Edinburgh Research Explorer CENP-B creates alternative epigenetic chromatin states permissive for CENP-A or heterochromatin assembly

CENP-B binds to CENP-B boxes on centromeric satellite DNAs (known as alphoid DNA in humans). CENP-B maintains kinetochore function through interactions with CENP-A nucleosomes and CENP-C. CENP-B binding to transfected alphoid DNA can induce de novo CENP-A assembly, functional centromere and kinetochore formation, and subsequent human artificial chromosome (HAC) formation. Furthermore, CENP-B also facilitates H3K9 (histone H3 lysine 9) trimethylation onalphoid DNA, mediatedby Suv39h1, at ectopic alphoid DNA integration sites. Excessive heterochromatin invasion into centromere chromatin suppresses CENP-A assembly. It is unclear how CENP-B controls such different chromatin states. Here, we show that the CENP-B acidic domain recruits histone chaperones and many chromatin modifiers, including the H3K36 methylase ASH1L, as well as the heterochromatin components Suv39h1 and HP1 (HP1 α , β and γ , also known as CBX5, CBX1 and CBX3, respectively). ASH1L facilitates the formation of open chromatin competent for CENP-A assembly on alphoid DNA. These results indicate that CENP-B is a nexus for histone modifiers that alternatively promote or suppress CENP-A assembly by mutually exclusive mechanisms. Besides the DNA-binding domain, the CENP-B acidic domain also facilitates CENP-A assembly de novo on transfected alphoid DNA. CENP-B therefore balances CENP-A assembly heterochromatin formation on satellite DNA.

At human centromeres, CENP-A assembles on a portion of a large repetitive DNA locus containing α-satellite DNA (alphoid DNA) (Willard and Waye, 1987). Most of the flanking satellite DNA regions ( pericentromeres) are occupied by heterochromatin (Sullivan and Karpen, 2004) enriched with Suv39h1-mediated trimethylation of histone H3 lysine 9 (H3K9me3), which recruits HP1 (HP1α, β and γ, also known as CBX5, CBX1 and CBX3, respectively). Heterochromatin is reported to have important roles in chromosome stability and maintenance of sister chromatid cohesion during chromosome segregation (Peters et al., 2001;Grewal and Jia, 2007). Thus, both centromere chromatin and heterochromatin are important for accurate chromosome segregation processes.
CENP-B is required for de novo CENP-A assembly to generate human artificial chromosomes (HACs) (Harrington et al., 1997;Ikeno et al., 1998) that segregate during cell division similarly to natural chromosomes (Tsuduki et al., 2006). When the CENP-B boxes on input alphoid DNA are mutated, HAC formation is diminished (Ohzeki et al., 2002). De novo CENP-A assembly also fails when alphoid DNA is introduced into CENP-B knockout (KO) mouse embryonic fibroblast (MEF) cells (Okada et al., 2007). In contrast, CENP-B is not required for the maintenance of established centromeres. No CENP-B box is found on the Y chromosome alphoid DNA (Masumoto et al., 1989;Earnshaw et al., 1991) or ectopically formed neocentromeres on chromosome arms (du Sart et al., 1997;Alonso et al., 2003). In addition, CENP-B KO mice are viable and fertile for at least one generation (Kapoor et al., 1998;Hudson et al., 1998;Perez-Castro et al., 1998).
Recent studies suggest a possible explanation for why CENP-B might not be essential for centromere maintenance and function. CENP-B helps stabilize kinetochore structure through interaction with CENP-A nucleosomes and CENP-C via its N-and C-terminal domains, respectively (Suzuki et al., 2004;Fachinetti et al., 2015;Fujita et al., 2015). However, CENP-C also binds to the C-terminal tail of nucleosomal CENP-A (Carroll et al., 2010;Fachinetti et al., 2013). Thus, the CENP-C-CENP-A direct interaction can bypass the requirement for CENP-B. In the absence of CENP-B, mutation of the C-terminal tail of CENP-A perturbs chromosome segregation because of the loss of the connection between CENP-C and centromeric chromatin (Fachinetti et al., 2015;Hoffmann et al., 2016). Therefore, CENP-B has important roles not only in de novo centromere formation, but also in centromere maintenance as a backup system.
In contrast to its function in CENP-A assembly and stabilization, CENP-B can also promote heterochromatinization. In CENP-B KO MEF cells containing human alphoid DNA integrated on a chromosome arm, CENP-B expression results in increased deposition of H3K9me3 on the alphoid DNA by Suv39h1 (Okada et al., 2007). Our previous studies using a synthetic alphoid DNA sequence containing a tetR binding site, tetO (alphoid tetO ), in the alphoid tetO HAC revealed that tethering of heterochromatin factors to the alphoid tetO HAC inhibits CENP-A assembly (Nakano et al., 2008;Cardinale et al., 2009;Ohzeki et al., 2012;Shono et al., 2015;Martins et al., 2016). Thus, CENP-B can promote alternative epigenetic chromatin states favoring either CENP-A assembly or heterochromatin formation on the same alphoid DNA. The underlying mechanism for this 'bidirectional' behavior is unknown.
In this study, we analyzed the function of CENP-B and identified proteins assembled by each CENP-B sub-domain. We identified not only the heterochromatin-promoting methyltransferase Suv39h1 and its binding protein HP1, but also new open chromatin-related modifiers and centromere proteins. Interestingly, one of the identified factors, H3 lysine 36 methyltransferase ASH1L, facilitated open chromatin formation competent for de novo CENP-A assembly on exogenous alphoid DNA in HT1080 cells. In addition, ASH1L depletion increased heterochromatin formation, but decreased CENP-A assembly on endogenous centromeric alphoid DNA in HeLa cells. ASH1L might be a key factor rendering CENP-B-bound alphoid chromatin competent for CENP-A assembly by avoiding excessive heterochromatinization.

