HDAC2/3 binding and deacetylation of BubR1 initiates spindle assembly checkpoint silencing

BubR1 acetylation is essential in spindle assembly checkpoint (SAC) signaling. Here we show that BubR1 deacetylation is a signal that initiates mitotic exit. Sustained BubR1 acetylation arrests the cells in metaphase, although chromosome congression is achieved. BubR1 deacetylation was coordinated with dephosphorylation in mitotic exit, suggesting the presence of a coordinated acetylation–phosphorylation code in mitotic signaling. Histone deacetylase (HDAC) 2 and 3 bound to acetylated BubR1 exclusively in mitosis and led to the polyubiquitination of BubR1. Subsequent degradation of BubR1 resulted in the disassembly of the mitotic checkpoint complex. Importantly, BRCA2 was required for HDAC2/3 association with acetylated BubR1 in nocodazole (Noc)‐arrested cells. Plk1, PP2A, P300/CBP‐associated factor (PCAF) and BubR1 were found in the mitotic BRCA2 complex, suggesting that BRCA2 acts as a signaling scaffold for BubR1 modification. Furthermore, we show that Plk1 is required for BRCA2 to localize at the prometaphase kinetochore (KT). Inhibition of Plk1 resulted in the loss of BRCA2 from the KT, and so did PCAF, consistent with the loss of BubR1 acetylation. Concordantly, BRCA2‐dysfunctional cells exhibited resistance to trichostatin A, which was restored when BRCA2 was introduced. That loss of Brca2 conferred resistance to various HDAC inhibitors was corroborated by the experiments in mouse pancreatic organoids. These results suggest that the BRCA2–BubR1 acetylation–deacetylation pathway is an important decision‐making point for the HDAC inhibitor response. Taken together, BRCA2 is a signaling platform for BubR1, and BubR1 deacetylation is a cue for SAC silencing.


Introduction
Individuals with germ-line mutations of the breast cancer susceptibility gene BRCA2 are predisposed to early-onset cancer. BRCA2-deficient cells exhibit genetic instability [1] and chromosome instability as well [2], manifested by the accumulation of DNA lesions, chromosome breaks, radial chromosomes, translocations and aneuploidy [3,4]. It has been revealed that BRCA2 is a multifaceted protein, functioning throughout the cell cycle: it is involved in homology-directed repair, stalled replication fork protection, telomere homeostasis and mitotic fidelity [4].
BRCA2-deficient cells display chromosome missegregation, which can be explained by the role of BRCA2 in fine-tuning the spindle assembly checkpoint (SAC). It recruits P300/CBP-associated factor (PCAF), an acetyltransferase that acetylates BubR1 in mitosis at a single lysine (K250), which is adjacent to D-box and KEN motif degrons [5]. This activity is critical in modulating anaphase-promoting complex/cyclosome (APC/ C) activity in mitosis [6]. Disruption of the interaction between endogenous BRCA2 and BubR1 in mice by the ectopic expression of the BubR1-binding domain of BRCA2 results in tumorigenesis without an apparent homology-directed repair defect [5], suggesting that chromosome missegregation can be oncogenic.
Concordantly, mice heterozygous for the acetylationdeficient BubR1 allele (replacement of the lysine 243 (K250 in humans) with arginine) exhibit a similar level and spectrum of tumorigenesis, when compared with transgenic mice with a disrupted endogenous BRCA2-BubR1 interaction [5], without additional genotoxic insults [7]. The results from the two different mouse models that interfered with K250 acetylation of BubR1 confirmed that BubR1 acetylation represents one axis of BRCA2's tumor suppressor mechanism. Moreover, further analysis of the mice heterozygous for BubR1 acetylation (K243R/+) taught us that BubR1 acetylation has a dual role: maintenance of the SAC and stabilization of chromosome-microtubule (MT) attachment [7].
Any chromosome unattached to the spindle in prometaphase generates a 'wait anaphase signal', which activates the SAC, which ultimately inhibits APC/C [8]. The SAC signal is initiated by Mad1 and Mad2 localization to the unattached kinetochore (KT). Mad2 undergoes a conformational change from open Mad2 to closed Mad2 and then forms a complex with BubR1, Bub3 and Cdc20, constituting the mitotic checkpoint complex (MCC) [9,10]. Sustained MCC binding to APC/C inhibits the metaphase-anaphase transition, and thus maintains the SAC.
