Suppression of centrosome duplication and amplification by deacetylases

Abstract

Centrosome duplication is controlled both negatively and positively by a number of proteins. The activities and stabilities of those regulatory proteins are in many cases controlled by posttranslational modifications. Although acetylation and deacetylation are highly common posttranslational modifications, their roles in the regulation of centrosome duplication had not been closely examined. Here, through focusing on the deacetylases, we investigated the role of acetylation/deacetylation in the regulation of centrosome duplication and induction of abnormal amplification of centrosomes. We found that the deacetylation event negatively controls centrosome duplication and amplification. Of the 18 total known deacetylases (HDAC1–11, SIRT1–7), ten deacetylases possess the activity to suppress centrosome amplification, and their centrosome amplification suppressing activities are strongly associated with their abilities to localize to centrosomes. Among them, HDAC1, HDAC5 and SIRT1 show the highest suppressing activities, but each of them suppresses centrosome duplication and/or amplification with its unique mechanism.

Introduction

The centrosome plays a key role in the assembly of bipolar spindles during mitosis, which is essential for balanced segregation of chromosomes into daughter cells. The centrosome is a non-membranous organelle composed of a pair of centrioles and surrounding protein aggregates. Like DNA, the centrosome duplicates once in each cell cycle, which is initiated around the time of S-phase entry and completed by the end of G₂ phase.^1,2 Centrosome duplication is highly regulated, and aberration in the regulatory mechanism can lead to multiple rounds of duplication in a single cell cycle, resulting in generation of ≥ 3 centrosomes (centrosome amplification).³ Centrosome amplification increases the frequency of defective mitoses (e.g., multi-polar spindle formation), leading to chromosome segregation errors. Numerous studies have shown that centrosome amplification occurs frequently in various types of cancers and is believed to be a major cause of chromosome instability in cancer cells.^3,4 Because centrosome duplication is a cell cycle-associated event, many cell cycle-regulatory proteins participate in the control of centrosome duplication both positively and negatively. The activities of those regulatory proteins as well as the proteins required for centrosome duplication are in many cases controlled by posttranslational modifications. To date, the studies on the role of posttranslational modifications in the regulation of centrosome duplication have been focused on phosphorylation and dephosphorylation, as many kinases (e.g., CDKs, polo-like kinases, Aurora A, etc.) participate in the regulation of centrosome duplication, and they themselves are often controlled by phosphorylation and dephosphorylation.⁵ Acetylation and deacetylation are equally common posttranslational modifications catalyzed by histone acetyltransferases (HATs) and histone deacetylases (HDACs).⁶ However, the role of acetylation and deacetylation in the regulation of centrosome duplication had not been closely studied. Acetylation occurs on the ε-amino group of lysine (Lys) residues, which eliminates positive charges, and thus potentially and profoundly affects the target protein’s structure and activity. By the same token, deacetylation can also impact their structures and activities. The acetylation/deacetylation event is also known to cross-talk with other posttranslational modifications. For instance, acetylation is known to frequently counteract ubiquitination of the protein either directly, by competing for the same target Lys residues, or indirectly, by altering the overall structure of the target proteins.⁷ In such a case, acetylation stabilizes the protein, while deacetylation destabilizes it.

Here we examined the role of the acetylation/deacetylation events in the regulation of centrosome duplication in cycling cells and induction of centrosome amplification (centrosome re-duplication) in arrested cells through focusing on the activities of deacetylases. In humans, there are total 18 deacetylases: HDAC1–11 and Sirtuin (SIRT)1–7. We found that the deacetylation event in general suppresses centrosome duplication and amplification. Of all the deacetylases, HDAC1, HDAC5 and SIRT1 were found to possess the strong activities to suppress centrosome amplification. However, each of these deacetylases suppresses centrosome duplication and/or amplification in a unique manner.

Results

Centrosomal proteins are acetylated

Although acetylation of α-tubulin is well-documented,⁸ it is not known whether other centrosomes-localizing proteins are acetylated. We thus examined acetylation of centrosomes-localizing proteins by co-immunostaining U2OS human osteosarcoma cells as well as Hel 299 human primary fibroblasts with anti-γ-tubulin and anti-acetyl-lysine (Ac-K) antibodies. The anti-Ac-K antibody-reactive signals were detected in unduplicated, duplicated and mitotic centrosomes of both U2OS and Hel 299 cells (Fig. 1A), indicating that centrosomal protein(s) are acetylated. We further examined the centrosomes isolated from the proliferating Hel 299 cells. The fractions from the discontinuous sucrose gradient fractionation were immunoblotted with anti-Ac-K, anti-γ-tubulin, anti-PCNA (for testing whether the centrosome preparation was contaminated with nucleus) antibodies (Fig. 1B). We detected several anti-Ac-K antibody-reactive protein bands in the centrosome enriched fraction (fraction 2), indicating that multiple centrosomal proteins are acetylated.