RESULTS
Exploration of factors that assemble on alphoid chromatin in a CENP-B-dependent manner CENP-B-dependent factors that assemble on alphoid DNA might be involved in bidirectional chromatin formation. We used a fluorescence microscopy-based interaction-trap (FMIT) assay (Shono et al., 2015;Ohzeki et al., 2016) to identify centromere proteins and histone modifiers recruited to ectopic alphoid tetO DNA by tetR-EYFP-CENP-B in the HeLa-Int-03 CENP-B KO cell line. This cell line was obtained by knocking out the CENP-B gene of HeLa-Int-03 to eliminate the influence of endogenous CENP-B on the 'prey' assembly (Ohzeki et al., 2012) (Fig. S1A-D). This assay detects the recruitment of proteins of interest (POI) at the ectopic chromosomal site within 1-2 days after tetR-EYFP-CENP-B transfection (Fig. 1A).
Evaluation of factors shown to promote endogenous CENP-A assembly on the HAC centromere Next, we evaluated whether each candidate obtained in the experiments shown in Fig. 1 had a positive or negative impact on centromeric CENP-A levels. Previous studies showed that endogenous CENP-A levels on the alphoid tetO HAC can be increased by tethering CENP-A deposition factors (e.g. HJURP) (Dunleavy et al., 2009;Foltz et al., 2009), CENP-A nucleosome stabilization factors and kinetochore proteins interacting with CENP-C to the HAC (Shono et al., 2015). Conversely, tethering of heterochromatin factors (e.g. Suv39h1) decreased CENP-A levels. We therefore evaluated whether the additional factors identified here affect CENP-A levels on the alphoid tetO HAC (HeLa-HAC-2-4) ( Fig. 2A,B).
We could not detect EYFP signals following transfection with constructs expressing full-length tetR-EYFP-ASH1L and tetR-EYFP-NSD1. These proteins were therefore sub-divided into regions including the C-terminal SET domain and the remaining N-terminal region. Besides tetR-EYFP-HJURP, tethering of tetR-EYFP-ASH1L 1901-2964 also enhanced the endogenous CENP-A signal on the alphoid tetO HAC (Fig. 2C,D). Conversely, tethering of the heterochromatin factors HP1α, HP1β, HP1γ and Suv39h1 decreased CENP-A, as previously reported (Shono et al., 2015;Martins et al., 2016). We therefore focused on ASH1L, which increases centromeric CENP-A levels, and HP1, which decreases centromeric CENP-A levels, in the following analyses. Fig. 1. Screening of proteins for CENP-B-dependent assembly on ectopic alphoid tetO DNA using a fluorescence microscopy-based interaction-trap (FMIT) assay. (A) Experimental design of the FMIT assay. Expression plasmids of a bait or prey are co-transfected into the CENP-B KO HeLa-Int-03 cell line, which has a synthetic alphoid repeat array containing tetO and CENP-B box sequences (alphoid tetO DNA) at an ectopic integration site. Colocalization of EYFP signal and the Halo tag-TMR ligand signal can be detected when an interaction exists between bait and prey and/or when the prey recognizes an alphoid tetO chromatin change induced by the bait. Negative and positive example images are shown in the upper right and lower right panels, respectively. In the positive case, tetR-EYFP-CENP-B Full recruits the Halo-tag fused to the CENP-B dimer domain (Halo-CENP-B 541-599). The yellow squares indicate the location of tetR-EYFP protein spots on the alphoid tetO DNA site (shown magnified in inset images). Scale bars: 5 µm. (B) Schematic structure of the bait (tetR-EYFP-CENP-B) and negative control (tetR-EYFP alone) used in this figure. DBD, DNA-binding domain; DDE, DDE-superfamily endonuclease domain; Acidic:1, acidic domain 1; Acidic:2, acidic domain 2; Dimer, dimerization domain. (C) Exploration of protein assembly on the ectopic alphoid tetO DNA integration site by CENP-B tethering. The bait and prey expression plasmids were co-transfected into the HeLa-Int-03 CENP-B KO cell line. The cells were fixed 1 d after transfection. More than 50 cells were analyzed to calculate the frequency of Halo positive EYFP-spots for each prey. The tetR-EYFP-CENP-B binds to not only tetO, but also to the CENP-B box. (D) CENP-B-independent (tetR-EYFP alone bait, blue bars) and CENP-B-dependent (tetR-EYFP-CENP-B bait, orange bars) assembly of Halo-tag fused histone modifiers, chaperones, CENPs and heterochromatin proteins was evaluated as the percentage of cells exhibiting colocalization of bait and prey on the ectopic alphoid tetO integration site. K4me, K9me, K27me and K36me indicate H3 lysine methyl transferases. K9de, K27de and K36de indicate H3 lysine demethylases. CBX, chromobox proteins; PcG, polycomb-group proteins; HDAC, histone deacetylases; HAT, histone acetyltransferases.
CENP-B facilitates ASH1L and HP1β localization to alphoid DNA ASH1L is a trithorax-group protein that methylates H3K36, and is reported to be mutually exclusive with polycomb group-mediated H3K27-methylated facultative heterochromatin (Schuettengruber   Miyazaki et al., 2013). ASH1L localization and function at centromeres has not previously been reported, but H3K36me2/3 has been reported to be present at centromeres (Bergmann et al., 2011;Bailey et al., 2016). Therefore, we first investigated whether endogenous ASH1L and HP1β localize to centromeres in a manner dependent on CENP-B binding [comparing wild-type CENP-B (CENP-B WT ) and CENP-B KO in HeLa-Int-03 cells] using anti-ASH1L or anti-HP1 antibodies combined with an anti-CENP-A antibody. HP1β is distributed across a rather wide area of the centromere, including the pericentromere, and does not localize as a focused spot like the CENP proteins and CCAN. Therefore, we developed the approach shown in Fig. 4A to map the distribution of ASH1L and HP1β fluorescence intensity at centromeres. We quantified and integrated the ASH1L and HP1β fluorescence intensity across a region from the pixel at the center of the CENP-A signal out to a radius of 11 pixels. We did this for each of the 45 centromeric CENP-A signals per cell. The distributions of fluorescence intensity of ASH1L and HP1β were both significantly higher in the CENP-A-proximal region in CENP-B WT cells. However, this gradient of localization was lost in CENP-B KO cells ( Fig. 4B-D).
Next, we demonstrated by immunostaining and ChIP analysis that CENP-B binding to an ectopic alphoid tetO DNA integration array also recruited endogenous ASH1L and HP1β in HeLa-Int-03 cells (array visualized by tethering tetR-EYFP, Fig. S3B,C). The ChIP recovery rate of ASH1L with alphoid DNA was reduced by 75% in the chromosome 21 centromere, and 80% in the integration site, in CENP-B KO cells compared to those in CENP-B WT cells (Fig. 4G). Similarly, the recovery rate of HP1β was reduced by 53% in the centromere and 47% in the integration site. In CENP-B KO cells, no significant decrease was detected in the total amount of cellular ASH1L and HP1β (Fig. S3A). Thus, ASH1L and HP1β assembly on alphoid DNA requires CENP-B binding.
These analyses also indicated that ASH1L and HP1β are not equally distributed around the centromeres of each cell. When CENP-B, ASH1L and HP1β were triple-stained and analyzed by the above method (Fig. 4A), it was found that ASH1L fluorescence was weak in cells where HP1β fluorescence was strong in the centromere region. Conversely, HP1β fluorescence was weak in cells where ASH1L fluorescence was strong in the centromere region ( Fig. 4E,F). We conclude that ASH1L and HP1β localization to centromeres appears to be mutually exclusive.

