The small ubiquitin-related modifier (SUMO) system of higher eukaryotes plays important roles in normal cell division, especially in chromosome segregation. However, only a few mitotic SUMO substrates have been identified in mammals. Here, we show that the mitotic kinase Aurora-B can be modified by SUMO. The E3 SUMO-protein ligase PIAS3 [protein inhibitor of activated STAT (signal transducer and activator of transcription)] dramatically enhanced poly-SUMO-2/3 conjugation of Aurora-B, whereas the SUMO-specific isopeptidase SENP2 (Sentrin/SUMO-specific protease) specifically deconjugated SUMO from Aurora-B. The Lys-202 residue on human Aurora-B was preferentially modified by SUMO, and enhancement of SUMOylation in cells facilitated Aurora-B autophosphorylation, which is essential for its activation. Conversely, SENP2-mediated deSUMOylation of Aurora-B down-regulated its autophosphorylation in cells and also impaired its re-activation in Aurora inhibitor VX-680-treated mitotic cells. Poly-SUMO-2 conjugation of Aurora-B occurred during the M phase of the cell cycle, and both SUMO-2 and PIAS3 were localized adjacent to Aurora-B in the kinetochores in early mitosis. Based on these results, we propose that Aurora-B is a novel mitotic SUMO substrate and that its kinase activity is fine-tuned by the SUMO system.
Small ubiquitin-related modifier (SUMO) proteins are one of the ubiquitin-like modifiers that become covalently conjugated to cellular proteins (Meulmeester & Melchior 2008). Budding yeast has one SUMO-protein SMT3 [suppressor of MIF (mitotic fidelity of chromosome transmission protein) two protein 3], whereas mammals have three SUMO homologs: SUMO-1, -2, and -3. Human SUMO-2 and -3 are almost identical to each other (referred to collectively as SUMO-2/3), and they are approximately 45% identical to SUMO-1. All three SUMOs are covalently conjugated to Lys residues within target proteins in a manner similar to the ubiquitination pathway. SMT3, as well as SUMO-2 and -3, can form chains because they possess internal consensus SUMO modification sites (Johnson & Gupta 2001; Tatham et al. 2001). Like ubiquitination, SUMOylation is also reversible; deSUMOylation is executed by Sentrin/SUMO-specific proteases (SENPs), which are SUMO-specific isopeptidases (Mukhopadhyay & Dasso 2007). Rapid conjugation/deconjugation of SUMO modification makes this modification highly dynamic. SUMOylation can affect substrates in at least two ways: an allosteric effect that induces substrate conformational change, and an effect that involves the creation/inhibition of a binding interface. Consequently, SUMOylation of substrates renders dramatic alterations in substrate intracellular localization, interactions, stability, and activity (Meulmeester & Melchior 2008).
The SUMO system plays important roles in many processes of higher eukaryotic cell life. Depletion of genes encoding SUMO or SUMO system enzymes is lethal for budding yeast and also leads to embryonic lethality in Caenorhabditis elegans, Drosophila melanogaster and mouse, mainly because of defects in chromosome segregation (Jones et al. 2002; Nacerddine et al. 2005; Dasso 2008; Dawlaty et al. 2008; Talamillo et al. 2008). Disruption of the corresponding genes in fission yeast results in viable but abnormal cells with severe growth defects and aberrant cell division (Tanaka et al. 1999; Dasso 2008). These genetic studies have shown the importance of the SUMO system for normal cell division, especially for chromosome segregation, through regulation of mitotic chromosome structure and spindle-kinetochore attachment. In addition, characteristic mitotic localizations of SUMOs have been reported: SUMO-1 localizes at the mitotic spindle and the spindle midzone, whereas SUMO-2/3 appears to localize at the centromeres and on condensed chromosomes (Zhang et al. 2008). To date, global analyses of SUMO substrates in several species have produced extensive lists of novel potential SUMO target proteins. However, only a few mitotic SUMO substrates have been identified in mammals: Ran GTPase-activating protein 1 (RanGAP1), DNA Topoisomerase IIα, centromere protein E (CENP-E), BUBR1 [Bub (budding uninhibited by benzimidazole)-related protein kinase 1], and Nuf2 (nuclear filament-containing protein) (Dasso 2008; Dawlaty et al. 2008; Zhang et al. 2008). SUMOylation of both RanGAP1 and DNA Topoisomerase IIα is regulated by RanBP2 (Ran-binding protein 2), one of the E3 SUMO-protein ligases (described herein as SUMO E3), and the outer kinetochore proteins CENP-E, BUBR1, and Nuf2 can be modified by poly-SUMO-2/3 conjugation during mitosis. SUMOylation appears to be required for targeting DNA Topoisomerase IIα and other mitotic SUMO substrates to the inner centromeres or to the kinetochores (and the mitotic spindle in the case of RanGAP1). Because a large number of different proteins are involved in diverse mitotic events, it is presumed that other, as yet unidentified proteins, are regulated by the SUMO system during mitosis.