Silencing of the SAC is the start of chromosome segregation, as APC/C becomes active, ubiquitinating cyclin B and securin for destruction. Compared with the mechanism of SAC signal generation and MCC formation, the signal that initiates SAC silencing is less well defined. As premature APC/C activation will lead to anaphase onset with unattached chromosomes, the mechanism that controls SAC silencing is essential for chromosome integrity. One important mechanism that is involved in exiting from mitosis is the role of TRIP13 [11][12][13] and p31 comet . TRIP13 and p31 comet are involved in disassembling the MCC by extracting Mad2 from the MCC, and thus are essential for SAC silencing [14,15].
In addition to the role of TRIP13 and p31 comet in extracting Mad2 for SAC inactivation, in vitro biochemical study showed that destruction of BubR1 silences SAC [16]. Importantly, the MCC was not maintained in mouse embryonic fibroblasts from K243R/+ mice because acetylationdeficient BubR1 was readily recognized by APC/C and destroyed in mitosis, resulting in the premature onset of anaphase with unequal chromosome segregation [7]. This result suggested that deacetylation of BubR1, which leads to the destabilization of BubR1, may be a cue for mitotic exit.
Here, we asked if specific deacetylases were involved in deacetylation of BubR1 in mitosis and silencing of the SAC. Then we asked whether BRCA2 was involved and what made BRCA2 function in mitosis. Finally, we asked whether the results have clinical implications in treating BRCA2-deficient cancers.

Results
Sustained acetylation of BubR1 results in prolonged chromosome congression, subsequent to SAC activation We asked if deacetylation of BubR1 could be a cue for SAC silencing. For this, mitotic progression in The statistical timing of NEBD to anaphase onset upon expression of various forms of BubR1. Endogenous BubR1 was depleted with siRNA (siBubR1) that targets 3 0 -UTR and mCherry-tagged BubR1 plasmids were transfected into HeLa cells, stably expressing H2B-GFP [6]. Cells were synchronized by thymidine block and then released for 6 h. Reversine (0.5 lM)-treated cases are marked. Cells were filmed with time-lapse microscopy in 5 min intervals for a total of 24 h (Movie S1, siControl; Movie S2, siBubR1; Movie S3, WT-BubR1; Movie S4, K250R-BubR1; Movie S5, K250Q-). Each dot represents a single cell event, and the bars mark the mean values (mean AE SEM; n > 50 cells in all cases). SPSS software was used and P values were obtained with Student's t test (A-C). Right, WB analysis of endogenous and mCherry-tagged BubR1. Anti-Mad2 and anti-actin WB analyses show that the siBubR1 has little off-target effect. (B) Representation of the timing from metaphase (congressed chromosomes) to anaphase onset (mean AE SEM; n > 50 cells in all cases). (C) The statistical timing of NEBD to anaphase onset upon expression of KARD3A mutant form of BubR1 (mean AE SEM; WT, n = 10; KARD3A, n = 28; P < 0.0001). (D) Captured images of representative cells expressing mCherry-tagged WT-BubR1 (Movie S3), K250R-(Movie S4), or K250Q-(Movie S5), KARD3A-(Movie S6), depleted of endogenous BubR1. Images are shown from NEBD to anaphase onset. Each frame is at a 5 min interval. The duration of metaphase (B) is marked with a bold line and anaphase onset is marked with a red arrow. (E) The effect of BubR1 acetylation status in chromosome congression is assessed by centromere alignment, marked with CENPB. mCherry-tagged CENP-B [52] was transfected into HeLa-FRT-BubR1 (Movies S7 and S10), -K250R (Movies S8 and S11) and -K250Q (Movies S9 and S12)-inducible cells. Expression of BubR1, K250R and K250Q was induced by doxycycline treatment. Cells were subjected to time-lapse video microscopy with 3 min intervals. The beginning of chromosome separation is marked with red arrows. Time 0, NEBD. Red, mCherry-CENP-B; Green, BubR1. Scale bar, 10 lm.  E   00  03  06  09  15  18  21  24  30  27  33  36  42  39  12   00  03  06  09  15  18  21  24  30  27  33  12  36   00  03  06  09  15  18  21  24  30  27  33  36  42  39  12  45   HeLa cells expressing wild-type, acetylation-deficient (K250R) and acetylation mimetic forms (K250Q) of BubR1 were compared and analyzed. To assess the effect of the various forms of ectopically expressed BubR1, endogenous BubR1 expression was silenced by siRNA that targets the 3 0 -UTR, and mCherry-tagged BubR1 expression plasmids were cotransfected. BubR1 acetylation deficiency results in premature degradation in mitosis through APC/C-Cdc20mediated ubiquitination [6]. When the mitotic timing was measured from nuclear envelope breakdown (NEBD) to anaphase onset, we corroborated that K250R expression leads to shortened mitosis and K250Q expression results in prolonged mitosis (Fig. 1A,D, see the time point marked with an arrow, and Movies S4 and S5). Interestingly, when the Mps1 inhibitor reversine [17] was added, the delay in mitosis from the expression of K250Q was eliminated ( Fig. 1A), indicating that BubR1 acetylation is not just a blockade of ubiquitination and degradation but is a cue that is coordinated with Mps1 kinase activity in SAC signaling (Fig. 1A, +Reversine).
