The Spatiotemporal Dynamics of Chromatin Protein HP1α Is Essential for Accurate Chromosome Segregation during Cell Division*

Background: HP1α is a heterochromatin protein essential for chromosome plasticity in mitosis. Results: HP1α localization to the centromere depends on two distinct structural determinants in interphase and mitotic cells. Conclusion: The centromere localization of HP1α is determined by its binding to H3K9me2/3 in interphase but to PXVXL motifs in mitosis. Significance: The context-dependent spatiotemporal dynamics of HP1α is essential for accurate mitosis. Heterochromatin protein 1α (HP1α) is involved in regulation of chromatin plasticity, DNA damage repair, and centromere dynamics. HP1α detects histone dimethylation and trimethylation of Lys-9 via its chromodomain. HP1α localizes to heterochromatin in interphase cells but is liberated from chromosomal arms at the onset of mitosis. However, the structural determinants required for HP1α localization in interphase and the regulation of HP1α dynamics have remained elusive. Here we show that centromeric localization of HP1α depends on histone H3 Lys-9 trimethyltransferase SUV39H1 activity in interphase but not in mitotic cells. Surprisingly, HP1α liberates from chromosome arms in early mitosis. To test the role of this dissociation, we engineered an HP1α construct that persistently localizes to chromosome arms. Interestingly, persistent localization of HP1α to chromosome arms perturbs accurate kinetochore-microtubule attachment due to an aberrant distribution of chromosome passenger complex and Sgo1 from centromeres to chromosome arms that prevents resolution of sister chromatids. Further analyses showed that Mis14 and perhaps other PXVXL-containing proteins are involved in directing localization of HP1α to the centromere in mitosis. Taken together, our data suggest a model in which spatiotemporal dynamics of HP1α localization to centromere is governed by two distinct structural determinants. These findings reveal a previously unrecognized but essential link between HP1α-interacting molecular dynamics and chromosome plasticity in promoting accurate cell division.

Heterochromatin protein 1␣ (HP1␣) is involved in regulation of chromatin plasticity, DNA damage repair, and centromere dynamics. HP1␣ detects histone dimethylation and trimethylation of Lys-9 via its chromodomain. HP1␣ localizes to heterochromatin in interphase cells but is liberated from chromosomal arms at the onset of mitosis. However, the structural determinants required for HP1␣ localization in interphase and the regulation of HP1␣ dynamics have remained elusive. Here we show that centromeric localization of HP1␣ depends on histone H3 Lys-9 trimethyltransferase SUV39H1 activity in interphase but not in mitotic cells. Surprisingly, HP1␣ liberates from chromosome arms in early mitosis. To test the role of this dissociation, we engineered an HP1␣ construct that persistently localizes to chromosome arms. Interestingly, persistent localization of HP1␣ to chromosome arms perturbs accurate kinetochore-microtubule attachment due to an aberrant distribution of chromosome passenger complex and Sgo1 from centromeres to chromosome arms that prevents resolution of sister chromatids. Further analyses showed that Mis14 and perhaps other PXVXL-containing proteins are involved in directing localization of HP1␣ to the centromere in mitosis. Taken together, our data suggest a model in which spatiotemporal dynamics of HP1␣ localization to centromere is governed by two distinct structural determinants. These findings reveal a previously unrecognized but essential link between HP1␣-interacting molecular dynamics and chromosome plasticity in promoting accurate cell division.
During mitosis, the duplicated genomes must be precisely and equally segregated into two daughter cells in order to inherit parental characteristics. Mitotic chromosome segregation is orchestrated by the dynamic interaction of spindle microtubules with the kinetochore. The kinetochore is a super proteinaceous complex that assembles on the centromere of each chromosome. The kinetochore contains two functional domains; the outer layer mediates kinetochore-microtubule connection, and the inner layer connects to centromeric DNA (1).
After DNA duplication in S phase, the cohesin complex (composed of Scc1/Rad21, Scc3/SA2, Smc1, and Smc3) associates with the sister chromatids and prevents their dissociation before mitosis. The release of sister chromatid cohesion is regulated by two different pathways. In prophase, cohesion of chromosome arms is released after SA2 phosphorylation by Plk1. However, centromere cohesion is protected by the PP2A phosphatase complex, which is recruited by Sgo1 and antagonizes Plk1 function. In metaphase, the protease separase is activated and cleaves the SA2 cohesin subunit, subsequently releasing the centromere cohesion, which allows the sister chromatids to segregate into the two daughter cells (23)(24)(25)(26).
The molecular mechanism underlying localization of HP1␣ and the precise role of HP1␣ dynamics in mitosis remain poorly understood. In this study, we have revealed that HP1␣ localization is regulated by different molecular mechanisms during interphase and the onset of mitosis. Using H2B-HP1␣ fusion protein analysis, we have demonstrated that HP1␣ dissociation from chromosome arms is required for the correct centromere loading of chromosome passenger complex (CPC), Sgo1, Mis14, and MCAK. In addition, we show that forced localization of HP1␣ to chromosome arms results in Sgo1-mediated stabilization of sister chromatid cohesion and enhances spindle elongation, suggesting that temporal control of HP1␣ dissociation from chromosome arms orchestrates dynamic sister chromatid cohesion and spindle geometry. Together, our study demonstrates that HP1␣ orchestrates a hierarchical interaction essential for accurate chromosome dynamics in mitosis.
