Histone Deacetylase Inhibitors Induce Premature Sister Chromatid Separation and Override the Mitotic Spindle Assembly Checkpoint

Histone deacetylase (HDAC) inhibitors (HDACI) exhibit antiproliferative activity against a variety of human cancers. They are currently undergoing clinical trials and the outlook is promising. However, the molecular basis of their antitumor properties remains poorly understood (1, 2). It was first suggested that HDAC inhibition induces the reexpression of silenced tumor suppressor genes by affecting the chromatin structure of their promoters (3). For example, treatment of cells with HDACI often induces the expression of the cyclin-dependent kinase (cdk) inhibitor p21^WAF1/CIP1 in a p53-independent manner (4, 5). The expression of a number of genes other than cell cycle regulator genes is also altered by HDACI treatment: they include genes involved in the induction of apoptosis, the activation of the immune response, and the inhibition of angiogenesis (6). However, gene profiling studies showed that the expression of a surprisingly small fraction of genes is affected by HDACI treatments and it remains uncertain whether these particular genes are sufficient to drive cancer cells towards differentiation, apoptosis, or growth arrest pathways (7–10). Therefore, it has been suggested that nontranscriptional targets may be equally important for HDACI antitumor activity (11–13). One potential nontranscriptional target is the centromere/kinetochore structure. Indeed, low-dose trichostatin A (TSA) treatment destabilizes the higher-order structure of pericentric heterochromatin, and leads to HP1 spreading and centromere subnuclear localization (14). Pericentric heterochromatin plays an important role in centromere function, and HDACI treatments may lead to kinetochore defects and aberrant mitosis. Accordingly, several studies report that HDACI treatment increases chromosomal instability characterized by chromatin bridges, chromosome breaks, and segregation defects (14–16). These features suggest that the nontranscriptional targets of HDACI include mitosis and the mechanisms that regulate its accuracy. Segregation deficiency is a common characteristic of HDACI-treated cells and of mitotic spindle assembly checkpoint (SAC) defects. We therefore examined whether HDACIs can target the SAC.

The SAC is a cellular mechanism that ensures faithful sister chromatid segregation during mitosis by delaying anaphase until all kinetochores have been attached to spindle microtubules (17). A defect in this function results in unbalanced chromosome segregation, and consequently, aneuploid daughter cells. Unattached kinetochores generate an inhibitory signal acting on the anaphase-promoting complex/cyclosome (APC/C), a multiprotein complex whose ubiquitin ligase activity is required to target mitosis-specific factors (such as cyclin B1 and securin) to proteasome-dependent degradation. The attachment of the last free kinetochore abrogates the inhibitory signal and thus permits the transition from metaphase to anaphase and subsequent exit from mitosis. Inhibition of the SAC by mutation of its components, haploinsufficiency, or down-regulation of one of its elements results in premature sister chromatid separation (PSCS) and APC/C target degradation despite of SAC activation conditions (18–21). Furthermore, long-term total SAC inhibition is not viable. Null mutations of the SAC component MAD2 is lethal (22) and extended inhibition of SAC by RNA interference affecting key SAC elements leads to cell death within a few generations (23). The most likely explanation for this lethality is the accumulation of nonviable aneuploid cells due to premature anaphase entry. Consequently, because the cell toxicity resulting from SAC inhibition is expected to be restricted to proliferating cells, the SAC is a putative therapeutic target for anticancer treatments (24). However, the only SAC down-regulation techniques currently known are genetic and, in the foreseeable future, unsuitable for application to human therapy.

To determine whether HDACIs can be used as pharmacologic tools to target the SAC, we investigated mitosis exit in HDACI-treated cells in which the SAC was activated. We found that HDACI treatment triggers transcription-independent mitosis exit in a large proportion of prometaphase-arrested cells. We monitored SAC inhibition by detection of the major mitosis exit markers, including sister chromatid separation and proteasome-dependent degradation of key APC/C targets.