Figure 1. Centrosome localizing proteins are acetylated. (A) U2OS and Hel 299 cells were co-immunostained with anti-γ-tubulin and anti-Ac-K antibodies and stained with DAPI for DNA. The arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 5 μm. (B) The fractions from the discontinuous sucrose gradient centrifugation of the centrosome preparatory lysates from Hel 299 cells were immunoblotted with anti-Ac-K and anti-γ-tubulin antibodies. The arrow points to the centrosome enriched fraction. (C) 293T cells were transfected with GFP-tagged centrin, septin 7 and Plk2 and co-immunostained with anti-γ-tubulin and anti-GFP antibodies, and stained with DAPI. The arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 5 μm. (D) 293T cells transfected with GFP-centrin, -septin 7 and -Plk2 were exposed to deacetylase inhibitors (TSA + NIA) for 12 h, and the cell lysates were immunoprecipitated with anti-GFP antibody. The immunoprecipitates were immunoblotted with anti-Ac-K antibody.

The proteomic analysis of the isolated centrosomes identified several acetylated proteins, including centrin, Polo-like kinase 2 (Plk2) and septin 7. Immunostaining of 293T cells transfected with green fluorescent protein (GFP)-tagged centrin, septin7 (SEPT7) and Plk2 showed that all these proteins localized to centrosomes (Fig. 1C). The centrosome localization of centrin and Plk2 have been shown previously.^9,10 SEPT7 is one of the septin family members, a highly conserved GTP-binding protein originally identified in a genetic screening of S. cerevisiae mutants defective in cytokinesis, and is known to play roles in cytokinesis, membrane trafficking and as structural scaffolds.^11,12 Although centrosome localization of SEPT7 had not been reported previously, our data show that SEPT7 localizes to centrosomes. The transfected cells were also cultured in the presence or absence of deacetylase inhibitors [trichostatin A (TSA): inhibitor for HDAC1–11, nicotinamide (NIA): inhibitor for SIRT1–7]. The lysates prepared from these cells were subjected to immunoprecipitation with anti-GFP antibody, and the immunoprecipitates were immunoblotted with anti-Ac-K antibody (Fig. 1D). GFP-centrin, -septin7 and -Plk2 were all acetylated (lane 1), and their levels of acetylation were noticeably increased when deacetylases were inhibited (lane 2). Based on these observations, we concluded that multiple centrosomally localized proteins are acetylated.

Some deacetylases localize to centrosomes and suppress centrosome amplification

The finding that multiple centrosomal proteins are acetylated raised the possibility that the acetylation/deacetylation reaction may occur at centrosomes, and enzymes responsible for acetylation/deacetylation may be present at centrosomes. Indeed, it has previously been shown that HDAC1 localizes to centrosomes.¹³ Thus, we tested all known deacetylases (HDAC1–11 and SIRT1–7) for centrosome localization. Due to the unavailability of antibodies that were suitable for immunostaining of some deacetylases, we expressed FLAG-tagged deacetylases in U2OS cells and co-immunostained with anti-γ-tubulin and anti-FLAG antibodies. The representative immunostaining images are shown in Figure 2, and the results are summarized in Table 1. HDAC1, -4, -10, -11 and SIRT1, -2 were found at both unduplicated and duplicated centrosomes. HDAC5 was found at unduplicated centrosomes, but not duplicated centrosomes. HDAC6 was found at some unduplicated centrosomes (~25% of unduplicated centrosomes), but not at duplicated centrosomes. The other deacetylases were not found at centrosomes.

Figure 2. Analysis of centrosomal localization of HDACs. To test whether HDACs localize to centrosomes, U2OS cells were transfected with each HDAC tagged with a FLAG-epitope, and co-immunostained with anti-γ-tubulin and anti-FLAG antibodies, followed by DAPI-staining for DNA. HDAC2, 3, 7, 8, 9 and SIRT3, 4, 5, 6, 7 were found to be negative for centrosome localization, and the images of the FLAG-HDAC2-transfected cells are shown as an example for the centrosome localization-negative staining. Arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 10 μm.

Table 1. Centrosome localization and centrosome amplification suppression of HDACs.

	Centrosome localization		Suppression of cent. amplification
HDACs	Unduplicated	Duplicated	*Level	Actual frequency(%)**
Control	NA	NA	NA	41.1 ± 2.3
HDAC1	+	+	++	16.8 ± 3.6
HDAC2	–	–	–	41.0 ± 1.0
HDAC3	–	–	+	34.5 ± 4.0
HDAC4	+	+	+	28.0 ± 1.6
HDAC5	+	–	++	17.6 ± 4.2
HDAC6	± *	–	+	34.0 ± 1.8
HDAC7	–	–	–	43.7 ± 3.0
HDAC8	–	–	+	34.8 ± 3.7
HDAC9	–	–	–	42.7 ± 1.6
HDAC10	+	+	+	29.4 ± 4.4
HDAC11	+	+	+	28.5 ± 5.4
SIRT1	+	+	++	20.0 ± 2.3
SIRT2	+	+	–	41.7 ± 1.9
SIRT3	–	–	–	40.8 ± 1.7
SIRT4	–	–	+	33.2 ± 1.3
SIRT5	–	–	–	39.6 ± 1.7
SIRT6	–	–	–	40.6 ± 1.7
SIRT7	–	–	–	37.7 ± 1.0

^* If the frequency of centrosome amplification is reduced to 70–85% of the control, suppression of centrosome amplification is labeled as +, and if it is reduced to < 50% of the control, suppression of centrosome amplification is labeled as ++. ^**Actual frequency of centrosomes amplification is shown as the means ± SD (see Fig. 3). *HDAC6 was found at unduplicated centrosomes in ~25% of cells. NA, not applicable.