Importantly, K250Q-expressing cells stayed in metaphase with congressed chromosomes for 45 min longer than WT, and 55 min longer than K250R-expressing cells (Fig. 1B,D, see bold lines, and Movies S3-S5). It should be noted that depending on the expression level of WT or K250Q, the timing in metaphase was affected in individual cells. Nevertheless, the statistical analysis of more than 60 cells (Fig. 1B) confirmed that K250Q expression achieves and maintains chromosome congression, even when the checkpoint was maintained [7]. The congression of chromosomes in K250Q-expressing cells was sustained while being delayed in metaphase ( Fig. 1D and Movie S5). This phenomenon may be the consequence of stabilized KT-MT attachment by the maintenance of acetylated BubR1 [7]. Consistent with this, expression of mutant BubR1 defective in phosphorylation at the KARD domain, which is phosphorylated by Plk1 and recruits PP2A-B56a to counteract Aurora B activity in KT-MT attachment [18], resulted in delayed mitosis ( Fig. 1C) with problems in chromosome congression (Fig. 1D, BubR1-KARD3A). These data are consistent with previous findings [19].
Expression of K250Q leads to sustained chromosome congression, yet chromosomes do not segregate. To corroborate that the congression is achieved upon K250Q expression, the movement of CENP-B, the centromere-binding protein located beneath the KT [20], was monitored. mCherry-tagged CENP-B was cotransfected in HeLa cells stably expressing Venus-tagged BubR1 (Movie S7), K250R (Movie S8) and K250Q (Movie S9), and the KT alignment was monitored. CENP-B alignment in metaphase was prolonged in K250Q-expressing cells ( Fig. 1E and Movie S12), compared with WT-BubR1 expressers ( Fig. 1E and Movie S10). In contrast, CENP-B in K250R-expressing cells hardly aligned ( Fig. 1E and Movie S11). Interestingly, BubR1 at the KT diminished before the completion of metaphase, as assessed by CENP-B segregation (Fig. 1E, Venus). The fluorescence intensity of the acetylation-mimetic form (K250Q, Movie S9) was stronger and remained aligned at the KT longer than that of the WT. K250R was never aligned and disappeared from the KT faster ( Fig. 1E and Movie S8) than the WT ( Fig. 1E and Movie S7). In this experimental setting, expression of K250Q eventually led to degradation of BubR1 from the KT and exit from mitosis ( Fig. 1E and Movie S9).
The effect of K250Q expression is similar to the cells lacking p31 comet , which functions in SAC silencing by titrating out the MCC component Mad2 [14,15,21], or APC15, which is required for rapid turnover of APC coactivator Cdc20 [22]. Taken together with the information from mice heterozygous for the acetylation-defective BubR1 allele (K243R/+) [7], BubR1 deacetylation may be another mechanism for SAC silencing that leads to MCC disassembly.

BubR1 deacetylation by HDAC2/3 leads to the ubiquitination of BubR1 and mitotic exit
Treatment of mitotic cells with hesperadin, the inhibitor for Aurora B [23], or the Mps1 inhibitor reversine [17] results in mitotic exit. To test the hypothesis that the deacetylation of BubR1 may be a cue for mitotic exit, we asked whether treatment with the inhibitor for Aurora B or Mps1 affected BubR1 acetylation. HeLa cells were synchronized in prometaphase by single thymidine block and release, followed by 10 lM paclitaxel treatment. Mitotic cells were enriched by mechanical shakeoff ( Fig. 2A, 0 h) and then released to progress into cell division by changing to fresh medium. Cell lysates for western blot (WB) were collected at 0, 1.5 and 3 h, respectively. WB analysis with an antibody specific to acetylated BubR1 (anti-AcK250) [6] confirmed that BubR1 in cells arrested in mitosis was acetylated as expected (Tax, 0 h; see the level of phospho-histone H3 (ph3), compared with attached cells (Att)].
