Report

Centromeric Aurora-B Activation Requires TD-60, Microtubules, and Substrate Priming Phosphorylation

See allHide authors and affiliations

Science  25 Jan 2008:
Vol. 319, Issue 5862, pp. 469-472
DOI: 10.1126/science.1148980

Abstract

The chromosome passenger complex (CPC) controls chromosome congression, kinetochore-microtubule attachments, and spindle checkpoint signaling during mitosis. Aurora-B kinase is the catalytic subunit of the CPC. To understand how a single kinase can regulate such diverse events, we have investigated the activation of Aurora-B and describe two distinct activation mechanisms. First, Aurora-B activation in vitro requires two cofactors, telophase disc–60kD (TD-60) and microtubules. TD-60 is critical to localize both the CPC and Haspin kinase activity to centromeres and thus regulates Aurora-B at several levels. Second, Aurora-B substrates can inhibit kinase activation, and this is relieved by phosphorylation of these substrates by the centromeric kinases Plk1 and Haspin. These regulatory mechanisms suggest models for phosphorylation by Aurora-B of centromeric substrates at unaligned chromosomes and merotelic attachments.

Aurora-B kinase is individually regulated at each centromere where it phosphorylates distinct sets of substrates at each mitotic stage to direct chromosome segregation (1). Aurora-B activity requires interaction with its partner INCENP (24) and (auto)phosphorylation of its T-loop (5, 6). Phosphorylation of an Aurora-B site on the C terminus of INCENP (S850) increases activity by a factor of 8 to 10. TD-60, another inner centromere protein, is required for centromeric targeting of the CPC and for chromosome congression (7). To understand how TD-60 and Aurora-B interact, we cloned, expressed, purified, and raised antibodies against Xenopus TD-60 (fig. S1, A and B) and showed that TD-60 immunoprecipitated from Xenopus extracts pulled down CPC subunits Dasra-A and INCENP (Fig. 1A and fig. S1D). Reciprocal immunoprecipitations with INCENP and Dasra-A confirmed the interaction.

Fig. 1.

TD-60 is required to congress chromosomes and localize CPC and Haspin kinase activity to the centromeres. (A) CPC and TD-60 interact in Xenopus CSF extracts. (B to D) Representative spindles from mock-depleted extracts (IgGΔ), TD-60–depleted extracts (TD-60Δ), or TD-60–depleted extracts rescued with recombinant TD-60 (TD-60Δ+rTD-60), stained with the antibodies indicated at the top of each panel. Scale bar, 10 μm.

To examine TD-60's role in recruiting the CPC during mitosis, we immunodepleted TD-60 from Xenopus cytostatic factor (CSF) extracts (fig. S1C), cycled them into S phase, and rearrested in the following metaphase, allowing spindle formation. TD-60–depleted extracts were impaired in their ability to align chromosomes to the metaphase plate (Fig. 1, B to D, and figs. S2A and S3A), as shown in cells (7). Phosphorylation of a well-defined chromatin substrate of Aurora-B kinase, serine 10 of histone-H3 (H3S10), was not affected by TD-60 depletion (Fig. 1D and fig. S2B). However, phosphorylation of histone-H3 on threonine-3 (H3T3), a known centromeric target of Haspin kinase (8, 9), required TD-60 (Fig. 1D). Loss of TD-60 also mislocalized Aurora-B and INCENP from the inner centromere (Fig. 1, B and C, and fig. S2B), as seen after small interfering RNA knockdown of TD-60incells (7). Adding recombinant TD-60 protein to depleted extracts restored congression and CPC localization to the centromere (Fig. 1, B and C, and fig. S2B). Therefore, TD-60 controls Aurora-B localization and the Haspin kinase activity at the centromere, but TD-60 is not required for Aurora-B activity on chromosome arms.