Cell culture and drug treatment. HeLa cells (human cervical carcinoma, obtained from American Type Culture Collection) were cultured in DMEM (Life Technologies) supplemented with l-glutamine, antibiotics, and 10% heat-decomplemented FCS (Invitrogen). In all experiments, growing cells were incubated in nocodazole (500 ng/mL) for 8 h and accumulated mitotic cells were harvested by mitotic shake-off (MSO) prior to other treatments. Nocodazole was permanently maintained in the medium when HDACIs or other drugs were added. For transcription inhibition and proteasome inhibition assays, α-amanitin (2 μg/mL) or MG132 (3 μmol/L) were added 1 h before TSA treatment; α-amanitin or MG132 were maintained in the culture medium after the addition of TSA. Unless otherwise stated, the TSA concentration was 200 ng/mL. All reagents were purchased from Sigma-Aldrich, except for MG132 (Biomol Int.).

Chromosome spreads. Metaphase spreads were done essentially as described (25) and DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI; 100 ng/mL) using Mowiol as the mounting medium. At least 500 cells were scored for each experimental point. All experiments were done at least in duplicate.

Immunoblotting. Cells were harvested, counted, lysed directly in Laemmli buffer and boiled for 5 min. Samples corresponding to 6 × 10⁴ cells were separated by SDS-PAGE and transferred to nitrocellulose membranes (Hybond-C Extra, Amersham); the membranes were probed with monoclonal anti–cyclin B1 antibodies (Santa Cruz), polyclonal anti-securin antibodies (Zymed), or anti-actin antibodies (Sigma) according to standard procedures. Western Lighting Chemiluminescence Plus (Perkin-Elmer) was used to visualize binding, according to the instructions of the supplier.

Cdk1 kinase assays and immunoprecipitation. Cdk1 immunoprecipitation followed by cdk1 kinase assays were as previously described (26). Assays were analyzed by autoradiography, and kinase activity was determined using a Phosphoimager (BAS-5000, FUJI). Cdk1 immunoprecipitates were analyzed by immunoblotting using a monoclonal anti-cdk1 antibody (Santa Cruz).

Immunofluorescence microscopy. Chromosome spread slides (Fig. 6B) were washed twice in PBS for 10 min, incubated with 0.1% Triton X-100, washed thrice in PBS, blocked with 5% FCS, and stained with rabbit polyclonal anti-H3 serine 10 phosphorylated antibodies (UBI) overnight at 4°C. Goat anti-rabbit conjugated-FITC antibodies (Biosource International) were used for secondary staining for 1 h at room temperature. For whole cell immunofluorescence (Figs. 5B and 6A), cells were grown on polylysine-treated coverslips, fixed in 3% formaldehyde, washed in PBS, incubated for 15 min with 100 mmol/L of glycine (pH 7), permeabilized with 0.1% Triton X-100, washed in PBS, blocked with 5% FCS, and incubated with rabbit polyclonal anti-lamin A/C antibodies (Cell Signaling), human CREST serum (a kind gift from Dr. C. Johanet, Hôpital Saint-Antoine, Paris), or rabbit polyclonal anti-MAD2 antibodies (Covance). Goat anti-rabbit conjugated-FITC antibodies (Biosource International) and goat anti-human conjugated-TRITC antibodies (Jackson ImmunoResearch) were used as secondary antibodies. In all experiments, DNA was stained with DAPI (100 ng/mL) and Mowiol was used as the mounting medium. Samples were examined under an epifluorescence microscope (Leica DMRM).