To gain an insight into the functional roles of the deacetylases in the regulation of centrosome duplication, we tested how ectopic expression of each deacetylase would affect centrosome re-duplication by the centrosome amplification assay. In this assay, cells are arrested in the centrosome duplication-permissive stages by exposure to hydroxyurea (HU) (DNA synthesis inhibitor). In these cells, DNA synthesis is inhibited, but centrosomes continue to re-duplicate, resulting in generation of amplified (≥ 3) centrosomes. U2OS cells were transfected with FLAG-tagged deacetylases, exposed to HU for 72 h, and the frequency of centrosome amplification was determined (Fig. 3A, the results are also summarized in Table 1). The frequencies of centrosome amplification were markedly reduced in the cells transfected with SIRT1, HDAC1 and HDAC5 and less pronouncedly in the cells transfected with HDAC3, -4, -6, -10, -11 and SIRT4. The rest of the deacetylases had no effect on centrosome re-duplication. Thus, some deacetylases possess the activities to suppress centrosome re-duplication. Moreover, there was no deacetylase that could promote centrosome re-duplication, indicating that the deacetylation event is in general suppressive to centrosome re-duplication. Noteworthily, there is a strong association between their abilities to localize to centrosomes and the activities to suppress centrosome re-duplication: all of the deacetylases that are found to localize to centrosomes showed the activities to suppress centrosome re-duplication, except SIRT2, while the deacetylases that fail to localize to centrosomes have either no or minimal activities for suppression of centrosome re-duplication except HDAC3, HDAC8 and SIRT4.

Figure 3. Analysis of HDACs for their activities to control centrosome re-duplication. (A) U2OS cells were transfected with FLAG-tagged HDACs. At 6 h post-transfection, cells were treated with HU (2 mM). After HU exposure for 72 h, cells were examined for their centrosome profiles. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. *p < 0.05 and **p < 0.01 compared with control. (B) U2OS cells were treated with either 100 ng/ml TSA (inhibitor of HDAC1–11) or 1 μM Ex-527 (inhibitor of SIRT1). After 12 h, HU was added to the media for 48 h, and the centrosome profiles were determined. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. **p < 0.01 compared with control (DMSO). Note: the frequency of centrosome amplification in the control is ~20% in the graph shown in (B) compared with ~40% in the graph shown in (A), because the duration of HU exposure is 48 h in (B) compared with 72 h in (A).

To confirm that the deacetylation event is suppressive for centrosome re-duplication, we pre-treated U2OS cells with two deacetylase inhibitors; TSA (inhibitor of HDAC1–11) and Ex-527 (inhibitor of SIRT1), followed by exposure to HU for 48 h, and the frequency of centrosome amplification was determined (Fig. 3B). Chemical inhibition of deacetylases resulted in promotion of centrosome re-duplication, further supporting that deacetylation is suppressive for centrosome re-duplication.

Analysis of HDAC1, HDAC5 and SIRT1 for suppression of centrosome duplication and amplification

Because SIRT1, HDAC1 and HDAC5 possessed the highest activities to suppress centrosome re-duplication, we extended our studies focusing on these three deacetylases to gain an insight of how HDACs suppress centrosome duplication and amplification. Since highly specific antibodies to these deacetylases are commercially available, we tested centrosome localization of the endogenous SIRT1, HDAC1 and HDAC5 (Fig. 4A). Similar to the findings obtained from the epitope-tagged deacetylases, HDAC1 and SIRT1 were found to localize to both unduplicated and duplicated centrosomes, and HDAC5 was found to localize to only unduplicated centrosomes. The immunoblot analysis of the fractions from the discontinuous sucrose gradient fractionation also showed that HDAC1, HDAC5 and SIRT1 were present in the centrosome enriched fraction (Fig. 4B).

Figure 4. The roles of HDAC1, HDAC5 and SIRT1 in the suppression of centrosome duplication and amplification. (A) U2OS cells were co-immunostained with anti-γ-tubulin (red) and anti-HDAC1, anti-HDAC5 or anti-SIRT1 (green) antibodies and then stained with DAPI (blue). The arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 10 μm. (B) The fractions from the discontinuous sucrose gradient centrifugation of the centrosome preparatory lysates from U2OS cells were immunoblotted with anti-HDAC1, -HDAC5, -SIRT1 and -γ-tubulin antibodies. The arrows point to the centrosome enriched fractions. (C) U2OS cells were transfected with the pSuper.retro.puro plasmids containing siRNAs targeting HDAC1, HDAC5 or SIRT1. As a control, the plasmid containing siRNA targeting luciferase (pS-Luc) was used. After puromycin selection, the surviving cells were pooled, and the lysates were immunoblotted with anti-HDAC1, anti-HDAC5 and anti-SIRT1 antibodies (a). The surviving cells were treated with HU for 48 h, and the centrosome profiles were determined (b). The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. *p < 0.05 and **p < 0.01 compared with control siRNA. (D) MSFs were used for studying the regulation of centrosome duplication in cycling cells. For the overexpression study, MSFs were serum-starved, and transfected with a FLAG-vector, -HDAC1, -HDAC5 or -SIRT1. At 24 h post-transfection, cells were serum-stimulated in the presence of BrdU. At indicated time points for total 28 h, cells were examined for centrosome duplication and BrdU incorporation (a, top two graphs). For the knockdown study, MSFs were transfected with siRNAs specific for HDAC1, HDAC5 or SIRT1. The pS-Luc was used as a control. At 24 h post-transfection, the successfully transfected MSFs were selected with puromycin, and examined for expression of HDAC1, HDAC5 and SIRT1 (b). The cells were serum-starved and serum-stimulated in the presence of BrdU. At indicated time points, cells were examined for centrosome duplication and BrdU incorporation (a, bottom two graphs). The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. *p < 0.05 and **p < 0.01 compared with the control cells.