ASH1L and HP1 assemblies exhibit mutually exclusive properties
We have shown that the CENP-B acidic region (403-556) recruits ASH1L and/or HP1 to an ectopic alphoid tetO array (Fig. 3) and that CENP-B also recruits both ASH1L and/or HP1 to endogenous centromeres (Fig. 4). However, the effects of ASH1L and HP1 on CENP-A chromatin assembly are diametrically opposed (Fig. 2C,D). We therefore asked how the binding of ASH1L or HP1β influences each other and centromere function.
Next, we determined the effect of tethering tetR-EYFP-HP1β or tetR-EYFP-ASH1L to the ectopic alphoid tetO array on the localization of endogenous ASH1L or HP1β. Tethering tetR-EYFP-HP1β reduced the binding of endogenous ASH1L, and vice versa (endogenous ASH1L or HP1β were reduced by 80%) (Fig. 5B-D). It thus appears that ASH1L and HP1β assembly might be mutually exclusive on alphoid DNA.
We then examined effects of tethering ASH1L or HP1β on centromere function. We measured the mitotic stability of the alphoid tetO HAC tethered with tetR-EYFP-ASH1L or tetR-EYFP-HP1β by counting EYFP HAC signals per nucleus (Fig. 5E). After tethering the control tetR-EYFP construct for 1 week, less than 8% of nuclei lacked a HAC signal (HAC=0) and over 83% of nuclei maintained one copy of the HAC per nucleus (HAC=1). In contrast, upon tetR-EYFP-HP1β tethering 28.8% of nuclei lost the HAC signal. Similarly, 47.1% of nuclei lost the HAC signal after tetR-EYFP-ASH1L tethering. It therefore appears that unbalanced assembly of ASH1L or HP1β on the HAC impairs centromere function.