Cell Culture, Transfection, and Synchronization-HeLa and 293T cells from the American Type Culture Collection (Manassas, VA) were maintained as subconfluent monolayers in Dulbecco's modified Eagle's medium (Invitrogen) with 10% fetal bovine serum (FBS; Hyclone) and 100 units/ml penicillin plus 100 mg/ml streptomycin (Invitrogen) at 37°C with 8% CO 2 . LAP-hMps1 HeLa stable cell lines were maintained in 0.1 g/l G418. HeLa cells were transfected by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's manual and synchronized at G 1 /S with 2.5 mM thymidine for 16 h, washed with phosphate-buffered saline (PBS) three times, and then cultured in thymidine-free medium for appropriate time intervals.
Immunofluorescence, Chromosome Spread, and Live Cell Imaging-For immunofluorescence staining, cells were seeded onto sterile, acid-treated, 12-mm coverslips in 24-well plates (Corning Inc.). At 24 -36 h after the aforementioned transfection, HeLa cells were rinsed for 1 min with PHEM buffer (100 mM PIPES, 20 mM HEPES, pH 6.9, 5 mM EGTA, 2 mM MgCl 2 , and 4 M glycerol) and permeabilized for 1 min with PHEM plus 0.1% Triton X-100, as described previously (32). Extracted cells were fixed in freshly prepared 3.7% paraformaldehyde in PHEM and rinsed three times in PBS. The cells were blocked with 0.05% Tween 20 in PBS (TPBS) with 1% bovine serum albumin (Sigma-Aldrich). These cells were incubated with the various primary antibodies in a humidified chamber for 1 h and then washed three times in TPBS. Primary antibodies were visualized with FITC-conjugated goat anti-mouse, -rabbit, or -human IgG; rhodamine-conjugated goat anti-human IgG; or Cy5conjugated goat anti-mouse IgG. DNA was stained with DAPI (Sigma-Aldrich). For microtubule staining, HeLa cells were fixed by 3.7% paraformaldehyde in PTEM buffer (100 mM PIPES, pH 6.8, 10 mM EGTA, 1 mM MgCl 2 , 0.2% Triton X-100) for 10 min, washed three times in PBS, and blocked as described above.
In general, to produce the chromosome spreads, mitotic cells obtained by selective detachment were incubated in PEM buffer (0.1 M PIPES, 2 mM EGTA, 1 mM MgSO 4 , pH 6.8) for 10 min. After attachment to the glass coverslips by centrifugation at 1,000 rpm for 5 min, the chromosomal spreads were fixed with 4% paraformaldehyde in PHEM for 10 min and blocked with 1% BSA.
HeLa cells expressing different kinds of plasmids entered mitosis at 7 h after thymidine release. Transfected cells grown on glass-based dishes (MatTek) were replaced by CO 2 -independent medium (Invitrogen) supplemented with 10% FBS and observed using the DeltaVision RT system (Applied Precision) at 37°C. The mCherry images were taken at 5-min intervals with an exposure time of 0.1 s. Images were analyzed with Softworx software (Applied Precision).
Image Acquisition and Processing, Fluorescence Intensity Quantification-Immunofluorescence images were collected on an inverted microscope (Olympus IX-70) with a ϫ60, numerical aperture 1.42 Plan Apo N objective.
Step sections (0.25 m) were acquired to generate three-dimensional image stacks. Olympus acquisition parameters, including exposure, focus, and illumination, were controlled by Softworx (Applied Precision). The three-dimensional image stacks were deconvolved and projected; subsequent analysis and processing of the images were performed by using Softworx, Photoshop, and Illustrator (Adobe). All statistical analysis was performed with GraphPad Prism V5 (GraphPad Software, Inc.).
Quantification of the levels of kinetochore-associated proteins was described previously (33). Briefly, the average pixel intensities within a 7 ϫ 7-pixel square positioned over a single kinetochore were measured, and the background pixel intensities of a 7 ϫ 7-pixel square positioned in a region of cytoplasm lacking kinetochores were subtracted. Maximal projected images were used for these measurements, and the pixel intensities at each kinetochore pair were then normalized against ACA pixel values to account for any variations in staining or image acquisition. All of the fluorescence intensity measurement was quantified by ImageJ (National Institutes of Health).

Identification of Structural Determinants for HP1␣ Localization to Heterochromatin in Interphase and
Mitosis-HP1␣ consists mainly of two functional parts: the N-terminal CD, which is responsible for binding to H3K9me2/3, and the C-terminal CSD, which allows homo-or heterodimerization (Fig. 1A). Mounting evidence demonstrates three characterized functional sites on HP1␣: Val-22, Ile-165, and Trp-174. It has been demonstrated that the V22M mutant prevents its binding to H3K9me2/3, the I165E mutation perturbs its dimerization capacity, and the W174A mutation disrupts its association with proteins containing the PXVXL motif ( Fig. 1B) (34). However, it is not known whether those three sites function synergistically or independently to orchestrate the molecular dynamics of HP1␣ during different phases within the cell cycle.
