Dysregulated behavior of cell cycle proteins and their control by ubiquitin E3 ligases is an emerging theme in human lung cancer. Here, we identified and characterized the activity of a novel F-box protein, termed FBXL2, belonging to the SCF (Skip-Cullin1-F-box protein) E3 ligase family. Ectopically expressed FBXL2 triggered G2/M-phase arrest, induced chromosomal anomalies and increased apoptosis of transformed lung epithelia by mediating polyubiquitination and degradation of the mitotic regulator, cyclin D3. Unlike other F-box proteins that target phosphodegrons within substrates, FBXL2 uniquely recognizes a canonical calmodulin (CaM)-binding motif within cyclin D3 to facilitate its polyubiquitination. CaM bound and protected cyclin D3 from FBXL2 by direct intermolecular competition with the F-box protein for access within this motif. The chemotherapeutic agent vinorelbine increased apoptosis of human lung carcinoma cells by inducing FBXL2 expression and cyclin D3 degradation, an effect accentuated by CaM knockdown. Depletion of endogenous FBXL2 stabilized cyclin D3 levels, accelerated cancer cell growth and increased cell viability after vinorelbine treatment. Last, ectopic expression of FBXL2 significantly inhibited the growth and migration of tumorogenic cells and tumor formation in athymic nude mice. These observations implicate SCFFBXL2 as an indispensible regulator of mitosis that serves as a tumor suppressor.
Lung cancer remains the leading cause of cancer deaths globally among men and women, and yet novel and selective metabolic targets for therapeutic intervention in this illness require further study (Weiderpass, 2010). The high proliferative capacity of lung neoplastic cells appears to be strongly linked to abberant regulation through critical checkpoints involved in cell cycle progression. D-type cyclins have integral roles in G1/S cell cycle progression (1995). Specifically, D-type cyclins function through four cyclin-dependent kinases (Cdks): Cdk 2, Cdk 4, Cdk 5 and Cdk 6. Active cyclin D/Cdk 4 and cyclin D/Cdk 6 partially phosphorylate retinoblastoma tumor suppressor protein (Rb) thereby reducing its binding to E2F, and thus promoting cell progression to the S-phase by allowing E2F-mediated activation of cyclin E gene transcription (Ohtani et al., 1995; Peeper et al., 1997). There are three major cyclin D family members: cyclin D1, cyclin D2 and cyclin D3, of which cyclin D1 has been well studied (Lew et al., 1991). As an important regulator for G1- to S-phase progression, cyclin D1 has an important role in multi-organ tumorogenesis (Zi and Agarwal, 1999; Zhang et al., 2005; Myung et al., 2006; Gautschi et al., 2007). Cyclin D1 degradation directly results in G1-phase arrest and thus might serve as a potential therapeutic target in neoplasia (Shan et al., 2009).
Cyclin D1 is degraded mainly through its site-specific ubiquitination and disposal by the proteasome. Specifically, the SCF (Skp1-Cullin1-F-box) E3 ligase family members FBXO4, FBXW8 and FBXO31 directly interact with cyclin D1 and mediate its ubiquitination (Lin et al., 2006; Okabe et al., 2006; Santra et al., 2009). However, more limited data are available regarding the function and molecular regulation of cyclin D3. Aside from its role in binding Cdk4 and Cdk6 to control G1- to S-phase transition, cyclin D3 may have other functions in cellular division and intracellular signaling (Tanguay et al., 2001). Cyclin D3 is ubiquitously expressed in several cell types, and also has a role during the G2/M-phase by interacting with CDK11p58 kinase during cell cycle progression. CDK11p58 promotes centrosome maturation and bipolar spindle formation, and its interaction with cyclin D3 preserves CDK11p58 activity. Thus, the cyclin D3–CDK11p58 interaction appears to be vital during mitosis and its abrogation could lead to G2 arrest (Zhang et al., 2002; Duan et al., 2010). Moreover, some studies associate cyclin D3 with metastasis of non-small cell lung carcinoma and with reduced survival (Papay et al., 2007; Sterlacci et al., 2010).
Ubiquitination of proteins brands them for degradation either by the proteasome or by the lysosome that regulates diverse processes (Tanaka et al., 2008). The conjugation of ubiquitin to a target protein is orchestrated by a series of enzymatic reactions involving an E1 ubiquitin-activating enzyme, ubiquitin transfer from an E1-activating enzyme to an E2-conjugating enzyme, and last, generation of an isopeptide bond between the substrate's ɛ-amino lysine and the carboxyl-terminus of ubiquitin catalyzed by a E3-ubiquitin ligase(Hochstrasser, 2000). Of the many E3 ligases, the SCF superfamily is among the best studied (Tyers and Willems, 1999). The SCF complex has a catalytic core complex consisting of Skp1, Cullin1 and the E2 ubiquitin-conjugating (Ubc) enzyme (Zheng et al., 2002; Cardozo and Pagano, 2004). The SCF complex also contains an adaptor receptor subunit, termed F-box protein, which targets hundreds of substrates through phospho-specific domain interactions (Cenciarelli et al., 1999). F-box proteins have two domains: an NH2-terminal F-box motif and a carboxyl-terminal leucine-rich repeat motif or WD repeat motif. The SCF complex uses the F-box motif to bind Skp1, whereas the leucine-rich/WD repeat motif is used for substrate recognition (Ilyin et al., 1999). Of the nearly 70 F-box proteins described, only ∼6 have defined roles in cellular processes (Skaar et al., 2009). Of note, the gene encoding the orphan F-box protein, FBXL2, was strongly repressed in human lung adenocarcinoma (Malard et al., 2007). Following its initial description (Ilyin et al., 1999), FBXL2 was shown to interact with the hepatitis C virus nonstructural protein 5A (NS5A), and this association was required for viral RNA replication (Wang et al., 2005). NS5A and the lipogenic enzyme, cytidylyltransferase, are the only known targets of FBXL2 and its effects within the cell cycle pathway have not been investigated (Chen et al., 2011).
