|Home | About | Journals | Submit | Contact Us | Français|
CHFR is an E3 ubiquitin ligase and an early mitotic checkpoint protein implicated in many cancers and in the maintenance of genomic stability. To analyze the role of CHFR in genomic stability, by siRNA, we decreased its expression in genomically stable MCF10A cells. Lowered CHFR expression quickly led to increased aneuploidy due to many mitotic defects. First, we confirmed that CHFR interacts with the mitotic kinase Aurora A to regulate its expression. Furthermore, we found that decreased CHFR led to disorganized multipolar mitotic spindles. This was supported by the finding that CHFR interacts with α-tubulin and can regulate its ubiquitination in response to nocodazole and the amount of acetylated α-tubulin, a component of the mitotic spindle. Finally, we found a novel CHFR interacting protein, the spindle checkpoint protein MAD2. Decreased CHFR expression resulted in the mislocalization of both MAD2 and BUBR1 during mitosis and impaired MAD2/CDC20 complex formation. Further evidence of a compromised spindle checkpoint was the presence of misaligned metaphase chromosomes, lagging anaphase chromosomes, and defective cytokinesis in CHFR knockdown cells. Importantly, our results suggest a novel role for CHFR regulating chromosome segregation where decreased expression, as seen in cancer cells, contributes to genomic instability by impairing the spindle assembly checkpoint.
Checkpoint with forkhead and RING finger domains (CHFR) is recognized as a novel mitotic stress checkpoint pathway regulator and biomarker for chemotherapeutic response of cancer cells to taxanes. It delays cells in prophase, before the mitotic spindle checkpoint, after exposure to microtubule poisons (i.e., nocodazole or paclitaxel) [1–7]. Subsequently, CHFR has been implicated in oncogenesis. CHFR expression is lost or decreased in many types of tumors compared to normal tissues, sometimes due to promoter hypermethylation [5,7–13]. Importantly, CHFR has strong tumor-suppressive functions in both knockout mouse model and in immortalized human mammary epithelial cells (IHMECs) and breast cancer cell lines [3,14]. Prolonged loss of CHFR expression led to abnormal chromosome complements (i.e., aneuploidy) in both models.
Aneuploidy is a hallmark of many cancers and may result from diverse mitotic defects including multipolar spindles secondary aberrant cytokinesis or centrosome amplification, sister chromatid cohesion defects, incorrect centromere attachment, or an impaired mitotic spindle assembly checkpoint (“spindle checkpoint”) . The spindle checkpoint prevents chromosome mis-segregation during cell division by delaying anaphase until the kinetochores of all sister chromatids are attached to the mitotic spindle microtubules. Spindle checkpoint gene mutations are rare, especially in breast cancers, but many cancer cells have an impaired or unregulated spindle checkpoint [15–17]. Although an accumulating amount of work has been focused on characterizing the spindle checkpoint, many questions remain. It is known that MAD2 and BUBR1 are recruited to unattached kinetochores early in mitosis. Through a poorly understood mechanism, MAD2, BUBR1, and BUB3 then sequester CDC20, a cofactor of the anaphase-promoting complex (APC), to inhibit its ubiquitin ligase activity and delay anaphase onset. Once all of the kinetochores are attached to the mitotic spindle, the inhibitory complex dissociates and CDC20 is allowed to bind to the APC, activating its ubiquitin ligase activity leading to the degradation of securin, separase, and other mitotic proteins to permit chromosome segregation (reviewed in Yu ). However, the factors or events that both initiate and stop the checkpoint-signaling pathway are unclear.
We hypothesized that the transient loss of CHFR would cause genomic instability by deregulating proteins important for mitotic spindle formation and function. To test this, we transiently decreased CHFR expression by siRNA in the genomically stable IHMEC cell line, MCF10A. Subsequent analysis of these cells revealed that they were more aneuploid and had increased expression of mitotic proteins including Aurora A, α-tubulin, and acetylated α-tubulin. In addition, we identified MAD2 as a novel CHFR interacting protein in which loss of CHFR expression led to impaired MAD2/CDC20 complex formation and the mislocalization of the key spindle checkpoint proteins BUBR1 and MAD2. Additional evidence for an impaired spindle checkpoint after decreased CHFR expression included the identification of four mitotic defects: misaligned metaphase chromosomes, lagging anaphase chromosomes, disorganized multipolar spindles, and defective cytokinesis.