To validate if HP1␣ localization to chromatin in interphase is totally dependent on the SUV39H1 pathway (35), aliquots of HeLa cells were transiently transfected to express GFP-HP1␣, followed by suppression of SUV39H1 activity either by siRNA-mediated knockdown or chemical inhibition using protocols that we established previously (33). As shown in Fig. 1C, transfection of siSUV39H1 resulted in an efficient repression of the endogenous SUV39H1 at 48 h post-transfection. Both the nuclear localization and the centromere localization of HP1␣ are evident in interphase cells treated with solvent DMSO. However, HP1␣ failed to localize to centromere in chaetocintreated cells or cells exposed to SUV39H1 siRNA (Fig. 1D). This result confirmed that the HP1␣ centromeric localization in interphase depends on SUV39H1 activity.
To further pinpoint the key residues responsible for HP1␣ localization, GFP-tagged HP1␣ V22M , HP1␣ I165E , and HP1␣ W174A mutants were generated, and their expression levels in HeLa cells were examined by Western blotting analyses (Fig. 1E). Our immunofluorescence analyses showed that HP1␣ V22M failed to localize to the centromeres in interphase cells (Fig. 1F, panel 2), which confirms that HP1␣ localization relies on H3K9me2/3. The HP1␣ I165E mutant was also unable to localize to centromeres (Fig. 1F, panel 3), suggesting that dimerization of HP1␣ is essential for a stable localization to the centromeres in interphase. Interestingly, the HP1␣ W174A mutant remained localized to the centromere region (Fig. 1F, panel 4).
To assess whether HP1␣ localization in mitosis depends on H3K9me2/3, HP1␣ localization was determined in the presence of chaetocin or SUV39H1 siRNA. Aliquots of HeLa cells were transiently transfected to express GFP-HP1␣ and followed with siRNA-mediated knockdown of SUV39H1 or chaetocin treatment as described previously (33). The transfected and treated cells were then subjected to an immunofluorescence study. As shown in Fig. 1G, HP1␣ localization to the centromeres was virtually unaltered by chaetocin treatment (panel 2), suggesting that centromeric localization of HP1␣ was not dependent on SUV39H1 activity in mitotic cells. In addition, HP1␣ localization to the centromeres was not abolished by depletion of SUV39H1 with siRNA ( Fig. 1G, panel 3). Thus, we conclude that SUV39H1-elicited H3K9me2/3 is not the structural determinant for HP1␣ localization to the centromere in mitotic cells.
We next sought to examine the respective roles of individual structural determinants in HP1␣ distribution in mitosis. To this end, aliquots of HeLa cells were transiently transfected to express wild type and HP1␣ point mutants, including GFP-HP1␣ V22M , GFP-HP1␣ I165E , GFP-HP1␣ W174A , and GFP-HP1␣ V22M/174A . These engineered proteins were expressed at comparable levels as judged by Western blotting analyses (Fig.  1E). As shown in the first two panels in Fig. 1H, the HP1␣ V22M mutant retained centromeric localization (panel 2), which is comparable with that of wild type HP1␣ (panel 1). This outcome is consistent with Fig. 1G, in which centromeric HP1␣ loading is independent of H3K9me2/3 in mitotic cells. Interest-ingly, GFP-HP1␣ I165E failed to localize at centromeres, suggesting that HP1␣ dimerization is required for a stable localization to the centromeres in both interphase and mitotic cells.
Surprisingly, in mitotic cells, the distribution of GFP-HP1␣ to centromeres was compromised by the W174A mutant, in contrast to its centromere-rich distribution in interphase cells. As shown in panel 4 (Fig. 1H), GFP-HP1␣ W174A exhibited a diffused distribution on entire chromosomes in mitotic cells. The diffuse HP1␣ chromosomal localization obscures whether there is residual HP1␣ localized to the centromeres and whether HP1␣ loading on chromosome arms is dependent on H3K9me2/3. To address these two concerns, we constructed a double mutant in which both Val-22 and Trp-174 were simultaneously mutated. As shown in Fig. 1H, GFP-HP1␣ V22M/W174A failed to localize to the chromosome arms or to the centromeres (panel 5), suggesting that the Val-22 cooperates with Trp-174 in loading HP1␣ W174A onto the chromosome. Thus, we conclude that both Val-22 and Trp-174 are required for accurate loading of HP1␣ onto chromosomes in mitosis.
The fact that mutation of the HP1␣ V22M mutant disrupted the centromeric localization of HP1␣ by further mutating Trp-174 led us to speculate that the loading of HP1␣ to the centromere is dependent on uncharacterized structural determinants containing the PXVXL motif instead of H3K9me2/3. To validate this hypothesis, we employed two chemical inhibitors to suppress SUV39H1 activity and Aurora B activity. As shown in Fig. 1I (panel 2), HP1␣ exhibits restricted distribution to centromeres but not to the entire chromosome, whereas suppression of Aurora B alone promotes chromosomal distribution of HP1␣. Thus, we conclude that the HP1␣ centromeric HP1␣ localization depends on H3K9me2/3 during interphase, but not during mitosis, which uses structural determinants containing the PXVXL motif.
