Aurora B is regulated by acetylation/deacetylation during mitosis in prostate cancer cells

Abstract

Protein acetylation has been implicated in playing an important role during mitotic progression. Aurora B kinase is known to play a critical role in mitosis. However, whether Aurora B is regulated by acetylation is not known. Using IP with an anti-acetyl lysine antibody, we identified Aurora B as an acetylated protein in PC3 prostate cancer cells. Knockdown of HDAC3 or inhibiting HDAC3 deacetylase activity led to a significant increase (P<0.01 and P<0.05, respectively) in Aurora B acetylation as compared to siLuc or vehicle-treated controls. Increased Aurora B acetylation is correlated with a 30% reduction in Aurora B kinase activity in vitro and resulted in significant defects in Aurora B-dependent mitotic processes, including kinetochore-microtubule attachment and chromosome congression. Furthermore, Aurora B transiently interacts with HDAC3 at the kinetochore-microtubule interface of congressing chromosomes during prometaphase. This window of interaction corresponded with a transient but significant reduction (P=0.02) in Aurora B acetylation during early mitosis. Together, these results indicate that Aurora B is more active in its deacetylated state and further suggest a new mechanism by which dynamic acetylation/deacetylation acts as a rheostat to fine-tune Aurora B activity during mitotic progression.—Fadri-Moskwik, M., Weiderhold, K. N., Deeraksa, A., Chuang, C., Pan, J., Lin, S.-H., Yu-Lee, L.-Y. Aurora B is regulated by acetylation/deacetylation during mitosis in prostate cancer cells.

Protein acetylation on lysine residues is a major post-translational modification governing a wide range of cellular processes (1, 2). Acetylation is mediated by histone acetyltransferases (HATs) and is dynamically and reversibly opposed by the action of histone deacetylases (HDACs) (3). HDACs play a role in epigenetic modification of chromatin structure in response to environmental changes (3–5) and are often elevated in various human cancers (6). Thus, HDACs have become attractive targets for anticancer therapeutic intervention.

We recently discovered that one of the HDACs, HDAC3, is localized on the mitotic spindle (7). Knockdown of HDAC3 by siRNA resulted in a collapsed spindle surrounded by a dome-like configuration of chromosomes (7). The collapsed spindle phenotype was exacerbated by incubation in cold temperatures, suggesting problems in kinetochore-microtubule attachment. Wild-type HDAC3, but not a deacetylase dead mutant HDAC3, rescued the collapsed spindle phenotype, which suggests that HDAC3 deacetylase activity is involved in spindle stability. Further, treatment of cells with apicidin, an HDAC inhibitor that shows selectivity toward HDAC3, resulted in aberrant mitosis and cytokinesis (8). These findings suggest that HDAC3 has unique functions in mitosis. The targets of HDAC3 in mitosis are unknown.

Mitosis is driven by post-translational modification of proteins, as transcription is silent and RNA translation is inhibited (9). Post-translational modification of proteins involving phosphorylation (10–13), methylation (1), ubiquitination (14), and small ubiquitin-like modifier (SUMO)ylation (15, 16) are well known in mitosis. However, protein modification by acetylation in mitosis is largely unexplored. To identify acetylated proteins in mitosis, we synchronized HeLa cells in mitosis and used an anti-acetyl-lysine antibody for immunoprecipitation (IP), followed by protein identification by mass spectrometry (8). We identified 51 unique nonhistone proteins involved in processes such as cell cycle regulation, chaperone function, and DNA damage repair. These findings support the hypothesis that acetylation is a prevalent event in mitosis.

Aurora B plays a critical role in mitotic progression. Aurora B (41 kDa) is the catalytic component of the chromosomal passenger complex (CPC) that includes the inner centromere protein INCENP, survivin and Borealin (17, 18). The CPC is involved in targeting Aurora B to various subcellular locations (17, 19–23) and modulates Aurora B kinase activity toward various substrates (17–19). Aurora B kinase activity is regulated by its association with the scaffold protein INCENP (18–20) and by post-translational modification of components of the CPC (17, 18, 22, 24, 25). Aurora B kinase activity is affected by phosphorylation (26), while ubiquitination (22, 27, 28) and SUMOylation (16, 29) have been shown to play a role in Aurora B/CPC localization but not in Aurora B kinase activity (16). Whether Aurora B is modified by acetylation is not known. In this study, we investigated Aurora B acetylation/deacetylation as a potential new mechanism that regulates Aurora B activity and function in mitosis.

MATERIALS AND METHODS

Antibodies and inhibitors

The following antibodies were used: acetyl lysine (rabbit, 1:2000; Millipore, Billerica, MA, USA), acetyl lysine (mouse, 1:1000; Millipore), Aurora B (mouse, 1:800; BD, Franklin Lakes, NJ, USA), Aurora B (rabbit, 1:2000; Abcam, Cambridge, MA, USA), HDAC3 (rabbit, 1:1500; GeneTex, Irvine, CA, USA), histone H3 (rabbit, 1:2000; Abcam), phospho-H3 (rabbit, 1:1000; Millipore), cyclin B1 (mouse, 1:800; BD), α-tubulin (rabbit, 1:2000; GeneTex), α-tubulin (mouse, 1:1000; Sigma-Aldrich, St. Louis, MO, USA), β-tubulin (tub2.1; mouse, 1:1000; Sigma), FLAG (mouse; Sigma), and CREST serum (human, 1:10,000 for immunofluorescence; Bill Brinkley, Baylor College of Medicine). Protease inhibitors, HDAC inhibitors suberoylanilide hydroxamic acid (SAHA) and apicidin, and nocodazole were purchased from Sigma-Aldrich. Monastrol, RO-3306, and MG132 were purchased from Calbiochem (EMD Biosciences, Gibbstown, NJ, USA).