MCF10A and HEK293 cells were obtained from the American Type Culture Collection (Manassas, VA) and cultured under recommended conditions. For CHFR knockdown, cells were untransfected (“mock”) or transfected with 2.0 µM of either a nontargeting siControl siRNA or a set of four siRNAs targeting CHFR using Dharmafect1 according to the manufacturer's instructions and were analyzed 72 hours later (siGENOME; Dharmacon RNA Technologies, Lafayette, CO). Cells were transfected with 6.0 µg of a Flag-tagged Aurora A construct (gift from Xiaochun Yu, University of Michigan) or a Flag-tagged CHFR using FuGENE 6 (Roche Applied Science, Indianapolis, IN) and lysates were harvested 24 hours later. Cells were treated with 15 µM of MG132 (Calbiochem, Gibbstown, NJ) for 10 hours and/or 200 ng/ml nocodazole (Sigma-Aldrich, St. Louis, MO) for 18 hours. To induce DNA damage, cells were treated with 0.3 µM aphidicolin for 24 hours.
MCF10A cells were treated with 50 ng/ml colcemid (Invitrogen, Carlsbad, CA) for 16 hours then collected and resuspended in a hypotonic solution of 2% KCl and 2% Na3C6H5O7 for 7 minutes at 37°C. Metaphase spreads were then prepared and stained with Giemsa as previously described . At least 25 metaphases were counted in triplicate for each sample.
Spectral karyotyping (SKY) analysis was performed according to the manufacturer's protocol (Applied Spectral Imaging, Vista, CA) and as previously described . Briefly, cells and slides were prepared as described previously, and unstained slides were aged in 2x SSC, treated with pepsin (Amresco, Solon, OH; 30 µg/ml in 0.01 N HCl), then rinsed with PBS. Slides were postfixed in 1% paraformaldehyde in PBS/MgCl2 and dehydrated in an ethanol series before and after denaturation in a 70% formamide/2x SSC solution. The denatured SKY probes (vial 1, SKY kit; Applied Spectral Imaging) were hybridized to the slides and incubated at 37°C for 2 days. After washings, antibodies (from vials 3 and 4, SKY kit) were added and incubated at 37°C for 1 hour each. The slides were counterstained with DAPI in an antifade solution. All images were acquired using an SD200 SpectraCube spectral imaging system (Applied Spectral Imaging) attached to a microscope (E800; Nikon, Belmont, CA) consisting of an optical head (a Sagnac interferometer) coupled to a multiline charge-coupled device camera (Hamamatsu, Bridgewater, NJ). Spectral Imaging (v. 2.6.1) and Sky View (v. 1.6.2; both from Applied Spectral Imaging) were used to acquire and analyze the images, respectively. The average of ten metaphases was used to create the consensus karyotype.
Whole-cell lysates were collected from approximately 80% confluent cultures. For samples analyzed for ubiquitination of α-tubulin, 2 mM N-ethylmaleimide (NEM; Sigma) was added to the lysis buffer. Western blots were prepared as previously described . Western blot membranes were blocked for 1 hour at room temperature and incubated overnight at 4°C in primary antibody. The following antibodies were used: a mouse CHFR mAb (1:500 dilution; Abnova Corp., Taipei City, Taiwan), a custom rabbit polyclonal antibody to the N-terminus of CHFR (1:1000), and a rabbit anti-Aurora A antibody (1.0 mg/ml, gift from Xiaochun Yu). A mouse anti-α-tubulin, mouse anti-acetylated α-tubulin, rabbit anti-Flag (all from Sigma-Aldrich), rabbit anti-α-tubulin (Cell Signaling Technology, Danvers, MA), rabbit anti-CDC20 (Novus Biologicals, Littleton, CO), and a rabbit anti-glutathione-S-transferase (GST; Santa Cruz Biotechnology, Santa Cruz, CA) were all used at 1:1000 dilutions. Anti-ubiquitin (1:100; Sigma) was also used to identify ubiquitinated α-tubulin. A mouse anti-MAD2 (1:500; BD Biosciences, San Jose, CA) and goat anti-MAD2 (1:250, antibody N19; Santa Cruz Biotechnology) were also used. An anti-glyceraldehyde-3-phosphate dehyrogenase antibody (1:10,000; Abcam, Cambridge, MA) and an HRP-conjugated β-actin antibody (1:10,000; Sigma-Aldrich) were used for loading controls. Blots were incubated in secondary antibody, anti-mouse:HRP or anti-rabbit:HRP, diluted in the blocking solution (both from Cell Signaling Technology). We used the SuperSignal West Pico chemiluminescent kit (Pierce Biotechnology, Rockford, IL) and exposed the blots to a BioMax XAR film (Eastman Kodak, Rochester, NY). Blots were stripped before reprobing with a different antibody. Where applicable, blots were analyzed from three experiments to verify expression changes. Densitometry was performed using an imaging system (FluorChem 8900; Alpha Innotech Corp., San Leandro, CA).