Aurora B inhibition (with hesperadin) or Mps1 inhibition (with reversine) after paclitaxel treatment led to the increase of dephosphorylated BubR1 ( Fig. 2A, second row, BubR1). In these cases, BubR1 acetylation markedly disappeared ( Fig  reduced ( Fig. 2A, ph3) upon hesperadin or reversine treatment, confirming that cells are exiting from mitosis. Treatment with 10 lM MG132 resulted in the maintenance of BubR1 acetylation ( Fig. 2A, MG132). Cyclin B level persisted upon MG132 treatment ( Fig. 2A, cyclin B), indicating that the cells are arrested in metaphase. When hesperadin or reversine was added along with MG132, acetylation of BubR1 was prolonged up to 1.5 h, but markedly diminished at 3 h from release. Meanwhile, WB to anti-BubR1 showed that dephosphorylated BubR1 increased. Owing to MG132 treatment, the total levels of BubR1, cyclin B and Cdc20 are maintained [22,24], indicating that the cells are still in mitosis, regardless of hesperadin or reversine treatment ( Fig. 2A, BubR1, cyclin B, Cdc20). Inhibition of Aurora B or Mps1 kinase results in the shutdown of KT signaling and exit from mitosis [22,25]. Therefore, the result suggests that BubR1 deacetylation is linked with dephosphorylation events and may function as a cue to exit mitosis.
Previous in vitro transfection experiments showed that class I HDACs bind to BubR1 [6]. However, the mitosis-specific association of deacetylases was not tested until now. We asked if class I HDACs bound to acetylated BubR1 in a mitosis-specific manner. For this, HeLa cells were treated or untreated with Noc and subjected to co-immunoprecipitation (co-IP) and WB. The result showed that HDAC2 and/or HDAC3 was capable of binding to BubR1 in Noc-treated mitotic cells, but not in untreated interphase cells. Interestingly, when reversine was added to MG132-treated mitotic cells, the interaction between BubR1 and HDAC2/3 decreased, suggesting that HDAC2/3 binds to the acetylated BubR1 in mitosis. Note that the acetylated BubR1 is also decreased (Fig. 2B). A similar test was done for SIRT2, as reports had suggested that SIRT2, an NAD + -dependent deacetylase, is implicated in control of cell division [26,27]. Also, SIRT2 was suggested to bind to BubR1 [28,29]. However, we did not observe a mitosis-specific binding between SIRT2 and BubR1 (data not shown). It is possible that SIRT2 binding to BubR1 is independent of mitosis: it may be involved in interphase, regulating aging [28,30]. SIRT2 deacetylation of lysine 668 is involved in aging [30], which therefore supports the notion that the outcomes of acetylation and deacetylation of these two sites are quite distinct [31]. However, we do not rule out the possibility that SIRT2 might be involved, because a report showed that SIRT2 also deacetylates K250 [29]. HDAC1 may function similarly in different cell types, as class I HDACs were all found to interact with BubR1 in a transfection experiment [6].
Then we asked if HDAC2 or HDAC3 association with BubR1 resulted in the ubiquitination of BubR1. Deacetylation (or the unacetylated form) of BubR1 results in APC/C-mediated ubiquitination and degradation [6,7]. Increasing amounts of FLAG-tagged HDAC2 or HDAC3 expression vectors were transfected, and the mitotic cell lysates were subjected to IP followed by WB. The result showed that the expression of HDAC2 or HDAC3 resulted in a marked increase of polyubiquitination in a dose-dependent manner (Fig. 2C). In the input control, the blot showed that the level of acetylated BubR1 decreased with the increasing amount of HDAC2/3 (Fig. 2C,  Ack250). Interestingly, the total BubR1 level increased in the same blot, indicating that the decrease of acetylated BubR1 is not merely due to protein degradation. Also, the result suggests that there may be a feedback mechanism to replenish BubR1 when BubR1 is continuously deacetylated and degraded to balance the checkpoint activation and silencing. When HDAC2 or HDAC3 is depleted in Noc-treated cells, the acetylated BubR1 level (AcK250) persisted, compared with control (Fig. 2D), corroborating that HDAC2/3 functions in mitosis to deacetylate BubR1. These results confirm that HDAC2 and HDAC3 (class I HDACs) bind to acetylated BubR1 and deacetylate it, leading to polyubiquitination and subsequent degradation.