Prometaphase-arrested cells undergo HDACI-induced PSCS in a transcription independent manner. In all experiments, mitotic SAC was activated by pretreatment with the microtubule-destabilizing agent nocodazole. Nocodazole treatment activates the SAC by preventing microtubule-kinetochore interaction. Once the SAC is activated, cells accumulate in prometaphase and, depending on the cell type and the nocodazole concentration, remain arrested between a few hours and several days. Under our culture conditions, >60% of HeLa cells were arrested in prometaphase following an 8-h nocodazole treatment and remained stably blocked in prometaphase for the following 24 h without significant mitosis exit or cell death. HeLa cells were treated with nocodazole for 8 h and mitosis-arrested cells were collected by MSO. Then, the cells were incubated with TSA, a powerful HDACI, for 16 h. Nocodazole was present in the culture medium during TSA treatment. A proportion of chromosome spreads from nocodazole/TSA–treated cells displayed a characteristic PSCS phenotype (Fig. 1A): sister chromatids were no longer attached at the centromere and the constriction point was not visible anymore. In addition, despite the extended nocodazole treatment (which is known to overcondense metaphasic chromosomes), chromosome arms in nocodazole/TSA–treated cells displaying PSCS seemed to be longer and less tightly condensed than in the controls. This PSCS phenomenon, also called “precocious anaphase,” was observed in ∼25% of nocodazole/TSA–treated cells, and all chromosomes in any one cell displayed an identical phenotype. No PSCS was detected in control nocodazole-treated HeLa cells.

TSA-induced PSCS was dose-dependent (Fig. 1B). In addition, the mitotic index of control nocodazole-treated cells was >80%, whereas that of TSA-treated cells declined in a TSA dose-dependent manner to <40% (Fig. 1C). Cell viability was assessed by trypan blue exclusion and cell counting. Under these treatment conditions, a few apoptotic cells (condensed fragmented nuclei as depicted on Fig. 5B, cells labeled “ap”) were detected although cell mortality was <5%. Therefore, specific mitotic cell mortality cannot account for the observed decrease in the mitotic index.

PSCS increase and mitotic index decrease induced by TSA were also observed in a human breast cancer cell line (MDA-MB-231) and in human primary fibroblasts and, thus, are not specific to the transformed HeLa cell line (Fig. 2).

To determine if the effects described above were specific for TSA or are found with other HDACIs, nocodazole-treated cells were incubated with various different HDACIs belonging to three different classes (Fig. 3A). Sodium butyrate and valproic acid are short-chain fatty acids, TSA is a hydroxamic acid and apicidin belongs to the cyclic tetrapeptide class of HDACIs (1). Sodium butyrate and apicidin induced PSCS (although to a lesser extent than TSA) but valproic acid did not (Fig. 3A,, top); however, all HDACIs tested decreased the mitotic index of nocodazole-treated cells (Fig. 3A , bottom). This suggests that HDAC inhibition in SAC-activated mitotic cells causes abnormal mitosis exit.

Although gene transcription in metaphase-arrested cells is almost completely silenced (27), it is possible that HDACI induced aberrant mitotic transcription. To determine whether the TSA-induced PSCS was due to the reactivation of transcription at loci normally silenced during mitosis, nocodazole-treated cells were incubated with α-amanitin (at a concentration known to completely inhibit transcription by RNA polymerase II) prior to TSA treatment (28). This inhibition of transcription had no effect on TSA-induced PSCS (Fig. 3B,, top), or on the mitotic index (Fig. 3B , bottom). These results suggest that HDAC inhibition overrides the prometaphase block in mitotic SAC-activated cells in a transcription-independent manner.

Key targets of APC/C are degraded in a proteasome-dependent manner following TSA treatment. SAC activation inhibits the ubiquitin ligase complex APC/C. Activation of APC/C at the metaphase-anaphase transition leads to polyubiquitination of cyclin B1 and securin, such that they are targeted to a proteasome-dependent degradation pathway. These two proteins are believed to be the only essential APC/C targets (29). Under conditions of sustained SAC activation (i.e., nocodazole treatment), APC/C targets accumulate and reach a high stable concentration.

To determine whether APC/C activity was affected by TSA treatment, the stability of APC/C targets was assessed by Western blotting using anti-cyclin B1 and anti-securin antibodies as probes. Cyclin B1 accumulated, as expected, in nocodazole-treated cells (Fig. 4, lane 2). TSA treatment of nocodazole-arrested cells resulted in a sharp decrease in the abundance of cyclin B1 (lane 3). Pretreatment with the proteasome inhibitor MG132 prevented this TSA-induced decrease (lane 4). Similar results were obtained for securin. The amount of cyclin B1 or securin detected in TSA-treated cells was intermediate between those observed in untreated and nocodazole-treated cells. These data show that TSA treatment induces a proteasome-dependent degradation of major APC/C targets.