We next tested how depletion of these deacetylases affects centrosome re-duplication. U2OS cells were transfected with siRNAs specific for SIRT1, HDAC1 and HDAC5. Expression of the respective protein was silenced to < 10% of the normal level (Fig. 4Ca). The cells were then exposed to HU for 48 h, and the frequencies of centrosome amplification were determined (Fig. 4Cb). Silencing of SIRT1, HDAC1 and HDAC5 all resulted in promotion of centrosome amplification, indicating that these deacetylases act as suppressors of centrosome re-duplication.

We next tested how depletion and overexpression of HDAC1, HDAC5 and SIRT1 affect centrosome duplication in cycling cells. For this experiment, we used primary skin fibroblasts from wild-type (wt) mice (MSFs) that show highly coordinated initiation of DNA replication and centrosome duplication. For the overexpression experiment, MSFs synchronized in a quiescent state by serum starvation were transfected with FLAG-HDAC1, -HDAC5 and -SIRT1, then serum-stimulated, and the rates of S-phase entry and centrosome duplication were determined (Fig. 4Da, top two graphs). Overexpression of HDAC1 and HDAC5 did not alter the rates of S-phase entry, while that of SIRT1 led to a delayed S-phase entry (Fig. 4Da, top-right graph). Overexpression of HDAC1 did not affect the rate of centrosome duplication, while that of HDAC5 and SIRT1 resulted in a marked delay in the initiation of centrosome duplication (Fig. 4Da, top-left graph), implicating HDAC5 and SIRT1 in the negative regulation of centrosome duplication in cycling cells. For the depletion experiment, the serum-starved MSFs were transfected with siRNAs specific for HDAC1, HDAC5 and SIRT1. Expression of each deacetylase was silenced to < 10% of the normal levels (Fig. 4Db). The transfected cells were serum-stimulated, and the rates of S-phase entry and centrosome duplication were determined (Fig. 4Da, bottom two graphs). There was no significant change in the rates of S-phase entry in cells silenced for HDAC1, HDAC5 and SIRT1 (bottom-right graph). Silencing of HDAC1 had minimal effect on the rate of centrosome duplication, while that of HDAC5 as well as SIRT1 resulted in accelerated initiation of centrosome duplication (Fig. 4Da, bottom-left graph). These results show that HDAC5 and SIRT1, but not HDAC1, are critically involved in the regulation of centrosome duplication in cycling cells in a suppressive manner.

The deacetylase activity is essential for HDAC1 and SIRT1, but not for HDAC5, to suppress centrosome amplification

To ask whether the deacetylase activity is important for HDAC1, HDAC5 and SIRT1 to suppress centrosome re-duplication, we tested the deacetylase-dead (DD) mutants of HDAC1 (His141→Ala), HDAC5 (His832, 833→Ala) and SIRT1 (His363→Tyr). The mutations introduced to these DD mutants did not affect their centrosome localization activities; HDAC1/DD and SIRT1/DD mutants localize to both unduplicated and duplicated centrosomes, and HDAC5/DD localizes to only unduplicated centrosomes (Fig. 5A). These DD mutants were tested for their abilities to suppress centrosome re-duplication (Fig. 5B). The HDAC1/DD and SIRT1/DD mutants failed to suppress centrosome re-duplication, indicating that the deacetylase activity is essential for HDAC1 and SIRT1 to suppress centrosome re-duplication. Surprisingly, the HDAC5/DD mutant suppressed centrosome re-duplication as effectively as wt HDAC5. To corroborate this finding, we tested two additional HDAC5 mutants, the deacetylase activity-negative HDAC5 mutant that lacks the entire deacetylase domain (HDAC5/NT) and a gain-of-function HDAC5 mutant (His1006→Tyr; HDAC5/H1006Y) that possesses a higher deacetylase activity than wt HDAC5 (Fig. 5C). Both NT and H1006Y mutants suppressed centrosome re-duplication at similar efficiencies with wt HDAC5 (Fig. 5D). Thus, HDAC5 suppresses centrosome re-duplication in a deacetylase activity-independent manner.