Overexpression of heterochromatin factors in ASH1Ldepleted cells reduces chromosomal stability
We investigated whether ASH1L depletion affects the chromatin status of alphoid DNA in HeLa cells using ChIP analysis. ASH1L siRNA treatment reduced total cellular ASH1L levels to 3.9% (Fig. 6B). ASH1L enrichment on the chromosome 21 centromeric alphoid DNA (alphoid chr.21 ) and the ectopic alphoid tetO array were both significantly decreased by ASH1L depletion (∼69% and 79%, respectively) (Fig. 6A). We also quantified H3 modifications, and detected a significant decrease in H3K36me2/3 levels and increased H3K9me3 and H3K27me2 levels following ASH1L depletion. Thus, the ASH1L depletion results showed the opposite effects to the effects of ASH1L tethering shown in Fig. 5A, and the effects were consistent. Interestingly, CENP-A levels on the centromeric alphoid chr.21 significantly decreased by 27% in ASH1L-depleted HeLa cells.
Next, we examined the effects of perturbing the balance of ASH1L and HP1s on natural chromosomes using siRNA depletion of ASH1L followed by overexpression of heterochromatin proteins. We then scored for the presence of micronuclei and lagging chromosomes ( Fig. 6C,D). Five days after transfection with ASH1L siRNA, cells were transfected with a plasmid expressing Halo-tagged heterochromatin proteins HP1α, HP1β, HP1γ or Suv39h1. In cells treated with control scrambled siRNA (siScr), the frequency of micronuclei and lagging chromosomes was not significantly increased by overexpression of the heterochromatin proteins (control Halo-tag alone, 4.6%; Halo-HP1α, 4.1%; Halo-HP1β, 4.2%; Halo-HP1γ, 4.6%; and Halo-Suv39h1, 4.5%). In contrast, after ASH1L depletion, the frequency of micronuclei was increased roughly twofold upon expression of the control Halo-tag alone (8.4%). Overexpression of Halo-tagged heterochromatin proteins further increased the micronucleus frequency (Halo-HP1α, 13.2%; Halo-HP1β, 11.6%; Halo-HP1γ, 14.1%; and Halo-Suv39h1, 13.2%). In CENP-B-depleted cells, the micronucleation frequency was slightly but significantly increased even by the control Halo-tag alone (6.3%), but not further increased by overexpressing Halo-tagged heterochromatin proteins. This can be explained by the lack of recruitment of heterochromatin proteins by CENP-B, which consequently prevents heterochromatin from being excessively assembled into centromeres.
Taken together, these results suggest that ASH1L may act on CENP-B chromatin to suppress the excessive assembly of HP1 and heterochromatin factors. Conversely, HP1 acts on CENP-B chromatin to suppress excessive assembly of ASH1L ( Fig. 5A-D). It thus appears that CENP-B and these recruiting factors create a 'bidirectional' alternative epigenetic chromatin balance that promotes accurate chromosome segregation (Fig. 6E).