HP1␣ Dissociation from Chromosome Arms Is Essential for Accurate Kinetochore Assembly-It is well established that HP1␣ dissociates from chromatin during entry into mitosis to facilitate chromosome dynamics. However, the physiological significance of this temporal regulation of HP1␣ distribution is less characterized. If HP1␣ dissociation from chromosome arms is essential for accurate mitosis, then persistent expression of HP1␣ on chromosome arms will perturb chromosome segregation in mitosis. To test our hypothesis, we generated fusion constructs containing H2B fused with HP1␣. As shown in Fig. 2A, aliquots of HeLa cells were transiently transfected to express mCherry-H2B-HP1␣ wild type and point mutants. The H2B-HP1␣ and H2B-HP1␣ W174A proteins exhibit a characteristic H2B distribution pattern on chromosome arms. If HP1␣ determines the localization of kinetochore components, persistent localization of HP1␣ would detour the localization of those components from kinetochore to chromosome arms. To this end, we employed an immunofluorescence assay to score the localization of 21 kinetochore components with distinct localization within the substructures of the kinetochore. As shown in Fig. 2B, six typical inner centromere proteins (i.e. Aurora B, INCENP, Borealin, Survivin, Sgo1, and Mis14) exhibit various degrees of concentration to chromosome arms in H2B-HP1␣expressing cells. As predicted, the abundance of the aforementioned proteins in the centromere was significantly reduced due to their relocation from centromeres to chromosome arms in mCherry-H2B-HP1␣-expressing cells ( Fig. 2C; p Ͻ 0.001). However, the W174A mutation blocks relocation of those proteins from the centromere to chromosome arms, demonstrating that the interaction of Trp-174 with the PXVXL motifs of those tested proteins determines their localization in mitotic cells. Although microtubule depolymerase MCAK was not associated with chromosome arms in mCherry-H2B-HP1␣-expressing cells, its centromeric localization was significantly reduced ( Fig. 2C; p Ͻ 0.001). As a control, the localization of other outer kinetochore proteins (i.e. outer kinetochore components Hec1, KNL1, CENP-E, BubR1, Mad1, SKAP, Ska1, and Zwint1 and inner kinetochore component CENP-H/I/L/U/ S/T) was not altered by the persistent localization of HP1␣ on the chromosome arms in H2B-HP1␣-expressing cells (Fig.  2B), 4 indicating that the overall structure of kinetochore was not grossly altered in H2B-HP1␣-expressing cells. Thus, we conclude that the spatial location of HP1␣ selectively governs the dynamics and precise distribution of CPC, Sgo1, Mis14, and MCAK via Trp-174.
It has been shown that loading of centromeric CPC depends on its association with Sgo1 upon Borealin phosphorylation by CDK1 (38). To assess whether the translocation of CPC to chromosome arms in H2B-HP1␣-expressing cells is a function of a direct interaction between HP1␣ and Borealin and whether such an interaction is regulated by CDK1 (39), aliquots of H2B-HP1␣-expressing cells were treated with CDK1 inhibitor RO3306 and control vehicle DMSO. As shown in Fig. 2D (panel  3), Aurora B relocates to chromosome arms in the presence of CDK1 inhibition, presumably due to the inhibition of CDK1elicited Borealin phosphorylation. Thus, the localization of HP1␣, and not Sgo1, provides the primary determinant for the spatial distribution of CPC in mitosis.
We next asked whether Aurora B kinase activity regulates CPC distribution in the presence and absence of HP1␣. To this end, aliquots of HeLa cells were transiently transfected to introduce HP1␣ siRNA. Forty-eight hours after the transfection, transfected cells were treated with 2.5 M ZM447439 or an equal volume of vehicle DMSO for 1 h before fixation and probing for Aurora B and ACA. As shown in Fig. 2E, the localization of CPC to the chromosome arms is negatively regulated by Aurora B kinase activity because chemical inhibition by ZM447439 promotes Aurora B retention on whole chromosome (panel 2). In order to examine how endogenous HP1␣ promotes chromosome Aurora B loading, we employed siRNA-mediated suppression of endogenous HP1␣ protein level. As shown in Fig. 2E (panel 4), suppression of HP1␣ attenuated the chromosomal localization of Aurora B localization (bottom panel, arrow). Thus, we conclude that localization of CPC on chromosome arms of mitotic cells is a function of HP1␣.
Because HP1␣ promotes the localization of CPC on chromosome arms, we next determined whether endogenous HP1␣ affects centromere-associated localization of CPC. To this end, endogenous HP1␣ was suppressed by siRNA. Because Haspin and Sgo1 are both known to promote centromeric CPC loading (40,41), Haspin and Sgo1 RNA interference were also performed as positive controls. As shown in Fig. 2F, HP1␣, Haspin or Sgo1 was efficiently suppressed by the corresponding siRNAs. Our quantitative immunofluorescence analyses show that the centromeric localization level of Aurora B was decreased to 57.1%, similar to that of Haspin-deficient cells (56.2%) and to that of Sgo1-deficient cells (60%) (Fig. 2, G and H, p Ͻ 0.001), indicating that HP1␣ is indeed required for centromeric CPC loading. Thus, we conclude that dynamic localization of HP1␣ is essential for accurate assembly of centromere/ kinetochore.