Cell division is precisely regulated by several post-translational modifications of proteins (including the above-mentioned SUMOylation), the best described of which is reversible phosphorylation. Accurate mitotic phase progression requires the appropriate phosphorylation of various proteins by mitotic kinases (Nigg 2001; Ferrari 2006). The Aurora family kinases are key mitotic kinases that are highly conserved from yeast to humans. There are three Aurora homologs (Aurora-A, -B, and -C) in human and mouse (Carmena et al. 2009). Although their homology at the protein level is >84%, their functions and subcellular localizations are distinct. Aurora-A is located at the centrosomes and mitotic spindle and is required for mitotic entry, centrosome maturation/separation, and spindle assembly (Barr & Gergely 2007). Aurora-B is the catalytic subunit of the chromosomal passenger complex (CPC). During the cell cycle, Aurora-B localizes on the inner centromeres until metaphase, where it regulates spindle-kinetochore attachment and chromatid cohesion. From anaphase onwards, Aurora-B translocates to the spindle midzone and equatorial cortex and subsequently accumulates in the midbody where it regulates cytokinesis (Andrews et al. 2003; Ruchaud et al. 2007). Both Aurora-A and -B can be fully activated by (auto) phosphorylation of their activation loop in a spatial- and time-specific manner. Interaction with a specific partner is required for activation; TPX2 [targeting protein for Xklp2 (Xenopus kinesin-like protein)], Ajuba and Bora for Aurora-A, and the other members of the CPC, including the scaffolding protein inner centromere protein (INCENP) and the targeting subunits survivin and Borealin, for Aurora-B (Carmena et al. 2009). However, Aurora-A/-B deactivation mechanisms remain unknown, apart from the involvement of dephosphorylation by protein phosphatase 1 and 2A and proteolytic and/or nonproteolytic ubiquitination (Carmena et al. 2009). Thus, to date, well-known regulatory modification of Auroras is reversible phosphorylation.
Mitotic functions of the SUMO system and the known cellular localization of SUMOs led us to examine whether the functions and/or the localization of Auroras is regulated by the SUMO system. In this study, we report SUMO system-mediated regulation of Aurora-B kinase activity during mitosis.
Aurora-B is SUMOylated in vitro and in cells
To determine whether Auroras can be SUMOylated, we first carried out in vitro SUMOylation assays. When GST-Aurora-A and -B purified from mammalian cells were incubated in the presence of SUMO-1, the SUMO-activating enzyme Aos1/Uba2 and the SUMO-conjugating enzyme Ubc9, high molecular weight species of both GST-Auroras were detected, indicating that SUMO-1 can be conjugated to both Auroras (Fig. 1A). We next investigated which of the SUMO paralogs could modify Auroras in cells. FLAG-tagged Aurora-B was co-transfected with either a His6-tagged SUMO-1 or SUMO-2 into HeLa cells. Each His6-SUMO-modified-protein was purified from transfected cells under denaturing conditions by pull-down using Ni-NTA beads and was subsequently examined by western blotting with an anti-Aurora-B antibody. High molecular weight Aurora-B species were detected in both His6-SUMO-1 and His6-SUMO-2 pulled down complexes, confirming that Aurora-B is SUMOylated in cells (Fig. 1B, left panel). Unlike the in vitro SUMOylation results, there were two major high molecular weight Aurora-B species in cells, suggesting multi- or poly-SUMO conjugation to Aurora-B in cells. We carried out the experiments by utilizing different molecular weight His6-SUMOs or His6-FLAG-SUMOs. The different molecular weight tags gave rise to different molecular weight outcomes between His6-SUMOs and His6-FLAG-SUMOs-moieties (Fig. S1 in Supporting Information), indicating that the higher-molecular bands between 58 and 80 kDa are SUMOylated Aurora-Bs. We also examined whether Aurora-A is SUMOylated under similar conditions. However, neither conjugation of SUMO-1 nor SUMO-2 to Aurora-A was detected in cells (data not shown).
Because Aurora-B plays an important role in chromosome segregation and cytokinesis, we examined the subcellular localization of SUMO-1 and SUMO-2 during mitosis using HeLa cells that stably expressed green fluorescence protein (GFP)-SUMO fusion proteins. As shown in Fig. 1C, both GFP-SUMO-1 and GFP-SUMO-2 were mainly detected at the centrosomes in metaphase cells and also localized to the kinetochores of mitotic chromosomes outside of the area where Aurora-B is localized (see chromosome spreads; Fig. 1C, arrows). In telophase, GFP-SUMO-2 accumulated within newly forming nuclei, and some GFP-SUMO-2 was present at the midbody, co-localizing with endogenous Aurora-B. In contrast, GFP-SUMO-1 was present at reforming nuclear envelopes and showed no midbody localization (Fig. 1C). From these results, Aurora-B could be SUMOylated during mitosis.
Aurora-B is modified by PIAS3-mediated SUMO-2 conjugation in cells
To investigate how Aurora-B SUMOylation is regulated in cells, we aimed to identify a SUMO E3 for Aurora-B. We focused on analysis of Siz (SAP and mIZ-finger domain)/PIAS family proteins, a Siz/PIAS (SP)-RING-type SUMO E3, because some members of this family are critical regulators of mitotic SUMO conjugation (Dasso 2008; Takahashi et al. 2008; Rytinki et al. 2009). First, five human PIAS family members were analyzed for their ability to interact with Aurora-B. Immunoprecipitation studies using HeLa cells transfected with each PIAS construct revealed that all five PIAS family members could associate with Aurora-B in cells (Fig. S2A in Supporting Information). We next tested which of the PIAS family members function as SUMO E3s toward Aurora-B in cells. Interestingly, although all of the PIAS E3s could bind Aurora-B, PIAS3 efficiently enhanced SUMO-2 conjugation of Aurora-B, but not SUMO-1 conjugation (Fig. 2A). High molecular weight SUMO-modified species of Aurora-B were detected, indicative of poly-SUMOylation (Fig. 2A, upper panel in the right). An SP-RING mutant of PIAS3 that lacks ligase activity (C334S) did not promote SUMO-2 conjugation of Aurora-B (Fig. S2B in Supporting Information), revealing that this effect was completely dependent on the SP-RING domain of PIAS3. Furthermore, the level of SUMO-2-conjugated Aurora-B was markedly decreased in cells depleted of endogenous PIAS3 using siRNA (Fig. 2B), showing that Aurora-B is a physiologically relevant target protein of PIAS3-mediated SUMO-2 conjugation in cells.