CENP-B affects the methylation-level of histone H3 on alphoid DNA in HT1080 cells
In this study, HeLa cells were used to demonstrate the balance of CENP-B-dependent ASH1L and HP1β assembly on alphoid DNA.
However, most previous studies, including ours, have used human fibrosarcoma cells (HT1080) to demonstrate de novo HAC formation and de novo CENP-A assembly with the introduced alphoid DNA. Therefore, we confirmed the CENP-B-dependent histone modification balance using alphoid DNA introduced into HT1080 cells. HT1080 cells exhibit much less heterochromatin assembly activity on alphoid DNA than HeLa cells (Ohzeki et al., 2012). A preliminary experiment showed that robust CENP-A and CENP-B assembly activities with introduced wild-type CENP-B box alphoid tetO DNA (which can bind to CENP-B) were detected at 2 weeks after transfection using ChIP-qPCR in HT1080 cells (Fig.  S4A). Conversely, mutant CENP-B box alphoid tetO DNA (which does not bind to CENP-B) did not show either CENP-A or CENP-B assembly. We therefore analyzed the H3 modification status in parallel with CENP-A chromatin formation on transfected alphoid DNA bearing either wild-type or mutant CENP-B boxes using a sensitive ChIP-competitive PCR analysis (Fig. 7B,C) at early time points after co-transfection (Ohzeki et al., 2002;Okada et al., 2007). The competitive PCR can accurately compare the abundance ratio between wild-type and mutant CENP-B box alphoid DNA (Fig. 7A).
Relative enrichment of CENP-A was detected on the wild-type CENP-B box alphoid tetO DNA 4 days after transfection (Fig. 7B,C). CENP-C assembly was also detected on the wild-type CENP-B boxes with slower kinetics at 1 week after transfection. Analysis in parallel also detected a comparatively early relative enrichment for H3K36me2 and H3K36me3 on the wild-type CENP-B boxes, within 4 days after transfection. A similar, albeit delayed, enrichment for H3K27ac was also detected at 1-2 weeks after transfection. Relative enrichments of H3K27me2 and me3, which have been reported to be mutually exclusive with H3K36 methylation (Yuan et al., 2011), were detected on the mutant CENP-B box alphoid tetO DNA 1-2 weeks after transfection (Fig. 7B,C). These results suggest that CENP-Bdependent methylation status changes occur earlier for H3K36me2/3 than for H3K27me2/3. In addition, these changes may include demethylation as well as methylation, because several demethylases were also identified in Fig. 1. These results can be explained if assembly of ASH1L or similar factors on alphoid DNA requires the presence of CENP-B. Indeed, we confirmed ASH1L localization at the centromere in HT1080 cells (Fig. S4B), and dominant assembly of ASH1L on transfected alphoid DNAs with CENP-B boxes continued for at least 2-4 weeks (Fig. S4C,D).
The CENP-B acidic domain and DNA-binding domain both assemble CENP-A de novo on exogenous alphoid DNA in HT1080 Fig. 3 shows that the CENP-B acidic domain recruits various factors, including ASH1L. Thus, we decided to investigate the involvement of the acidic domain on de novo CENP-A assembly. To analyze the mechanisms by which CENP-B promotes de novo centromere formation, we attempted to identify CENP-B domains involved in CENP-A chromatin formation on transfected alphoid DNA. However, this was complicated because the CENP-B Nterminal DNA-binding domain (DBD) is both essential for CENP-B box binding and also interacts with CENP-A nucleosomes to induce de novo CENP-A assembly (Okada et al., 2007;Fujita et al., 2015).
To bypass the requirement for CENP-B DBD-CENP-B box interaction, we tethered tetR-fused CENP-B to synthetic alphoid DNA containing a tetO sequence (alphoid tetO DNA). Bacterial artificial chromosomes (BACs) containing alphoid tetO DNA arrays carrying the wild-type or mutant CENP-B boxes were introduced into HT1080 cells stably expressing tetR-EYFP (control) or tetR-EYFP-CENP-B in the presence or absence of doxycycline (Dox+, no tethering; Dox−, tethering) (Fig. 8A). CENP-A assembly was detected on wild-type CENP-B box alphoid tetO DNA in cell lines expressing either tetR-EYFP or tetR-EYFP-CENP-B in the presence or absence of doxycycline. This assembly reflects interactions between the endogenous CENP-B DBD and CENP-B box DNA.
In contrast, mutant CENP-B box alphoid tetO DNA failed to assemble CENP-A in cells expressing either tetR-EYFP or tetR-EYFP-CENP-B in the presence of doxycycline. However, when doxycycline was absent (allowing tetR binding to the tetO sequences), CENP-B and CENP-A assembly were readily detected on the alphoid tetO DNA (Fig. 8A). Thus, CENP-B tethering to the alphoid tetO DNA via the tetR-tetO interaction can bypass the requirement for CENP-B boxes in de novo CENP-A assembly.
Next, we used the tetR tethering assay to assess CENP-A assembly activity promoted by various CENP-B sub-domains. Mutant CENP-B box alphoid tetO DNAs were transfected into cell lines stably expressing tetR-EYFP fused to a series of CENP-B deletions. Significant CENP-A assembly was detected with fusions carrying the N-terminal DBD and also fusions carrying the acidic domains (amino acids 403-556) (Fig. 8B). CENP-A assembly activity was low following tethering of the CENP-B dimer domain, but significantly above background. This is likely explained by interactions with endogenous full-length CENP-B (Fig. 3B) (Kitagawa et al., 1995). Thus, both the DBD and acidic domain of CENP-B can promote de novo CENP-A assembly on transfected alphoid DNA.
A HAC formation assay (Ikeno et al., 1998;Ohzeki et al., 2002) confirmed that CENP-A assembly via CENP-B tethering can induce functional centromere formation on exogenous mutant CENP-B box alphoid tetO DNA. When transfections were done in the absence of doxycycline, HACs were formed in two of 34 transformed HT1080 cell lines expressing tetR-EYFP-CENP-B Full (5.9%). In contrast, in the presence of doxycycline, no HAC was detected in 39 cell lines (Fig. 8C). No HACs were detected in HT1080 cell lines expressing either tetR-EYFP-CENP-B DBD or tetR-EYFP-CENP-B acidic domain . Thus, tetR-fused, full-length