Immunoprecipitations (IPs) were completed according to the manufacturer's instructions using the protein G immunoprecipitation kit (Sigma). Briefly, whole-cell lysates were combined with 5 to 10 µl of the specified antibody (mouse IgG1 isotype control from BD Biosciences, mouse anti-α-tubulin, mouse anti-Flag M2 antibody, or goat anti-MAD2) and diluted in the supplied 1x IP buffer and incubated for at least 2 hours at 4°C. Then, 50 µl of Protein G beads were added to the lysate/antibody mix and were incubated overnight at 4°C. After washes in 1x IP buffer and 0.1x IP buffer, the immunoprecipitated lysates were boiled in the columns in 40 µl of 2x Laemmli's loading buffer then eluted by centrifugation and analyzed by Western blot analysis.
A GST:CHFR fusion construct was created using the pGEX2T vector (Amersham, Piscataway, NJ) and was expressed in the DH5α strain of Escherichia coli. Logarithmic E. coli cultures were collected in lysis buffer (2.5 mM PMSF in 1.0% Triton X-100 with a protease inhibitor cocktail from Roche), sonicated, and then cleared by centrifugation. One milligram of E. coli lysates was combined with 50 µl of washed Glutathione Sepharose 4B beads (Amersham) for 2 hours at 4°C. Then, 1 mg of whole-cell lysates from MCF10A cells was added to the beads and incubated overnight at 4°C. After washes with NTEN200 buffer (20 mM Tris-HCl, 1 mM EDTA, 0.5% NP-40, 25 µg/ml PMSF, and 200 mM NaCl), the bound proteins were eluted with 10 mM glutathione and collected by centrifugation. Isolated proteins were identified by Western blot analysis.
Yeast two-hybrid screening was performed following the manufacturer's instructions (Clontech Laboratories, Inc., Mountain View, CA). Briefly, the full-length cDNA of human CHFR was subcloned into pGBKT7 to express the fusion peptide with Gal4BD. pGBKT7-CHFR was introduced into the yeast strain AH109, which was then transformed with a 293T cell cDNA library cloned into the pACT2 vector. A total of 1 x 107 transformants were plated on SD/-Leu/-Trp/-His agar medium containing 5 mM 3-aminotriazole, and the colonies that grew in the selective medium were assayed for β-galactosidase activity. Plasmid DNA was extracted from the positive clones and was sequenced.
In two-chambered slides, 3 x 105 MCF10A cells were plated in each chamber, and were then fixed in 4% paraformaldehyde and blocked in 5% milk, 1.0% BSA in 0.025% TBS-Triton X-100. Staining was performed using an anti-α-tubulin antibody (1:100; Sigma), a rabbit anti-Aurora A antibody (1:50; Cell Signaling Technology), or an anti-Histone H3-phospho-Ser28 antibody (1:100; Upstate-Millipore, Billerica, MA), all of which were hybridized in blocking buffer overnight at 4°C. Slides were hybridized with an anti-mouse:Alexa Fluor 594 or an anti-rabbit:Alexa Fluor 488 secondary antibody (Invitrogen Molecular Probes, Carlsbad, CA) diluted to 1:200 in blocking buffer. Samples were preserved with ProLong Gold antifade mounting medium with DAPI (Invitrogen Molecular Probes). For colocalization of CHFR andMAD2, slides were prepared as described previously using a monoclonal mouse anti-CHFR antibody (Abnova Corp.) at a 1:50 dilution and a goat anti-MAD2 antibody (antibody N19; Santa Cruz Biotechnology) at a 1:50 dilution. Anti-mouse:Alexa Fluor 594 and anti-goat:Alexa Fluor 488 secondary antibodies were used at a 1:100 dilution (both from Invitrogen Molecular Probes).