About 10% of endogenous TD-60 was sufficient to localize Aurora-B to centromeres (Fig. 2, A to D). Chromosome congression defects were still seen in 60% of spindles from the partially depleted extracts (Fig. 2, B and C, and fig. S3A). Although approximately 80% of the endogenous amount of centromeric Aurora-B properly localized to the centromere, phosphorylation on its T-loop was reduced by 97% compared with immunoglobulin G (IgG)–depleted extracts (Fig. 2B, D). Similarly, 70% of INCENP localized to centromeres, but staining of centromeres with the α-pS850 antibody was reduced by 98% (Fig. 2, C and D). We conclude that a small amount of TD-60 is sufficient to localize the CPC to the centromere, but a greater amount is required to activate centromeric Aurora-B kinase.

Fig. 2.

Partial depletion of TD-60 reveals requirement for TD-60 in Aurora-B activation. (A) Immunoblot showing that 10% of endogenous TD-60 remains after depletion. (B and C) Residual endogenous TD-60 after depletion is sufficient to localize Aurora-B (B) and INCENP (C) to centromeres but not to activate Aurora-B kinase as measured by phosphorylation of the T-loop (B) or INCENP-S850 (C). (D) Quantification of mean integrated fluorescence intensity for (B) and (C). N = 10 nuclei, 16 centromeres per nuclei. Error bars represent SEM. Scale bar, 10 μm.

The activity of immunoprecipitated CPC was not affected by incubation with physiological concentrations (11 nM) of recombinant TD-60 (Fig. 3A and fig. S4A). Aurora-A kinase, an Aurora-B paralog, is stimulated by microtubules (10). Combinations of taxolstabilized microtubules (10 μM) and TD-60 (7.2 μM, 180 pmol) stimulated Aurora-B kinase activity at concentrations where neither microtubules nor TD-60 had an effect (Fig. 3A). Aurora-B and full-length INCENP (fig. S4B, AI) bind microtubules, consistent with the previously identified microtubule-binding activity in the central region of INCENP (11, 12). However, there is at least one additional microtubule-binding site in either Aurora-B (60-361) or a small C-terminal fragment of INCENP (790-856) (fig. S4B, AI790-856). Recombinant TD-60 independently bound microtubules. We conclude that both Aurora-B/INCENP and TD-60 directly interact with microtubules and that all four components are required for kinase activation.

Fig. 3.

TD-60 and microtubules are required to activate Aurora-B kinase. (A) Anti-CPC IP kinase activity was measured by γ32PO4 incorporation onto MCAK(2-263) after the indicated treatments. (B) Microtubules (MTs) and TD-60 are required to activate Aurora-B/INCENP after phosphatase treatment. (C) Time course of experiment in (B). (D) Parallel time points from the experiment in (C) were immunoblotted with antibodies to phospho-T-loop to follow Aurora-B activation.

To study Aurora-B activation in a defined system, we developed a three-step in vitro assay using highly purified components (Fig. 3B). Because Aurora-B phosphorylates both its own T-loop and the C terminus on INCENP, it is fully active when expressed with INCENP and purified from Escherichia coli. To study Aurora-B activation, we first dephosphorylated recombinant AI790-856 (13) by λ-phosphatase. Second, λ-phosphatase was inhibited with sodium orthovanadate (vanadate) and the kinase was incubated with presumptive activators. Third, kinase activity was assayed upon addition of histone H3 as a substrate with γ32P-ATP (adenosine triphosphate) (Fig. 3, B and C). After inactivation by phosphatase treatment, Aurora-B could not activate itself after 30 min (Fig. 3, B to D). Although microtubules on their own partially stimulated the kinase, a time course demonstrated synergistic activation by a combination of TD-60 and microtubules (Fig. 3, C and D). Together the two cofactors stimulated Aurora-B kinase activity by a factor of more than 100 and showed cooperative kinetics with a 2.5-min lag-phase during which the T-loop of Aurora-B was autophosphorylated (Fig. 3, C and D). This cooperativity suggests positive feedback by Aurora-B kinase in trans.

