Histone H3 Thr-3 Phosphorylation by Haspin Positions Aurora B at Centromeres in Mitosis

The chromosomal passenger complex (CPC)—which contains Aurora B, inner centromere protein (INCENP), Survivin, and Borealin—is found on chromosome arms in prophase, concentrates at inner centromeres during prometaphase, and transfers to the central spindle at anaphase (1). Aurora B phosphorylates several substrates at these locations, including histone H3 Ser¹⁰ (H3S10ph) on chromosome arms, mitotic centromere-associated kinesin (MCAK) at inner centromeres, centromere protein–A Ser⁷ (CENP-AS7ph) at outer centromeres, and the KNL1/Mis12 complex/Ndc80 network at kinetochores (1–7). Current models suggest that centromeric Aurora B responds to lack of tension across sister kinetochores that are incorrectly attached to the spindle. Bipolar kinetochore attachment forces may pull kinetochore substrates away from inner centromeric Aurora B, which would lead to substrate dephosphorylation, selective stabilization of correct microtubule attachments, and eventually satisfaction of the spindle checkpoint (fig. S1) (8). Despite its central importance, it is not understood how Aurora B accumulates at centromeres (9–11).

Immunofluorescence microscopy of mitotic cells shows that phosphorylated histone H3 Thr³ (H3T3ph) and Aurora B localize similarly at inner centromeres (Fig. 1A and fig. S2) (12). We therefore tested whether Haspin, which is responsible for generating H3T3ph in mitosis (13, 14), is required for CPC localization. When Haspin was subjected to RNA interference (RNAi) in HeLa cells, a marked reduction in Aurora B at centromeres (greater than fivefold) and an increase on chromosome arms were observed (Fig. 1, B and C). In contrast, Haspin depletion had little effect on the amounts of CPC subunits in mitosis and did not disrupt CPC formation (fig. S3). Similar results were obtained in a human osteosarcoma cell line (U2OS cells), where Haspin RNAi delocalized all CPC components from centromeres (Fig. 2A and fig. S4). Microinjection of H3T3ph-specific antibody into mitotic LLC-PK pig kidney cells caused a similar delocalization of Aurora B (Fig. 1, D and E), indicating that Haspin acts through H3T3ph to position the CPC during mitosis and not through effects at other cell cycle stages.

• Download high-res image
• Open in new tab
• Download Powerpoint

• Download high-res image
• Open in new tab
• Download Powerpoint

To determine whether the reduced CPC accumulation at centromeres after Haspin depletion was functionally significant, we first focused on MCAK, a microtubule-depolymerizing kinesin whose inner centromeric localization is dependent on Aurora B (3, 4, 15). Both Aurora B and Haspin RNAi reduced the centromeric enrichment of MCAK in ≥70% of U2OS cells (Fig. 2, B to D) and HeLa cells (fig. S5). Artificial retargeting of Aurora B to centromeres with CENP-B–INCENP (8) largely restored MCAK localization in Haspin-depleted cells (fig. S6). Disruption of CPC activity also compromises the spindle checkpoint response, particularly to low doses of taxol (16–20). Consistent with this, HeLa cells transfected with Haspin small interfering RNA (siRNA) were less efficiently arrested in mitosis by taxol than controls were (Fig. 2E). Thus, Haspin appears to influence MCAK localization and checkpoint signaling at centromeres by controlling Aurora B localization.

Two experiments suggest that these effects are not a consequence of defects in chromosome cohesion caused by Haspin RNAi (14). First, Haspin overexpression, which increases cohesion between chromosome arms (14), also influenced Aurora B localization (fig. S7). Second, loss of cohesion induced by depletion of Sgo1 (14, 21, 22) rendered Aurora B localization more diffuse, but it remained enriched at centromeres in >80% of nocodazole-treated cells, and centromeric MCAK was little changed (Fig. 2, A to D). Therefore, as reported elsewhere (14, 21–24), loss of centromeric CPC and MCAK is not an inevitable consequence of cohesion loss.

In contrast, Haspin RNAi had minor effects on total CENP-AS7ph and H3S10ph (fig. S8A), although both were reduced by Aurora B RNAi, as expected (2, 5, 16). Haspin RNAi also did not detectably alter total Thr²³² autophosphorylation (fig. S8B) or in vitro kinase activity of immunoprecipitated Aurora B (fig. S8C). H3S10ph, CENP-AS7ph, and Aurora B T232ph immunofluorescence remained present in individual U2OS and HeLa cells in which Aurora B and MCAK were delocalized (Fig. 2F and fig. S8, D to F). Thus, Aurora B kinase activation and some of its chromosomal functions appear less sensitive to delocalization of the CPC from inner centromeres than localization of MCAK.