For BUBR1 and MAD2 localization, cells were fixed in 4% paraformaldehyde then permeabilized for 5 minutes in 0.5% Triton X-100 dissolved in 1x PBS. Slides were blocked in 5% milk in 0.1% TBS-Triton X-100 then hybridized with an anti-BUBR1 antibody (Sigma-Aldrich) or an anti-MAD2 antibody (BD Biosciences) at a 1:200 dilution in blocking buffer overnight at 4°C. Slides were hybridized with an anti-mouse:Alexa Fluor 594 secondary antibody (1:500; Invitrogen) diluted in blocking buffer for 1 hour at room temperature then preserved with ProLong Gold antifade mounting medium with DAPI (Invitrogen Molecular Probes).
We used a compound Leica DMRB microscope (W. Nuhsbaum, Inc., McHenry, IL) and either a x63 or a x100 objective lens. An external Leica EL6000 light source was used for immunofluorescence images. Images were recorded using a Retiga 2000R 12-bit digital camera and QCapture Pro v5.1 software (QImaging, Surrey, Canada). For the analysis of CHFR and MAD2 colocalization, a confocal microscope (FV-500; Olympus, Center Valley, PA) was used with a x40 objective lens and FluoView software (Olympus).
Images were processed for resolution, magnification, and gamma settings using Adobe Photoshop CS2. We used the analysis of variance test for statistical significance, and P < .05 was considered significant. Error bars depict the SE from triplicate experiments. One asterisk (*) indicates P ≤ .05 and two asterisks (**) indicate P ≤ .001.
We previously reported that the stable loss of CHFR expression by shRNA in IHMECs led to increased aneuploidy . Further analysis of these cells by SKY revealed two distinct cell populations—minimally more aneuploid or near tetraploid (Figure 1A compared to Figure 1, B and C). The parental karyotype of MCF10A cells, determined from eight metaphase spreads, was as follows: 48, XX,1qhph,+del(1)(p?),t(3;9)(p14;p21),+del(7)(q?),i(8)(q10),t(3;5) (p?;?). The consensus karyotype of five metaphase cells from the minimally greater aneuploid population of MCF10A cells expressing CHFR shRNA was as follows: 47~50,XX,+X,t(1;2)(q?;q?),t(3;9) (p14;p21),der(6)t(6;19)(p?;?),+del(7)(q?),t(3;5)(p?;?), der(11)t(8;11) (?;p?), t(15;18)(?;p?),+20. The consensus karyotype from five metaphase cells for the near tetraploid population was as follows: 81~95, XXXX,1,t(1;2)(q?;q?),der(2)t(1;2)(q?;q?),3,t(3;9)(p14;p21)x2,5, der(6)t(6;19)(p?;?)x2, +del(7)(q?),9,t(3;5)(p?;?),10,der(11)t(8;11) (?;p?)x2,13,der(15)t(15;18)(?;p?)x2,17,18x2,+20x2, 22. MCF10A cells with CHFR shRNA often gained chromosomes 20 and X and had four novel chromosomal translocations, namely, t(1;2), t(6;19), t(8;11), and t(15;18) (Figure 1, B and C), suggesting that CHFR may regulate genomic stability through multiple mechanisms.
To determine whether the genomic instability was a byproduct of prolonged culture after CHFR knockdown, MCF10A cells were transiently transfected with a pool of four siRNAs targeting CHFR (“MCF10A:CHFR-siRNA cells”) and were analyzed for aneuploidy and chromosome breakage. CHFR expression was decreased by at least 80% 72 hours after transfection as detected by Western blot analysis (Figure 1D). We observed no chromosome breaks on metaphase spreads after treatment with aphidicolin to induce DNA damage (data not shown). However, 32% of MCF10A:CHFR-siRNA cells showed increased aneuploidy, typically having 49 to 59 chromosomes, compared to less than 5% of the mock transfected and nontargeting (“siControl”) negative control counterparts 72 hours after transient transfection (Figure 1, E and F; P ≤ .001). This indicated that CHFR-associated aneuploidy occurs quickly and is not simply a result of prolonged cell culture conditions. Given this, we wanted to elucidate the mechanism(s) by which aneuploidy occurred in IHMECs that had lost CHFR expression.