Cell culture, synchronization, and HDAC3 knockdown

Human prostate cancer 3 (PC3)-mm2 prostate cancer cells were cultured in minimal essential medium (Gibco; Invitrogen, Carlsbad, CA, USA) containing 10% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA, USA) and penicillin/streptomycin (Gibco) in 5% CO₂. Cells were synchronized by a 2 mM thymidine block, released into fresh medium for 10 h, and treated with 100 ng/ml nocodazole for 16 h for early mitotic cells. To inhibit HDAC3 activity, cells were incubated with 100 nM apicidin for the last 3 h in the synchronization process (8, 30). For protein knockdown, cells were transfected with small interference oligo against luciferase (siLuc) control or small interference oligo against HDAC3 (siHDAC3) using Oligofectamine (Invitrogen) for 48–72 h, as described previously (7).

IP and immunoblotting

Cells were lysed in RIPA buffer (100 mM NaCl; 20 mM Tris, pH 8.0; 1.5 mM EDTA; 5 mM EGTA; 0.1% Triton X-100; and 5% glycerol) containing protease (mammalian protease inhibitor and 0.1 M PMSF) and phosphatase inhibitors (5 mM Na3VO₄, 5 mM NaF, serine/threonine phosphatase, and tyrosine phosphatase inhibitors). SAHA (5 μM) was added in the lysis buffer and in all subsequent steps. Lysates were subjected to IP (1 μl antibody/mg lysate), SDS-PAGE, and Western blot as described previously(8, 30).

Microscopy and fluorescence intensity measurements

Cells were plated in 12-well dishes overnight, washed in 2× PHEM (60 mM K-PIPES; 25 mM HEPES, pH 6.9; 10 mM EGTA; and 4 mM MgSO₄) and fixed with 4% paraformaldehyde for 20 min at room temperature. Cells were extracted with 0.5% Triton X-100 in phosphate-buffered saline (PBS) for 15 min, blocked in 5% bovine serum albumin in PBS, and incubated with primary antibodies overnight at 4°C, followed by 1 h with secondary antibodies (Invitrogen) at room temperature. Coverslips were mounted (ProLong Gold with antifade and DAPI; Invitrogen), and images were acquired using a Nikon TE2000 widefield microscope system (Nikon, Lewisville, TX, USA) (7, 31, 32). Confocal images were acquired as z stacks with a step size of 0.4 μm using a ×60 oil/1.40-NA objective.

For fluorescence intensity measurements of phospho-S10 histone H3 on chromatin, cells were stained for pS10H3 (1:1000, rabbit), Aurora B (1:500, mouse), and DNA (DAPI). A region of interest (ROI) encircling the entire DNA was outlined, using the NIS-Elements AR 3.0 software (Nikon) ROI automatic detection function, and corrected for background noise. The average intensity values were determined for pS10H3 and Aurora B. The pS10H3 signals were then normalized against the Aurora B signals and expressed as a ratio.

Live-cell imaging

PC3 cells were plated on a sterile 35-mm glass-bottomed dish (VWR, Radnor, PA, USA), grown for 2 d, stained with 200 nM Hoechst dye (Invitrogen) for 1 h, and released into fresh medium containing DMSO or 100 nM apicidin. Multiple image capture points were selected on a Nikon BioStation IM (Nikon). Images were acquired every 3–4 min for ≥3 h using phase-contrast and UV filters and ×20 magnification, compiled using NIS-Elements 3.1 software (Nikon), and presented using Adobe PhotoShop CS (Adobe Systems, San Jose, CA, USA) as described previously (32).

Aurora B IP kinase assay

PC3 cells (1×10⁶) were transfected with either siLuc or siHDAC3 oligos for 24 h, and then cotransfected with FLAG-Aurora B and its activator myc-INCENP for another 24 h. Immunoprecipitated FLAG-Aurora B was incubated with 1 μg recombinant histone H3 (Sigma) in the presence of 5 μCi [³²P]-γ-ATP (25 Ci/mmol; MP Biomedicals, Solon, OH, USA) in vitro for 30 min at 30°C, as described previously (33). Cells were also cotransfected with either EGFP-Aurora B (34) or FLAG-Aurora B (33), myc-INCENP, and FLAG-HDAC3 (7). Immunoprecipitated EGFP-Aurora B or FLAG-Aurora B was used in IP kinase assays. The reactions were resolved on SDS-PAGE, transferred to a filter, analyzed by autoradiography, and then immunoblotted for Aurora B and stained with Ponceau S to determine equal H3 input. The phospho-H3 signals were divided by total H3 and normalized against Aurora B levels by densitometric analysis.

Statistical analysis

Data were confirmed in multiple independent experiments, quantified by Student's t test, and expressed as means ± se or sd. Values of P < 0.05 were considered statistically significant.