Figure 5. HDAC1 and SIRT1, but not HDAC5, suppress centrosome re-duplication in a deacetylase activity-dependent manner. (A) The FLAG-tagged deacetylase-dead (DD) mutants of HDAC1 (H141A), HDAC5 (H832A, H833A) and SIRT1 (H363Y) were transfected into U2OS cells. The transfected cells were co-immunostained with anti-FLAG and anti-γ-tubulin antibodies, and stained with DAPI. The arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 10 μm. (B) FLAG-tagged HDAC1/DD, HDAC5/DD and SIRT1/DD were transfected into U2OS cells. Cells were then exposed to HU for 72 h, and their centrosome profiles were examined. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. *p < 0.05 compared with control. (C) The schematic diagrams of the HDAC5/NT mutant that lacks the deacetylase domain and HDAC5/H1006Y “gain-of-function” mutant that has a higher deacetylase activity than wt HDAC5 are shown. (D) FLAG-HDAC5/NT, -HDAC5/H1006Y and -wt HDAC5 were transfected into U2OS cells. Cells were then exposed to HU for 72 h and their centrosome profiles were examined. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. *p < 0.05 and **p < 0.01 compared with control.

The centrosome localization and centrosome amplification suppressing activities of HDAC5 are controlled by phosphorylation

Because the activities and properties of deacetylases are often controlled by phosphorylation,¹⁴ we tested whether phosphorylation of HDAC1, HDAC5 and SIRT1 affects their activities to suppress centrosome amplification. HDAC1 is known to be phosphorylated on Ser421 and Ser423 residues.¹⁵ HDAC5 is known to be phosphorylated on Ser259 and Ser498 by calcium/calmodulin-dependent protein kinase (CaMK).¹⁶ SIRT1 is known to be heavily phosphorylated over 20 Ser and Thr residues by a number of kinases, including DRYK1A/3, JNK1, CDK1 and CK2. Although the nascent form of SIRT1 is already active, these phosphorylations have been shown to increase the deacetylase activity of SIRT1 in general.^17-19 Of these phosphorylated residues, the roles of phosphorylation on Ser559 and Ser661 by CK2 are currently unknown.²⁰ However, CK2 localizes to centrosomes²¹ and is implicated in induction of centrosome amplification.²² We generated the non-phosphorylatable HDAC1, HDAC5 and SIRT1 mutants by replacing the respective Ser residues to Ala (HDAC1/2SA, HDAC5/2SA and SIRT1/2SA) (Fig. 6A). Using these mutants, we first tested whether phosphorylation affects their activities to suppress centrosome re-duplication. The non-phosphorylatable mutants tagged with FLAG-epitopes were transfected into cells, and the transfected cells were subjected to the centrosome amplification assay (Fig. 6B). HDAC1/2SA and SIRT1/2SA suppressed centrosome amplification as efficiently as their respective wild-types (see Fig. 3A for data reference). We also tested SIRT1 mutants with some other phosphorylated Ser/Thr residues replaced to Ala, and they also suppressed centrosome amplification as efficiently as the wild-type (data not shown). Thus, phosphorylation modifications of HDAC1 and SIRT1 are not essential for suppressing centrosome amplification. However, for SIRT1, it remains possible that phosphorylations of other residues may play roles in the fine-tuning of the deacetylase activities to control centrosome duplication during the cell cycle. In contrast, HDAC5/2SA failed to suppress centrosome amplification (the results are summarized in Fig. 6D), indicating that the centrosome amplification suppressing activity of HDAC5 is critically controlled by phosphorylation.

Figure 6. Phosphorylation is required for HDAC5 to localize to centrosomes and suppress centrosome amplification. (A) Generation of non-phosphorylatable mutants of HDAC1 (HDAC1/2SA), HDAC5 (HDAC5/2SA) and SIRT1 (SIRT1/2SA). (B) FLAG-tagged HDAC1/2SA, HDAC5/2SA and SIRT1/2SA were transfected into U2OS cells. Cells were then exposed to HU for 72 h and their centrosome profiles were examined. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. **p < 0.01 compared with control. (C) FLAG-tagged HDAC1/2SA, HDAC5/2SA and SIRT1/2SA were transfected into U2OS cells. The transfected cells were co-immunostained with anti-γ-tubulin and anti-FLAG antibodies, followed by DAPI-staining. Arrows point to centrosomes. The insets show the magnified images of the centrosome areas. Scale bar: 10 μm. (D) Summary of the centrosome localizing and centrosome amplification suppressing activities of the HDAC1, HDAC5 and SIRT1 non-phosphorylatable mutants.

We next tested whether phosphorylation affects centrosome localization of HDAC1, HDAC5 and SIRT1. To this end, the non-phosphorylatable mutants tagged with a FLAG epitope were transfected into U2OS cells, and the transfected cells were co-immunostained with anti-γ-tubulin and anti-FLAG antibodies (Fig. 6C). Similar to their respective wild-types, HDAC1/2SA and SIRT1/2SA localized to both unduplicated and duplicated centrosomes. We also tested SIRT1 mutants with other phosphorylated Ser/Thr residues replaced to Ala, and they showed similar patterns of centrosome localization with the wild-type (data not shown). In contrast, HDAC5/2SA failed to localize to centrosomes (the results are also summarized in Fig. 6D). Thus, the centrosome localizing activities of HDAC1 and SIRT1 are independent of their phosphorylation states, but phosphorylation plays a critical role for HDAC5 to localize to centrosomes. This observation implies that the failure of HDAC5/2SA to suppress centrosome amplification may be due to the failure to localize to centrosomes.

