E allele containing a mutant phosphodegron that can be activated by phosphorylation but not isomerized by Pin1 (23). Cyclin E P382I replaces the second proline of the proline-proline bond in wild-type cyclin E with isoleucine. We have shown that this mutant, unlike the wild type, is efficiently targeted for ubiquitylation and turnover by SCF^Fbw7α and does not require Pin1 for degradation (23). The mutant protein was tagged with the fluorophore mCherry so that it could be distinguished from endogenous cyclin E (expression of mCherry-tagged alleles is shown in Fig. S5C in the supplemental material). Whereas mCherry-cyclin E exhibits nucleolar accumulation comparable to that of wild-type cyclin E (Fig. 5D, row I, and E), mCherry-cyclin E P382I shows greatly reduced nucleolar localization (Fig. 5D, row II, and E), similar to the product of an allele with a completely inactivated phosphodegron, cyclin E T380A (3) (Fig. 5D, row IV, and E). However, unlike cyclin E T380A, cyclin E P382I is not defective in ubiquitin-mediated degradation (23). This is most likely because cyclin E P382I can be multiubiquitylated by SCF^Fbw7α and targeted for proteosomal degradation in the nucleoplasm. We found that RNAi-mediated silencing of Fbw7α restored nucleolar localization of cyclin E P382I (Fig. 5D, row III, and E), confirming that this mutant is competent to bind Fbw7γ and localize to the nucleolus, provided that it cannot be intercepted by SCF^Fbw7α (Fig. 5D). On the other hand, cyclin E T380A shows no nucleolar localization with or without silencing of Fbw7α (Fig. 5D, row V, and E), consistent with a requirement for binding Fbw7γ for nucleolar translocation.

Fig 5.

Prolyl isomerization of the cyclin E phosphodegron is required for nucleolar translocation of cyclin E. (A) Saos2 cells were transfected with siRNA targeting Pin1 or with a scrambled control siRNA. After 32 h, cells were treated with aphidicolin and cultured for an additional 16 h. Cells were washed, fixed, and analyzed by immunofluorescence microscopy by staining with cyclin E antibody (red) or NPM antibody (green). DAPI (blue) was used to visualize nuclei. Bar, 10 μm. (B) Percentage of cells with cyclin E-positive nucleoli after Pin1 knockdown compared to the scrambled control. Error bars represent standard deviations (n = 3). (C) HEK293A cells transfected with siRNAs targeting Pin1, Fbw7, or control siRNA against GFP. Cells were lysed 48 h after transfection, total nuclear and nucleolus-enriched fractions were extracted, and proteins were separated by SDS-PAGE followed by immunoblotting with cyclin E antibody and antibody to the nucleolar marker nucleolin. (D) Saos2 cells were transfected with mCherry-tagged wild-type cyclin E (I), mCherry-cyclin E T380A (IV), or mCherry-cyclin E P382I (II). mCherry-cyclin E P382I- and mCherry-cyclin E T380A-transfected cells were also simultaneously depleted of Fbw7α by siRNA-mediated silencing (III and V). Cyclin E was visualized by mCherry intrinsic fluorescence (red) and NPM by immunofluorescence analysis using anti-NPM antibodies (green). Note the relocalization of the cyclin E P382I mutant (which does not require Pin1 for degradation) (II versus III), but not the T380A mutant (which is resistant to Fbw7-dependent degradation) (III versus IV), to nucleoli upon depletion of Fbw7α. (E) Quantification of mCherry-cyclin E-positive nucleoli. The bar graph represents the percentage of cells with mCherry-cyclin E-positive nucleoli. Error bars represent standard deviations (n = 3). Representative stainings are shown in panel D.

Fbw7γ requires nucleophosmin to localize cyclin E into the nucleolus.