It was previously shown that mouse embryonic fibroblasts (MEFs) and tissues from CHFR knockout mice overexpressed Aurora A kinase . To assess this and to provide confirmatory evidence in our human model, we performed Western blot analysis and found that MCF10A:CHFR-siRNA cells had much greater Aurora A expression compared to the control cells (Figure 2A). We also found that Flagtagged Aurora A could interact with endogenous CHFR in MCF10A cells by IP (Figure 2B). The physical interaction of these two proteins, combined with Aurora A overexpression in MCF10A:CHFR-siRNA cells, substantiates previously published observations that Aurora A is a target for CHFR-mediated ubiquitination for degradation .
Early in mitosis, Aurora A localizes to centrosomes where it mediates their maturation and separation and spindle formation . We found that Aurora A properly localized to the centrosomes during metaphase, as evidenced by the two distinct foci that colocalized to the spindle poles, and less than 1.5% of control cells had greater than two Aurora A foci (Figure 2, C, subpanels b, d, f, and h, and D). However, although Aurora A was properly localized, in 16% of MCF10A: CHFR-siRNA cells, more than two Aurora A foci were detected, indicating increased Aurora A expression and suggesting centrosome amplification (Figure 2, C subpanels j and l, and D; P < .05).
In MCF10A:CHFR-siRNA cells, the mitotic spindle was often more condensed with poor polar microtubule formation (Figure 2C, subpanels i and l). Because CHFR is best known for its role in delaying mitotic entry due to stress on the microtubules, we hypothesized that CHFR may interact with, and possibly regulate, tubulin proteins. We performed a GST pull down using a purified GST:CHFR fusion protein and lysates from MCF10A cells. We found that CHFR interacts with α-tubulin when the MCF10A cells were previously treated with nocodazole (Figure 3A). In addition, this was confirmed by coimmunoprecipitation experiments with a Flag-tagged CHFR construct expressed in HEK293 cells in which IP of both Flag-CHFR and α-tubulin were able to coimmunoprecipitate the other protein. However, the interaction was not dependent on nocodazole treatment using this method (Figure 3B).
To determine whether CHFR can ubiquitinate α-tubulin, we treated MCF10A cells that had been transfected with control or CHFR siRNAs with the proteasome inhibitor MG132 with or without concomitant treatment with nocodazole. After IP for α-tubulin and immunoblot analysis for ubiquitin, we found that CHFR is able to ubiquitinate α-tubulin during nocodazole exposure, as evidenced by the loss of ubiquitin signal in MCF10A:CHFR-siRNA cells (Figure 3C, lane 3 vs lane 6). Western blot analysis confirmed that CHFR can regulate α-tubulin because there was a reproducible increase in α-tubulin protein levels in MCF10A:CHFR-siRNA cells (Figure 3, D and E; P < .05). The amount of acetylated α-tubulin, a key component of the mitotic spindle, in MCF10A:CHFR-siRNA cells was consistently double that of controls (Figure 4, C and D; P < .05).
Because CHFR is an E3 ubiquitin ligase, it likely interacts with several proteins to regulate mitotic events. In an effort to identify novel CHFR interacting partners, we performed a yeast two-hybrid screen using a full-length human CHFR cDNA as bait. A 293T cell cDNA library was screened by using the Gal4-based yeast two-hybrid system. By screening 1 x 107 clones, we found nine positive clones. Among nine positive clones, three clones encoded the E2 ubiquitinconjugating enzyme UBE2N, further confirming the effectiveness of our yeast two-hybrid screening. We also identified Epsin, SKAP55, an uncharacterized protein encoded by cDNA FLJ14707, cone-rod homeobox protein (CRX), and electron transfer flavoprotein alpha subunit (ETFA) as additional CHFR interacting proteins. One clone encoded the full length of MAD2, a spindle checkpoint protein that ensures proper chromosome segregation.
To confirm an interaction between CHFR and MAD2 in vivo, we performed immunofluorescence with confocal microscopy to visualize endogenous CHFR and MAD2 proteins. These two proteins colocalized in the cytoplasm of interphase cells and strongly colocalized in mitotic cells, particularly during metaphase and anaphase (Figure 4A, subpanels c and d). Of note, both proteins exhibited brighter staining in mitotic cells, indicating that their expression is higher during mitosis when compared to interphase cells.We also performed coimmunoprecipitation experiments by expressing a Flag-tagged full-length CHFR cDNA in HEK293 cells. By immunoprecipitating Flag:CHFR and by immunoblot analysis for MAD2, we found additional evidence supporting an interaction between CHFR and MAD2 (Figure 4B).