Calmodulin (CaM) (16.7 kD) is a highly conserved calcium-sensing protein that antagonizes some proteinases and modulates stability of regulatory proteins (Rhoads and Friedberg, 1997). CaM binds its targets in a calcium-bound (holoCaM) or calcium-1-free (apopCaM) form, and thus its interactions with partners may be calcium dependent or calcium independent (Rhoads and Friedberg, 1997). Many CaM-binding proteins harbor recognition motifs characterized by a basic amphipathic helix, moderate to high helical hydrophobic moment and a net positive charge. Other motifs described include an IQ motif (I/LQXXXRGXXXR), a 1-8-14, and 1-5-10 CaM-binding motif (Rhoads and Friedberg, 1997). As a major intracellular calcium receptor, CaM exerts multiple modes of regulatory control within the cell cycle (Kahl and Means, 2003). For example, CaM appears essential for Cdk4 activity and nuclear accumulation of cyclin D1-Cdk4 during the G1-phase (Taules et al., 1998). CaM also binds to the centrosome protein CP110, and CaM RNA interference impairs the latter stages of cytokinesis leading to the formation of binucleate cells (Tsang et al., 2006). Interestingly, CaM also directly interacts with cyclin E and mediates calcium-sensitive G1/S-phase transition (Choi and Husain, 2006; Choi et al., 2006). We showed that CaM binds and stabilizes cytidylyltransferase required for phospholipid synthesis during cellular membrane formation suggesting that its combinatorial effects regulate processes intrinsic to cellular growth and repair (Chen and Mallampalli, 2007).
Here, by using a SV40-tumorogenic cell line (Wikenheiser et al., 1992), we show that cyclin D3 is polyubiquitinated by FBXL2 and degraded within the proteasome, and that this process is attenuated by CaM. Cyclin D3 stability is not controlled by FBXL2 targeting of a phosphodegron typical of F-box proteins. Rather, FBXL2 recognizes and binds in a calcium-independent manner to an IQ molecular signature also recognized by CaM. SCFFBXL2-mediated cyclin D3 ubiquitination and degradation was sufficient to induce mitotic arrest in the SV40 transformed cell line and inhibited growth of human adenocarcinoma xenografts in nude mice indicative of ability of this orphan F-box protein to exert antitumor activity.
Ectopic expression of FBXL2 induces G2 arrest and degradation of cyclin D3
We investigated FBXL2 activity on cell cycle progression first using a transformed murine lung epithelial (MLE) cell line and then extended studies to use of human adenocarcinoma cells. Overexpression of FBXL2, unlike mutant FBXL2 (N100) lacking its F-box domain, results in accumulation of rounded, apoptotic cells (Figure 1a). These morphological changes after ectopic expression of FBXL2 were associated with cell cycle arrest, as evidenced by reduced DNA synthesis and a significant increase in a cell population within the G2/M-phase that accumulated in a time-dependent manner (Figure 1b, Supplementary Figure S1). To investigate the mechanism by which FBXL2 induced G2/M arrest, we analyzed immunoreactive levels of 15 cell cycle regulatory proteins after FBXL2 overexpression. Surprisingly, we observed that the levels of cyclin D2 and cyclin D3, two G1/S-phase regulators, markedly decreased (Figure 1c). In contrast, levels of other G1 cyclins and the G2/M regulators cyclin A and cyclin B were unaffected. Ectopic FBXL2 expression did not alter levels of other Cdks or negative control proteins, but did decrease levels of a known substrate, CCT (Chen et al., 2011). Of note, ectopic expression of dominant negative mutant FBXL2N100 modestly increased immunoreactive cyclin D2 and cyclin D3 levels (Supplementary Figure S2). RNAi-mediated knockdown of cyclin D3, but not cyclin D2, reproduced in part the effects of FBXL2 overexpression with regard to G2 arrest in cells (Figure 1d and Supplementary Figure S3). Cyclin D3 silencing resulted in significant increases (∼50%) in the G2-phase, whereas knockdown of cyclin D2 showed limited effects (Supplementary Figure S4A). Knockdown of cyclin D3, but not cyclin D2, promotes apoptosis (Supplementary Figure S4B). RNAi-mediated knockdown of FBXL2 had no effect on the G2-phase, but tended to increase the S-phase and reduced the G1-phase (Supplementary Figure S5). These results were somewhat marginal (∼30%), but suggest that in MLE cells, cyclin D3 as a D-type cyclin also triggers G1–S-phase progression. Alternatively, FBXL2 might exert a more complex effector role with other yet unidentified substrates during cell cycle progression, some of which may be required for S-phase entry or egress.