It has been proposed that binding to the nucleolar protein nucleophosmin (NPM) mediates nucleolar accumulation of Fbw7γ (31). Indeed, cancer-linked mutations of NPM that cause delocalization from the nucleolus also delocalize Fbw7γ (31). We therefore analyzed the localization of cyclin E in NPM nullizygous mouse embryonic fibroblasts (MEFs) in order to determine the role of NPM in cyclin E nucleolar localization. Note that NPM^−/− MEFs require inactivation of p53 for survival. Immunofluorescence analysis indicated that cyclin E nucleolar localization was normal in p53^−/− MEFs (Fig. 6). However, deletion of NPM eliminated nucleolar localization of cyclin E, consistent with the idea that NPM-mediated nucleolar translocation of Fbw7γ may also translocate bound cyclin E into the nucleolus. Restoring NPM to NPM^−/− MEFs also re-establishes cyclin E nucleolar localization (Fig. 6).

Fig 6.

Cyclin E requires nucleophosmin (NPM) for nucleolar translocation. (A) p53^−/− control (top) and dKO (p53^−/− and NPM^−/−) (middle) and 3× Myc-NPM-transfected dKO (bottom) MEFs were treated with aphidicolin (2 μM) to arrest cells in early S phase and epoxomicin to inhibit proteasomal activity and stained with cyclin E antibody (green). Nucleoli were visualized by staining with anti-UBF or anti-Myc tag antibody (red). 3× mMyc-NPM corresponds to NPM with three tandem Myc epitope tags. Bar, 10 μm. (B) The percentage of cells with cyclin E positive nucleoli comparing p53^−/− and dKO MEFs treated as described for panel A.

Fbw7γ-mediated sequestration of cyclin E in the nucleolus diminishes replicative stress resulting from cyclin E overexpression.

Why might degradation-related ubiquitylation of cyclin E be carried out in the nucleolus? Indeed, this poses a topological problem, as proteasomes are excluded from the nucleolus (32), and therefore, any multiubiquitylated protein would have to be exported out of the nucleolus for proteasomal degradation. One reason might be that translocation of cyclin E into the nucleolus upon phosphorylation is a rapid means of separating cyclin E-Cdk2 from substrates, equivalent functionally to inactivation. In order to test this idea, we expressed F-box-deleted alleles of Fbw7α and Fbw7γ, respectively, in Saos-2 cells. As shown in Fig. 3, both alleles confer dominant negative phenotypes, leading to overexpression of cyclin E. However, whereas Fbw7αΔF promotes nuclear accumulation of cyclin E, Fbw7γΔF causes nucleolar accumulation. One of the hallmarks of cyclin E overexpression is replicative stress leading to the appearance of γH2AX and reduced rates of replication (33, 34). Cells expressing either Fbw7αΔF or Fbw7γΔF were synchronized in S phase using an aphidicolin block and then released for 2 h (Fig. 7). Cells were either then fixed for immunofluorescence analysis or pulse-labeled with the thymidine analog ethynyldeoxyuridine (EdU) for two-dimensional fluorescence-activated cell sorting (FACS) analysis. Whereas control nuclei exhibited low levels of γH2AX, nuclei expressing Fbw7αΔF exhibited high levels, indicating replicative stress (Fig. 7A and B). Nuclei expressing Fbw7γΔF exhibited low levels of staining similar to the control (Fig. 7A and B). Likewise, Fbw7αΔF expressing populations showed a reduced number of S-phase cells actively replicating DNA compared to the control, whereas Fbw7γΔF cells did not show a reduction (Fig. 7C). Furthermore, even for cells actively replicating DNA in the Fbw7αΔF population, the rate of replication was decreased, based on the reduced staining intensities of many of the cells (Fig. 7C). Western blots of extracts from cells treated in parallel indicated that accumulation of cyclin E was equivalent when Fbw7αΔF and Fbw7γΔF were overexpressed (Fig. 7D). Unlike expression of Fbw7γΔF, siRNA-mediated silencing of Fbw7γ causes nuclear accumulation of cyclin E and exclusion from the nucleolus, similarly to silencing of Fbw7α. We therefore analyzed γH2AX and DNA replication in S-phase populations silenced for either Fbw7α or Fbw7γ. Reduction of either Fbw7 isoform led to an increase in γH2AX levels and a decrease in cells actively replicating DNA (Fig. 7E, F, and G), indicative of replicative stress. We therefore conclude that Fbw7γ-mediated sequestration of cyclin E in the nucleolus has biological consequences. Specifically, separation of cyclin E-Cdk2 from its targets abrogates the effects of cyclin E overexpression in terms of markers of replicative stress.