In the previous experiment, Aurora-B/INCENP was incubated with cofactors without nucleotide first, and then the substrate and γ32P-ATP were added last. The order of addition of components affected the reaction. We preincubated three of four components with unlabeled ATP and then added the fourth component and γ32P-ATP to measure kinase activity. Preincubation of microtubules, TD-60, and the CPC with ATP eliminated the lag phase, which suggests that these components are sufficient to activate Aurora-B. All conditions that included the histone-H3 substrate in the preincubation step prevented kinase activation (Fig. 4A). The experiment in Fig. 4A was performed with the CPC complex (fig. S1E), and similar results were seen using AI790-856 (fig. S5A). We saw similar effects when myelin basic protein was the substrate (fig. S5B). These data demonstrate that preincubation of substrates with inactive kinase inhibits subsequent kinase activation by microtubules and TD-60.

Fig. 4.

Phosphorylation of substrates by other mitotic kinases reverses substrate inhibition of Aurora-B activation. (A) Order of addition regulates Aurora-B activation in vitro. (B) Histone-H3 phosphorylation on pT3H3 permits Aurora-B activation. (C) Binding of either TD-60 or Aurora-B/INCENP to a histone-H3 tail peptide or the same peptide phosphorylated on T3. (D) Plx1 was incubated with MCAK and ATP, followed by addition of Aurora-B/INCENP and indicated cofactors. Phosphorylation of MCAK by Aurora-B was measured by immunoblot with indicated phosphospecific antibodies. (E) Model for Aurora-B activation by TD-60, microtubules, and substrate priming.

If substrates prevent Aurora-B activation, there must be mechanism(s) to reverse this inhibition. We hypothesized that phosphorylation of Aurora-B substrates by other kinases would reverse the inhibitory effect. Peptides of the histone-H3 N terminus were preincubated with TD-60, followed by the addition of the CPC, microtubules, and unlabeled ATP. Activity was measured after addition of γ32P-ATP and myelin basic protein substrate for 2 min. Preincubation with an unphosphorylated histone-H3 peptide potently inhibited kinase activation (Ki-50 nM) (Fig. 4B). However, when this peptide was phosphorylated on T3, the kinase was activated with myelin basic protein (Fig. 4B) or MCAK-S196 (fig. S7A) as a substrate. Thus, H3 tails inhibit kinase activation unless they are phosphorylated on T3. We propose that Haspin phosphorylation of T3 at the centromere facilitates Aurora-B activation.

Unphosphorylated substrates presumably bind either Aurora-B/INCENP (fig. S1F) or TD-60 to inhibit activation. The unphosphorylated H3 peptide bound Aurora-B/INCENP much more strongly than did the phosphorylated peptide (Fig. 4C). TD-60 poorly bound either peptide at the concentrations tested. These data suggest that unphosphorylated substrates directly bind Aurora-B/INCENP to prevent activation.

TD-60 is required for H3S10 phosphorylation in vitro but not in vivo. We tested the requirement for cofactors (TD-60 and microtubules) and priming phosphorylation on a physiological centromeric substrate. The microtubule depolymerase MCAK is phosphorylated by Aurora-B on S196 and T95 in the centromere in prometaphase and on T95 on chromosome arms in prophase (1417). Microtubules poorly stimulated total kinase activity on MCAK(2-263), and TD-60 did not further stimulate activity (fig. S5C).

We determined whether a priming phosphorylation could restore TD-60 and microtubule cooperative activation of Aurora-B on MCAK. Another centromeric kinase, the Xenopus Polo kinase (Plx1), phosphorylates MCAK. Plx1 immunoprecipitated from mitotic Xenopus extracts phosphorylated MCAK as well as Aurora-B in this assay (fig. S6B). Further analysis of Plx1 phosphorylation on MCAK mapped the majority of Plx1 sites to the N-terminal region (2 to 116) (fig. S6, A and C).