We next examined cyclin D3 half-life after FBXL2 overexpression and knockdown. Ectopic FBXL2 expression significantly decreased t1/2 of cyclin D3, whereas knockdown of FBXL2 markedly increased cyclin D3 t1/2 (Figure 1e). Consistent with these results, FBXL2 overexpression did not significantly alter CCND2 or CCND3 steady-state messenger RNA levels (Supplementary Figure S6). To further assess the specificity of these effects, a random F-box protein (FBXL12) from the FBXL family was selected and ectopically expressed in cells demonstrating lack of ability of this protein to disrupt cyclin D3 stability (Supplementary Figure S7). Moreover, loss and gain-of-function studies of FBXL2 did not alter cyclin D1 t1/2 (Supplementary Figure S8). Last, ectopic expression of FBXL2 results in a reduction in endogenous cyclin D2 and cyclin D3 levels in other cell lines (Supplementary Figure S9).
FBXL2 targets cyclin D3 for ubiquitination during mitosis
To test whether FBXL2 interacts with cyclin D3, we performed co-immunoprecipitation (co-IP) experiments. MLE cells were first synchronized separately to G1-, S- and the G2-phase; mitotic cells were also collected through shake-off. Cells were immediately lysed and subjected to FBXL2 intraperitoneally (i.p.) Cyclin D3 and FBXL2 immunoreactive levels oscillated during the cell cycle; low levels of endogenous cyclin D3 linked to high levels of FBXL2 tended to occur during mitosis (Figure 2a, left panel, and input, (M)). Also, cyclin D3 in the FBXL2 immunoprecipitates was only detected in mitotic cells (Figure 2a). Thus, endogenous FBXL2 protein targets cyclin D3 during mitosis. When a panel of cells was immunostained for cyclin D3 and counterstained with 4,6-diamidino-2-phenyl indole to visualize the nucleus, the fluorescent intensity representing endogenous cyclin D3 levels was lower in identified cells undergoing mitosis when compared with cells in interphase (Figure 2b). In other experiments, synchronized cells were subjected to cyclin D3 co-IP; the immunoprecipitates were then resolved in SDS–PAGE and probed with ubiquitin. Polyubiquitinated products (a high molecular weight smear) of cyclin D3 were detected during mitosis (Figure 2c, arrows). Hence, endogenous FBXL2 targets these cyclins for polyubiquitination not during interphase but with mitosis. We next determined the subcellular pathway for FBXL2-mediated degradation of cyclins. Addition of the proteasomal inhibitor, MG132, to cells led to the accumulation of cyclin D3 protein whereas this was not seen with the lysosomal inhibitor, leupeptin (Figure 2d). Importantly, inclusion of purified SCFFBXL2 with the full complement of E1 and E2 enzymes plus ubiquitin was sufficient to generate polyubiquitinated cyclin D3 species in vitro (Figure 2e).
Cyclin D3 is polyubiquitinated within its C-terminus
To determine the ubiquitination acceptor site within cyclin D3, deletional and candidate approaches were used that suggested that Lys268 might be a functionally relevant molecular site (Figure 3a, data not shown). Thus, we examined polyubiquitination and stability of a Lys268R mutant in cells (Figure 3b). MG132 treatment triggered appearance of polyubiquitinated wild-type cyclin D3; in contrast, the proteasomal inhibitor failed to increase accumulation of the cyclin D3 mutant, suggesting that Lys268 is a putative ubiquitination site for cyclin D3 (Figure 3b). The Lys268R mutant exhibited significantly extended t1/2 compared with the wild-type cyclin (Figure 3c). Co-expression of FBXL2 with cyclins resulted in the degradation of wild-type cyclin D3, but not the Lys268R mutant (Figure 3d). Using in vitro ubiquitination assays in which wild-type or mutant cyclin D3 were reacted with the purified ubiquitin SCFFBXL2 complex, the Lys268R mutant was not ubiquitinated (Figure 3e). Importantly, after expression of mutant cyclin D3, ectopically expressed FBXL2 failed to induce efficient G2/M arrest (Figure 3f).
FBXL2 and CaM both vie for cyclin D3 docking
CaM binds and protects some regulatory proteins and is needed for cell cycle progression (Kahl and Means, 2003). Cyclin D3 harbors two potential CaM-binding IQ motifs within its NH2-terminus suggesting that these motifs may be required for CaM interaction (Figure 3a). Wild-type or NH2-terminal deletion cyclin D3 were transfected in cells, cell lysates were then applied to CaM–sepharose beads to test protein interaction. The pull-down experiments show that only wild-type cyclin interacts with CaM optimally with inclusion of ethylenediaminetetraacetic acid, and this interaction was significantly disrupted by even low micromolar calcium concentrations (Figure 4a, top panel). A cyclin D3 variant devoid of the amino-terminus failed to interact with CaM, supporting the premise that an IQ motif within this region is required for molecular interaction between the cyclin D3 and CaM (Figure 4a). We next identified which of the two potential IQ motifs within cyclin D3 are required for CaM binding. Glu100 of cyclin D3 was essential for CaM binding (Figure 4a, lower panel). Interestingly, FBXL2 also failed to interact with the NH2-terminal deletion cyclin D3 (Figure 4b, top panel). Importantly, FBXL2 also utilizes this molecular site (Glu100) within this motif to target cyclin D3 (Figure 4b, lower panel). Co-expression of FBXL2 with cyclins resulted in the degradation of wild-type cyclin but not the cyclin D3Q100A point mutant, and failed to induce efficient G2/M arrest (Figures 4c and d). For confirmation, in vitro ubiquitination assays demonstrated that the point mutant of cyclin D3 (Q100A) was not ubiquitinated (Figure 4e), and this variant exhibited a significantly longer t1/2 compared with wild-type cyclin D3 (Figure 3f). FBXL family proteins contain leucine-rich repeats for substrate targeting and residues 80–423 contain 12 leucine-rich repeats that display extensive internal homology (Figure 4g). In mapping studies, cell lysates expressing his-tagged FBXL2 truncation mutants were co-purified with GST-cyclin D3 using his-pull downs. The data show that deletion of the last five LLRs (C250) or the last two LLRs (C350) markedly disrupted FBXL2–cyclinD3 interaction. Thus, cyclin D3 binds FBXL2 within its last two LLR domains (350–423).