Because the involvement of PIAS3 in mitotic events has not been reported, we investigated the mitotic localization of PIAS3 using HeLa cells expressing GFP-tagged PIAS3. As reported previously (Kotaja et al. 2002), GFP-PIAS3 displays a speckled pattern of nuclear distribution during interphase (data not shown). Interestingly, like GFP-SUMOs, nuclear GFP-PIAS3 appeared as paired dots on both sides of Aurora-B from prophase to metaphase (Fig. 2C, arrows), indicating that GFP-PIAS3 has an expression pattern that is characteristic of kinetochore localization. We confirmed these results using HeLa cells co-expressing GFP-PIAS3 together with a mCherry fusion human CENP-A (Fig. S2C in Supporting Information). The localization of GFP-PIAS3 to kinetochores was more obvious than that of GFP-SUMOs. It is noteworthy that the GFP-PIAS3 kinetochore signals disappeared with chromosome alignment on the metaphase plate (Fig. 2C and Fig. S2C in Supporting Information) (see Discussion). In late mitosis, accumulation of GFP-PIAS3 within newly forming nuclei was observed when the cells were fixed with formaldehyde (data not shown). When we analyzed cells from which the bulk of the soluble proteins was removed by extraction, GFP-PIAS3 did not show any characteristic localization pattern from anaphase onward, and only a small fraction of GFP-PIAS3 was present on the spindle midzone/midbody (Fig. 2C). Thus, combining mitotic SUMO-2 distribution (Fig. 1C) with these observations suggested that PIAS3-mediated SUMO-2 conjugation of Aurora-B may occur at the centromere/kinetochore early in mitosis and/or at the midbody late in mitosis.
SUMO-2 conjugation of Aurora-B occurs from mitosis to the early G1 phase
We next confirmed that SUMO-2 conjugation of Aurora-B occurs in mitotic cells. The levels of SUMO-2-modified Aurora-B at different cell cycle stages were analyzed using HeLa cells that co-expressed Glu-Glu-tagged Aurora-B and His6-tagged SUMO-2 and that were synchronized at the G1/S boundary using a double-thymidine block method. High molecular weight smears of Glu-Glu-Aurora-B, ranging from 80 to 175 kDa, were detected by Western blotting from 9 to 14 h after release from the block (Fig. 2D, upper panel in the left, and Fig. S3 in Supporting Information), which corresponds to late G2 to early G1 phases of the cell cycle (data not shown). This smear pattern was specific for Aurora-B and was because of the fact that the total level of SUMOylated-proteins was more abundant in log phase than in late G2 to early G1 phases. We also determined whether endogenous Aurora-B was modified by SUMO-2 during mitosis, using log phase or mitotic HeLa cells expressing His6-tagged SUMO-2, with or without co-expression of myc3-tagged PIAS3. More pronounced high molecular weight smears of Aurora-B were observed in the His6-SUMO-2 pulled down complexes in the presence of PIAS3 during mitosis than during the log phase of growth (Fig. 2E, upper panel in the left). These results show that conjugation of Aurora-B by (poly-) SUMO-2 occurs mostly in mitosis/early G1 phases. However, only a small subpopulation of Aurora-B is conjugated (discussed below).
Aurora-B Lys-202 is the principal SUMO-2 conjugation site
To further clarify the role of SUMO-2 conjugation of Aurora-B, we next determined the SUMOylation site(s) in Aurora-B. SUMOylation of proteins occurs at specific Lys residues, which, in most cases, are embedded in a consensus sequence motif, ψ-Lys-Xaa-Glu/Asp, in which ψ is a large hydrophobic residue (Ulrich 2008). Aurora-B has a single consensus motif surrounding Lys-202 (Fig. 3A). It is noteworthy that this motif is almost completely conserved from yeast to humans (Fig. 3A). To examine whether Lys-202 of Aurora-B is the acceptor site for SUMO conjugation, the effect of mutation of this Lys residue to Arg (K202R) on in vitro SUMOylation of GST-Aurora-B was assayed. SUMO-2 conjugation to GST-Aurora-B was abolished by this mutation (Fig. 3B, left panel), showing that Lys-202 is a potential SUMOylation site.