In addition, we studied the localization of two critical spindle checkpoint proteins, BUBR1 and MAD2, during mitosis. Normally, both proteins have a punctate staining pattern early in mitosis, reflecting their localization to unattached kinetochores. When MAD2 is recruited to the kinetochore, it undergoes a conformation change that increases its affinity for CDC20 binding . Staining becomes diffuse later in mitosis after all chromosomes are attached to the mitotic spindle and the two proteins dissociate from CDC20 and the kinetochores, thereby allowing CDC20 to activate APC and allow the initiation of anaphase. Normal BUBR1 and MAD2 staining patterns were observed in negative control cells in metaphase (Figure 4, C and D). In contrast, MCF10A:CHFR-siRNA cells demonstrated diffuse BUBR1 and MAD2 staining early in metaphase, indicating that both proteins were not properly localizing to the kinetochores of unattached chromosomes (Figure 4, C and D, right panels).
The spindle checkpoint can be activated by treating cells with microtubule-targeting poisons like nocodazole. As mentioned previously, MAD2 recruited to the kinetochores binds to CDC20 to sequester it from the APC and prevent anaphase onset, especially in response to microtubule-targeting drugs. Due to the mislocalization of MAD2 in MCF10A:CHFR-siRNA cells described previously, we hypothesized that MAD2/CDC20 complex formation would be compromised in CHFR knockdown cells. To test this, we coimmunoprecipitated CDC20 with MAD2 in nocodazole-treated control and MCF10A: CHFR-siRNA cells.We discovered that significantly lowered CHFR expression impaired MAD2/CDC20 complex formation in nocodazoletreated MCF10A cells, as evidenced by the decrease in CDC20 signal in MCF10A:CHFR-siRNA cells compared to controls despite equal amounts of MAD2 being immunoprecipitated (Figure 4E).
Given the evidence that CHFR may participate in the spindle checkpoint by interacting with MAD2 and regulating its function and localization, we performed immunofluorescence to visualize chromosomes during mitosis to determine whether the spindle checkpoint was functioning to properly segregate sister chromatids.Nearly 25% of MCF10A:CHFR-siRNA cells had metaphase chromosomes not properly located to the metaphase plate, which was much higher than the approximately 3% frequency observed in the control cells (Figure 5, A and B). Lagging chromosomes and chromosome bridges were also observed during anaphase in MCF10A:CHFR-siRNA cells (Figure 5C). This suggested that the spindle checkpoint was disrupted in cells with decreased CHFR expression.
One potential outcome of chromosome nondisjunction is impaired cytokinesis, resulting in binucleated cells and potential tetraploidy . As noted previously, nearly 13% of cells with stably decreased CHFR by shRNA were binucleated and tetraploid (Figure 1A and data not shown). In fact, 6% of transiently transfected MCF10A:CHFR-siRNA cells were also binucleated when analyzed immunofluorescence, suggesting possible tetraploidy, compared to only approximately 1% of negative control cells (Figure 5, D and E; P < .05). This was confirmed by the occasional tetraploid MCF10A: CHFR-siRNA metaphase spread when cells were assessed for aneuploidy.
The work presented here indicates that CHFR is extremely important for the maintenance of genomic stability in mammary epithelial cells. Our results support and help explain the previously published findings of aneuploidy in MEFs from Chfr null mice and IHMEC lines [3,14]. The observed chromosome rearrangements that we noted spectral karyotyping likely resulted from prolonged culture, and the disruption of DNA damage response genes secondary to the aneuploidy, which we have shown can develop within a few days after decreased CHFR expression. To the contrary, the presence of additional chromosomes with a numeric change in chromosome number, or aneuploidy, in cells treated with siRNA against CHFR provides powerful evidence that CHFR is required for genomic stability through proper chromosome segregation during mitosis. Furthermore, the lack of chromosome breaks on the metaphase spreads from MCF10A cells transiently transfected with siRNA to decrease CHFR mRNA and protein suggested that CHFR might not participate directly in the DNA damage response induced by aphidicolin. This conclusion is supported by previous studies in which CHFR expression did not alter the DNA damage response after treatment with other genotoxic reagents [1,22].