RESULTS

Knockdown of HDAC3 and inhibition of HDAC3 activity result in mitotic defects in prostate cancer cells

We found previously that HDAC3 is localized on the mitotic spindle (7) and regulates mitotic progression in HeLa cells (8). Here, we examined whether this unusual localization of HDAC3 also occurs in PC3 cells. In prometaphase, metaphase, and anaphase, HDAC3 is localized on the mitotic spindle, with a higher concentration found near the spindle poles (Fig. 1A). In telophase, HDAC3 becomes more concentrated on the outer edge of the spindle midzone adjacent to the segregating chromosomes. As the nuclear envelope reforms during late mitosis and cytokinesis, HDAC3 reenters the nucleus, where it is known to regulate gene transcription. These results show that HDAC3 localization is dynamic in mitotic prostate cancer cells.

Figure 1.

HDAC3 knockdown results in mitotic defects in PC3 prostate cancer cells. A) Unperturbed PC3 cells undergoing mitosis were stained for tubulin (green), HDAC3 (red), and DNA (blue). Bar, 5 μm. B) Cells were transfected with siLuc control or siHDAC3 oligos for 72 h. Lysates were immunblotted for HDAC3, followed by β-tubulin as a loading control. Percentage distribution (mean±se) of cells in different phases of mitosis and cytokinesis was analyzed. C) Unperturbed PC3 cells were incubated with Hoechst dye to stain the DNA for 1 h and released into fresh medium containing either DMSO or 100 nM apicidin. Images were acquired every 3–4 min for 3 h on a Nikon BioStation and are shown as either black and white images for contrast or phase-contrast images where DNA is pseudocolored in green. Time at NEB was set as 0 h:00 min. Arrowheads indicate uncongressed chromosomes; enlarged in insets. Progression through early mitosis in the apicidin-treated cell was delayed relative to the control cell. Time from NEB to metaphase-to-anaphase transition was followed for 40–75 individual cells from 2 independent experiments (means±sd; left panel). An average transit time (means±se) was determined for cells that have completed metaphase-to-anaphase transition (right panel). n = cells followed live. n.d., not detected. P values are as shown. D) Cells were treated with siRNA oligos, synchronized in mitosis with 100 ng/ml nocodazole, released into fresh medium containing 20 μM MG132 as indicated, and stained for DNA (blue), tubulin (green), and a centromere marker CREST (red). Arrowheads indicate uncongressed chromosomes. Percentage distribution (mean±se) of cells containing either 0–4 or >4 uncongressed chromosomes was determined from 3 independent experiments. n = cells counted. Scale bars = 5 μm (A, D); 10 μm (C). *P < 0.01.

Next, we investigated the role of HDAC3 in mitotic PC3 cells by analyzing the effects of HDAC3 knockdown. HDAC3 levels were significantly reduced in siHDAC3-treated cells relative to siLuc-treated control PC3 cells (Fig. 1B). Under this condition, a significant increase was found in prometaphase cells in the HDAC3-deficient cells relative to control cells, which suggests a delay in mitotic progression.

We further analyzed early mitotic progression by treating cells for a short time with a small-molecule HDAC inhibitor, apicidin, that shows selectivity for HDAC3 (8, 35). PC3 cells were released into 100 nM apicidin and the time from nuclear envelope breakdown (NEB) to metaphase-to-anaphase transition at the single-cell level was observed by time-lapse live-cell imaging. In individual control cells (n=75), 64% of cells reached the metaphase-to-anaphase transition between 60–90 min after NEB, with an average transit time of 70 min (Fig. 1C). In contrast, in apicidin-treated cells (n=40) only 38% of the cells progressed from NEB to metaphase-to-anaphase transition during the same time period. Over 60% of apicidin-treated cells took >90 min to progress from NEB to metaphase-to-anaphase transition (Fig. 1C). Using only those apicidin-treated cells that exhibited clear metaphase-to-anaphase transition (n=28), the average transit time was found to be 90 min, a significant 20-min delay, when compared to control cells (Fig. 1C). These results suggest that inhibition of HDAC3 deacetylase activity delayed early mitotic progression. Given that during mitosis, transcription is silent (7, 8, 36) and translation is largely inhibited (8), the changes observed in mitotic progression after apicidin treatment are consistent with HDAC3 exerting an effect on post-translational protein modification and not gene transcription as previously reported (7, 8, 36). Taken together, these results support that HDAC3 levels (Fig. 1B) as well as HDAC3 deacetylase activity (Fig. 1C) are required for proper progression through early mitosis.

During prometaphase, chromosomes are captured by the mitotic spindle to align at the metaphase plate. Problems in this process could result in chromosome congression errors that can be examined during metaphase. We treated siHDAC3-transfected PC3 cells for 16 h with nocodazole, a microtubule depolymerizer that arrests cells in a prometaphase-like state, to enrich for mitotic cells (Fig. 1D), then released the cells for 3 h into MG132, a proteasome inhibitor, to prevent metaphase-to-anaphase transition (31, 32). We found that 83 ± 0.4% of control cells contained a tight band of chromosomes aligned at the metaphase plate. About 17 ± 0.4% of control cells contained >4 uncongressed chromosomes at the metaphase plate that may be related to the highly metastatic nature of the PC3 cells (37). In HDAC3-deficient cells, the number of cells containing >4 uncongressed chromosomes (Fig. 1D, arrowheads) increased to ∼50 ± 0.007%, a mitotic error also observed during live-imaging of apicidin-treated PC3 cells (Fig. 1C, arrowheads and insets). Many chromosomes were spread out along the mitotic spindle and even adjacent to the spindle poles in HDAC3-deficient cells. Furthermore, the mitotic spindle was often smaller, which is consistent with the collapsed spindle phenotype previously observed in HDAC3-deficient HeLa cells (7, 8). These studies suggest that HDAC3 plays a role in chromosome alignment at the metaphase plate during mitosis in prostate cancer cells.