To directly test whether priming phosphorylation by Plx1 activates MCAK as an Aurora-B substrate, we measured the kinetics of Aurora-B kinase activity on two different Aurora-B sites on MCAK(2-263) by immunoblots using antibodies specific to phosphorylated residues T95 and S196 (1417). In a control reaction in which MCAK was incubated with Aurora-B, there was little activity on MCAK (Fig. 4D and fig. S7C, MCAK only). Preincubating Plx1 with MCAK before the addition of Aurora-B produced robust phosphorylation on the MCAK-T95 site but little activity on the MCAK-S196 site. MCAK-T95 phosphorylation was seen in all conditions where Plx1 was preincubated with MCAK, which suggests that this is the key event to allow Aurora-B to phosphorylate this site. We believe the stimulation is caused by Plx1 phosphorylation of MCAK on unidentified sites, because Plx1 robustly phosphorylates the N terminus of MCAK (fig. S6, A to C) but does not phosphorylate TD-60 (fig. S6B), MCAK-S196, or MCAK-T95 (fig. S7B).

After phosphorylation with Plx1, MCAK-S196 phosphorylation required the same cofactors as histone-H3. The greatest activation was seen when Plx1 was preincubated with MCAK and then both microtubules and TD-60 were added (Fig. 4D and fig. S7C). Preincubation of MCAK with Plx1 followed by either TD-60 or microtubules alone produced substantial activity, although neither gave as robust activity as when both cofactors were present.

Our data are consistent with the following model of Aurora-B activation (Fig. 4E). Aurora-B requires a rotation to bring the N- and C-terminal domains into register (13), which we propose is the step catalyzed by TD-60 and microtubules to allow T-loop autophosphorylation. Active kinase phosphorylates activation sites on inactive CPCs, generating positive feedback. Substrates inhibit activation because they compete with the T-loop for the active site or block domain rotation. Priming phosphorylation dissociates substrates from Aurora-B, allowing activation.

Our data suggest mechanisms of cross-talk between three major mitotic kinases. CDK1 phosphorylation of substrates recruits Plx1, which then drives local Aurora-B activation by priming substrates. Moreover, Haspin phosphorylation of H3T3 might establish a centromeric histone code that ensures that high Aurora-B activity is restricted to centromeres in prometaphase.

During prophase Aurora-B regulates global events such as changes to mitotic chromatin, whereas in prometaphase each mitotic chromosome is autonomously controlled by Aurora-B activity to generate spindle checkpoint signals and correct kinetochore-microtubule attachments. Currently the best example of these phenomena is the regulation of MCAK by Aurora-B. Can the interplay of substrate inhibition and cofactor availability to control Aurora-B activity explain the spatial and temporal changes seen in MCAK regulation? In G2/Prophase, the CPC may use Plx1 priming phosphorylation to phosphorylate MCAK-T95 throughout the nucleus (16, 17). In anaphase, the CPC, TD-60, and Plx1 are colocalized on midzone microtubules spatially restricting MCAK-S196 phosphorylation to this region (14). In prometaphase, phosphorylation of MCAK-S196 is enriched on the centromeres of unaligned chromosomes and also at centromeres of chromosomes that have a kinetochore attached to both poles (merotelic attachments) (18). This chromosome autonomous regulation may be explained by the requirement of a physical contact between microtubules and Aurora-B during activation. Centromeres on unaligned chromosomes may have frequent contact with microtubules nucleated by centrosomes, whereas the kinetochores of aligned chromosomes could protect inner centromeres from microtubules nucleated from the poles (fig. S7D). An exception would be a merotelic attachment where microtubules are brought into contact with inner centromeres even on chromosomes at the metaphase plate (1820). In this way, Aurora-B provides information to centromeres on the location of each chromosome on the spindle as well as its microtubule attachment status.

Supporting Online Material

www.sciencemag.org/cgi/content/full/319/5862/469/DC1

Materials and Methods

Figs. S1 to S7

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article