TSA treatment induces classical mitosis exit markers. Our data suggests that HDAC inhibition leads to SAC dysfunction. To confirm that prometaphase-arrested cells indeed exit mitosis in the presence of HDACI, we examined classical markers of mitosis exit including cdk1 inactivation, histone H3 serine 10 dephosphorylation, nuclear membrane reformation, and release of SAC components from the kinetochore.

To determine whether TSA-induced degradation of cyclin B1 led to a decrease of cdk1 activity, cdk1 was immunoprecipitated and its kinase activity was measured with histone H1 as a substrate. TSA treatment induced a substantial decrease in the cdk1 activity in nocodazole-treated cells, although the amounts of cdk1 protein recovered in the immunoprecipitates from treated and untreated samples were comparable (Fig. 5A). Coomassie staining of the gel showed that reduced H1 phosphorylation signal following TSA treatment was not due to reduced amounts of histone H1 recovered in the assay (Fig. 5A, “H1 Total”). Proteasome inhibition in nocodazole/TSA–treated cells did not rescue the cdk1 activity (Fig. 5A), although the addition of MG132 blocked TSA-induced cyclin B1 degradation (Fig. 4). This can be explained by the finding that cdk1 and polyubiquitinated cyclin B1 were dissociated by a nonproteolytic function of the proteasome (30). Thus, in the presence of MG132, polyubiquitinated cyclin B1 is not degraded and can be detected by immunoblotting, but is not associated with cdk1. Quantification experiments showed that cdk1 activity was three times higher in control nocodazole-treated cells than in nocodazole/TSA–treated cells (Fig. 5A , bottom).

Mitotic index measurements (Fig. 1C) showed that TSA treatment was associated with more nuclei-like figures displaying decondensed chromatin but were unable to indicate whether a nuclear envelope was reformed around the chromatin. To resolve this question, nocodazole/TSA–treated cells and control (nocodazole alone) cells were immunolabeled for lamin A/C (Fig. 5B). Most TSA-treated nuclei-like figures were positive for lamin A/C, indicating that a nuclear envelope had reformed in these cells.

SAC inhibition and subsequent anaphase entry are characterized by the release of SAC components from the kinetochores. We examined kinetochore localization of MAD2, a major downstream SAC effector, by immunofluorescence. Kinetochores were labeled with a human CREST autoimmune serum, which recognizes the kinetochore proteins CENP-A, CENP-B, and CENP-C, whereas MAD2 was detected with a specific antibody. Characteristic CREST-positive kinetochore dots were observed in both control and nocodazole/TSA–treated mitotic cells, indicating that TSA treatment does not lead to major kinetochore damage, although MAD2 was found to be displaced from kinetochores in approximately one third of nocodazole/TSA–treated mitotic cells (Fig. 6A). MAD2 antibodies failed to label chromosome spreads, so we could not determine whether the mitotic cells that had lost MAD2 from kinetochores were the same as those displaying PSCS. Nevertheless, the respective proportions of cells exhibiting these phenomena were similar enough to assume that this was indeed the case.

We also tested the loss of histone H3 serine 10 (H3S10) phosphorylation. During normal mitosis, H3S10 becomes phosphorylated in prophase and remains phosphorylated until early anaphase. In metaphase-arrested cells (SAC activated), this phosphorylation was maintained. Chromosome spreads from control (nocodazole alone) and nocodazole/TSA–treated cells were immunostained for H3S10 phosphorylation (Fig. 6B). All chromosome spreads from nocodazole-treated control cells were positive for H3S10 labeling. In contrast, in nocodazole/TSA–treated cells, chromosome spreads displaying a normal phenotype (i.e., attached sister chromatids) were labeled by the H3S10-specific antibody, whereas chromosome spreads displaying PSCS were negative for H3S10 labeling. Figure 6B shows one normal, H3S10-positive, spread and, within the same microscope field, a H3S10-negative spread displaying a TSA-induced PSCS (arrow). Loss of H3S10 phosphorylation following TSA treatment has been previously reported using immunoblot techniques (31). However, our chromosome spread immunostaining approach allowed us to show that cells which displayed reduced TSA-induced H3S10 phosphorylation were the same as those which displayed PSCS. This indicates that cells presenting a TSA-induced PSCS are engaged into exit from mitosis. Taken together, our findings show that HDAC inhibition renders the SAC ineffective.