We next determined whether the Lys-202 residue-mediated SUMO-2 conjugation of FLAG-Aurora-B transfected into cells. High molecular weight Aurora-B smears were sharply reduced in the K202R Aurora-B mutant (Fig. 3B, upper panel in the right, lane 2), revealing that Aurora-B is mainly poly-SUMOylated on Lys-202 in cells. The remaining SUMOylated Aurora-B bands were eliminated by further Lys-to-Arg substitutions at residues 85 and 87 of Aurora-B (K202R + K85/87R) (Fig. 3B, upper panel in the right, lanes 2 vs. 3). Thus, these results show that Lys-202 of Aurora-B is the preferential SUMOylation site that additional Lys residues can also be conjugated by SUMO, and that PIAS3 induces poly-SUMO-2 conjugation of Aurora-B on these Lys residues.
To ensure that the KR mutants of Aurora-B retain kinase activity similar to wild type, the wild-type and KR mutants of Aurora-B were bacterially expressed together with the IN-box segment of their CPC partner, INCENP (Sessa et al. 2005), and were then purified. Their kinase activities were measured using a GST-tagged fragment of histone H3, which is a well-known physiological substrate of Aurora-B. All Aurora-B KR mutants failed both to autophosphorylate and to phosphorylate histone H3 and the IN-box segment; nevertheless, the wild-type Aurora-B was fully activated, and the wild-type and KR mutants of Aurora-B could bind to the IN-box segment (Fig. 3C, Fig. S4 in Supporting Information, and data not shown). These results show that all of these KR mutants had lost their kinase activity.
SUMO-specific isopeptidase SENP2 deconjugates SUMO-2 from Aurora-B
Because the Aurora-B KR mutants had no kinase activity, it was impossible to use these KR mutants to explore the functional significance of SUMO conjugation of Aurora-B in cells. Therefore, to analyze the functional significance of Aurora-B SUMOylation, we instead used SENPs in deSUMOylation assays. First, we asked which SENPs might be involved in the deSUMOylation of Aurora-B. We assayed the ability of five FLAG-tagged mammalian SENP family members (SENP1, SENP2, SENP3, SENP5, and the mouse SENP2 isotype Smt3IP2/Axam2) to deSUMOylate co-transfected, FLAG-tagged Aurora-B in cells (Nishida et al. 2001; Mukhopadhyay & Dasso 2007). Immunoprecipitation experiments showed that all five mammalian SENP family members associated with Aurora-B in cells (Fig. S5A in Supporting Information). However, only over-expressed SENP2 could eliminate SUMO-2–modification of Aurora-B (Fig. 4A, upper panel, lanes 4 vs. 2). Interestingly, over-expression of the other four SENP family members enhanced SUMO-2 conjugation of Aurora-B albeit in the absence of PIAS3 over-expression. The high molecular weight species of SUMO-2-modified Aurora-B induced by each of these SENPs are almost identical to each other and are further remarkably similar to those induced during M phase (Fig. 4A, upper panel, lanes 3 and 5–7, compared with Fig. 2D). These effects of over-expression of the other four SENP family members might be because of an increased level of free intracellular SUMO because of their promotion of the deSUMOylation of other substrates, indicating that these SENPs do function in cells but fail to deSUMOylate Aurora-B. We further assessed SENP2-catalyzed deSUMOylation of Aurora-B in vitro using recombinant proteins. The SUMO-2 modification of GST-Aurora-B was almost completely eliminated by the addition of GST-fused wild-type SENP2, whereas it was unaffected by GST-SENP2 in the presence of NEM, an inhibitor of SUMO-specific isopeptidases, or by a catalytically inactive mutant of SENP2 (C549A) (Fig. S5B in Supporting Information). This result indicates that SENP2 specifically and effectively deconjugates SUMO-2 from Aurora-B.
We utilized SENP2-mediated SUMO-2-deconjugation of Aurora-B to investigate the role of Aurora-B SUMO-2 conjugation. First, the effects of SENP2 over-expression on the mitotic localization of Aurora-B were investigated. As shown in Fig. 4B, immunofluorescence studies revealed that the centromere and midbody localization of endogenous Aurora-B was not affected by over-expression of FLAG-tagged SENP2. We also observed a defect in chromosome congression and in the generation of lagging chromosomes during telophase as well as incomplete cytokinesis in SENP2 over-expressing cells (Fig. 4B, and data not shown). These effects are consistent with a previous work (Zhang et al. 2008).
SUMO-2 conjugation facilitates autophosphorylation of Aurora-B during mitosis
The principal SUMO conjugation site of Aurora-B, Lys-202, resides in the catalytic loop region and is part of the substrate-binding pocket. Lys-202 is also adjacent to the autophosphorylation site Thr-232 in the activation loop and to the ATP binding pocket (Fig. S6 in Supporting Information). This structural information led us to ask whether SUMOylation of Aurora-B on Lys-202 could affect its kinase activity.
Because autophosphorylation of Aurora-B on Thr-232 is required for its full activation, we examined whether autophosphorylation and/or kinase activity of Aurora-B was important for its SUMOylation. FLAG-Aurora-B that was activated by co-expression of the myc3-tagged IN-box, a FLAG-Aurora-B kinase-dead mutant (KD), and FLAG-wild-type Aurora-B without a co-expressed IN-box were similarly modified by PIAS3-mediated SUMO-2 conjugation in cells (Fig. 5A, upper panel in the left), showing that neither autophosphorylation nor Aurora-B kinase activity is required for its SUMOylation. In addition, SUMO-2-conjugated phospho-Aurora-B was also detected (Fig. 5A, upper panel in the right), indicating that autophosphorylation of Aurora-B does not hinder its SUMOylation.