We were able to confirm the previously published finding that Chfr can regulate Aurora A expression in the mouse also holds true in humans . Aurora A is amplified and overexpressed in many cancers, including breast cancer, and overexpression in cultured human cells leads to transformation [20,23]. In addition, a transgenic mouse overexpressing Aurora A in the mammary epithelium leads to tumor formation and genomic instability . The chromosome missegregation phenotypes in MCF10A:CHFR-siRNA cells were also highly reminiscent of MEFs that overexpress Aurora A . CHFR was recently characterized as a tumor suppressor and, as shown here, many of its genomic instability phenotypes resemble Aurora A overexpression; therefore, we propose that one major mechanism by which CHFR inhibits oncogenesis may be through its negative regulation of Aurora A [3,14]. Novel drugs currently are being generated that target the Aurora kinases . Because decreased CHFR expression has been linked to sensitivity to microtubule-targeting drugs, future studies may find a synergistic effect when taxanes and Aurora kinase inhibitors are both used for treatment.
These findings also indicate that CHFR may play a role in regulating α-tubulin turnover or stability, especially after microtubule stress. This is the first clue as to how the “CHFR checkpoint” responds to microtubule poisons, although an unidentified signaling cascade is also likely to be involved in this checkpoint. The ubiquitination and possible degradation of α-tubulin may be necessary to remove those α/β-tubulin dimers that are targeted by microtubule poisons. In unstressed cells, CHFR may also be required for proper spindle formation because it seems to regulate the amount of acetylated α-tubulin, an important component of the mitotic spindle. Aurora A kinase is also required for proper spindle formation, supposedly through its positive regulation of a protein called HURP . HURP is required for both chromosome congression and alignment and for the polymerization and stabilization of microtubules during mitotic spindle formation. Therefore, the capacity of CHFR to control spindle formation may be through its upstream regulation of Aurora A, although it may also be due to CHFR's capability to ubiquitinate α-tubulin and control the amount of acetylated α-tubulin that is available for use during spindle assembly.
One of the characteristics of stabilized microtubules is the acetylation of α-tubulin on residue lysine 40. Acetylated α-tubulin is associated with decreased microtubule turnover and is localized to the mitotic spindle, centrosomes, and the mitotic midbody [28,29]. An increase in acetylated α-tubulin, such as that observed here, would likely result in overstabilized microtubules that would hinder mitotic spindle movement or would prevent its proper formation. This may help to explain why CHFR-negative cells are more sensitive to taxanes. The cellular strain of the excess of overstabilized acetylated microtubules, combined with stress induced by microtubule poisons, may enable the cell to surpass a threshold of tolerable stress that would result in apoptosis. This hypothesis is supported by reports of a synergistic effect on both apoptotic response and microtubule stabilization, as indicated by acetylated α-tubulin, when endometrial cancer cells are treated with both the histone deacetylase inhibitor (HDI) trichostatin A and paclitaxel . Interestingly, some of the targets of HDIs are also tubulin deacetylase proteins, such as HDAC6 and SIRT2 [31,32]. Recent studies also show that treating cells with HDIs down-regulates Aurora A expression . Future clinical studies may find that the synergistic effect between HDIs and taxanes may be different in CHFR-positive versus CHFR-negative cancer cells.
The finding that CHFR knockdown results in increased amounts of acetylated α-tubulin is particularly interesting because another protein that has been found to initiate a “CHFR checkpoint-like” response to microtubule poisons is SIRT2, a tubulin and histone deacetylase . SIRT2 overexpression is a phenocopy of CHFR overexpression in regards to the regulation of mitotic entry and response to mitotic stress. Therefore, hypothetically, decreased SIRT2 expression should resemble decreased CHFR expression in both the response to mitotic stress and the amount of acetylated α-tubulin in the cell. Future studies should determine whether the increase in acetylated α-tubulin after decreased CHFR expression is due to SIRT2 or through the activation of Aurora A-regulated HURP.