HDAC3 depletion compromises kinetochore attachment to mitotic spindles

Errors in chromosome alignment could result from problems in kinetochore-microtubule attachment (38–40). In metaphase, properly attached kinetochores will be oriented in the direction of the long axis of the mitotic spindle, with an average of <10° angle between a kinetochore pair and the spindle long axis (41). Confocal microscopy analysis showed that in control cells, >90% of the kinetochore pairs, shown as CREST-positive double foci, exhibited a shallow 0–15° angle relative to the spindle long axis, with an average spindle angle of 7.3 ± 0.55° (Fig. 2A, inset). In contrast, >50% of the kinetochore pairs in HDAC3-deficient cells were oriented at a steeper angle, from 16 to >30° relative to the spindle long axis, with an average spindle angle of 22.3 ± 1.80°. These results suggest that HDAC3 depletion resulted in kinetochore misorientation at the metaphase plate.

Figure 2.

HDAC3 knockdown perturbs kinetochore attachment to spindle microtubules. A) PC3 cells were transfected with siLuc or siHDAC3 oligos for 72 h, stained for β-tubulin (green), CREST (red), and DNA (not shown), and analyzed by confocal microscopy. Kinetochore orientation was measured as the angle (α) between a kinetochore pair and the long axis of the mitotic spindle that is set as 0° angle (41). Only kinetochore pairs in the same focal plane were analyzed. Distribution of kinetochore pair angles from 0–15, 16–30, and >30° angle relative to the spindle long axis is shown. Average ± se angles (α)of 130 kinetochore pairs examined from 5 cells/condition are also shown. *P < 0.001. B, C) PC3 cells were transfected with siLuc or siHDAC3 oligos for 60 h, treated with either nocodazole for 16 h (B) or 100 μM monastrol for 16 h (C), released into MG132, and incubated on ice for 10 min as indicated. Cells were stained for β-tubulin (green), CREST (red), and DNA (blue). Percentage mitotic spindles (mean±se) exhibiting stable and bipolar vs. unstable and disorganized morphology after cold treatment was analyzed. n = number of spindles counted in 3 independent experiments. Scale bars = 5 μm. *P < 0.05, **P < 0.02.

Kinetochore misorientation at the metaphase plate may be due to a defect in kinetochore formation or problems in kinetochore-microtubule attachments. Because kinetochore formation is not affected by HDAC3 depletion (7), we analyzed microtubule attachments to kinetochores in HDAC3-deficient cells. As one approach, we tested the stability of the mitotic spindle in response to a brief treatment with cold temperature (7). Spindle microtubules that are properly attached to kinetochores are resistant to cold treatment, while those microtubules not attached to kinetochores are selectively destroyed by a brief incubation on ice (42). Cells were synchronized using nocodazole, followed by release into MG132 to enrich for cells in metaphase. Although cold treatment moderately affected spindle morphology, the overall structure of the kinetochore-attached bipolar spindle remained robust in control cells (Fig. 2B). In contrast, cold treatment resulted in a highly disorganized and unstable spindle in HDAC3-deficient cells, suggesting that spindle microtubules are not properly attached to kinetochores after HDAC3 depletion. We further employed a different cell synchrony protocol to validate this observation. HDAC3-knockdown cells were synchronized with monastrol, a small-molecule inhibitor of the Eg5 kinesin motor (kinesin 5; ref. 43) that arrests cells in prometaphase with monopolar spindles. The monopolar spindles achieve bipolarity when cells are released from the monastrol blockade (31). Under this condition where spindle microtubules remained intact, we also observed a significant increase in unstable mitotic spindles that tended to collapse onto the spindle poles after cold treatment in HDAC3-deficient cells (Fig. 2C). We also note that the chromosome miscongression phenotype was exacerbated after monastrol treatment in HDAC3-deficient cells, in agreement with the congression errors observed in nocodazole-treated HDAC3-deficient cells (Figs. 1D and 2B). Taken together, HDAC3 depletion compromises kinetochore-microtubule attachment and spindle stability in prostate cancer cells.

Aurora B is acetylated during mitosis

A key kinase responsible for correcting kinetochore-microtubule misattachments is Aurora B (39, 40, 44). Interfering with Aurora B functions has been shown to lead to incorrect kinetochore-microtubule attachments and chromosome misalignment at the metaphase plate (38, 45). The mitotic phenotypes observed in HDAC3-deficient cells (Figs. 1 and 2) overlap with phenotypes observed in Aurora B dysfunctional cells, raising the possibility that a key target affected by HDAC3 depletion may be Aurora B. To examine whether Aurora B is acetylated in mitosis, we first transfected FLAG-Aurora B into PC3 cells, synchronized the cells in mitosis, and harvested cells by a “mitotic shake-off” (32). Cell lysates were immunoprecipitated with anti-FLAG antibody, followed by immunoblotting with anti-acetyl-lysine antibody and reblotting for Aurora B. We found that FLAG-Aurora B was acetylated in mitosis (Fig. 3A, left panel). Using a similar mitotic shake-off approach, we next examined endogenous Aurora B and found that endogenous Aurora B was acetylated in mitotic PC3 cells (Fig. 3A, right panel). We further confirmed Aurora B acetylation by a reciprocal IP, where mitotic cell lysates were immunoprecipitated with anti-acetyl-lysine antibody, followed by blotting for Aurora B (Fig. 3B). These results show that Aurora B is acetylated in mitotic PC3 cells.