We next investigated the effects of SUMOylation on autophosphorylation and/or kinase activity of Aurora-B. FLAG-tagged Aurora-B was co-transfected with or without the myc3-IN-box, His6-SUMO-2, and wild-type or C334S-PIAS3 as shown in Fig. 5B. Whole cell lysates were subsequently immunoblotted with the anti-phospho-Aurora-B antibody. Weak autophosphorylation of Aurora-B was detected when the IN-box was co-expressed with Aurora-B (Fig. 5B, lane 2). Surprisingly, Aurora-B autophosphorylation was increased by the additional co-expression of SUMO-2, and this effect was markedly enhanced by further co-expression of wild-type, but not the SP-RING mutant C334S, of PIAS3 (Fig. 5B, lanes 4, 6 and 10). Conversely, co-transfection of siRNA directed against PIAS3, or co-expression of SENP2, resulted in a significant reduction of autophosphorylated Aurora-B, albeit in the presence of both the IN-box and SUMO-2 (Fig. 5C, lane 4 and D, lane 6). These results suggest that PIAS3-mediated SUMO-2 conjugation increases the level of Aurora-B autophosphorylation in cells.
We next explored the consequences of SENP2-mediated SUMO-2 deconjugation on endogenous Aurora-B autophosphorylation in mitosis. There was no difference in anti-phospho-Aurora-B antibody staining between control and SENP2 over-expressing cells in prometaphase (Fig. 6A), indicating that SUMO-2 conjugation is dispensable for the initial activation of Aurora-B. We next evaluated the effect of SENP2-mediated SUMO-2 deconjugation on the recovery of Aurora-B autophosphorylation after its inhibition by treatment of the cells with VX-680, a specific Aurora inhibitor. Autophosphorylation of Aurora-B was completely abolished by VX-680 in both control and SENP2 over-expressing cells (Fig. 6B). At 15 min after washout of VX-680, Aurora-B immunofluorescent autophosphorylation signals were again observed. However, the recovery of Aurora-B autophosphorylation was significantly impaired in SENP2 over-expressing cells (average % of signal intensity 53% vs. 101% in control cells) (Fig. 6C, C′ and D). As mentioned above, SENP2 over-expression has no effect on the mitotic localization of Aurora-B (Fig. 4B); thus, the reduction in Aurora-B autophosphorylation in SENP2 over-expressing cells was not because of mislocalization of Aurora-B.
Collectively, these findings raise the possibility that SUMO-2 conjugation promotes autophosphorylation of Aurora-B (i.e. its activation) in cells. Even though SUMO-2 conjugation is not essential for the initial activation of Aurora-B, SUMOylation enhances Aurora-B activation during mitosis. However, these effects could not be confirmed using the SUMOylation-resistant KR mutants because of their lack of kinase activity (see Fig. 3C). Therefore, we cannot rule out the possibility that the above effects are indirectly generated by SUMOylation and/or deSUMOylation of cellular substrates other than Aurora-B.
SUMO-2 conjugation of Aurora-B per se does not directly affect its kinase activity in vitro
Considering that Aurora-B is activated by in trans phosphorylation (Rosasco-Nitcher et al. 2008), we speculated that only a few SUMO-2-conjugated Aurora-B molecules would lead to a significant enhancement of total Aurora-B autophosphorylation. We therefore examined whether SUMOylation of Aurora-B directly activates Aurora-B kinase activity. For this experiment, we assayed the kinase activity of nonmodified, or SUMO-2 modified GST-Aurora-B, using GST-histone H3(5–15) as a substrate. In the absence of the IN-box, the kinase activity of SUMO-2-modified and nonmodified Aurora-B was both extremely low, displaying only faint autophosphorylation compared to nonmodified Aurora-B in the presence of the IN-box, and no phosphorylation of histone H3 was detected (Fig. 7A, lanes 2 and 3), indicating that SUMO-2 modification of Aurora-B itself did not enhance autophosphorylation at Thr-232 and its kinase activity under these conditions.
When we carried out in vitro kinase assays using activated Aurora-B with the IN-box, SUMOylation did not additionally enhance Aurora-B kinase activity (Fig. 7B, lanes 1 and 2). Because it was possible that Aurora-B was already fully activated by co-expression of the IN-box, therefore masking any potential effect of SUMOylation on activation, we next dephosphorylated the purified Aurora-B using λ-phosphatase and then examined the effect of Aurora-B SUMOylation on re-activation of Aurora-B kinase activity. However, even under this condition, Aurora-B kinase activity was not altered by SUMOylation (Fig. 7B, lanes 4 and 5). These results show that SUMO-2 conjugation of Aurora-B does not directly affect its kinase activity in vitro. We next determined whether SUMO-2 modification of Aurora-B influences its association with other CPC components (in vitro translated-INCENP, survivin and Borealin). However, no differences were observed in the association of SUMO-2-modified or nonmodified Aurora-B with any CPC component (Fig. 7C). We therefore concluded that SUMOylation of Aurora-B per se does not directly participate in modulation of CPC complex formation in vitro.
The importance of the SUMO system for cell division, especially for chromosome segregation, has been revealed by genetic and cell biological studies. Here, we report that the mitotic kinase Aurora-B can be modified by poly-SUMO-2/3 during M phase. We have further determined that Aurora-B SUMOylation is regulated by the SUMO E3 PIAS3 and the SUMO-specific isopeptidase SENP2, and that SUMOylation facilitates, rather than directly induces, Aurora-B kinase activity.