Chromosomal Arm HP1 Regulates Sister Chromatid Separation by Recruiting Sgo1-It has been reported that Sgo1 localization to chromosome arms is determined by HP1␣ (27). A recent study from the Ishizaka group (19) has demonstrated that HP1␣ and HP1␥ but not HP1␤ are required for cohesion of the chromosomal arms. To determine whether HP1␣ affects chromosomal arm cohesion through Sgo1 and CPC, we first examined whether HP1␣ forcibly localized to chromosomal arms inhibits segregation of sister chromatids. Cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A were synchronized at the G 1 /S phase. At 7 h after G 1 /S release, cells were treated with nocodazole for 3 h. Chromosome spreads were then prepared and examined under a fluorescence microscope. As shown in Fig. 3A, the squashed chromosomes could be divided into five groups: (i) intertwined (most sister chromatids are less condensed and intertwined); (ii) hypocondensed and unseparated (all sister chromatids are less condensed and unseparated); (iii) unseparated (most sister chromatids are unseparated); (iv) arm open (most sister chromatids are separated in chromosome arms but keep connected in the centromere); and (v) hypercondensed (all sister chromatids are highly condensed and unseparated). Chromosome squash from mCherry-H2B-expressing cells exhibits typical separated sister chromatids (mainly group iv and v phenotypes). Surprisingly, most squashed chromosomes from mCherry-H2B-HP1␣-expressing cells exhibit well separated sister chromatids (mainly group i, ii, and iii phenotypes), suggesting that ectopic expression of chromosome arm-localized HP1␣ disrupted normal chromosome cohesion, leading to premature separation of sister chromatids during mitosis. As a control, the majority of sister chromatids in H2B-HP1␣ W174A -expressing cells were arm-opened but unseparated (mainly group ii, iii, and v phenotypes) (Fig. 3B).
We next sought to examine whether the perturbation of sister chromatid cohesion seen in mCherry-H2B-HP1␣-expressing cells is due to an alteration of Sgo1 spatiotemporal dynamics. To this end, we examined the chromosome spreads from cells expressing Sgo1 WT and HP1␣ binding-deficient Sgo1 V453E (36) in the absence of endogenous Sgo1. As shown in Fig. 3, C and D, most sister chromatids were unseparated (56.5% of chromosome arms open and 40.5% of arms closed) in the control group. Suppression of endogenous Sgo1 resulted in premature separation of arms and centromeric cohesion of sister chromatids, which is consistent with previous reports (17). Reexpressing Sgo1 WT prevented premature sister chromatid separation (44.5% of chromosome arms separated but 52.5% of arms unresolved). However, re-expressing Sgo1 V453E was unable to prevent premature separation of sister chromatids (52.0% of chromosome arms open, 21.5% of arms completely separated, and 26.5% of arms closed). These results indicate that Sgo1 plays a functional role in safeguarding centromeric cohesion that depends on its association with HP1␣. Thus, we conclude that chromosome arm-associated HP1␣ recruits Sgo1 and thereby protects the cohesion of sister chromatids.
Although it has been reported that Aurora B is involved in the cohesion of chromosome arms (26), the underlying molecular mechanisms are not well understood. To delineate the aforementioned mechanisms, chromosome spreads were prepared to determine whether Aurora B is involved in the HP1␣-Sgo1 pathway. As shown in Fig. 3E, suppression of Sgo1 or combined HP1␣ϩ␥ resulted in a defect in sister chromatid cohesion (Fig. 3, E-G), which is consistent with several previous reports (19,26,42). Inhibition of PLK1 or Aurora B protected chromosome arm cohesion and led to closed chromosome arms (Fig. 3, E-G; p Ͻ 0.01). In addition, treatment with BI2536 was able to sustain sister chromatid cohesion in the absence of Sgo1 or HP1␣ϩ␥ (Fig. 3, E-G) because the inhibition of SA2 phosphorylation by PLK1 promotes cohesion (24). In contrast, inhibition of Aurora B by ZM447439 did not attenuate the loss of sister chromatid cohesion resulting from the suppression of Sgo1 or HP1␣ϩ␥ (Fig. 3, F-H). Collectively, these data suggest that the mechanisms by which Aurora B and PLK1 orchestrate the separation of sister chromatid cohesion are different and that Aurora B-elicited regulation of sister chromatid separation requires HP1␣-Sgo1 interaction.
HP1␣ Dissociation from Chromosome Is Essential for Faithful Spindle Length-After having examined the impact of persistent expression of HP1␣ on chromosome plasticity and centromere protein targeting, we then asked whether persistent expression of HP1␣ on chromosome arms affects spindle geometry. The spindle lengths were measured in cells express-  HeLa cells were treated with Sgo1 siRNA or HP1␣ϩ␥ siRNA for 24 or 48 h, respectively, and exposed to 2.5 M ZM447439 or 100 nM BI2536 for 10 h before centrifugation to generate chromosome spreads. Aurora B kinase inhibitor ZM447439 and PLK1 kinase inhibitor BI2536 were added to the cells after thymidine release, and 10 h after this treatment, chromosomes were spread, fixed, and then stained with ACA and DAPI. Scale bars, 10 m.  ing H2B-HP1␣. As shown in Fig. 4, A and B, the spindle length in H2B-HP1␣-expressing cells increased to 15.5 Ϯ 4.8 m compared with 12.0 Ϯ 2.6 m in cells expressing H2B. In contrast, spindle length is 12.3 Ϯ 4.1 m in cells expressing H2B-HP1␣ W174A , similar to that observed in control cells. The increased spindle length did not result from anaphase spindle elongation events because the cyclin B level remains unaltered (Fig. 4C).