Understanding the molecular mechanisms of the antiproliferative properties of HDACIs is a major issue both in terms of anticancer therapy and in our knowledge of basic cellular functions. We investigated the integrity of the mitotic SAC in cells treated with HDACIs and found that HDAC inhibition results in SAC being overridden.

Previous reports suggest a role for HDACs in SAC regulation but the various studies led to apparently contradictory conclusions. For example, it has been reported that HDAC inhibition in mitosis prevented sister chromatid centromere separation, suggesting that HDACIs trigger the SAC abnormally (15). In contrast, other authors reported that HDACIs induced a down-regulation of cyclin B1–dependent kinase activity (32), suggesting that HDACI treatment leads to checkpoint inhibition. Accordingly, Dowling et al. described premature mitosis exit, H3S10 dephosphorylation, and diminished BUBR1 kinetochore localization following TSA treatment (31). These results clearly plead in favor of the existence of a TSA-induced SAC defect. However, recent data show that a slow but continuous degradation of cyclin B induces an apparent mitosis exit in spite of the presence of a functional and active SAC (33). In this particular situation, other APC/C targets are not degraded and MAD2 and BUBR1 remain attached to kinetochores. With the noteworthy exception of BUBR1 partial loss from kinetochores, data reported in ref. (31) could be attributed to a TSA-induced cyclin B degradation alone. BUBR1 delocalization described in ref. (31) is incomplete, and thus, it is not clear whether the remaining kinetochore-bound BUBR1 was sufficient to maintain the SAC or not. Conflicting reports exist on the amount of BUBR1 required to activate the SAC. One study reported that 25% of the normal amount of BUBR1 led to SAC defects (34), although in another study, heterozygous mice expressing only 29% of the normal amount of BUBR1 showed no SAC malfunction (35). Therefore, although the data reported by Dowling et al. represent the strongest evidence of a TSA-induced SAC malfunction to date, one cannot formally exclude that TSA induced cyclin B depletion–like defects and diminished BUBR1 kinetochore localization without inactivating the SAC. The finding by Cimini et al. (15) that TSA treatment induces persistent sister chromatid cohesion in anaphase would support this hypothesis. Thus, the primary consequences of SAC inhibition being the activation of APC/C and the subsequent proteasome-dependent degradation of its only known essential targets: cyclin B and securin, it seems essential to examine sister chromatin cohesion and securin status in order to assess definitively whether or not SAC function is altered. Our results show that TSA treatment induces securin degradation and sister chromatid separation. Therefore, the abnormal persistent chromatid cohesion observed in ref. (15) following TSA treatment was more likely the consequence of a TSA-induced impaired chromatin topological organization (chromatid entanglement) rather than a failure to cleave cohesins.

We found that when nocodazole-treated cells (to induce SAC activation) were subsequently treated with TSA, 25% of the mitotic cells displayed a characteristic PSCS phenotype which resembles that observed when the SAC is inhibited by antibody injection (18), RNA interference (20, 21), or by genetic means (19). Sister chromatids that have separated prematurely are considered a hallmark of a defective checkpoint (19). We also showed that TSA treatment induced the proteasome-dependent degradation of cyclin B1 and securin, two essential APC/C targets. Finally, we monitored exit from mitosis by different and independent means including testing for loss of cdk1 activity, histone H3 serine 10 dephosphorylation, reformation of the nuclear envelope, and MAD2 release from kinetochores. All tests done indicated that TSA treatment induced exit from mitosis in SAC-activated cells. Thus, our results show that HDAC inhibition overrides the SAC.