Figure 3.

Aurora B is acetylated in mitotic prostate cancer cells. A) IP of Aurora B, followed by anti-acetyl-lysine (AcK) blot. Left panel: PC3 cells were transfected with FLAG-Aurora B, treated for 16 h with 100 ng/ml nocodazole, and harvested by a mitotic shake-off (32). Lysates were immunoprecipitated with anti-FLAG or IgG control antibodies, immunoblotted with an anti-AcK antibody, and reblotted for Aurora B. Right panel: mitotic PC3 cells were prepared by a thymidine block, release, and nocodazole treatment, as described in Materials and Methods, and lysed under reducing conditions with dithiothreitol in RIPA buffer, as described previously (8, 30). Cell lysates were diluted 10-fold before IP for Aurora B. The blot was probed as above. B) IP AcK followed by Aurora B blot. Mitotic cells were prepared as in A. Lysates (0.26 mg) were immunoprecipitated with anti-AcK or anti-IgG antibodies and immunoblotted for Aurora B. HC, heavy chain; LC, light chain. C) PC3 cells were transfected with siLuc or siHDAC3 oligos for 48 h. Lysates (0.5 mg) were immunoprecipitated and analyzed as in B. Acetylated Aurora B as a fraction ± se of total Aurora B was calculated from 3 independent experiments. *P < 0.01. D) Mitotic cells were prepared as in B, except that either DMSO or 100 nM apicidin was added for the last 3 h. Lysates (0.8 mg) were analyzed as in C. Acetylated Aurora B/total Aurora B (mean± se) was calculated from 2 independent experiments. **P < 0.05.

Aurora B acetylation is regulated by HDAC3

To determine whether Aurora B acetylation may be regulated by HDAC3, we transfected PC3 cells with siRNA oligos against HDAC3 for 48 h. In HDAC3-deficient cells, Aurora B acetylation was significantly increased relative to siLuc-transfected control cells (Fig. 3C), suggesting that the acetylation status of Aurora B is regulated by HDAC3 levels. Total Aurora B levels remained essentially unchanged, while HDAC3 levels were reduced in the HDAC3-knockdown cells. To specifically examine HDAC3 activity on Aurora B in mitosis, we treated mitotic cells with 100 nM apicidin, which inhibits HDAC3 deacetylase activity (8, 35). The presence of apicidin significantly enhanced Aurora B acetylation (Fig. 3D), suggesting that acetylation of Aurora B is regulated by HDAC3 activity in mitotic cells. Total Aurora B and HDAC3 levels remained unchanged with apicidin treatment. Taken together, these studies suggest that acetylation of Aurora B is modulated by HDAC3 in PC3 cells.

Aurora B associates with HDAC3 in mitotic cancer cells

We next asked whether Aurora B could interact with HDAC3 in mitotic cells. PC3 cells were synchronized in mitosis, and Aurora B or HDAC3 was immunoprecipitated from mitotic cells. We detected HDAC3 in the Aurora B immunoprecipitated complex but not in the IgG control (Fig. 4A, top panel). Conversely, we detected Aurora B in the HDAC3 immunoprecipitated complex in a reciprocal IP, but not in the IgG control (Fig. 4A, bottom panel). These results indicate that endogenous Aurora B associates with HDAC3 in mitotic PC3 cells.

Figure 4.

Aurora B associates with HDAC3, and its kinase activity is regulated by HDAC3. A) PC3 cells were synchronized in mitosis as in Fig. 3B. Top panel: lysates (1 mg) were immunoprecipitated with anti-Aurora B or anti-IgG antibodies, followed by immunoblotting for HDAC3 and reblotting for Aurora B. A lower band was sometimes observed in the Aurora B input lane (see also right panel). Bottom panel: reciprocal IP was performed, where HDAC3 was immunoprecipitated, followed by immunoblotting for Aurora B and reblotting for HDAC3. Similar data were obtained in 3 independent experiments. B) IP kinase assay. PC3 cells were transfected with either siLuc or siHDAC3 oligos for 24 h and then cotransfected with FLAG-Aurora B, as described in Materials and Methods. Aurora B kinase activity (mean±se) on histone H3 in vitro was determined as ³²P-H3 signals (autorad)/total H3 (Ponceau) and normalized against immunoprecipitated Aurora B (blot) from 2 independent experiments. *P = 0.029. C) Cells were cotransfected with EGFP-Aurora B and FLAG-HDAC3 for 36 h. Lysates (1 mg) were immunoprecipitated with anti-AcK antibodies and immunoblotted for EGFP-Aurora B, FLAG-HDAC3, or tubulin. Acetylated EGFP-Aurora B/total EGFP-Aurora B (mean±se) was calculated from 2 independent experiments. *P < 0.05. D) Cells were transfected with EGFP-Aurora B for 36 h and treated with either DMSO or 2 μM ZM447439, an Aurora B kinase inhibitor, for 5 h before EGFP-Aurora B was immunoprecipitated for an IP kinase assay as in B. E) FLAG-Aurora B cotransfected with FLAG-HDAC3 was employed in an IP kinase assay. Aurora B kinase activity (mean±se) on histone H3 was determined as in B from 2 independent experiments. *P = 0.003. F, G) PC3 cells were transfected with siLuc or siHDAC3 oligos for 48 h and then treated with either nocodazole for 16 h (F) or 9 μM RO-3306 for 16 h followed by MG132, as indicated (G). Cells were stained for pS10H3 (red), Aurora B (green), and DNA (blue). Scale bars = 5 μm. n = number of cells counted in 2 independent experiments. pS10H3 staining over chromosomes was quantified as the ratio (mean±se) of pS10H3/Aurora B signals, as described in Materials and Methods. *P < 0.0001, **P < 0.001.