Fig 7.

Nucleolar translocation of cyclin E reduces replication stress caused by cyclin E overexpression. (A) Saos2 cells were treated with doxycycline to induce expression of F-box deleted alleles (ΔF) of Fbw7α and Fbw7γ as indicated. After 8 h, cells were treated with aphidicolin (2 μM) and cultured for an additional 16 h. Cells were washed and released from aphidicolin arrest for 2 h, fixed, and analyzed by immunofluorescence microscopy by staining with γH2AX antibody (red) or FBW7 antibody (green). DAPI (blue) was used to visualize nuclei. (B) Relative average intensities of γH2AX Saos2 cells after doxycycline-induced expression Fbw7αΔF or Fbw7γΔF, compared to cells with empty vector. (C) Saos2 cells treated with doxycycline and aphidicolin as described for panel A were pulse-labeled for 15 min with EdU for two-dimensional FACS analysis (right). The percentage of cells synthesizing DNA with standard deviations is shown (left). (D) Western blots of extracts of cells treated as described for panels A and B, indicating relative levels of cyclin E. (E) Saos2 cells were transfected with siRNAs targeting Fbw7α and Fbw7γ or a scramble sequence as indicated. After 32 h, cells were treated with aphidicolin (2 μM) and cultured for an additional 16 h. Cells were washed, released from aphidicolin arrest for 2 h, fixed, and analyzed by immunofluorescence microscopy by staining with γH2AX antibody (red). DAPI (blue) was used to visualize nuclei. (F) Relative average intensities of γH2AX staining in Saos2 cells after transfection with isoform-specific Fbw7 siRNAs or scrambled control as described for panel E. (G) Saos2 cells treated as indicated for panel E were pulse-labeled for 15 min with EdU for two-dimensional FACS analysis. The percentage of cells synthesizing DNA (with standard deviations) is shown.

The pattern of nucleolar staining as cells traverse S phase suggests that nucleolar translocation of cyclin E is more rapid than cyclin E degradation.

Translocation of cyclin E into the nucleolus for ubiquitylation as a means of rapid inactivation would only be justified as a strategy if translocation occurred more rapidly than ubiquitin-mediated proteolysis. Although we did not test the rates of these processes directly, observation of partitioning of cyclin E between the nucleus and nucleolus as a function of progression through S phase suggests that translocation is more rapid than degradation. HEK293A cells synchronized using a double thymidine block and released in the absence of proteasome inhibitors were fixed at intervals and stained for cyclin E as cells traversed S phase. Whereas initially, in early S phase, cyclin E exhibits a uniform distribution within the nucleus, it is progressively depleted from the nucleoplasm and accumulates in the nucleoli as cells progress through S phase (Fig. 8) and overall cyclin E levels decline (35–37), consistent with this hypothesis.

Fig 8.

Cyclin E accumulates in the nucleolus as cells progress through S phase. (A) Examples of cyclin E staining patterns observed in S-phase HEK293A cells. (B) HEK293A cells were subjected to a double thymidine block and then released at time zero. Cells were fixed at 2-h intervals up to 8 h and stained using anti-cyclin E antibody. Nuclei were then scored based on the pattern observed. The graph shows the percentages of nuclei with both nuclear and nucleolar staining and with only nucleolar staining, shown in panel A. At early times, the remainder of nuclei exhibited only nuclear staining, and at the latest time point, most of the remaining nuclei exhibited no cyclin E staining. n = 300 for each time point. Error bars represent standard deviations. (C) Flow-cytometric analysis of DNA content in an experiment whose results are shown in panel B. Cells were fixed and stained with propidium iodide. The ordinate corresponds to the number of events, and the abscissa shows DNA content (fluorescence intensity).