In every experiment done, HDACI treatment induced PSCS in only part of the cell population. Immunostaining experiments in nocodazole/TSA–treated cells with an antibody specific for a hyperacetylated form of histone H3 showed that histone H3 was hyperacetylated in all cells, ruling out the possibility that TSA did not act on a particular cell subpopulation (data not shown). However, this “partial” effect is consistent with the finding that TSA treatment led to only partial proteasome degradation of APC/C key targets. Although the reason for this heterogeneous response is not clear, a similarly incomplete response has also been observed in other SAC inhibition studies. For example, MAD2 knockdown in HeLa and HaCat cells induces PCSC in ∼50% of the cells (20). Similarly, mutations of the gene encoding the SAC component BUBR1 in the mosaic variegated aneuploidy syndrome (36) leads to PSCS in only a fraction of the cells (37).

Not all HDACI were as effective in inducing PSCS. TSA was the best inducer, sodium butyrate and apicidin were intermediate, and valproic acid did not induce PSCS. However, all HDACI induced a similar decrease in the mitotic index in nocodazole-treated cells. This apparent discrepancy may be due to the differences in the spectra of action of these drugs. Although little is known about the potential for each drug to inhibit a given subset of HDACs, TSA is one of the most potent HDACIs and is believed to act on both class I and class II HDACs. Thus, we propose that at least two distinct HDACs are required in mitosis: one necessary to maintain the SAC and sensitive to all the drugs tested, and the other required for anaphase progression and sensitive to TSA but insensitive to valproic acid (and partially sensitive to the two other drugs tested). Under this hypothesis, nocodazole/TSA–treated cells escape SAC activation, but a proportion remains blocked in anaphase, allowing the observation of PSCS. Conversely, nocodazole/valproic acid–treated cells would exit mitosis without pausing in anaphase. In this case, PSCS would not be observed but the mitotic index would decrease, in accordance with our observations.

In addition to PSCS, TSA treatment led to partial decondensation of mitotic chromosomes. This finding was restricted to cells displaying PSCS suggesting that the observed decondensation is a consequence of mitosis exit rather than a direct effect of chromatin hyperacetylation. Similar chromosome morphologic changes have been observed in HeLa and HaCat cells in which the key SAC component MAD2 has been knocked down (20). Thus, this incomplete decondensation phenotype is likely to be a direct consequence of TSA-induced SAC overriding.

The precise nature of the antiproliferative mechanism(s) of HDACIs remains a subject of debate. Sustained SAC inhibition leads to the death of proliferative cells within a few generations (23) and our results show that HDAC inhibition results in SAC inhibition making HDACIs the only clinically validated pharmacologic SAC inhibitors known thus far. We propose that this SAC-inhibiting activity may contribute to the antiproliferative properties of HDACIs. Histone acetylation is involved in transcription, so it is plausible that HDAC inhibition leads to the reexpression of antiproliferative genes silenced in tumor cells. However, a number of nonhistone proteins are regulated by acetylation and the possibility of an HDACI antiproliferative mechanism independent from transcription has emerged (11–13). Our results show that HDACI-driven SAC inhibition is transcription-independent, and thus, raise the possibility that the targets of HDACIs in anticancer therapy could be, at least in part, nontranscriptional.

Grant support: Association pour la Recherche sur le Cancer (grant no. 4466), Ligue Contre le Cancer (Comités Départementaux de l'Hérault and de l'Aveyron and Comité Régional du Languedoc-Roussillon), and GEFLUC (Languedoc-Roussillon). G. Eot-Houllier is supported by a Ligue Nationale Contre le Cancer fellowship. G. Fulcrand is supported by a Ligue Régionale Contre le Cancer Research Ph.D. scholarship (Comité du Languedoc Roussillon). L. Magnaghi-Jaulin is an investigator of the Centre National de la Recherche Scientifique.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank C. Johanet and J. Piette for the gift of reagents and J.M. Vanacker, J.M. Brondello, and J. Piette for discussions and critical reading of the manuscript.