BRCA2 serves as a mitotic platform for BubR1 signaling BRCA2 fine-tunes the SAC by acting as a scaffold for BubR1 and acetyl-transferase PCAF association at the KT [5]. Therefore, we asked if BRCA2 was required for the association of HDACs with BubR1 as well. BRCA2 depletion led to a marked decrease of acetylated BubR1 (AcK250) in Noc-treated mitotic cells (Fig. 3A). As BubR1 is acetylated at K250 exclusively in mitosis, AcK250 was barely detected, and hence the depletion of BRCA2 had little effect in asynchronously growing cells (Fig. 3A, left panel). In IP and WB, the association of BubR1 with HDAC2 or HDAC3 decreased in mitosis. The BubR1 association with HDAC2/3 was barely observed in asynchronous cells (Asn), and thus there was no effect of BRCA2 in interphase (Fig. 3A, right panel). IP with monoclonal antibody specific to AcK250 (mAcBubR1) and subsequent WB confirmed that BRCA2 depletion in Noc-arrested cells interfered with HDAC2/3 association with acetylated BubR1 (Fig. 2B).
When we examined the BRCA2 immune complex, we found that BRCA2 was associated with BubR1, Plk1 and PP2A only in mitotic cells (Fig. 3C, Noc). In comparison, PCAF association with BRCA2 was found both in mitotic and non-mitotic cells (Fig. 3C).
The previous report showed that PCAF binds to BRCA2 in interphase as well [32], which is consistent with the data shown here. However, it should be noted that the outcome of PCAF binding to BubR1 in interphase is quite different from mitotic binding. Acetylated BubR1 recruits PP2A-B56a and antagonizes Aurora B activity on Ndc80, which stabilizes KT-MT attachments for anaphase onset [7]. The finding that the PP2A catalytic subunit and Plk1 were found in the BRCA2 immune complex, together with BubR1, suggests that all of these proteins may be found in a complex on the platform of BRCA2. We do not rule out the possibility that the proteins bind to mitotic BRCA2 independently. In this regard, this hypothesis needs further investigation.
Plk1 is required for BRCA2 to act as a mitotic scaffold for BubR1 acetylation Entry into mitosis is regulated by timely controlled phosphorylation events by M phase kinases. As Plk1 phosphorylates BRCA2 at multiple sites [32,33], and because Plk1 also phosphorylates BubR1 and stabilizes KT-MT attachment [18,34], we tested whether Plk1 activity is required for BRCA2 to function in mitosis.
Similarly, treatment of cells with BI 2536, a small molecule inhibitor specific for Plk1 [35][36][37], also failed to pull down BubR1 and Plk1 from BRCA2 IP (Fig. 4C, BI). Accordingly, BubR1 acetylation was lost upon BI 2536 treatment (Fig. 4B,C, anti-AcK250 in input control). Note that the BRCA2 immune complex from Noc-treated cells consistently involves BubR1 and Plk1 (Fig. 4C, Noc). Plk1 and BubR1 binding in mitosis is well established [18,19,38], and BRCA2 and  Twenty-four hours later, Asn or mitotic cells with Noc were harvested. Lysates were then subjected to IP with anti-HDAC2 and anti-HDAC3, followed by WB with antibodies indicated. (B) Noc-treated cells were subjected to IP with monoclonal antibody to Ack250 (mAcBubR1), followed by WB with antibodies indicated. (C) Mitotic HeLa cells were collected by Noc treatment and shake-off. Asynchronously growing HeLa cells (Asn) and Noc-treated cells were subjected to IP with anti-BRCA2 antibody, followed by WB with antibodies indicated. IP with rabbit IgG was included for negative control. Three percent of the total cell lysates were employed for input control. Plk1 association is known, albeit in interphase cells [4,32,33]. The data shown here indicate a mitosisspecific binding of Plk1 and BRCA2. Therefore, the results altogether suggest the possibility that Plk1-BRCA2-BubR1 may form a prometaphase-specific trimolecular signaling cascade. It could be argued that inhibition of Plk1 interferes with entry into mitosis, and therefore the loss of complex formation between BRCA2 and BubR1. Also, it needed to be assessed whether Plk1 affected BRCA2-BubR1 binding at the prometaphase KT. Cells were first Noc arrested, then challenged with BI 2536 for 1 and 4 h, respectively. In IP, followed by WB, inhibition of Plk1 kinase activity led to loss of both Plk1 and BubR1 from the BRCA2 complex (Fig. 4D). Simultaneously, BubR1 acetylation was markedly reduced (Fig. 4D, input, AcK250).