DISCUSSION

Rapid sequestration of phosphorylated cyclin E in the nucleolus.

Our data suggest that nucleolar translocation can functionally inactivate cyclin E, thus serving as a mechanism for rapid sequestration. Specifically, whereas overexpression of cyclin E in the nucleus causes replicative stress, forced sequestration of overexpressed cyclin E into the nucleolus abrogates this phenotype. There is ample precedent for nucleolar sequestration of nuclear proteins in order to spatially separate them from nucleoplasmic and even cytoplasmic targets in various regulatory contexts (38–40). Translocation into the nucleolus may occur more rapidly than multiubiquitylation, justifying this sequestration strategy. This idea is supported by the observation of cells with nucleolar accumulation of cyclin E even in the absence of proteasome inhibitors (Fig. 1C and D). Presumably, these are mid- to late-S-phase cells that have already cleared most of the cyclin E from their nucleoplasm but have not yet completed processing it in the nucleolus. Indeed, by using immunofluorescence analysis of cyclin E in synchronized populations, we have shown that cyclin E is progressively depleted from the nucleoplasm and accumulates in the nucleoli as cells progress through S phase (Fig. 8) and overall cyclin E levels decline (35–37), consistent with the hypothesis that nucleolar translocation occurs more rapidly than multiubiquitylation. Note that a previous study provided evidence that cyclin E accumulates in nucleoli at the G₁-S boundary rather than later in S phase. We do not know the reason for the discrepancy, but that analysis was carried out using a bladder carcinoma-derived cell line, which might exhibit unusual kinetics of cyclin E accumulation and degradation (29). Once the protein is sequestered from functional targets, the temporality of cyclin E ubiquitylation would be irrelevant. When finally reintroduced into the nucleoplasm, the multiubiquitylated protein would be instantly targeted by the proteasome, consistent with low levels of multiubiquitylated cyclin E in nucleolus-depleted fractions (Fig. 4).

Most of the known nuclear targets of SCF^Fbw7 contain a phosphodegron with an analogous double proline bond, including c-Myc, c-Jun, SREBP, NF-κB2, and PGC1α (6–8, 11, 12, 15–17). We speculate that a similar system of nucleolar sequestration and ubiquitylation based on proline-proline cis-trans isomerization constitutes a feature of the degradation pathway of these proteins as well. Indeed, it was shown in one study that SCF^Fbw7γ is critical for the ubiquitylation and degradation of c-myc (5), consistent with this model.

Differential interactions with Fbw7 isoforms.

The nucleolar localization of cyclin E requires the Pin1/SCF^Fbw7α-dependent noncanonical isomerization of the proline-proline bond in the primary cyclin E phosphodegron. This cis-trans isomerization event is required for efficient binding and ubiquitylation by SCF^Fbw7γ. It is this interaction that likely mediates translocation of cyclin E into the nucleolus based on the localization signals intrinsic to Fbw7γ. However, this model suggests that the role of Fbw7α is to interact with a suboptimal conformation of the cyclin E phosphodegron, which is presumed to be the default state after translation. Based on the crystal structure of a peptide corresponding to the cyclin E phosphodegron bound to Fbw7, a high-affinity interaction depends on the Pro-Pro peptide bond being in the trans configuration (20). Therefore, the structure of the phosphodegron in native cyclin E must favor the cis conformation. This conformation would allow only a low-affinity interaction because the second phosphate at position 384 (pSer384) would not be capable of interacting with a basic surface on the phosphodegron interaction site of Fbw7, required for a high-affinity interaction (20). Presumably the α-specific domain of Fbw7α, which binds Pin1 (23), is responsible in some unknown manner for stabilizing this weak interaction, for promoting Pin1-mediated noncanonical cis-trans isomerization of the proline-proline bond, and for release of isomerized cyclin E without ubiquitylation so that it can then bind to and be processed by SCF^Fbw7γ. It has been shown that Pin1 binding to cyclin E depends on phosphorylation of Ser384 (but not Thr380) (30), perhaps providing the additional binding energy required for forming a high-affinity interaction in a ternary complex and preventing a direct high-affinity interaction between cyclin E and Fbw7α. If the second proline in the phosphodegron is replaced with a different amino acid, this system is completely bypassed, and SCF^Fbw7α is competent to ubiquitylate cyclin E with high efficiency (23).