HDAC1 suppresses centrosome amplification via downregulation of cyclin A expression

Because the mechanisms underlying centrosome amplification in arrested cells and centrosome duplication in cycling cells share many common features, the proteins that suppress centrosome amplification in arrested cells usually act as suppressors of centrosome duplication in cycling cells, and vice versa. As shown in Figures 3 and 4, HDAC1 possesses the activity to suppress centrosome amplification (re-duplication) in the cell cycle-arrested cells, but not centrosome duplication in the cycling cells. We have previously shown that centrosome duplication in cycling cells and centrosome re-duplication in cells arrested in late G₁/early S-phase requires the CDK2-cyclin E activity, while centrosome re-duplication in cells arrested in late S and G₂ phases requires the CDK2-cyclin A activity.²³ Because we used HU to arrest cells for the centrosome amplification assay, which arrests the majority of cells in late S and G₂, we hypothesized that HDAC1-mediated deacetylation might lead to downregulation of CDK2-cyclin A, resulting in suppression of centrosome re-duplication in the HU-arrested cells, but not centrosome duplication in cycling cells. In support of this hypothesis, HDAC1 has been shown to repress the E2F-dependent transcription of cyclin A through interaction with pRb/p130.^24,25 We thus examined the expression level of cyclin A in cells overexpressing HDAC1. FLAG-HDAC1 and the control vector plasmid were transfected into cells, and the cell lysates prepared from the transfected cells were immunoblotted for cyclin A as well as other cyclins and CDK2 (Fig. 7Aa). In consistent with the previous studies, we found that the level of cyclin A was decreased to ~10% of the normal level by HDAC1 overexpression, while there was no obvious change in the levels of other cyclins (cyclin B, D, E) nor CDK2. Thus, HDAC1-associated reduction of cyclin A is highly specific. It is known that acetylation often competes with ubiquitination for the same lysine residues, and in such a case, deacetylation promotes ubiquitination and subsequent ubiquitin-dependent protein degradation.⁷ To exclude the possibility that deacetylation by HDAC1 may promote proteasome-dependent degradation of cyclin A, we compared the levels of cyclin A between the control and FLAG-HDAC1-transfected cells in the presence of MG132 proteasome inhibitor (Fig. 7Ab). The reduction of the cyclin A levels occurred in the FLAG-HDAC1-transfected cells even when proteasomes were inhibited, excluding the possibility that HDAC1 may promote proteasome-dependent degradation of cyclin A. The quantitative real-time PCR analysis of the total RNA prepared from the control and HDAC1-transfected cells revealed that the level of cyclin A mRNA was markedly decreased (> 50% reduction compared with the control cells) (Fig. 7Ac)., implying that HDAC1 represses the transcription of cyclin A.

Figure 7. HDAC1 suppresses centrosome amplification via downregulation of cyclin A expression. (A) U2OS cells were co-transfected with FLAG-HDAC1 along with pBabe-puro at 15:1 ratio. A FLAG vector plasmid was used as a control. The transfected cells were treated with HU and puromycin for 72 h and the lysates prepared from surviving cells were subjected to immunoblot analysis with anti-FLAG, anti-cyclin A, anti-cyclin E, anti-cyclin D, anti-cyclin B, anti-CDK2 and anti-β-actin antibodies (a). The transfected cells after 72 h HU treatment were also treated with MG132 (10 μM) for 5 h before preparation lysates, and the lysates were immunoblotted with anti-FLAG, anti-cyclin A and anti-β-actin antibodies (b). Total RNA prepared from the transfected cells were reverse-transcribed to cDNA, and real time-PCR was performed to determine the relative mRNA level of cyclin A. 18S RNA was used as the internal control. The results are shown in the graph as means ± SD (arbitrary units) from three independent experiments. *p < 0.05 compared with control (c). (B) U2OS cells were treated with either HU (2 mM) or mimosine (0.5 mM) for 72 h. Cells were then subjected to flow cytometric analysis. (C) U2OS cells were co-transfected with HDAC1 along with GFP as a transfection marker at 15:1 ratio. A vector plasmid was used as a control. At 6 h post-transfection, cells were treated with either HU or mimosine for 72 h, and the centrosome profiles of the GFP-positive cells were determined. The results are shown in the graph as means ± SD from three independent experiments. For each experiment, > 200 cells were examined. **p < 0.01 compared with the vector control.

When cells are exposed to HU, 60–70% of cells become arrested in late S and G₂ phases (Fig. 7B), in which centrosome re-duplication depends on the presence of cyclin A. In contrast, when cells are exposed to mimosine, an inhibitor of initiation of DNA replication,²⁶ 60–70% of cells become arrested in late G₁ and early S phases (Fig. 7B), in which centrosome re-duplication depends on the presence of cyclin E.²³ If our hypothesis that HDAC1 suppresses centrosome amplification through downregulation of cyclin A is correct, HDAC1 should fail to suppress centrosome amplification in cells arrested with mimosine. To test this, cells transfected with FLAG-HDAC1 were subjected to the centrosome amplification assay using HU and mimosine (Fig. 7C). In the HU-arrested cells, HDAC1 effectively suppressed centrosome amplification. However, in the mimosine-arrested cells, HDAC1 failed to suppress centrosome amplification. Thus, HDAC1 is unable to suppress centrosome amplification in the late G₁/early S-arrested cells that requires CDK2-cyclin E, while HDAC1 can suppress centrosome amplification through downregulation of cyclin A in the late S/G₂-arrested cells that requires CDK2-cyclin A, further supporting that HDAC1 suppresses centrosome amplification through downregulation of cyclin A.