Because MCAK regulates microtubule dynamics and spindle length (43) and forced expression of H2B-HP1␣ liberates the localization of MCAK from the kinetochore, we sought to examine directly whether the reduction of MCAK at the kinetochore accounts for an increased spindle length. To perform these studies, we engineered a fusion protein containing MCAK and the coiled-coil domain of Hec1 (amino acids 261-642, termed Hec1C, as illustrated in Fig. 4D) (44) to allow for a stable kinetochore localization of MCAK independent of HP1␣ regulation. The fusion proteins were expressed as correct sizes and at comparable levels (Fig. 4E). We next tested whether the expression of Hec1C-MCAK can override the alteration of spindle geometry in H2B-HP1␣-expressing cells. To this end, aliquots of HeLa cells were transiently transfected to express both Hec1C-MCAK and H2B-HP1␣. As shown in Fig. 4F, Hec1C-MCAK but not GFP-MCAK localized to the kinetochore in H2B-HP1␣-expressing cells, consistent with our experimental design. In addition, the localization of Hec1C-MCAK to the kinetochore was not affected by the Aurora B inhibitor ZM447439 (Fig. 4G). We next examined whether Hec1C-MCAK restored spindle length in H2B-HP1␣-expressing cells. As shown in Fig. 4, H and I, expression of Hec1C-MCAK dramatically shortened the spindle length from 15.5 Ϯ 4.8 to 13.8 Ϯ 3.6 m, suggesting that the liberation of kinetochore-associated MCAK by H2B-HP1␣ expression accounts for the spindle elongation. Because Aurora B-elicited phosphorylation inhibits MCAK depolymerase activity (45,46), we anticipated that the addition of Aurora B inhibitor would further shorten spindle length in the cells expressing both Hec1C-MCAK and H2B-HP1␣. Consistent with our prediction, inhibition of Aurora B further decreased the overall spindle length to 11.9 Ϯ 4.1 m. (Fig. 4, H and I). Thus, we conclude that spatial localization of HP1␣ is critical to orchestrate an accurate spindle length and geometry.
HP1␣ Dissociation from the Chromosome Arms Is Essential for Faithful Mitotic Progression-It has been previously demonstrated that centromeric Aurora B activity regulates the kinetochore localization of Mps1/TTK, a key spindle checkpoint protein kinase (47). Because the H2B-HP1␣ expression induced decreased Aurora B level at the centromere, we asked whether reduction of Aurora B level was sufficient to alter kinetochore localization of Mps1 or spindle assembly checkpoint activation. As shown in Fig. 5A, expression of H2B-HP1␣ results in a brief decrease of kinetochore-localized Mps1. Statistical analyses indicated that this alteration is significant ( Fig. 5B; p Ͻ 0.05), suggesting that decreased centromeric Aurora B in H2B-HP1␣expressing cells attenuates the kinetochore localization of Mps1. This reduction of Mps1 localization was released when Trp-174 is mutated, supporting the critical role of HP1␣ as an upstream determinant for functional kinetochore assembly, such as stable MPS1 localization.
The perturbation of spindle geometry by forced localization of HP1␣ on chromosome arms has prompted us to evaluate whether the expression of H2B-HP1␣ interferes with accurate mitotic progression. To this end, aliquots of HeLa cells were transiently transfected to express mCherry-H2B, mCherry-H2B-HP1␣, and H2B-HP1␣ W174A , followed by real-time imaging analyses. In cells expressing H2B-HP1␣, most cells entered anaphase 50 min after nuclear envelope breakdown, similar to that observed for cells expressing H2B or HP1␣ W174A (Fig. 5, C  and D), suggesting that forced expression of H2B-HP1␣ does not induce significant delay in mitotic progression or cause mitotic arrest. Careful examination of chromosome dynamics revealed that the fidelity of chromosomal segregation was seriously altered in mCherry-H2B-HP1␣-expressing cells. As shown in Fig. 5, D and E, 39% of the H2B-HP1␣-expressing cells segregated sister chromatids without achieving metaphase alignment, and 27% of the H2B-HP1␣-expressing cells exhibited chromatid bridges in anaphase, suggesting that the spindle checkpoint has been compromised. Furthermore, 25% of the H2B-HP1␣-expressing cells failed to complete cytokinesis (Fig.  5D, panel 3), suggesting that HP1␣ orchestrates a hierarchical interaction essential for accurate chromosome dynamics in mitosis. Consistent with this notion, HP1␣ W174A expression does not induce severe chromosome segregation defects (Fig. 5,  D and E).