Aurora B kinase activity is modulated by HDAC3

We then examined whether Aurora B interaction with HDAC3 affects Aurora B kinase activity in vitro. PC3 cells were transfected with siLuc or siHDAC3 oligos and then cotransfected with FLAG-Aurora B. Aurora B was immunoprecipitated with an anti-FLAG-antibody and its kinase activity toward the substrate histone H3 was assessed by in vitro kinase assays (33). HDAC3 knockdown resulted in a significant 34 ± 0.05% decrease in histone H3 phosphorylation using Aurora B immunoprecipitated from HDAC3-deficient cells relative to control cells (Fig. 4B). These results suggest that HDAC3 depletion reduces Aurora B kinase activity in vitro.

To further examine HDAC3 effects on Aurora B kinase activity in vitro, we employed a gain-of-function approach by cotransfecting Aurora B with FLAG-HDAC3 (7) to determine whether overexpression of HDAC3 would reduce Aurora B acetylation and enhance Aurora B kinase activity. Overexpression of HDAC3 resulted in a significant 35 ± 0.08% decrease in Aurora B acetylation levels (Fig. 4C). The decrease in Aurora B acetylation is accompanied by a significant ∼20% increase in Aurora B kinase activity as determined by IP kinase assays in vitro (Fig. 4E). Immunoblotting with anti-FLAG antibody confirmed that equal levels of Aurora B were immunoprecipitated and the presence of HDAC3 only in the FLAG-HDAC3 transfected sample. As a control, Aurora B kinase activity toward H3 in vitro was inhibited by 2 μM ZM447439, a small-molecule inhibitor of Aurora B (ref. 46 and Fig. 4D), confirming the specificity of Aurora B in the IP kinase assays in (Fig. 4B, E). These results suggest that HDAC3 overexpression reduces Aurora B acetylation in vivo and enhances Aurora B kinase activity in vitro.

We next examined the effect of HDAC3 on Aurora B kinase activity in vivo. PC3 cells were transfected with either siHDAC3 or siLuc oligos for 64 h, treated with nocodazole for the last 16 h, and stained for Aurora B and its chromatin substrate histone H3 (phospho-S10H3). We observed a significant decrease in pS10H3 levels in HDAC3-deficient cells relative to control cells (Fig. 4F). We further employed a different cell synchrony protocol to validate this observation. Unlike the nocodazole treatment, which depolymerizes microtubules, HDAC3-knockdown cells were synchronized with the Cdk1 inhibitor RO-3306 at the G₂/M transition (47) and then released into early mitosis in the presence of MG132 to prevent progression beyond metaphase. Under this condition where spindle microtubules remained intact, we also observed a significant decrease in pS10H3 levels after HDAC3 knockdown (Fig. 4G). These results suggest that HDAC3 depletion reduces Aurora B kinase activity toward its substrate histone H3 in vivo.

Aurora B and HDAC3 colocalize in prometaphase

Because Aurora B and HDAC3 can be found in the same biochemical complex (Fig. 4A), we asked at which stage in mitosis do Aurora B and HDAC3 interact. Unperturbed mitotic PC3 cells were stained for Aurora B and HDAC3. As expected, Aurora B localized to the centromeres, which appeared as foci on the chromosomes in prometaphase and metaphase cells, and as part of the chromosomal passenger complex in the midzone in anaphase and the midbody in telophase cells (Fig. 5A). Interestingly, Aurora B and HDAC3 colocalized primarily at the kinetochore-microtubule interface in prometaphase cells, in particular, on uncongressed chromosomes that had not aligned at the metaphase plate (Fig. 5A, yellow in insets and cartoon). As Aurora B is known to correct improper kinetochore-microtubule attachments during prometaphase and metaphase (38, 39), the colocalization of Aurora B with HDAC3 suggests that HDAC3 may regulate Aurora B function at the kinetochore-microtubule interface in early mitosis.

Figure 5.

Aurora B colocalizes with HDAC3 at the kinetochore-microtubule interface in prometaphase. A) Unperturbed growing PC3 cells were stained for HDAC3 (red), Aurora B (green), and DNA (blue). Aurora B and HDAC3 costained (yellow) at the kinetochore-microtubule interface, in particular, on uncongressed chromosomes (insets and cartoon) in prometaphase cells. B) To enrich for prometaphase-like cells, PC3 cells were treated with nocodazole for 16 h to depolymerize microtubules. Cells were stained as in A. C) Confocal images were taken of cells prepared as in B, with DNA omitted for clarity. A single optical section is shown for 3 representative cells. Insets: Aurora B and HDAC3 costaining (yellow) at the kinetochore-microtubule interface. Scale bars = 5 μm.