Nocodazole s i C o n t r o l s i B R C A 2 s i C o n t r o l s i B R C A 2 Asn Noc s i C o n t r o l s i B R C A 2 s i C o n t r o l s i B R C
Next, the localization of BRCA2 to KTs was assessed. An immunofluorescence assay coupled with metaphase chromosome spread showed that BI 2536 treatment abolished BRCA2 at the KT (Fig. 4E, BI 2536), whereas Noc-treated control cells displayed colocalization of BRCA2 and BubR1 at the KT (Fig. 4E, Noc), as reported [5]. Cotreatment of BI 2536 with Noc (Noc + BI 2536) also abolished BRCA2 at the KT. These data suggest that Plk1 is required for BRCA2 to localize at the KT. This also showed that Plk1 inhibition interfered with the BRCA2 association with BubR1 at outer KTs, whereas BubR1 localization was not affected (Fig. 4E,  F). Taken together, we postulate that the mitotic function of BRCA2 at the KT requires Plk1.

BRCA2-deficient cells are resistant to HDAC inhibitors
The observation that BRCA2 serves as a platform for modulating SAC signaling, particularly on BubR1, led us to think that BRCA2 status might affect the sensitivity to histone deacetylase (HDAC) inhibitors. To test this idea, cancer cell lines that possessed intact BRCA2 or with BRCA2 absent (Fig. 5A) were subjected to a viability test, based on the 3-(4, 5-dimethyl-thiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT) assay after treatment with pan-HDAC inhibitor (HDACi) trichostatin A (TSA) for 2 or 3 days. Capan-1, which is truncated with BRCA2 [39], exhibited far less sensitivity to TSA, compared with BRCA2-positive cell lines, HeLa, GM847, U2OS, and WI-38 (Fig. 5B,C). Meanwhile, Capan-1 displayed similar level of sensitivity to Plk1 inhibitor BI 2536, when compared with BRCA2-positive cells (Fig. 5D). These results suggest that the absence of BRCA2 imposed on the cells reduced sensitivity towards HDACi.
To corroborate that BRCA2 status specifies the sensitivity to HDACi, BRCA2 was depleted from HeLa or WI-38 cells by transfection with siRNA. The result showed that the absence of BRCA2 significantly reduced the sensitivity to TSA, both in HeLa and WI-38 cells (Fig. 5E). Concordantly, ectopic expression of BRCA2 in Capan-1 cells restored the sensitivity to TSA, comparable to HeLa cells (Fig. 5F,G). Taken together, these results suggest that the status of BRCA2, intact or dysfunctional, is an important factor in determining the sensitivity to HDACi. We suggest that this is because BRCA2 serves as the scaffold for mitotic signaling on BubR1 acetylation. As BRCA2 works downstream of Plk1, BRCA2 deficiency did not affect the sensitivity towards Plk1 inhibitor.
As the result from the cancer cell lines could be due to the different genetic backgrounds, we decided to utilize mouse pancreatic organoids to confirm that the Brca2 status is related to the response to HDACi.
Organoids are self-assembled three-dimensional epithelial structures that are allowed to expand long term [40]. As dysfunctional BRCA2 is associated with pancreatic cancer development [41], we decided to culture and test the drugs in pancreatic organoids. Pancreatic organoids are regarded as stem cell cultures of the ductal epithelial cells [42]. We generated mouse organoids from conditional Brca2 knockout mice (Brca2 F11/F11 ) [5,43] that were bred with endoplasmic reticulum (ER)-Cre-expressing transgenic mice (CAGGC-Cre). In this mouse, Brca2 exon 11 can be deleted in a tamoxifen-inducible way. Mouse pancreas organoids were cultured with or without the treatment of 4-hydroxytamoxifen (4-OHT). To expand our scope of applicability in targeting the BRCA2-BubR1 complex with HDACis, we used different types of HDACis (Fig 6A,B). The result showed that Brca2 mutant mouse organoids were resistant to pan-HDACis TSA and 17 lM suberoylanilide hydroxamic acid (SAHA) [44]. Unexpectedly, Brca2-deleted organoids were resistant to LMK-235, the HDACi reported to have effects on class II HDACs [45]. Control organoids were resistant to the treatment of depsipeptide FK-228 (10 nM), which has epigenetic effects and is a class I HDACi [46], while Brca2-depleted organoids responded mildly (Fig. 6A). The efficiency of the depletion of Brca2 in organoids after 4-OHT treatment was assessed by genomic PCR (Fig. 6B). The reason why Brca2deficient organoids are resistant to class II HDACi, and less to the class I type, is not fully understood. It could be because HDACi usually have pleiotropic effects as they are also involved in epigenetic modulation. Nevertheless, Brca2-deficient cells and organoids are resistant to HDACi.

Discussion
We showed that HDAC2/3 binds to acetylated BubR1 and leads to ubiquitination and degradation of BubR1. Acetylation-deficient BubR1, K250R, prematurely disappears from the KT, accompanied by chromosome missegregation.