Discussion

Analysis of deacetylases (HDAC1–11 and SIRT1–7) for their abilities to localize to centrosomes and activities to control centrosome amplification revealed that HDAC1, -4, -5, -6, -10 and SIRT1, -2 localized to centrosomes and HDAC1, -3,- 4, -5, -6, -8, -10, -11, SIRT1 and -4 possessed the activities to suppress centrosome re-duplication. Importantly, none of these HDACs was found to promote centrosome amplification, indicating that the deacetylation event is, in general, suppressive for centrosome amplification. Moreover, there is an association between the activity to suppress centrosome amplification and the ability to localize to centrosomes, suggesting that many deacetylases may function at centrosomes to suppress centrosome amplification. The exceptions to this are HDAC3, HDAC8, SIRT2 and SIRT4. HDAC3, HDAC8 and SIRT4 do not localize to centrosomes, but possess the activity to suppress centrosome re-duplication. HDAC8 has previously been implicated in the suppression of centrosome duplication: depletion of HDAC8 leads to centriole splitting, an initial event of centrosome duplication.²⁷ Because centrosome duplication is a cell cycle-dependent process, many non-centrosomal events that link the mitogenic signaling and centrosome duplication are equally critical for the initiation of centrosome duplication, and HDAC3, HDAC8 and SIRT4 likely target such non-centrosomal events. In contrast, SIRT2 localizes to centrosomes but does not possess the activity to suppress centrosome amplification. SIRT2 has been shown to localize to centrosomes²⁸ and is implicated in the control of mitotic progression.²⁹ SIRT2 (as well as HDAC6) has also been shown to deacetylate α-tubulin,^30-32 which leads to stabilization of microtubules (e.g., axonemal microtubules)^8,31 and is also implicated in cytoplasmic trafficking.^33,34 Thus, centrosomal SIRT2 may control centrosome function and/or behavior rather than duplication.

Among all the deacetylases, HDAC1, HDAC5 and SIRT1 possess the higher activities to suppress centrosome amplification than the others. HDAC1 was found to suppress centrosome amplification in the late S and G₂-arrested cells in a deacetylation activity-dependent manner. However, neither overexpression nor depletion of HDAC1 had any significant effect on centrosome duplication during the cell cycle. We have previously shown that centrosome duplication in cycling cells requires the activity of CDK2-cyclin E, while centrosome re-duplication in the late S and G₂-arrested cells requires the activity of CDK2-cyclin A.²³ We found that HDAC1 suppresses centrosome amplification in the late S and G₂ cells via downregulation of cyclin A expression. Indeed, it has been shown that HDAC1 binds to pRb2/p130, and this interaction increases the ability of pRb2/p130 to inhibit transcription of the E2F-dependent cyclin A promoter.²⁴ Moreover, HDAC1 fails to suppress centrosome amplification in the cells treated with mimosine, which is known to arrest cells in late G₁ and early S phases, where the activity of CDK2-cyclin E is required for centrosome re-duplication. For this reason, HDAC1 is able to suppress centrosome amplification in the late S and G₂-arrested cells, but not centrosome duplication in cycling cells and centrosome re-duplication in the late G₁ and early S-arrested cells.

In contrast to HDAC1, HDAC5 was found to suppress both centrosome duplication in cycling cells and centrosome amplification in arrested cells. However, it does so in a deacetylation activity-independent manner. Although this is unexpected, the deacetylation-independent activities of HDACs have been reported previously for HDAC5 as well as HDAC7. For instance, HDAC5 and -7 possess autonomous repressor functions that are apparently independent of their deacetylase activity.³⁵ More recently, it has been shown that HDAC7 physically associates with Runx2 and represses Runx2 transcriptional activity in a deacetylase-independent manner.³⁶ Thus, HDAC5 probably suppresses centrosome duplication/amplification through physical interaction with other protein(s) that control centrosome duplication/amplification. Through examination of the effect of phosphorylation of HDACs for their activities to suppress centrosome amplification, we found that non-phosphorylatable HDAC5 mutant failed to localize to centrosomes, and also failed to suppress centrosome amplification. Thus, phosphorylation (Ser259 and/or Ser498), which is catalyzed by CaMK,¹⁶ is critical for centrosome localization of HDAC5, and centrosome localization may be critical for HDAC5 to suppress centrosome amplification.