To further examine Aurora B and HDAC3 colocalization in prometaphase, we enriched for prometaphase-like cells with a 16 h nocodazole treatment. Under this condition, HDAC3 was found on the depolymerized spindle remnants (Fig. 5B) as described previously (7). Many chromosomes located adjacent to the spindle remnants showed colocalization of Aurora B and HDAC3 (Fig. 5B, yellow in inset). We further examined this interaction using confocal microscopy. Three representative images were chosen to show that Aurora B and HDAC3 colocalized at the kinetochore-microtubule interface (Fig. 5C, yellow in insets). Taken together, Aurora B and HDAC3 are transiently colocalized at the kinetochore-microtubule interface during prometaphase, raising the possibility that HDAC3 modification of Aurora B may be a dynamic process to modulate Aurora B function at the kinetochore-microtubule interface.

Aurora B undergoes dynamic acetylation/deacetylation in mitotic cells

We next examined the time course of Aurora B acetylation during this early phase of mitosis. PC3 cells were synchronized in prometaphase as indicated, released into fresh medium to allow cells to re-enter mitosis, and analyzed every 30 min for 90 min (Fig. 6A). Aurora B was found to be acetylated in prometaphase-like cells (Fig. 6A, time 0). Interestingly, the levels of acetylated Aurora B decreased between 30–60 min after nocodazole release, even as total Aurora B levels increased in the same time interval. Acetylated Aurora B levels increased again at 90 min after nocodazole release (Fig. 6B), at a time corresponding to metaphase-to-anaphase transition as evidenced by the decline in total Aurora B (Fig. 6A) and cyclin B1 (Fig. 6C) levels. We note that the timing of metaphase-to-anaphase transition following nocodazole release (90 min) is a little longer than the timing of metaphase-to-anaphase transition observed in unperturbed mitotic cells (70 min) (Fig. 1C). This difference is likely due to the time required for microtubules to regrow and establish a bipolar mitotic spindle after nocodazole washout (48). This experiment shows that Aurora B acetylation is highly dynamic during progression from prometaphase to metaphase in early mitosis.

Figure 6.

Aurora B undergoes acetylation/deacetylation in mitosis. A) PC3 cells were synchronized in early mitosis as indicated and harvested at 0, 30, 60, and 90 min after nocodazole release. Lysates (0.2 mg) were immunoprecipitated with anti-AcK or IgG antibodies and immunoblotted for Aurora B. B) Quantification of acetylated Aurora B as a fraction ± se of total Aurora B from 4 independent experiments. *P = 0.02. C) Quantification of cyclin B1 levels relative to tubulin loading control (mean±se) normalized to time 0 for the experiments in B. *P = 0.004. D) Model. Aurora B undergoes dynamic acetylation/deacetylation that acts as a rheostat to regulate Aurora B kinase activity during mitosis. In this model, Aurora B is acetylated by an as-yet unidentified HAT, and its deacetylation is regulated by HDAC3. The deacetylated form of Aurora B exhibits higher kinase activity. In prometaphase, Aurora B and HDAC3 transiently colocalize at the kinetochore-microtubule interface on congressing chromosomes. Localized deacetylation of Aurora B by HDAC3 maintains a high level of kinase activity for correcting aberrant kinetochore-microtubule attachments. At metaphase, Aurora B and HDAC3 do not colocalize at aligned chromosomes, possibly to prevent destabilizing correct kinetochore-microtubule attachments. Arrow indicates Aurora B (gray) at centromere; arrowheads indicate HDAC3 (black) on mitotic spindle; loops represent sister chromosomes.

DISCUSSION

Our studies demonstrate for the first time that the mitotic kinase Aurora B is modified by acetylation during mitosis, and that Aurora B acetylation plays a role in early mitosis. We show that the level of acetylated Aurora B is regulated by the histone deacetylase HDAC3, and that Aurora B is more active in its deacetylated state. Furthermore, Aurora B undergoes dynamic acetylation/deacetylation during mitosis. Our studies thus provide a new mechanism for Aurora B regulation in mitosis.

Aurora B acetylation/deacetylation in early mitosis

Our studies suggest that Aurora B modification by acetylation is another regulatory mechanism in early mitosis. Unlike SUMOylation (16) and ubiquitination (27, 28), which regulate Aurora B subcellular localization, acetylation of Aurora B appears to regulate Aurora B kinase activity in early mitosis. We found that Aurora B associates with HDAC3 and transiently colocalizes with HDAC3 at the kinetochore-microtubule interface on congressing chromosomes during prometaphase (Fig. 5A), a period corresponding with robust Aurora B kinase activity. It has been shown that Aurora B is highly enriched at misaligned chromosomes (49) and incorrectly attached chromosomes (44) to regulate the resolution of improper kinetochore-microtubule attachments. Interestingly, HDAC3 colocalization with Aurora B at this stage likely deacetylates Aurora B locally to maintain a high level of kinase activity for error correction at the kinetochore-microtubule interface. Consistent with this notion, in HDAC3-deficient cells Aurora B kinase activity is reduced (Fig. 4B) and Aurora B-dependent processes are compromised (Figs. 1–3), including kinetochore-microtubule attachment, mitotic spindle stability during cold treatment, and chromosome congression at the metaphase plate.