Meanwhile, K250Q expression leads to sustained BubR1 at the KT with prolonged metaphase. Therefore, HDAC2/3 binding and deacetylation of BubR1 at K250 lead to the disassembly of the MCC, which is critical in SAC silencing and mitotic exit (Fig. 6C). SIRT2 deacetylase was also shown to bind and deacetylate K250 of BubR1 in vitro [29], but failed to show that SIRT2 expression led to BubR1 ubiquitination. As SIRT2 also acts on the C terminus of BubR1 and functions in controlling aging [30], it is possible that SIRT2 preferentially functions in interphase, and only partially in mitosis. Nevertheless, with compelling lines of evidence, we suggest that HDAC2/3 binding and deacetylation of BubR1 is a signal for SAC silencing and mitotic exit. For BubR1 acetylation and deacetylation, BRCA2 is required. Furthermore, Plk1 is required for BRCA2 to localize at prometaphase KTs and bind to BubR1. Based on the presented data, it is reasonable to think that BRCA2-deficient cells will be relatively resistant to HDACi, while response to Plk1 inhibitor will not be affected. Indeed, that was the case. We have shown this in cultured cell lines and syngenic mouse pancreas organoids (Figs 5 and 6A).
BubR1 links KT-MT attachment to SAC activity [47], and Plk1 phosphorylation on the KARD domain at the C terminus of BubR1 is critical for the stabilization of KT-MT attachment [18]. When the KARD domain is phosphorylated by Plk1, BubR1 can recruit PP2A-B56a to the KT, which now antagonizes Aurora B kinase activity. As Aurora B destabilizes KT-MT attachment by phosphorylating the Knl1-Mis12-Ndc80 network, which forms the platform for MT attachment at the KT, Aurora B's kinase activity on Knl1-Mis12-Ndc80 is essential for error correction [48,49]. Now when amphitelic attachment is achieved, Plk1  phosphorylates BubR1, recruits phosphatase activity and antagonizes Aurora B activity, stabilizing MT attachment to the KT (Fig. 6C, illustrated in regular line arrow).
Interestingly, mouse cells deficient in BubR1 acetylation (K243R/+) exhibited failure in the stabilization of KT-MT attachment, accompanied by elevated Ndc80 phosphorylation and failure in recruiting PP2A-B56a [7]. The result had suggested that BubR1 acetylation at the N terminus cross-talks with the C terminus phosphorylation. This idea that BubR1 acetylation cross-talks with phosphorylation was tested and proven in this study. BubR1 is deacetylated when cells exit from mitosis ( Fig. 2A). When Aurora B or Mps1 inhibitor is treated with MG132, BubR1 is deacetylated, although the cells are in mitosis with sustained cyclin B level ( Fig. 2A). From these results, we suggest an interesting module in SAC signaling in which there may be a BubR1 phosphorylation-acetylation code, including deacetylation and dephosphorylation, which is also illustrated in the model (Fig. 6C).
It is interesting to note that the BRCA2 immune complex includes Plk1, BubR1, PCAF and PP2A (Fig. 3C). How the complex is formed has not been determined yet: whether they are in a complex together or undergo a dynamic change. Nevertheless, BubR1-HDAC2/3 binding in mitosis was lost without BRCA2. Therefore, we postulate that BRCA2 may be a signaling scaffold that brings acetylase, deacetylase and phosphatase to BubR1. We suggest that BRCA2's participation in mitosis reinforces BubR1's function in mitosis, the stabilization of KT-MT attachment and SAC maintenance (Fig. 6C, bold arrow). Furthermore, BubR1 deacetylation is a cue for MCC disassembly. In prometaphase, Plk1 may act in both ways: directly phosphorylating BubR1 (Fig. 6C, regular arrows on left) and enabling BRCA2 to function in mitosis that reinforces BubR1 activity (Fig. 6C, bold arrow). Thus, Plk1-BRCA2-BubR1 forms a signaling triangle, ultimately leading to proper chromosome segregation (Fig. 6C).
Cancer cells are characterized by unlimited cell growth, and therefore the control of mitosis was suggested to be effective for cancer treatment. However, clinical trials of mitotic kinase inhibitors, including Plk1 inhibitors are controversial. On the other hand, HDACi control cell division and fate but because they control epigenetic gene expressions at the global level, HDACi have pleiotropic effects [50]. It should be noted that HDACi target non-histone proteins as well, as shown here with BubR1. Interestingly, HDACi treatment often leads to mitotic arrest [51], suggesting that BubR1 may have been an important target. Various HDACi were developed and tested, but due to their pleiotropic activities on tumor cells and invading immune cells, clinical application is complicated. In this study, we showed that BRCA2-deficient cells exhibit reduced sensitivity to HDACi. Taking this information into account, assessing the status of BRCA2 will be beneficial in HDACi application. Collectively, the results shown here have many implications in precision medicine.