We tested whether H3T3ph plays a direct role in CPC binding at inner centromeres. Biotinylated peptides encompassing residues 1 to 21 of histone H3 with various side-chain modifications were used in “pull-down” experiments from mitotic HeLa cell lysates. Survivin bound strongly to H3(1-21) peptides containing H3T3ph but poorly to unmodified H3(1-21), H4(1-21), or peptides containing H3S10ph or H3T11ph (Fig. 3A and fig. S9A). Borealin, which directly binds Survivin (11, 25), also bound preferentially to H3T3ph peptides (Fig. 3A). Enrichment of Aurora B on H3T3ph peptides was detectable, but weaker than that of Survivin or Borealin (fig. S9A), perhaps because not all CPC components are associated in cell lysates (25). Trimethylation of H3 at Lys⁴ adjacent to T3ph (H3T3phK4me3) strongly diminished association of CPC components with H3T3ph peptides (Fig. 3A and fig. S9A). Two different recombinant forms of the CPC [the full complex, and a subcomplex containing INCENP(1-58), Survivin, and Dasra A, a Xenopus homolog of Borealin (xISD)] also bound preferentially to H3T3ph peptides, which confirmed that H3T3ph enhances direct binding of the CPC to H3 almost 20-fold (Fig. 3, B and C).

• Download high-res image
• Open in new tab
• Download Powerpoint

Purified recombinant human Survivin also preferentially bound H3T3ph peptides (Fig. 3D, fig. S9B), consistent with previous reports that Survivin is a key factor targeting the CPC to centromeres (9, 19, 20, 26). Similar to previous observations (20, 27), in HeLa cells depleted of endogenous Survivin, the mutation D70A/D71A (in which Asp⁷⁰ and Asp⁷¹ are replaced by Ala) (7) within the BIR domain prevented Survivin from concentrating at centromeres in mitosis, causing it to become diffusely distributed on chromatin (fig. S10, A to C). Survivin-D70A/D71A still localized to the central spindle, however (fig. S10D). Recombinant Survivin-D70A/D71A showed no binding to H3 peptides in vitro (Fig. 3D), which suggested that the failure of Survivin-D70A/D71A to concentrate at centromeres is due to a failure to recognize H3T3ph.

We then tested whether Survivin-D70A/D71A mimicked the effects of Haspin depletion on CPC functions. Indeed, Survivin-D70A/D71A supported CPC formation (fig. S11A) and substantial Aurora B autophosphorylation, H3S10 phosphorylation, and CENP-AS7 phosphorylation but did not restore normal localization of Aurora B or MCAK to inner centromeres (Fig. 4, A to C, and fig S11, B and C) or support spindle checkpoint arrest in the presence of low doses of taxol (fig. S12A). Survivin-D70A/D71A rescued cytokinesis failure caused by Survivin depletion (fig. S12B), but it did not allow normal correction of improper spindle-kinetochore attachments, and an increase in chromosome alignment defects was evident compared with that in cells expressing wild-type Survivin (fig. S12, C and D). Similarly, in chicken DT40 cells expressing Survivin-D72A/D73A (equivalent to D70A/D71A in human Survivin) and in human cells containing Survivin mutated at other sites in the BIR domain, the CPC failed to localize to centromeres, and defects in the spindle checkpoint response to taxol were reported, yet mitotic progression and cytokinesis could still occur (19, 20). Finally, microinjection of H3T3ph-specific antibody into LLC-PK or Xenopus S3 cells also caused mitotic defects consistent with loss of Aurora B activity at centromeres, while leaving intact its function in cytokinesis (Fig. 4D and movies S1 to S4), which further supported a role for H3T3ph in centromeric CPC function.

• Download high-res image
• Open in new tab
• Download Powerpoint

Our results are consistent with a model in which H3T3ph generated by Haspin enhances the binding of the CPC to centromeric chromatin and suggest a molecular basis for the importance of the Survivin BIR domain (9, 19, 20, 27). BIR domains in other molecules use the equivalent of residue Asp⁷¹ (D71) to bind the N-terminal Ala of the N-terminal “inhibitor of apoptosis (IAP)–binding motif” (IBM) of proteins such as Smac/DIABLO (28). Survivin binds Smac/DIABLO weakly, but the relevance of this interaction is unclear (29, 30). Instead, the IBM-binding region of the Survivin BIR domain may recognize the N-terminal Ala-Arg-Thr-Lys sequence of histone H3 when Thr³ is phosphorylated, which raises the possibility that other BIR domains act as phospho-specific protein recognition modules.

H3T3ph does not appear to be required for chromatin association or activity of the CPC per se. Rather, H3T3ph contributes to accurate positioning of the CPC at inner centromeres. In this respect, the CPC may resemble heterochromatin protein HP1, which binds histone H3 methylated at Lys⁹ but can also interact with chromatin by Lys⁹ methylation–independent mechanisms (31). In addition, H3T3ph prevents H3 from competitively inhibiting Aurora B autophosphorylation, which could facilitate Aurora B activation at inner centromeres (32). Therefore, Haspin might act through H3T3ph to both position and modulate activation of Aurora B at centromeres.

Loss of inner centromeric CPC accumulation has similar differential effects on centromeric Aurora B functions whether caused by Haspin depletion or Survivin-D70A/D71A mutation. Our results imply that proximity to inner centromeric Aurora B is not the sole determinant of substrate phosphorylation at centromeres, likely because substrates have different susceptibilities to phosphorylation and dephosphorylation by Aurora B and phosphatases (fig. S1). It may be valuable to further test the idea that Aurora B target sites are “programmed” to respond differently to microtubule-dependent displacement from the inner centromere, which would allow sophisticated regulation of kinetochore and centromere functions in response to tension.