The effects of cyclin E expression level on Fbw7 isoform usage.

It has been reported that HCT116 cells with deletions of sequences encoding the γ-specific domain of Fbw7γ are not defective in cyclin E degradation (27). However, in the same study, it was found that HEK293A cells, also used for some experiments in the present study, require both Fbw7α and Fbw7γ for degradation (23, 26), as we have shown previously. In an attempt to resolve this apparent discrepancy, we systematically tested the Fbw7 isoform requirements for a number of commonly used cell lines and found a clear stratification into two groups: one that requires only Fbw7α, as is exemplified by HCT116, and one that requires both Fbw7α and Fbw7γ, as described for HEK293A (26). The first group included U2OS, DLD1, and HeLa in addition to HCT116, whereas the second group included MCF7 and Saos2 (used in the current study), in addition to HEK293A, as well as nontransformed human diploid fibroblasts and hTERT-immortalized human mammary epithelial cells. The parameter that shows an absolute correlation with a requirement for Fbw7α only is highly elevated levels of cyclin E expression, which is characteristic of some tumor-derived cell lines. Low, presumably normal, levels of cyclin E expression correlate with a requirement for both Fbw7α and Fbw7γ. Indeed, by ectopically overexpressing cyclin E in immortalized mammary epithelial cells, it was possible to abrogate the requirement for Fbw7γ (26). One mechanism whereby cyclin E overexpression might bypass the requirement for Fbw7γ is the use of a second, low-affinity phosphodegron. This phosphodegron, which is centered around Thr62 of cyclin E, cannot bind the Fbw7 β-propeller with the same high affinity as the primary phosphodegron, which is centered around Thr380, but it does not contain a Pro-Pro bond, which would require isomerization (3, 20). Therefore, high cyclin E concentrations might drive sufficient interaction of this low-affinity degron with SCF^Fbw7α to ubiquitylate cyclin E in the nucleoplasm without participation of SCF^Fbw7γ.

How does cyclin E translocate into the nucleolus?

Nucleophosmin (NPM) has been shown to form a complex with Fbw7γ and to mediate its nucleolar translocation (31). Loss of NPM leads to nuclear localization of Fbw7γ but nucleolar exclusion. An AML-associated mutation that delocalizes NPM to the cytoplasm also delocalizes Fbw7γ to the cytoplasm, indicating that NPM dictates the subcellular location of Fbw7γ. Therefore, it is conceivable that NPM translocates Fbw7γ-substrate complexes into the nucleolus, possibly accounting for the patterns observed with cyclin E. The fact that silencing of Fbw7γ leads to actual nucleolar exclusion of cyclin E (Fig. 2A) suggests that localization of cyclin E in the nucleolus is not passive and that Fbw7γ has an active role in cyclin E nucleolar translocation. It has been proposed alternatively that Fbw7γ nucleolar localization is mediated by binding to the pseudosubstrate Ebp2, which is itself targeted to the nucleolus (41). However, the conclusions used to establish this relationship depend largely on Ebp2 overexpression. RNAi-mediated silencing of Ebp2 prevented Fbw7γ localization to the nucleolus, consistent with this model, but also perturbed nucleolar structure and was highly toxic to the cells. Therefore, a direct role for Ebp2 in Fbw7γ nucleolar localization is uncertain.

Pin1 and Fbw7 stability.

It was recently reported that Pin1 promotes Fbw7 autoubiquitylation and degradation in a phosphorylation-dependent manner (42). Specifically, depletion of Pin1 leads to accumulation of Fbw7 and a reduction of abundance of at least some SCF^Fbw7 substrates. In contrast, cyclin E accumulates under conditions of Pin1 depletion (23, 30) (Fig. 5), suggesting that the requirement for cyclin E phosphodegron isomerization supersedes any effect on Fbw7 levels.