Except for Borealin, mitotic SUMO-2/3 conjugation of CPC components has not been detected (Klein et al. 2009). In this study, we showed in vivo SUMO-2 conjugation of the catalytic subunit of CPC, Aurora-B (Figs 1B, 2A, 2D, 2E, 4A, 5A, Figs S1 and S2B in Supporting Information) as well as its in vitro SUMOylation using tagged-Aurora-B protein purified form mammalian cells (Figs 1A, 3B, 7 and Fig. S5B in Supporting Information). However, we could not confirm in vitro SUMO-2 conjugation of Aurora-B using bacterially expressed Aurora-B, even when co-expressed with the IN-box and in the presence of PIAS3 (data not shown). This finding suggests that SUMOylation of Aurora-B requires other post-translational modification(s) and/or other co-factor(s). Indeed, phosphorylation of a SUMO substrate has been reported to enhance its SUMOylation (Bossis & Melchior 2006). However, at least autophosphorylation of Aurora-B at Thr-232 is dispensable for its SUMOylation (Fig. 5A). Further investigation is required to explore in detail the factor(s) that enhances Aurora-B SUMOylation.
We also noted that, even though maximal SUMO-2/3 conjugation of Aurora-B was reached at M phase, it was confined to only a small subpopulation of Aurora-B (Fig. 2D,E). Our interpretation is as follows; PIAS3, the SUMO E3 for Aurora-B, clearly localizes adjacent to Aurora-B at the kinetochores during early mitosis (from prophase to prometaphase) (Fig. 2C), and this localization facilitates Aurora-B SUMO-2/3 conjugation. However, we observed that, during chromosome congression, PIAS3 was steadily eliminated from the kinetochores and had completely disappeared from the kinetochores of accurately aligned chromosomes at metaphase (Fig. 2C and Fig. S2C in Supporting Information). In addition, de-SUMOylation by SENP2 may allow the mitotic SUMO modification of Aurora-B spatially and temporally dynamic. This dynamic regulation of SUMOylation contributes to the fine-tuning of Aurora-B activity at the centromeres. Kinetochore localization of PIAS3 is also of interest because its localization pattern is very similar to that of mitotic spindle assembly checkpoint (SAC) proteins MAD2 and BUB1. These key SAC components are recruited to spindle-unattached kinetochores but are released/reduced upon attachment of kinetochore microtubules (Musacchio & Salmon 2007). These findings strongly suggest the participation of PIAS3 in SAC and a function of the SUMO system in chromosome segregation.
There are two important issues that remain to be addressed regarding Aurora-B SUMOylation; (i) does Aurora-B SUMOylation directly facilitate its activation?; (ii) how does SUMOylation facilitate Aurora-B activation? Although we defined the SUMOylation sites in Aurora-B (Fig. 3B), unfortunately, its SUMO-deficient mutants (KR mutants) did not have any kinase activity in vitro. To avoid the co-purification of contaminating kinase(s) including endogenous Aurora-B in mammalian cells, we examined the kinase activity of bacterially purified recombinant Aurora-B. Under the same condition, the wild-type Aurora-B was fully activated, and the wild-type and KR mutants of Aurora-B could bind to the IN-box segment (Fig. 3C and Fig. S4 in Supporting Information). Our empirical data indicate that the kinase activity of Aurora family kinases is often lost even when a single amino acid is substituted, probably due to a resulting conformational change. Thus, the kinase-defective KR mutant is not appropriate to study the in vivo effects of SUMOylation on Aurora-B. For this reason, we utilized SENP2-catalyzed deSUMOylation to analyze Aurora-B SUMOylation in vivo instead of KR mutants and clearly showed that autophosphorylation of Aurora-B was dramatically enhanced in the presence of SUMO-2 and PIAS3 (Fig. 5B,C), whereas it was reduced by SENP2 expression in vivo (Figs 5D and 6). However, it remains unclear whether these effects were mediated by direct SUMOylation of Aurora-B or indirectly by PIAS3 and SENP2 modulation of other substrates. We also could not observe any effects of Aurora-B SUMOylation on its kinase activity or on its interaction with the other CPC components in vitro (Fig. 7). Thus, indirect modulation of Aurora-B kinase activity by other SUMO substrates is a definite possibility. SUMOylation of Aurora-B was reported, whereas this manuscript was in preparation (Fernandez-Miranda et al. 2010). The major SUMOylation site of Aurora-B that we defined is in excellent agreement with this report. However, it is noteworthy that the other report used the KR mutant of Aurora-B to reveal in vivo functions of Aurora-B SUMOylation.