Unlike HDAC1 and HDAC5, SIRT1 was found to participate in the regulation of centrosome duplication and amplification in a more expected manner: SIRT1 suppresses centrosome duplication in cycling cells and centrosome amplification in arrested cells in a deacetylase activity-dependent fashion. For instance, overexpression of SIRT1 results in suppression of centrosome duplication and amplification, but the deacetylase activity-dead SIRT1 mutant fails to do so. Moreover, knockdown of SIRT1 results in promotion of centrosome duplication and amplification. Further analyses of the molecular mechanisms underlying the SIRT1 (as well as HDAC5 and other HDACs)-associated suppression of centrosome duplication and amplification, including identification of their target proteins, will clearly advance our understanding of how centrosome duplication is controlled in cycling cells and how amplified centrosomes are generated in cancer cells, which is currently underway in our laboratory. In recent years, there has been an effort to develop HDAC inhibitors as a cancer treatment. One model for the mechanism by which the compounds work is that the increase in histone acetylation leads to activation and repression of key genes whose expression/repression either inhibits tumor growth or induces tumor cell death.³⁷ Our present findings show that inhibition of certain HDACs promotes centrosome amplification, which destabilizes chromosomes and, hence, increases the risk of acquisition of further malignant phenotypes and genesis of secondary tumors. Further studies of the roles of HDACs in the control of centrosome duplication and amplification will provide important information for developing therapeutic HDAC inhibitors with minimal effect on centrosome abnormality.

Materials and Methods

Cells and transfection

U2OS, Hel 299, 293T and MSFs were maintained in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin. Transfection was performed using either FuGENE HD (Roche) or Lipofectamine 2000 (Invitrogen).

Plasmids, siRNA sequences, primers and antibodies

For plasmids encoding siRNAs for HDACs, the 19 nt target sequences were placed into the pSuper.puro.retro vector (OligoEngine). Human and mouse HDAC1: 5′-GCCTCACCGAATCCGCATG -3′; human HDAC5: 5′-GCCTGGTGCTGGATACAAA-3′; mouse HDAC5: 5′-GCAAGCATTCTACAACGAT-3′; human SIRT1: 5′-GTTGGATGATATGACACTG-3′; mouse SIRT1: 5′-CAGGTTGCAGGAATCCAAA-3′. The sequence targeting luciferase (5′-CATCACGTACGCGGAATAC-3′) was used as a negative control. The primers used to amplify cyclin A in real-time PCR assay are: 5′-CCATACCTCAAGTATTTG CCA TCA-2’ (forward) and 5′-AGCTTTGTCCCGTGACTGTGT-3′ (reverse). The antibodies used in this study are: anti-γ-tubulin (TU-30, Santa Cruz), anti-centrin antibody (a gift from Dr. J. Salisbury), anti-Ac-K (ST1027, Calbiochem; 9441 sec, Cell Signaling), anti-FLAG (M2 and F7425, Sigma), anti-HDAC1 (07–443, Upstate), anti-HDAC5 (07–045, Upstate), anti-SIRT1 (sc-459, Santa Cruz; ab7343–100, Abcam), anti-GFP (Roche; sc-8334, Santa Cruz), anti-PCNA (sc-56, Santa Cruz), anti-cyclin A (sc-751, Santa Cruz), anti-cyclin E (sc-198, Santa Cruz), anti-cyclin D (sc-753, Santa Cruz), anti-cyclin B1 (sc-752, Santa Cruz), and anti-CDK2 (sc-6248, Santa Cruz).

Isolation of centrosomes

Centrosomes were isolated from Hel 299 and U2OS cells by the discontinuous sucrose gradient centrifugation method as described previously.³⁸

Immunoblot analysis

Cells were lysed in lysis buffer [1% SDS, 1% NP-40, 50 mM Tris (pH 8.0), 150 mM NaCl, protease inhibitor cocktails]. The lysates were boiled for 5 min and cleared by centrifugation at 4°C. The samples were denatured in sample buffer [2% SDS, 10% glycerol, 60 mM Tris (pH 6.8), 5% β-mercaptoethanol, 0.01% bromophenol blue], resolved by SDS-PAGE and transferred onto membrane. The blots were incubated in blocking buffer [5% (w/v) nonfat dry milk in Tris-buffered saline + Tween 20 (TBS-T)] for 1 h, and incubated with primary antibodies for 16 h at 4°C, followed by incubation with horseradish peroxidase-conjugated secondary antibodies for 1h at 25°C. The antibody-antigen complex was visualized by ECL chemiluminescence (Pierce), and quantified by Quantity One (BioRad).

Indirect immunofluorescence

Cells growing on coverslips were fixed with 100% methanol at -20°C or with 10% formalin at 25°C for 20 min. For examination of centrosomal localization of HDACs, cells were briefly extracted with extraction buffer [0.75% Triton X-100 and 2 mM EGTA in 5 mM PIPES (pH6.7)] for ~15 sec, followed by washing with PBS prior to fixation. Without the brief extraction, signals of HDACs at centrosomes were often difficult to detect. Fixed cells were permeabilized and blocked by 10% normal goat serum in PBS for 1 h. Cells were then incubated with primary antibodies for 1 h with either Alexa Fluor 488- or 594-conjugated secondary antibodies for 1 h. Cells were than stained for DNA with 4’, 6-diamidino-2-phenylindole (DAPI). For immunostaining of centrosomes for the study of centrosome duplication and amplification, we used both anti-γ-tubulin and anti-centrin antibodies, which gave similar results. The BrdU-incorporation assay was performed using the 5-Bromo-2’-deoxy-uridine Labeling and Detection Kit II (Roche) according to the manufacturer’s instruction.

Real-time PCR analysis

Total RNA was isolated from cells with TRIzol reagent (Invitrogen). Reverse transcription and quantitative real-time PCR were performed as described previously.³⁹

Disclosure of Potential Conflicts of interest

No potential conflicts of interest were disclosed.