We report here the identification of a novel CHFR interacting protein, MAD2. Although we found no evidence that MAD2 is a ubiquitination target of CHFR (data not shown), we noticed that when CHFR expression was drastically reduced and the cells were treated with nocodazole to induce the checkpoint response, there was a slight decrease in MAD2 protein levels (Figure 4E). This suggests that another ubiquitin ligase is possibly degrading MAD2 prematurely, or to a greater degree, in the absence of CHFR. Because decreased CHFR expression impairs MAD2/CDC20 complex formation, we hypothesize that, in this scenario, CDC20 is prematurely activating the APC complex to initiate anaphase and ubiquitinate target mitotic proteins, such as MAD2. Alternately, MAD2 expression itself may be down-regulated in nocodazole-treated cells with decreased CHFR expression. Further studies are warranted to clarify the mechanism(s) of impaired MAD2 function and expression in cells with decreased CHFR expression.
With an impaired spindle checkpoint, cells with decreased CHFR expression could enter anaphase without all of their chromosomes localized to the metaphase plate and the sister chromatids attached to the mitotic spindle, leading to the appearance of lagging chromosomes and unequal chromosome segregation among the two daughter cells, such as that reported here. One potential outcome of improper chromosome segregation is the abortion of cytokinesis, resulting in binucleated cells and tetraploidy, which was also observed in this work . Of interest, our work strongly agrees with previous findings that the yeast orthologs of CHFR, DMA1 and DMA2, also function in regulating the spindle checkpoint and cytokinesis [35,36].
A recent report indicated that one isoform of CHFR (Accession No. AF_170724; the same isoform was used in this article) contains a KEN box motif, which targets proteins to the CDH1-APC complex for degradation. This further supports a role for CHFR in regulatingmitosis and the spindle checkpoint . Although we found that a KEN box deletion mutant of CHFR did not impair CHFR/MAD2 coimmunoprecipitation (data not shown), further studies are warranted to determine whether the KEN box in CHFR is required for spindle checkpoint function and/or degradation by the CDH1-APC complex.
We also discovered that CHFR overexpression is toxic to many breast cell lines independent of the method of transfection or retroviral transduction (both transient and stable; data not shown), which is why HEK293 cells were used to express the Flag-tagged CHFR construct used in coimmunoprecipitation experiments. This suggests that CHFR expression must be tightly regulated—too much is toxic whereas too little causes genomic instability and tumorigenesis. This is reminiscent of other mitotic checkpoint proteins, such as MAD2, in that both too little and too much of the protein are deleterious . Recently, this finding was also reported in HCT116 and RKO colon cancer cell lines . Determining the mechanism(s) causing CHFR overexpression toxicity likely will answer many of the questions that remain about the function of CHFR.
These findings have led us to propose a model for how CHFR may regulate genomic instability and/or tumorigenesis (Figure 6). We suggest that decreased or lost CHFR expression causes overexpression of Aurora A and both unmodified and acetylated α-tubulin, and the mislocalization of MAD2. Aurora A overexpression could lead to centrosome amplification, an impaired spindle checkpoint, and possibly defective mitotic spindle formation, leading to aneuploidy and impaired cytokinesis. The mislocalization of MAD2 also causes an impaired spindle checkpoint response. The increase in acetylated α-tubulin could cause stress on the mitotic spindle. Both pathways would lead to genomic instability, contributing to tumorigenesis. As indicated by the generality of this model, much research remains to elucidate the role of CHFR in regulating mitosis and genomic instability.
Many reports have indicated that CHFR plays an important role in carcinogenesis and has tumor-suppressive qualities. An abundance of evidence indicates that CHFR mRNA expression is decreased in many cancer types, often due to promoter methylation (reviewed in Privette and Petty ). In addition, a knockout mouse model and cell culture models of lost or decreased expression, respectively, indicate that CHFR has tumor-suppressive qualities [3,14]. Additional evidence for CHFR's role in tumorigenesis is that it is important for cell cycle regulation, chemotherapeutic response to taxanes, and cellular proliferation [3,6,39] (reviewed in Privette and Petty ). Here, we present novel evidence for an additional tumor-suppressive function of CHFR—maintaining genomic stability by regulating the mitotic spindle assembly checkpoint. Cancer often develops in concert with the loss of cell cycle regulation and genomic instability; CHFR may function in both processes.
The authors thank Esther Peterson for helpful suggestions and Sally Camper for sharing her Leica DMRB microscope.
1This work was supported by a Department of Defense Breast Cancer Research Program Fellowship, #BC050310, and by the National Institutes of Health (NIH) National Research Service Award #5-T32-GM07544 from the National Institute of General Medicine Sciences to L.M.P. and an NIH National Cancer Institute grant RO1CA072877 to E.M.P.