In metaphase cells, we note that HDAC3 staining is diminished on the mitotic spindle closest to the metaphase plate (Fig. 1A), such that HDAC3 is no longer colocalized with Aurora B at the centromere (Fig. 5A). The lack of HDAC3 and Aurora B colocalization at metaphase is consistent with the observed increase in Aurora B acetylation at 90 min postnocodazole release (Fig. 6), a time that corresponds with metaphase-to-anaphase transition. Recent studies have indicated that Aurora B levels at centromeres rapidly decrease once chromosomes are aligned at the metaphase plate to avoid destabilizing the correct kinetochore-microtubule attachments (40, 49, 50). It is possible that the lack of HDAC3 at the metaphase plate would permit Aurora B acetylation that helps to reduce Aurora B kinase activity, in addition to the decrease in Aurora B levels at centromeres of aligned chromosomes (40, 49, 50). Thus, acetylation/deacetylation of; Aurora B may act as a rheostat to fine-tune local Aurora B kinase activity in early mitosis (Fig. 6D). The 20–30% reduction in Aurora B kinase activity observed with increased Aurora B acetylation as well as the modest ∼20% increase in Aurora B kinase activity associated with decreased Aurora B acetylation may be a reflection of such fine-tuning of Aurora B kinase activity by acetylation/deacetylation. How dynamic acetylation/deacetylation crosstalks with phosphorylation of Aurora B to regulate the phosphorylation gradient of kinase activity centered at the centromere (40, 50, 51) during prometaphase to metaphase progression merits further analysis.

HDACs in mitosis

HDAC3 is a well-known chromatin protein and a component of the N-CoR/SMRT corepressor complex that regulates gene transcription (1, 52). However, HDAC3 effects in mitosis are independent of gene transcription (7, 36). In mitosis, the entire HDAC3 corepressor complex is relocated from the chromatin to the mitotic spindle to regulate two functions, spindle stability and kinetochore-microtubule attachments (7). In this study, we identified that at least one of the HDAC3 targets is Aurora B, which can explain defects in spindle stability and kinetochore-microtubule attachments in HDAC3-deficient cells (Fig. 2 and ref. 7). The reduced kinase activity of Aurora B in HDAC3-deficient cells may contribute to the weakened mitotic spindle, as the CPC is required for mitotic spindle assembly and stability (53–55). However, HDAC3 may have other targets. We observed that HDAC3 is enriched on the microtubules at the outer edge of the spindle midzone in telophase, suggesting that HDAC3 may have other targets at this location. In addition, our recent proteomics analysis identified a wide variety of nonhistone proteins that are acetylated during mitosis (8). Interestingly, we note that Aurora B was not found in the acetylproteome analysis in mitotic HeLa cells (8) or in a similar analysis in mitotic PC3 cells (unpublished results), likely due to the dynamicity in Aurora B acetylation during mitosis. How the acetylation/deacetylation of protein networks coordinates mitotic progression awaits further analysis.

HDAC3 appears to be the main deacetylase involved in mitosis (7, 8, 36), as knockdown of HDAC1 or HDAC2 did not result in mitotic defects (36, 56, 57). HDAC3 has been suggested to deacetylate histone H3 tails for optimal H3 S10 phosphorylation by Aurora B (58). Thus, the decrease in H3 S10 phosphorylation in HDAC3 knockdown cells could be attributed to a combination of hyperacetylation of histones (59), as well as reduced kinase activity of Aurora B. Further studies to identify the HAT that modifies Aurora B will be useful to determine whether the HAT is acting directly on Aurora B to reduce its kinase activity and/or on H3 to hyperacetylate and thereby attenuate H3 phosphorylation observed in vitro and in vivo. HDAC3 knockdown has also been reported to affect methylation of H3 lysines K4 and K9 that leads to mitotic defects, but the results have not been consistent, possibly due to the dynamic nature of acetylation/deacetylation throughout mitosis, the duration of HDAC3 knockdown, and other differences in experimental protocols (36, 57, 58). In our studies, HDAC3 knockdown does not affect kinetochore formation, as Aurora B (this study), Plk1, CENP-E, and the checkpoint proteins Mad2 and BubR1 are properly localized at the kinetochores (7). In contrast, treatment with pan-HDAC inhibitors led to more severe mitotic phenotypes, including kinetochore assembly defects, loss of CPC and spindle checkpoint proteins at the kinetochore, errors in kinetochore-microtubule attachment, premature sister chromatid separation, and impaired microtubule dynamics (8, 56, 60–63). Thus, the mitotic defects induced by pan-HDAC inhibitors are likely due to a selective inhibition of HDAC3 and other HDACs yet to be identified.

In summary, our data suggest that Aurora B is modulated by dynamic acetylation/deacetylation in mitosis. Acetylation/deacetylation is likely to act as a rheostat to fine-tune Aurora B kinase activity in correcting kinetochore-microtubule attachments, regulating mitotic spindle stability, and promoting chromosome congression in early mitosis. Our study supports the emerging concept that protein acetylation (1, 2, 64, 65) is a prevalent form of post-translational modification in cell cycle regulation. Acetylation likely interacts with other post-translational modifications (10–15) to regulate the activities and/or localization of multiprotein complexes (66) in mitotic regulation. As HDAC inhibitors for transcriptional regulation are being tested for cancer treatments (6, 67), the involvement of acetylation/deacetylation as a novel biochemical control in mitotic progression is an area that may have clinical implications.