Cyclin D3 polyubiquitination is largely phosphorylation-independent
Traditional mechanisms of cyclin degradation appear to involve signals involving phosphorylation of Thr286 (Diehl et al., 1997). To determine whether FBXL2:cyclin D3 interaction is phosphorylation-dependent, mitotic cell lysate from MLE cells was treated with either vehicle or phosphatase followed by co-IP of endogenous FBXL2. The results indicate that cyclin D3 associates with FBXL2 in both vehicle and phosphatase-treated immunoprecipitated samples (Supplementary Figure S10A); thus, FBXL2 interaction with cyclin D3 is phosphorylation-independent. Similar results were observed in A549 cells (Supplementary Figure S10B). Further, both wild-type and phosphatase-treated cyclin D3 were prone to in vitro ubiquitination by FBXL2 (Supplementary Figure S11). Additional studies interrogated the corresponding T283 molecular site within cyclin D3. Both wild-type cyclin D3 and a cyclin D3 Thr283A mutant-bound FBXL2 and FBXL2 ubiquitinated both substrates in vitro (Supplementary Figure S12). Last, although the Thr283A mutant displayed modestly extended protein t1/2 compared with wild-type cyclin D3, only a double mutant (Q100A Thr283A) exhibited significant stability (Supplementary Figure S13). The results do not exclude cyclin D3 ubiquitination by other F-box proteins, but suggest strongly that FBXL2 partakes in its turnover and that this involves an IQ motif.
SCFFBXL2-directed ubiquitination of cyclin D3 is antagonized by CaM
CaM overexpression significantly increased cyclin D3 half-life, whereas CaM knockdown decreased the stability of the protein (Figure 5a). Similar to ectopic FBXL2 expression and CaM repression (Lu et al., 1992), CaM silencing resulted in G2/M arrest, whereas overexpression of CaM increased the proportion of cells within the S-phase (Figure 5b). CaM also directly interacts with FBXL2, as several bulky and hydrophobic residues within the F-box amino-terminus appear important for CaM interaction (data not shown). We next tested the functionality of mutant FBXL2 proteins. Cells transfected with either wild-type FBXL2 or FBXL2F79A (that does not bind CaM, data not shown) effectively decreased cyclin levels (Figure 5c, left panel) and resulted in G2/M arrest (Figure 5d). When cells were pre-infected with a replication-deficient adenovirus expressing CaM (Adv-CaM) and then transfected with these FBXL2 plasmids, CaM gene transfer was not able to totally rescue cyclin D3 protein levels (Figure 5c right panel), nor prevent G2/M arrest when cells were transfected with FBXL2F79A (Figure 5d). These results suggest that in addition to competition for IQ motif occupancy within cyclins, CaM's ability to directly bind and sequester FBXL2 to protect cyclin D3 might also be functionally important.
We also investigated the FBXL2, CaM and cyclin interactions during cell cycle progression. Lysate from synchronized cells were subjected to FBXL2, cyclin D3 i.p., the immunoprecipitates were then resolved in SDS–PAGE and probed with CaM. There was significant loss of FBXL2/CaM binding during mitosis and CaM associates with cyclin D3 at the highest levels during the S-phase with limited, if any, binding during mitosis (Figure 5e). As both CaM and FBXL2 interact with cyclins through the same IQ motif, we next tested the concept that there exists intermolecular competition between FBXL2 and CaM for cyclin D3 binding. This was analyzed using pull-down experiments in the presence or absence of calcium in which FBXL2 was immobilized on beads and used as bait for cyclins and CaM (Figure 5f). Three negative controls were included: (1) FBXL2-agarose alone was assayed to control for cyclin and CaM contamination; (2) Cyclin D3 and CaM with calcium were run over empty talon beads, eluted and proteins were resolved by SDS–PAGE followed by cyclin and CaM immunoblotting to ensure that associations were FBXL2-specific; and (3) V5 immunoblotting was used as a loading control, to ensure that pull-down experiments had equivalent amounts of FBXL2. The results indicate that not only does FBXL2 directly interact with cyclin D3 and CaM, but that excess CaM disrupts FBXL2 interaction with cyclin D3 (Figure 5f). Interestingly, excess calcium promotes cyclin D3 binding to FBXL2. By using a synthetic cyclin D3 peptide (LQLLGTVCLL) encoding the putative CaM-binding motif, we observed tight binding between CaM and cyclin D3 (Kd=0.31 μM, Figure 5g) using isothermal calorimetry.