Alternatively, it is possible that some novel activation factors that specifically interact with SUMO-conjugated Aurora-B exist in cells. The importance of noncovalent SUMO binding has been recently emphasized, and a class of SUMO-interacting motifs (SIM) has been identified (Kerscher 2007; Meulmeester & Melchior 2008; Perry et al. 2008; Ulrich 2008). Indeed, CENP-E localization to kinetochores depends on an internal SIM that specifically binds to SUMO-2/3 chains rather than on direct SUMO-2/3 conjugation of CENP-E (Zhang et al. 2008). Moreover, a recent report showed that the inner kinetochore protein CENP-I is modified by poly–SUMO-2/3 conjugation, and that this modification allows CENP-1 to be degraded by RNF4, a RING-type ubiquitin-protein isopeptide ligase E3, which contains multiple SIMs and targets poly-SUMOylated proteins for proteasomal degradation (Perry et al. 2008; Geoffroy & Hay 2009; Mukhopadhyay et al. 2010). SENP6-mediated SUMO-2/3-deconjugation of CENP-I prevents its degradation, and the balance between SUMOylation/de-SUMOylation controls inner kinetochore assembly (Mukhopadhyay et al. 2010). These findings point to a contribution of poly-SUMO-2/3 chains and SIM-containing proteins to kinetochore function. Localized kinase activity of Aurora-B is required for spindle-kinetochore attachment in a tension-dependent manner and ensures chromosome biorientation on the spindle (Sandall et al. 2006). Thus, we cannot exclude the existence of an unknown factor(s), which specifically interacts with the SUMO-conjugated form of Aurora-B and facilitates its activation at the inner centromeres and/or kinetochores. We are currently searching for such factor(s). Because regulators of Aurora-B SUMOylation (PIAS3/SENP2) are distinct from those for Borealin (RanBP2/SENP3), and moreover because SUMO-2/3 conjugation of Borealin does not affect the localization or complex formation of the CPC (Klein et al. 2009), their SUMOylation must be involved in the regulation of protein–protein interaction other than interaction between components of the CPC. Future studies are necessary to clarify this point.
It is well established that the major factors that regulate Aurora-B are the other CPC components, although other activation factors have also been reported. For example, telophase disk-60 kDa (TD-60), another chromosomal passenger protein, together with microtubules, is required for full activation of Aurora-B in vitro (Rosasco-Nitcher et al. 2008). Moreover, chromatin-induced local activation of Aurora-B at the centromere has been reported, which is apparently initiated by Borealin-mediated local clustering of Aurora-B on chromatin (Kelly et al. 2007). Hence, it is likely that some SIM-containing microtubule-related or chromatin-associated proteins specifically and positively act on SUMOylated-Aurora-B during mitosis. Additional modifications and/or factors, most likely mediated by the three regulatory subunits of the CPC, would facilitate fine-tuning of Aurora-B activity at the appropriate time and place. The highly dynamic SUMOylation of Aurora-B described in this paper is one effective modification for such fine-tuning.
Monoclonal anti-Aurora-A (M11-17), anti-Aurora-B (H7-4), and anti-phospho-Aurora-B (Thr-232) (pAB2.1) antibodies have been described (Honda et al. 2000; Li et al. 2004; Yasui et al. 2004). The following commercial antibodies were used: monoclonal anti-myc, anti-GST (glutathione S-transferase) (BioAcademia); anti-β-actin (AC-15), and anti-FLAG (M2) (Sigma); anti-cyclin B1 (GNS1), and anti-PIAS3 (C-12) (Santa Cruz Biotechnology); polyclonal anti-SUMO-2/3 (ABGENT); phospho-histone H3(Ser-10) (Cell Signaling Technology); anti-FLAG, anti-myc (BioAcademia); horseradish peroxidase-conjugated anti-mouse (Rockland) or anti-rabbit IgG (Bethyl Laboratories); and Alexa Fluor 488- or 594-labeled goat anti-mouse or anti-rabbit IgG (Molecular Probes).
Expression and purification of recombinant proteins
His6-SUMO-1, -2, and each FLAG-SENP family member were constructed as described previously (Nishida & Yamada 2008). Human PIAS family members were inserted into myc3/pcDNA3. Point mutations within Aurora-B, PIAS3, and SENP2 were engineered using standard double PCR mutagenesis. The plasmids expressing human Aurora-A, -B, and the IN-box segment of INCENP (783–918 amino acids) have been described (Li et al. 2004; Ohashi et al. 2006). For retrovirus production, cDNAs were inserted into altered versions of pMXs, pMXs-GFP, or pMXs-mC that contained green fluorescent protein (GFP) or monomeric Cherry (mCherry) at the 5′ end of the respective cloning site (Ban et al. 2009). The hexa-histidine (His6)-tag Xenopus SUMO-specific E2-conjugating enzyme Ubc9 (ubiquitin-conjugating enzyme 9) and the His6-mouse SUMO-activating enzyme subunit Uba2 were expressed in Escherichia coli BL21(DE3) using the pET vector. The GST-mouse SUMO-activating enzyme subunit Aos1 and GST-SENP2 were expressed in E. coli using the pGEX vector. Purification of proteins was carried out as described previously (Ohashi et al. 2006). The GST tag of mouse Aos1 was cleaved by thrombin (GE Healthcare).
Cell culture, transfection, and retroviral infection
All cells were grown in Dulbecco’s modified Eagle’s medium (Nissui) supplemented with 10% (v/v) fetal bovine serum (Cell Culture Technologies). HeLa S3 cells were transfected with each plasmid using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Cells were synchronized in G1/S phase using a double-thymidine block method, as described previously (Ban et al. 2009). The cells shown in Fig. 2D,E were transfected with each plasmid during the second thymidine block. Stealth siRNA (small interfering RNA) against PIAS3 (5′-AAUGAUAAGAGAUUCAUAGGGAGCC-3′) was purchased from Invitrogen. siRNA was transfected into HeLa S3 cells using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instructions. The cells shown in Figs 2B and 5C were first transfected with siRNA and were transfected with the indicated plasmids 24 h later. Retrovirus production has been described (Ban et al. 2009).