(E) Influence of ASH1L or
HP1β tethering on the chromosome (HAC) stability. HACs were visualized as tetR-EYFP fluorescence spots. Initially, the tetR-EYFP fusion-expressing alphoid tetO HAC-bearing cells were cultured in the presence of doxycycline (without tethering condition). The tetR-EYFP-fused protein was then tethered to HAC by washing doxycycline out. A time course experiment after doxycycline washout (6 h to 1 week) was performed. HACs were counted as the number of tetR-EYFP spots. Gray, orange and blue bars indicate the percentage of cells containing ≥2, 1 or 0 HACs, respectively. More than 50 cells with detectable EYFP signal were analyzed for each assay. Results are mean±s.e.m. (n=3 experiments). Scale bars: 5 µm.
CENP-B tethered to mutant CENP-B box alphoid DNA via tetO retains the ability to assemble de novo CENP-A competent for functional centromere formation, but either the N-terminal DBD or the acidic domain alone does this with reduced effectiveness. Further analysis is needed to determine whether either domain alone has the ability to assemble functional centromeres at low efficiency.
Furthermore, to investigate whether ASH1L is involved in the de novo CENP-A assembly pathway promoted by the CENP-B acidic domain in HT1080 cells (Fig. 8B), we depleted ASH1L and used ChIP-qPCR to quantify CENP-A assembly following tethering of tetR-EYFP-CENP-B 403-556 to alphoid tetO DNA with mutant CENP-B boxes. Levels of de novo CENP-A assembly and H3K36 methylation on transfected mutant CENP-B box alphoid tetO DNA were significantly decreased in the ASH1Ldepleted cells (Fig. S5B). Since anti-CENP-B antibody showed no enrichment of mutant CENP-B box alphoid tetO DNA, it is obvious that endogenous CENP-B is not involved in this CENP-A assembly reaction. Importantly, no significant change was detected in total cellular CENP-A levels following ASH1L depletion to 6.3% of control levels (Fig. S5A). These results suggest that ASH1L is critical also for de novo CENP-A assembly via the CENP-B acidic domain in HT1080.

CENP-B is a remarkable protein that can promote 'open' or 'closed'
chromatin states on alphoid DNA, with profound consequences for centromere activity and kinetochore assembly. Here, we have shown that two independent CENP-B domains can stimulate de novo CENP-A assembly on introduced alphoid DNA. In addition to the N-terminal DNA-binding domain, which can bind to CENP-A nucleosomes (Fujita et al., 2015;Fachinetti et al., 2015), the CENP-B acidic domain (amino acids 403-556; Earnshaw, 1987b) facilitated assembly of the H3K36 methyltransferase ASH1L, thereby inducing an open chromatin state competent for CENP-A assembly or exchange. Thus, the CENP-A assembly mechanism involving the acidic domain is quite different from that involving the DBD. Importantly, full-length CENP-B containing both domains could efficiently induce functional centromere and HAC formation on transfected alphoid DNA.