Ectopic expression of FBXL2 induces apoptosis
Ectopic expression of FBXL2 triggered an increase in apoptosis in human lung adenocarcinoma (A549) cells by fluorescence-activated cell sorting (FACS) analysis using annexin V staining (Figures 6a and b). Viable cells were also quantified (Figure 6c), showing that FBXL2 markedly inhibited the growth of A549 cells in culture. A549 cells transfected with small interfering RNA (siRNA) targeting CaM or FBXL2 also displayed differential cell numbers, as FBXL2 knockdown promoted growth of A549 cells in culture, consistent with the ability of FBXL2 knockdown to increase S-phase (Supplementary Figure S5). We further tested the ability of lung cancer chemotherapeutic agents to induce FBXL2 levels. Of the agents examined, only vinorelbine, a drug that causes G2/M arrest in lung cancer, increased FBXL2 levels and decreased cyclin D3 content (Figure 6e). The kinetics of vinorelbine's effects revealed that by 15 h sufficient induction of FBXL2 and decreased levels of cyclin D3 were observed (Figure 6f). A549 cells were also transfected with siRNA against FBXL2 or CaM for 48 h, before exposure to vinorelbine for additional 48 h. Apoptosis was measured by FACS analysis using annexin V staining (Figure 6g) and quantified (Figure 6h). The results demonstrate that CaM knockdown accentuated, whereas FBXL2 silencing reduced A549 cells apoptosis induced by vinorelbine.
Ectopic expression of FBXL2 inhibits tumorogenicity
Initially, we assessed tumor cell migration using a would-healing assay. FBLX2 overexpression inhibited A549 cell migration, whereas co-expression of a proteolytically resistant cyclin D3 was able to restore cell migratory activity (Figure 7a). We also transfected A549 cells with a plasmid encoding doxycycline-inducible FBXL2. Doxycycline increased levels of immunoreactive FBXL2 that in turn induced cyclin D3 degradation coupled with an increase in cellular apoptosis (Figure 7b). As a complementary in vivo model, we assessed tumorogenicity after implanting A549 cells under either stable or inducible expression of FBXL2 in athymic nude mice. Stable expression of FBXL2 significantly reduced tumor size compared with the control implants (Figure 7c). Similar results were also obtained by using the inducible system for FBXL2 expression (Supplementary Figure S14). Importantly, when tumor tissues were collected from four control and four FBXL2 mice at the end point and analyzed, immunoblotting showed significant decreases in cyclin D3 protein levels coupled with increased levels of FBXL2 protein (Figure 7d).
The precise role of the majority SCF-based E3 ligases, including F-box proteins, has not been established. The observation that FBXL2 is repressed in human lung adenocarcinoma (Malard et al., 2007) raises the possibility that it might regulate molecular programs involved in neoplasia, such as cell cycle progression. Here, we show that both ectopically expressed and endogenous FBXL2 regulate viablity and proliferation of transformed or tumorogenic epithelia by ubiquitin-mediated degradion of cyclin D3. We also observe opposing activities of SCFFBXL2 and a ubiquitous calcium sensor, CaM, on cyclin D3 polyubiquitination and degradation. FBXL2 functions as the receptor component of a prototypical SCF ubiquitin E3 ligase that targets a CaM-binding signature; this represents a unique property of FBXL2 as other F-Box proteins recognize phosphoserine or phosphothreonine sites within target substrates (Liu et al., 1999; Hansen et al., 2004; Watanabe et al., 2005). For example, the well-studied family member cyclin D1 harbors a key site (Thr286) for recognition by the E3 ligase subunits FBXO4, FBXW8 and FBXO31 (Lin et al., 2006; Okabe et al., 2006; Santra et al., 2009). Although this Thr site is conserved among the D cyclins and might also be used as a targeting signal by cyclin D3 for other F-box proteins, FBXL2 did not utilize this signature for its substrate. This was evidenced by the ability of both dephosphorylated cyclin D3 and a related point mutant (cyclin D3T283A) to bind and be ubiquitinated by FBXL2 (Supplementary Figures S10–S12). Rather, FBXL2 docks within a consenus IQ signature within its substrate to facilitate ubiquitination. This was evidenced by ectopic expression of point mutants of cyclin D3 within the IQ motif that were resistant to ubiquitination and were sufficient to extend cyclin D3 half-life (Figure 4f). Whether FBXL2 uses this mode of targeting to other substrates requires additional investigation, but these results provide the first evidence that this F-box protein appears to be a major regulator of cyclin D3 lifespan and thus might serve as a key growth inhibitory signal.
That F-box protein-mediated ubiquitination and proteasomal degradation of cyclin D3 would result in G2/M arrest was unexpected, as these cyclins predominantly regulate G1/S transition, and knockdown of cyclin D3 has been shown to induce G1 arrest (Sicinska et al., 2003). However, cyclin D3 may have a dual role in also mediating G2/M-phase progression (Fang et al., 2002; Zhang et al., 2002). Thus, in lymphoblastic cells that selectively and highly express cyclin D3, its knockdown was expected to result in G1 arrest given the role of D-type cyclins in G1–S progression. However, in our studies, MLE cells express all three D-type cyclins, and overexpression of FBXL2 selectively downregulated cyclin D3, but not cyclin D1; hence because of cyclin D redundancy G1/S-phase progression is preserved and G2/M arrest was observed. In addition to this redundancy, the G1/S-phase may also be less prone to FBXL2-induced blockade because of lower FBXL2 levels during interphase and higher levels of CaM and cyclin D3.