To further elucidate the chromosome alignment defects in H2B-HP1␣-expressing cells, we carefully examined the kinetochore-microtubule attachment by immunofluorescence. As shown in Fig. 5F, overexpression of H2B-HP1␣, but not of H2B or of H2B-HP1␣ W174A , perturbed chromosome alignment efficiency. Those non-aligned chromosomes exhibit an appearance reminiscent of merotelic attachment (enlarged box). To ascertain the precise kinetochore-microtubule attachment on chromosomes with forced localization of HP1␣, aliquots of HeLa cells were transiently transfected to express mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A for 24 h, followed by a chilling on ice for 10 min to destabilize non-kinetochore microtubules (32). The transfected and chilled cells were then fixed and stained for tubulin and centromere markers. As shown in Fig. 5G, kinetochore microtubules in H2B-expressing cells remain intact and connected to the centromere. The magnified montage shows that spindle microtubules are captured by centromeres (top). However, many chromosomes were scattered around the spindle pole with unstable kinetochore fibers (Fig. 5G, middle) relative to those of the control cells, suggesting that HP1␣ dissociation from the chromosome arm during the onset of mitosis is essential for orchestrating a stable kinetochore-microtubule attachment for subsequent chromosome alignment at the equator. Consistent with this notion, kinetochore microtubules remain intact and stable in H2B-HP1␣ W174A -expressing cells (Fig. 5G, bottom).
Our real-time imaging analyses show that forced expression of H2B-HP1␣ promotes premature anaphase, as shown by sister chromatids segregating without achieving metaphase alignment. In addition, chromatid bridges in anaphase-like cells were readily apparent in H2B-HP1␣-expressing cells. To ascer-tain whether H2B-HP1␣-expressing cells are able to enter anaphase, we carried out immunofluorescence microscopic analyses of fixed H2B-HP1␣-expressing cells. Those fixed cells were stained with centromere marker ACA. As shown in Fig.  5H, sister chromatids are intertwined and unresolved in these H2B-HP1␣-expressing cells. In contrast, sister chromatids are fully segregated in cells expressing H2B or H2B-HP1␣ W174A , indicating that spatiotemporal dynamics of HP1␣ is essential for accurate mitotic progression.

DISCUSSION
In this study, we reported that the localization of HP1␣ in interphase and mitosis is orchestrated by distinct structural determinants. The centromere localization of HP1␣ is determined by its binding to H3K9me2/3 in interphase but to PXVXL motifs in mitosis. This HP1␣-dependent localization is essential for Aurora B kinase activity in the centromere and accurate chromosome segregation in mitosis. Further analyses showed that other PXVXL-containing proteins, such as Mis14, are involved in determining the localization of HP1␣ to the centromere in mitosis. Our data suggest a model in which spatiotemporal dynamics of HP1␣ localization to the centromere is governed by two distinct structural determinants.
During interphase, HP1␣ is broadly localized across a broader range of chromatin structures, including centromeres. However, its binding and localization to the chromosomal arm is eliminated in mitosis (21,22). H3K9me2/3 of histone H3 determines HP1␣ localization to heterochromatin in interphase cells via the CD of HP1␣. However, at the onset of mitosis, the majority of HP1␣ dissociates from chromosome arms and maintains a centromeric localization via PXVXL motifcontaining proteins through its CSD. Previously, we reported that centromeric H3K9me2/3 levels rise during mitosis (33). However, the centromeric localization of HP1␣ during mitosis is not exclusively dependent on H3K9me2/3 but also requires proteins containing the PXVXL motifs, such as INCENP or Mis14 (36,37). In addition, dimerization of HP1␣ is essential for its centromeric localization.
Although the dissociation of HP1␣ from chromosome arms during mitosis is well established (34,48), its physiological functions still remained unclear. By expressing persistent chromosomal arm-localized H2B-HP1␣, we observed that chromosomal arm HP1␣ disrupted the correct centromere/kinetochore localization of CPC, Sgo1, Mis14, and MCAK. These experimental data suggest that the physiological significance of HP1␣ release from chromosomal arms during mitosis is to promote accurate centromere/kinetochore assembly. Aurora B negatively regulates not only the localization of HP1␣ on the chromosomal arms but also its own localization to the chromosomal arms. Our studies suggest that there is an Aurora B loop on chromosomal arms consisting of Aurora B-H3S10ph-HP1-Borealin/INCENP-Aurora B and that inhibiting Aurora B activity prevents CPC and HP1␣ from dissociating. On the other hand, similar to Haspin and Sgo1, HP1␣ also affects centromeric CPC loading (40,41,49). Although the mechanism for this is currently elusive, we suggest that HP1␣ affects centromeric CPC assembly through providing a physical bridge for CPC to transport from the chromosomal arms to the pericentromere and then to the centromere.
This study has demonstrated that chromosomal arm-localized HP1␣ is responsible for Sgo1 localization. In addition, we show that HP1␣ is involved in sister chromatid arm cohesion protection through an HP1-Sgo1 pathway. Aurora B is also involved in cohesion dissociation of chromosomal arms (26), but its role and mechanism of action are also unknown. Our studies, however, suggested that Aurora B may exert its cohesion disassociation function through the HP1-Sgo1 pathway and that this molecular mechanism is different from that of Plk1. This mechanism is likely because we have observed that sister chromatid cohesion remained intact even in the absence of Sgo1 or HP1␣ when PLK1 was inhibited or SA2 was unphosphorylated. In contrast, sister chromatid cohesion was no longer maintained in the absence of Sgo1 or HP1␣ϩ␥ with low Aurora B activity. Thus, we reasoned that inhibition of Aurora B activity alone can protect sister chromatid cohesion due to an active H3K9me2/3-HP1-Sgo1 axis.