Functionally antagonistic histone modifications induced by CENP-B binding
We previously reported that CENP-B facilitates H3K9 methylation at an ectopic alphoid DNA integration site on chromosome arms in MEF cell lines (Okada et al., 2007). Here, we have confirmed that CENP-B binding promotes Suv39h1 and HP1 assembly and H3K9 methylation on alphoid DNA at both ectopic integration sites and endogenous centromeres in human cell lines. Importantly, CENP-B-induced H3K9 methylation only occurs subsequent to CENP-A and CENP-C assembly on exogenous alphoid DNA (Ohzeki et al., 2012). Remarkably, H3K36 methylation, a mark associated with active transcription, was also found on alphoid DNA arrays and occurred with similar timing to CENP-A and CENP-B assembly. Our results demonstrate that ASH1L and/or HP1 recruited by the CENP-B acidic domain assemble mutually exclusively and dynamically on alphoid DNA to 'bidirectionally' generate alternative epigenetic chromatin, whose balance is important to maintain centromere function (Figs 4-6).
CENP-B binding to the CENP-B box is inhibited by DNA methylation, which is typically associated with transcriptionally inactive chromatin (Tanaka et al., 2005;Okada et al., 2007). Centromeric chromatin states might be dynamic and/or exchangeable with each other during cell cycle progression, differentiation or senescence. Suv39h1 expression levels and H3K9me3 levels on chromosomes in HeLa cells are much higher than those in HT1080 cells (Ohzeki et al., 2012). Interestingly, levels of CENP-B at centromeres also differs between cell types and during senescence. Moreover, the balance of heterochromatin formation at centromeres, as revealed by HP1 and CENP-A accumulation, can change as human embryonic fibroblast cells undergo senescence (Maehara et al., 2010).
Antagonistic 'open' and 'closed' histone modifications could be distributed on different sub-regions of the long alphoid DNA repeats. Indeed, alphoid DNA sub-regions forming centromere chromatin can vary on the same chromosome (as observed for human chromosome 17; Maloney et al., 2012). Mapping the chromatin state of sub-regions of the alphoid DNA is challenging due to the highly repetitive nature of the DNA. For better understanding of the correlation between the centromeric chromatin states and kinetochore assembly, it is important to clarify the mechanisms by which CENP-B binding to specific repetitive DNA structures can produce such radically different functional chromatin states.

The CENP-B acidic domain creates a bidirectional chromatin 'switch'
Our screening identified a number of factors recruited to chromatin by the CENP-B acidic domain (CENP-A was unusual, in that it was also recruited by the DBD). How do acidic domains recruit such various factors to chromatin? We previously suggested that acidic regions might unfold chromatin to allow access of other proteins, interact with histones to recruit other associated factors to chromatin or directly interact with other chromatin proteins (Earnshaw, 1987b). Indeed, many histone chaperones have acidic domains, and subsequent studies have supported the proposal that interaction with basic histone residues is important to promote proper association of histones and DNA (Tyler, 2002). The histone chaperone FACT affects nucleosome structure by interaction between its acidic domain and histones (Tsunaka et al., 2016). The CENP-B acidic domain promotes H3.3 assembly on alphoid DNA (Shono et al., 2015;Morozov et al., 2017; and data not shown), suggesting that it produces a chromatin state competent for histone replacement. This is consistent with older studies in which polyglutamic acid was able to promote nucleosome assembly from isolated histones in vitro (Stein et al., 1979). It is possible that, in addition to modifying chromatin structure, targeting of various chromatin-related factors might also involve direct interactions with the acidic domain. In future studies, it will be important to map interactions between the CENP-B acidic domain and client proteins by structural analysis.
Chromatin modifications induced by the CENP-B acidic domain promote structural changes that effectively produce a 'bidirectional' state of alphoid DNA chromatin. On the one hand, CENP-B can facilitate CENP-A assembly by recruiting ASH1L and other centromere-related proteins or by replacing histonesfor example, with H3.3. H3.3 has been suggested to be an important placeholder for CENP-A replenishment, but is also important for maintenance of heterochromatin (Dunleavy et al., 2011;Shono et al., 2015;Múller and Almouzni, 2017). On the other hand,  CENP-B facilitates heterochromatinization in non-centromeric regions (especially at the ectopic alphoid DNA integration site or pericentromere, that is, in the absence of the epigenetic CENP-A mark) by recruiting Suv39h1 and HP1. These results suggest that CENP-B can switch its activity according to the presence or absence of the functional centromere, possibly because the CENP-B DBD interacts with CENP-A nucleosomes.
H3.3 recruitment and H3K36 methylation downstream of ASH1L recruitment by the CENP-B acidic domain might promote transcriptional activity and subsequent CENP-A assembly. Alternatively, because H3K36me2/3 typically arises downstream of transcription, centromeric transcription promoted by another mechanism might promote interactions between CENP-B and its associated factors to create the open state of centromere chromatin that is required for kinetochore assembly and stability. It is possible that as-yet-undiscovered functional interactions between CENP-B and the RNA polymerase II machinery might be at the heart of the bidirectional ability of CENP-B to promote heterochromatic or transcriptionally active chromatin states.
We conclude that CENP-B regulates CENP-A chromatin not only by interacting with CENP-A and CENP-C, but also by recruiting histone modifiers. Future studies will explore how histone modifications associated with contrasting chromatin states interact with each other on alphoid DNA sub-regions and how this bidirectional chromatin 'switch' is controlled by CENP-B and other centromere-specific factors.