Materials and methods
Statistical analysis SPSS 22 Statistics software (IBM corp., Armonk, NY, USA) was used for all statistical analysis. Student's t test was used. The mean AE SEM is shown wherever required.

Constructs, antibodies and siRNA
Various BubR1 constructs were generated by PCR and subcloned into pcDNA3.1-mCherry for expression in cultured cells. K250R and K250Q BubR1 mutants were generated by site-directed mutagenesis using pcDNA3.1-mCherry BubR1 as the template.
The following antibodies were purchased: anti-mCherry

Microscope image acquisition, live imaging and processing
Microscopy images of fixed cells were acquired with a Del-taVision system (Applied Precision/ GE Healthcare, Issaquah, WA, USA), equipped with a 1009 objective lens (Olympus, Tokyo, Japan). The images were obtained with 0.2 lm distanced optical sections on the z-axis. Each section was deconvoluted and projected into one image using the SOFTWORX software (Applied Precision, Issaquah, WA, USA) as previously described [6,7]. For live cell imaging, cells were monitored using the UPlanFLN 409/NA 1.30 oil lens on a DeltaVision microscope equipped with a chargecoupled device camera (Photometrics, Tuscon, AZ, USA) in a CO 2 chamber at 37°C (Applied Precision). The cells were seeded in a glass-bottom dish containing culture media and images were acquired 3 or 5 min intervals as two sections with 7-lm z steps using 409, 1.35 NA 0.10 mm WD objective lens and maximally projected.

Immunoprecipitation and western blot
Cells were lysed in NETN buffer (150 mM NaCl, 1 mM EDTA, 20 mM Tris, pH 8.0, and 0.5% NP-40) supplemented with protease inhibitors for IP and WB analysis. Cells were incubated on ice for 20 min and supernatants were collected after centrifugation at 4°C, 290 g. Cell lysates were pre-cleared for 1 h at 4°C with protein A/G beads and subjected to IP with indicated antibodies at 4°C overnight. Immune complexes were precipitated by incubating with Protein A/G beads for 2 h. Before elution with 39 SDS sample buffer, immunoprecipitated beads were washed three times with the lysis buffer.

Immunofluorescence on metaphase chromosome spreads
For chromosome spreads coupled with immunofluorescence assays, cells were incubated with 200 ngÁmL À1 Noc for 4 h. To harvest the cells, medium was removed and cells were washed with PBS, then pelleted. Cells were swelled in 0.2% KCl and 0.2% trisodium carbonate solution and cytospun at 145 g for 5 min. The slides were subjected to immunofluorescence assay as described [6].

Cellular viability
Cells were seeded into 96-well plates (3 9 10 3 cells per well) and incubated overnight. The following day, cells were treated with increasing concentration of TSA (T5882; Sigma-Aldrich) or BI 2536 (S1109; Selleckchem) and incubated for 2 or 3 days. Cell viability was evaluated by MTT (Sigma-Aldrich) assay. The amount of dye formed was quantified using an ELISA plate reader at 540 nm.

Mouse pancreas organoid culture and HDACi treatment
Mouse protocol and animal experiments were approved by the Institutional Animal Care and Use of Committee (IACUC) of Seoul National University (SNU-130219-4-9). Mice pancreas organoids were isolated and cultured from conditional Brca2 knockout mice, Brca2 F11/11 [5,43], that was crossed with ER-Cre mice (CAGGCre-ER, The Jackson Laboratory, Bar Harbor, ME, USA). Mouse pancreas was lysed in digestion solution containing collagenase and DNase I and ductal cells were collected as described previously [40]. Collected ductal cells were washed with cold PBS and seeded onto Matrigel matrix (Corning, NY, USA). Cells were supplemented with culture medium containing stem cell growth factors as described previously [40]. When cells started to form organoids, they were treated with 400 lM 4-OHT (Sigma-Aldrich) to knock-out the BRCA2 gene. Mouse organoids were re-plated onto 24-well plates and were treated with varying doses of TSA (Sigma-Aldrich), LMK-235 (Selleckchem), FK-228 (Selleckchem) and SAHA (Sigma-Aldrich). Organoids were imaged with a Zeiss Axio Observer (Carl Zeiss AG, Oberkochen, Germany) after 5 days of drug treatment. recipient of the SNU PhD fellowship and HK is a POSCO Science fellow.