Immunofluorescence analyses of HeLa cells were carried out as described previously (Ban et al. 2009). For the preparation of metaphase chromosome spreads using cytospin, nocodazole (30 ng/mL, final concentration) was added to the culture medium 3 h before harvesting the cells. Mitotic cells were collected using the shake-off procedure, were treated with 75 mm KCl at 37 °C for 30 min, and were then centrifuged onto coverslips at 500 g for 10 min in a cytocentrifuge (KS-4000, KUBOTA). The chromosome spreads were fixed for 15 min with 3.7% (v/v) formaldehyde in PBS, washed twice with PBS, and were then permeabilized for 5 min with 0.2% (v/v) Triton X-100 in PBS. After three washes with PBS, samples were processed for immunofluorescence as above. Nuclei and chromosomes were stained with DAPI (4′,6-diamidino-2-phenylindole). Samples were observed under a confocal microscope FLUOVIEW FV1000 (Olympus). To study re-activation of Aurora-B in cells, the Aurora kinase inhibitor VX-680 (600 nm, final concentration, Kava Tech.) was added to the culture medium for 4 h; the cells were then washed and incubated in fresh medium for 15 min prior to fixation.
In vitro SUMOylation/deSUMOylation assays
For assay of in vitro SUMOylation, GST fusion Auroras that were expressed in HeLa S3 cells were purified with glutathione-Sepharose beads (GE) and were then mixed with 0.28 μg of purified SUMO-activating enzyme (mouse Aos1/His6-mouse Uba2), 0.5 μg of purified His6-Xenopus Ubc9, and 1.3 μg of SUMO-1 or SUMO-2 (Boston Biochem Inc.). The mixture was incubated at 25 °C for 1 h in the presence of 50 mm Tris–HCl, pH 7.4, 5 mm MgCl2, 2 mm DTT (dithiothreitol), and 2 mm ATP in a final volume of 50 μL. The resin was washed three times with wash buffer [10 mm Tris–HCl, pH 7.4, 3 mm MgCl2, 200 mm NaCl, 0.2 mm PMSF, and 0.1% (v/v) Nonidet P-40]. The SUMOylated proteins on the resin were subjected to SDS–PAGE and were detected by Western blotting with the indicated antibodies. For in vitro deSUMOylation, pre-SUMOylated GST-Aurora-B, trapped on the resin as described earlier, was incubated for 1 h at 30 °C with 1 μg of purified GST-SENP2 in 10 mm Tris–HCl, pH 8.0, 150 mm NaCl, and 1 mm DTT in a final volume of 50 μL with or without 20 mm NEM (N-ethylmaleimide). The resin was washed once with wash buffer and was then subjected to SDS–PAGE. GST-Aurora-B was detected by Western blotting using the anti-Aurora-B antibody.
In vivo SUMOylation/deSUMOylation assays
His6-tagged SUMOs were transiently transfected into HeLa S3 cells together with the constructs indicated in the text. Two days after transfection, Ni-NTA (Ni2+-nitrilotriacetic acid) agarose bead (Qiagen) pull-downs of cell lysates were carried out as described previously (Nishida & Yamada 2008).
In vitro kinase assays
The indicated purified kinase was incubated with a bacterially purified GST-histone H3(5–15) kinase substrate for 20 min at 30 °C in 20 mm Tris–HCl, pH 7.5, 10 mm MgCl2, 0.5 mm DTT, 0.1 mm EDTA, and 0.1 mm ATP in a final volume of 30 μL. The reaction products were separated by SDS–PAGE and were detected by Western blotting with the appropriate antibody. To study re-activation of Aurora-B in vitro, GST-Aurora-B, which was purified from HeLa S3 cells that co-expressed the myc3-IN-box, was pretreated with λ-phosphatase (400 units; New England Biolabs) for 30 min at 30 °C in 50 mm Tris–HCl, pH 7.5, 100 mm NaCl, 2 mm MnCl2, 2 mm DTT, 0.1 mm EDTA, and 0.01% (v/v) Brij 35 in a final volume of 50 μL. After inhibition of λ-phosphatase with 1 mm Sodium orthovanadate (Na3VO4), the resin was washed three times with wash buffer. SUMOylation, and subsequently kinase assays, was then carried out using these resins as described above.
Immunoprecipitation and in vitro binding assays
Preparation of cell lysates and immunoprecipitation procedures were described previously (Ohashi et al. 2006). In vitro binding assays were carried out using GST-Aurora-B purified from HeLa S3 cells and myc-tagged INCENP, survivin and Borealin generated by in vitro transcription/translation using the TNT Quick Coupled T7 kit (Promega), as described (Ban et al. 2009).
We are grateful to Dr. T. Kitamura (Tokyo University) for the Plat-E cells and pMXs retroviral vectors and Dr. H. Saitoh (Kumamoto University) for Ubc9, Uba2 and Aos1 expression vectors. We thank Drs. H. Saitoh and K. Tanaka (Kwansei Gakuen University) for critical reading of the manuscript and valuable suggestions. This study was supported by a grant-in-aid for Scientific Research (to R.B. and T.U.). The authors would like to acknowledge the technical expertise of Center for Integrated Research in Science, Shimane University.