As above, cyclin D3 is also a multi-functional protein that directly interacts with and confers activity for CDK11p58 (Zhang et al., 2002; Duan et al., 2010), a key cyclin-dependent kinase that controls centrosome maturation and bipolar spindle formation (Petretti et al., 2006). As with ectopic FBXL2 in cells, knockdown of CDK11p58 results in G2 arrest and apoptosis; significant CDK11 depletion results in misaligned and lagging chromosomes, permanent mitotic arrest and cell death (Hu et al., 2007). Hence, SCFFBXL2-directed ubiquitination and degradation of cyclin D3 would potentially impair its association with CDK11p58 and reduce its activity. One additional function of CDK11 is to recruit Polo-like kinase 4 (PLK4) and Aurora A to the centrosome that also regulate mitotic events and chromosomal stability (Petretti et al., 2006). Knockout of PLK4 or expression of a defective mutant Aurora protein also results in apoptosis (Glover et al., 1995; Rosario et al., 2010). Collectively, these results suggest that one explanation for G2/M-phase delay and apoptosis might involve SCFFBXL2 inactivation of CDK11p58 by ubiquitination and depletion of cyclin D3. This mechanism would dislocate PLK4 and Aurora A causing cell cycle arrest (Figure 8). In support of this, ectopic expression of FBXL2 does not decrease CDK11 protein levels but reduces assembly of the cyclin D3: CDK11p58 complex and binding of PLK4 and Aurora A (data not shown). There also exist some differing effects on cycle progression after silencing cyclin D2 or cyclin D3. Knockdown of cyclin D3, but not cyclin D2, resulted in G2/M arrest. As both cyclins are highly conserved and exert some redundant functions, a more modest phenotype observed with cyclin D2 knockdown might be because of compensation by cyclin D3. This would occur especially if lower levels of cyclin D2 are present in MLE cells compared with cyclin D3 as seen with A549 cells and CHO cells (Supplementary Figure S9).
We have uncovered functionally distinct domains that govern molecular interplay between the effectors, FBXL2 and CaM, and the putative target, cyclin D3. First, cyclin D3 harbors a canonical IQ motif typical of calcium-independent CaM-binding proteins (Figure 4a). Accordingly, both FBXL2 and CaM target this signature within cyclin D3 and bind in the absence of calcium (Figures 4a and 5e). Structural analysis predicts a α-helix for the IQ signature within cyclin D3 (LQLLGTVCLL). Interestingly, cyclin D1 contains a modified sequence (LQLLGATCMF) with substitution of a Thr for a hydrophobic residue at position seven. This could explain our findings that ectopic expression or knockdown of FBXL2 does not affect cyclin D1 levels (Supplementary Figure S8). Alternatively, this could be because cyclin D3 is the dominant D-type cyclin within the MLE cells, thus prone to FBXL2 targeting.
Our data indicate that calcium availability differentially regulates the interactions between the F-box protein, CaM and the FBXL2 substrate. In our previous studies, the presence of calcium did not alter the molecular interaction between CaM and cytidylyltransferase (Chen and Mallampalli, 2007) and yet CaM binding with cyclin D3 was completely interrupted by even 1 μM calcium. In both cases, calcium promotes FBXL2 interaction with its substrates, although FBXL2 still associates with cyclin D3 in the absence of calcium. Physiologically, these molecular interactions might depend upon subcellular compartmentalization of binding partners and availability of calcium signals within these compartments. For example, cytidylyltransferase is an amphitrophic enzyme that largely exists in the cytoplasm in lung cells whereas cyclins are nuclear (Ridsdale et al., 2001). Hence, cytidylyltransferase would be predicted to be protected by CaM in settings when calcium levels are very low, but during sepsis when cytosolic calcium currents fluctuate, it might be prone to ubiquitination by calcium-activated SCFFBXL2 (Chen et al., 2011). In the nucleus, the physiological levels of calcium will also likely regulate these molecular interactions during cell division. Although cyclin D3 binds CaM in the absence of calcium, a modest increase in calcium (∼1000 nM) almost totally disrupts cyclin D interaction with CaM (Figures 4a and 5e) and calcium increases cyclin D3 binding to FBXL2 (Figure 5f). However, during the G1/S-phase, CaM binds cyclin D3 (Figure 5e) when low levels of calcium (20–40 nM) are typically present in the nucleus providing an environment conducive for these interactions (Korkotian and Segal, 1996). Moreover, during prophase, the nuclear envelope is disrupted, chromosomal condensation occurs and nuclear contents are transiently exposed to higher calcium concentrations within the cytosol (200–1000 nM) (Brown and Shoback, 1984; Pszczolkowski et al., 1999). The prediction is that these higher calcium concentrations would trigger CaM dissociation from cyclin D3 to enhance cyclin vulnerablity for SCFFBXL2-mediated ubiquitination. Here, calcium might have a more important role in destabilizing cyclin D3 by releasing CaM and recruiting SCFFBXL2 to the IQ motif.