The forced chromosomal localization of HP1␣ also prevents kinetochore MCAK assembly. The reason for this could be decreased Aurora B levels because it has been shown that Aurora B activity promotes kinetochore MCAK localization (45,46). On the other hand, the decreased MCAK may be a factor responsible for the elongated spindle because forced kinetochore localized MCAK was only able to partially rescue the elongated spindle induced by H2B-HP1␣ expression. Obviously, it is possible that some other depolymerizing mechanism may also be involved in this process. Our experimental results presented here are the first to demonstrate that the chromatinbinding protein can affect spindle geometry and length via a FIGURE 5. HP1␣ dissociation from the chromosome arms is essential for faithful mitotic progression. A, representative images of HeLa cells stably expressing LAP-Mps1 that were transfected with mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A for 24 h and then fixed and stained with ACA and DAPI. Scale bars, 10 m. B, statistical analyses of relative LAP-Mps1 immunofluorescence intensity at the kinetochore (in each case, n Ͼ 50 kinetochores from more than 20 cells; mean Ϯ S.E. (error bars); *, p Ͻ 0.05). C, quantification of the time from NEB to anaphase onset in HeLa cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A . HeLa cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A were imaged live, and time was recorded between the start of nuclear envelope breakdown and the onset of anaphase. D, representative real-time images of HeLa cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A . Sister chromatid missegregation or failure of segregation or bridge can be seen in mCherry-H2B-HP1␣-expressing cells. Scale bars, 10 m. E, quantification of phenotypic changes observed in cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A , as shown in D. F, representative images of HeLa cells transfected with mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A . 7 h after released from thymidine block, cells were treated with 10 M MG132 for 2 h, and then stained with ACA and DAPI. Note that incorrect attachment is observed in some mCherry-H2B-HP1␣-expressing cells. Scale bars, 10 m. G, representative images of HeLa cells expressing mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A . Cells were cold-treated before fixation by chilling on ice for 10 min, fixed, and stained with ACA, anti-tubulin antibody, and DAPI. Scale bars, 10 m. H, HeLa cells expressing mCherry-H2B-HP1␣ showed severe sister chromatids segregation abnormalities. HeLa cells transfected with mCherry-H2B, mCherry-H2B-HP1␣, or mCherry-H2B-HP1␣ W174A were fixed 10 h after release from thymidine block and then stained with ACA and DAPI. Scale bars, 10 m. hierarchical interaction at the kinetochore-chromosome interface.
Aurora B activity is required for the effective kinetochore localization of Mps1 (47). Forced chromosomal localization of HP1␣ lowered centromeric CPC loading. As a consequence, this caused a decreased Mps1 kinetochore localization. In excellent agreement with the decreased Aurora B and Mps1 signal, we observed a high frequency of anaphase-lagging chromosomes (Fig. 5E), indicating compromised spindle assembly checkpoint function. In H2B-HP1␣-expressing cells, mitotic progression was not disturbed; however, sister chromatid alignment and segregation defects were obvious. The alignment defects may be due to incorrect kinetochore-MT attachment, and segregation defects may be caused by uncondensed or intertwined sister chromatids.
In summary, we have demonstrated that the structural determinants responsible for HP1␣ localization to the centromere are different between interphase and the onset of mitosis and that accurate localization of HP1␣ is essential for faithful mitotic progression. Furthermore, its dissociation from the chromosomal arms is required for sister chromatid segregation, chromosome condensation, spindle length control, chromosome alignment, and faithful segregation. However, a number of questions still need to be resolved, including the function of centromeric HP1␣. Although studies by Kang et al. (36) suggest that it is dispensable, our results show that knocking down HP1␣ levels and pulling HP1␣ toward the chromosomes both inhibit the centromeric assembly of CPC. Currently, it is unclear how centromere-associated HP1␣ affects the upstream localization of CPC at the mitotic centromere. Further questions that still need to be addressed by future studies include the following. (i) How was MCAK centromeric localization liberated by H2B-HP1␣ expression? (ii) Are there any other microtubule depolymerases involved in the elongated spindle regulation induced by H2B-HP1␣ expression? (iii) How is chromosome condensation affected in H2B-HP1␣-expressing cells? (iv) Despite the fact that we have shown that the H2B-HP1␣ W174A mutation abolishes the phenotypes associated with H2B-HP1␣ expression, what are the other proteins that regulate spindle geometry via interaction with HP1␣ Trp-174? (v) Chromosome arm-localized HP1␣ orchestrates the cohesion between sister chromatid arms via recruiting Sgo1; do other cohesion protection proteins, such as Pds5 or Sororin, contribute to this process? The answers to all of the aforementioned questions and molecular delineation of underlying mechanisms will better our understanding of HP1␣ functional roles in mitotic progression.