Cell staining and Halo-fused protein labeling
Cells were fixed with 80% acetone at −20°C for 10 min or 2.5% formaldehyde (Wako; 063-04815) at room temperature for 10 min. Formaldehyde-fixed cells were permeabilized with 0.5% Triton X-100 in PBS. Fixed cells were blocked in 2% BSA in PBS for 30 min. Cells were incubated at 37°C for 1 h with each of the primary and secondary antibodies. For chromosome spreads, HT1080 tetR-EYFP-CENP-B N21 cell line (Fig. 8C) was treated with 350 nM TN-16 (Wako) (Kitagawa et al., 1995) for 3 h in the culture medium. Mitotic cells were harvested, incubated in 0.075 M KCl for 10 min on ice, and then spread on a cover-glass using a Cytospin3 centrifuge (Shandon). The subsequent immunostaining and fluorescence in situ hybridization (FISH) were performed to confirm HAC formation according to a previously reported method (Ikeno et al., 1998;Ohzeki et al., 2002). For Halo-tag labeling, cells were cultured in the presence of 10 nM Halo-TMR-Ligand (Promega; G8251) or 10 nM Halo-Biotin-Ligand (Promega; G8281).

Immunoblotting
Immunoblotting was performed as previously described (Shono et al., 2015) with slight modifications. To separate and detect ASH1L protein, whole cell extracts were electrophoresed using a Multi Gel II mini 5 system (CosmoBio; 443138). HiMark™ Pre-Stained protein standard (ThermoFisher; LC5699) was used as molecular weight marker.

Microscopy and quantification of images
For quantification analysis of CENP-A (Fig. 2), z-stack images were acquired with a spacing of 0.22 µm to cover an entire nuclear signal on an Axio Observer.Z1 microscope (Zeiss) equipped with a CSU-X1 confocal scanner unit (Yokogawa), iXon3 DU897E-CS0 camera (Andor) and Plan-Apochromat 100×/1.46 oil lens (Zeiss) using Andor iQ2 software (Andor). Quantitative analysis of acquired images was performed as previously described (Shono et al., 2015). Other cell images were acquired on an Axio Observer.Z1 (Zeiss) equipped with a LSM700 scanning module and an Objective Plan-Apochromat 63×/1.46 oil lens (Zeiss) using ZEN 2009 software (Zeiss). For quantification analysis in Fig. 4, z-stack images were acquired with a spacing of 0.36 µm, and with total depth of 5.4 µm. Signal intensity was acquired for each pixel in a 22×22 pixel region around the spot from a z-stack maximum intensity projection cell image using Fiji software (Schindelin et al., 2012). The pixel intensities located in a symmetric position with respect to the x-axis and y-axis were summed up. The integrated pixel intensities at the same distance from the spot were averaged.

ChIP assay and qPCR or competitive PCR
The ChIP assay was performed as previously described (Fujita et al., 2015) with slight modifications. The cells were trypsinized, harvested in a centrifuge tube, washed once with PBS, and fixed in 1.0% formaldehyde analyzed cell lines' column shows the number of colonies of cells isolated under selection using G418. The fate of the introduced BAC DNA was analyzed by FISH using a BAC probe. BAC DNA was detected either as an entity independent from the host chromosome (HAC) or as part of the host chromosome (integration). Two HAC cell lines were obtained. In these two cell lines, HAC signals were detected as a single HAC per cell in 92.3% and 90% of cells (n>20 cells). No integration signal was observed on the host chromosomes in these two HAC cell lines.

Figure S1
Effectiveness evaluation of FMIT assay on the screening proteins for assembling on ectopic alphoid tetO DNA in a CENP-B-dependent manner.
(A) Schematic drawing of strain construction. HeLa-Int-03 has an ectopic integration site of alphoid tetO DNA repeats. This cell line was established in the previous study (Ohzeki et al., 2012).