Our data suggest that cyclin D3 availability required for cell cycle progression will also depend on the relative binding affinities between FBXL2, CaM and their targets. Each of these proteins was demonstrated to interact in vitro (Figure 5f). Of note, despite ectopic expression of an Adv5-CaM in cells, this was inefficient at restoring cyclin D3 levels when co-expressing an FBXL2 mutant that lacks ability to interact with CaM. These results suggest that CaM's ability to act as a decoy to directly antagonize and sequester the F-box protein represents a mechanistically relevant interaction in addition to FBXL2 and CaM intermolecular competition for occupancy within the cyclin IQ motif. However, our isothermal calorimetry studies demonstrating very low binding constants between CaM and cyclin D3 (Kd=0.31 μM, (Figure 5g)) versus relatively higher values between CaM and FBXL2 (Kd=0.81 μM, data not shown) suggest that CaM competition with FBXL2 might be a more functionally relevant mechanism in vivo.
The data presented here suggest that tight interplay between F-box protein FBXL2 and CaM will regulate mitotic events through control of cyclin D3 abundance. CaM has a vital role in centrosome formation during mitosis by interacting with the centrosome protein, CP110; mutant CP110 that lacks ability to bind CaM leads to failure of cytokinesis (Tsang et al., 2006). Counter-intuitively, CaM fails to protect cyclin D3 during mitosis thereby potentially contributing to SCFFBXL2-induced ubiquitination and degradation. As stated above, cyclin D3-dependent CDK11p58 activity is also essential for mitosis, but excessive CDK11p58 levels repress cellular proliferation and induce apoptosis (Duan et al., 2010). Hence, the dissociation of CaM from cyclin D3 and its targeting by the SCFFBXL2 complex during the transition to mitosis might be an exquisite mechanism to balance CDK11p58 levels thereby regulating cell proliferation.
Materials and methods
The sources of the transformed MLE cell line, CaM, Erk, LPCAT1 and GST antibodies, were described previously (Chen and Mallampalli, 2007; Ray et al., 2010). Purified SCFFBXL2 was purchased from Abnova (Walnut, CA, USA). Purified bovine CaM, ubiquitin, E1, E2, MG132, leupeptin and cycloheximide were purchased from Calbiochem (La Jolla, CA, USA). Rabbit polyclonal ubiquitin, cyclin and Cdk sampler kits were purchased from Cell Signaling (Danvers, MA, USA). Mouse monoclonal cyclin D2 and cyclin D3 antibodies were from Abcam (Cambridge, MA, USA). Doxycycline was purchased from Clontech (Mountain view, CA, USA). Lipofectamine 2000, mouse monoclonal V5 antibody, 4,6-diamidino-2-phenyl indole nuclear staining kits, the pcDNA3.1D cloning kit, E. coli One Shot competent cells, the pENTR Directional TOPO cloning kits and the Gateway mammalian expression system were from Invitrogen (Carlsbad, CA, USA). BD TALON purification and buffer kits, and FACS kits were purchased from BD Biosciences (San Jose, CA, USA). The F-box proteins cDNA were purchased from OpenBiosystems (Huntsville, AL, USA). Nucleofector transfection kits were from Amaxa (Gaithersbury, MD, USA). Adenoviral constructs encoding CaM were generated as described (Chen and Mallampalli, 2007). Immobilized protein A/G beads were from Pierce (Rockford, IL, USA). Cell viability based on adenosine triphosphate generation was assayed using a CellTiter-Glo Luminescent Cell Viability kit from Promega (Madison, WI, USA). Annexin V staining kits and the complete proteasome inhibitors were from Roche (Madison, WI, USA). Lambda protein phosphatase was from New England Biolabs (Ipswich, MA, USA). Goat polyclonal FBXL2 antibody, scrambled RNA and siRNAs were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Rabbit polyclonal FBXL2 antibody was custom made from Rockland Immunochemicals Inc. (Gilbertsville, PA, USA). All DNA sequencing was performed by the University of Pittsburgh DNA Core Facility.
MLE cells were cultured in Dulbecco's Modified Eagle Medium-F12 (Gibco, Carlsbad, CA, USA) supplemented with 2–10% fetal bovine serum (DMEM-2 or 10). Cells in some studies were synchronized using serum starvation (DMEM-F12) for 48 h or treatment with nocodazole or aphidicolin. In other studies, cells were treated with leupeptin or MG132 at 1:2000 dilution. For half-life studies, cells were treated with cycloheximide (40 μg/ml) at different time points. Cells lysates were prepared by brief sonication in 150 mM NaCl, 50 mM Tris, 1.0 mM ethylenediaminetetraacetic acid, 2 mM dithiothreitol, 0.025% sodium azide and 1 mM phenylmethylsulfonyl fluoride (Buffer A) at 4 °C.
Expression of recombinant protein and RNAi
All plasmids were delivered into cells using nucleofection or lipofectamine 2000 (Chen and Mallampalli, 2009; Agassandian et al., 2010). Cellular expression of green fluorescent-tagged plasmids using this device was achieved at >90%. For overexpression of CaM, 4 × 106 cells were plated in 100 mm dishes for 24 h, then infected with Adv-CaM or an empty vector (Adv-Con) at MOI=40 for 12 h followed by FBXL2 plasmid expression. For siRNA studies, 1 × 106 cells were transfected using lipofectamine 2000 with 10 μg of RNA and harvested after an additional 48 h.
Co-IP and binding assays
In all, 250 μg of total protein from cell lysates was precleared with 20 μl of protein A/G beads for 1 h at 4 °C. Also, 5 μg of primary antibody was added for 18 h incubation at 4 °C. Forty microliters of protein A/G beads was added for an additional 6 h of incubation. Beads were slowly centrifuged and washed five times using 50 mM HEPES, 150 mM NaCl, 0.5 mM EGTA, 50 mM NaF, 10 mM Na3VO4, 1 mM phenylmethylsulfonyl fluoride, 20 μM leupeptin and 1% (v/v) Triton X-100 (RIPA) buffer, as described (Mallampalli et al., 2000). The beads were heated at 100 °C for 5 min with 80 μl of protein sample buffer before SDS–PAGE and immunoblotting. For CaM-binding assays, CaM sepharose beads were incubated with V5-FBXL2 or cyclin D3-transfected cell lysates (50 μg), with or without Ca2+ at 4 °C for 2 h. Eluted products were processed for SDS–PAGE and immunoblotting as described (Chen and Mallampalli, 2007).
Cells (2 × 105) were plated at 70% confluence on 35 mm MetTek glass bottom culture dishes (MetTek, Ashland, MA, USA). Immunofluorescent cell imaging was performed on a Nikon A1 confocal microscope (Nikon, Melville, NY, USA) using 405 nm, 458 nm, 488 nm, 514 nm or 647 nm wavelengths. All experiments were done with a 60 × oil differential interference contrast objective lens. Cells were washed with PBS and fixed with 4% paraformaldehyde for 20 min, then exposed to 15% BSA, 1:500 primary antibodies and 1:1000 Alexa 488 or Alexa 647-labeled goat antimouse or rabbit secondary antibody sequentially for immunostaining.
In vitro ubiquitin conjugation assay
The ubiquitination of V5-cyclin D3 was performed in a volume of 25 μl containing 50 mM Tris pH 7.6,. 5 mM MgCl2, 0.6 mM dithiothreitol, 2 mM adenosine triphosphate, 1.5 ng/μl E1, 10 ng/μl Ubc5, 10 ng/μl Ubc7, 1 μg/μl ubiquitin (Calbiochem), 1 μM ubiquitin aldehyde, 4–16 μl of purified Cullin1, Skp1, Rbx1 and in vitro synthesized FBXL2. Reaction products were processed for V5 immunoblotting.
Quantitative RT–PCR, cloning and mutagenesis
Total RNA was isolated and reverse transcription was performed followed by quantitative real-time PCR with SYBR Green qPCR mixture as described (Butler and Mallampalli, 2010). All mutant constructs were generated using PCR-based approaches using appropriate primers or site-directed mutagenesis. PCR-based approaches were used to clone FBXL2 into Plvx and Plvx-Tight vectors (Clontech) for constitutive or inducible expression of FBXL2 in A549 cells. To generate lentivirus encoding FBXL2, Plvx-FBXL2 and Plvx-Tight-FBXL2, Plvx-TetOn plasmids were co-transfected with Lenti-X HTX packaging plasmids into 293FT cells following the manufacturers’ instructions. After 72 h, virus was collected and titered using a p24 Rapid Titration Kit (Clontech). A549 cells were either infected with lentivirus encoding Plvx-FBXL2 for constitutive FBXL2 expression or infected with lentivirus encoding Plvx-Tight-FBXL2 and Plvx-TetOn for inducible FBXL2 expression.
Cell cycle and apoptosis analysis
Transfected cells were incubated with BrdU (20 μM) for 40 min, fixed and stained following manufacturer's protocols (BD Biosciences, Sparks, MD, USA). FACS samples were analyzed with the AccuriC6 system. DNA content was analyzed using FCS3 express software (De Novo Software, Los Angeles, CA, USA). When analyzing cell cycle, a gate for 7AAD was set to exclude polyploidy cells. Otherwise cells were counted and the percentage of cells with 2N, 4N and 8N DNA content was expressed as a percentage of total cells. Cells were also stained with Annexin V for 15 min following the manufacturer's protocol (Roche). Apoptotic cells were counted, and apoptotic cells were expressed as a percentage of total cells.
Cell migration assays
A549 cells were plated in 6-well plates followed by transfection of the plasmids. After 24 h, a thin line was created on the cell monolayers by scratching using a pipet tip. Cells were observed under light microscopy and photos taken over time and migratory activity representative of three individual wells was analyzed by ImageJ software (NIH, Bethesda, MD, USA).
Nude/Nude mice (purchased from Charles River, Wilmington, MA, USA) were acclimated at the University of Pittsburgh Animal Care Facility, and maintained according to all federal guidelines and under the University of Pittsburgh Institutional Animal Care and Use Committee (IACUC)-approved protocols. Mice were deeply anesthetized with ketamine (80–100 mg/kg i.p. and xylazine (10 mg/kg i.p.), followed by i.p. injection of 5 × 106 A549 cells (100 μl) into the left shoulder. To induce FBXL2 expression, 2 mg/ml of doxycycline was added to the drinking water. Mice were closely monitored every 3 days; tumor volume was calculated using a formula length × width × height × π/6.
Statistical comparisons were performed with the Prism program, version 4.03 (GraphPad Software Inc., San Diego, CA, USA) using an ANOVA 1 or an unpaired 2 t-test with P<0.05 indicative of significance.
We thank AF Stewart for critical review of the manuscript and helpful suggestions. This material is based upon work supported, in part, by the US Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Biomedical Laboratory Research and Development. This work was supported by a Merit Review Award from the US Department of Veterans Affairs and National Institutes of Health R01 grants HL096376, HL097376 and HL098174 (to RKM). The contents presented do not represent the views of the Department of Veterans Affairs or the United States Government.
About this article
Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)