Ccp1-Ndc80 switch at the N terminus of CENP-T regulates kinetochore assembly

Significance Precise chromosome segregation relies on kinetochores. How kinetochores are precisely assembled on centromeres through the cell cycle remains poorly understood. Centromeres in most eukaryotes are epigenetically marked by nucleosomes containing the histone H3 variant, CENP-A. Here, we demonstrated that Ccp1, an anti–CENP-A loading factor, interacts with the N terminus of CENP-T to promote the assembly of the outer kinetochore Ndc80 complex. This work further suggests that competitive exclusion between Ccp1 and Ndc80 at the N terminus of CENP-T via phosphorylation ensures precise kinetochore assembly during mitosis. In addition, CENP-T is critical for Ccp1 centromeric localization, which in turn regulates CENP-A distribution. Our results reveal a previously unrecognized mechanism underlying kinetochore assembly through the cell cycle.

How the spatial relationship between the CENP-A nucleosome and CENP-T particles in centromeres is regulated remains unclear.
The fission yeast Schizosaccharomyces pombe contains large regional centromeres and is considered to be a model system for centromere study. The CENP-A homolog, Cnp1, is enriched in centromere cores, which are surrounded by pericentromeric heterochromatin (28)(29)(30). CENP-A Cnp1 nucleosomes nucleate kinetochore assembly. Mislocalization of CENP-A Cnp1 results in severe chromosome segregation defects in fission yeast (28,(31)(32)(33)(34). Fission yeast also contains the CENP-T homolog, Cnp20, which associates with centromeres throughout the cell cycle. The same as in higher eukaryotes, CENP-T Cnp20 in S. pombe is essential for viability (35).
Recently, Ccp1, a nucleosome assembly protein (NAP) family protein, has been shown to play an important role in antagonizing the loading of CENP-A in fission yeast (36). Ccp1 forms a homodimer and is enriched at centromeres. Ccp1 acts as a key player in balancing CENP-A and histone H3 levels in the region (36). How Ccp1 regulates the CENP-A level in centromeres remains elusive. Interestingly, its centromere localization is cell cycle regulated. Ccp1 is dissociated from centromeres at the onset of mitosis and reassociates with centromeres at the end of mitosis (36,37). The biological importance of the cell cycle-dependent interaction between Ccp1 and centromeres is unknown.
Here using mass spectrometry, we found that Ccp1 interacts directly with CENP-T Cnp20 in fission yeast. We further identified a conserved Ccp1-interaction motif (CIM) at the N terminus of CENP-T Cnp20 , which is adjacent to the Ndc80 receptor motif. We demonstrated that CIM is important for Ccp1 localization. Furthermore, our data suggested that CDK1-mediated phosphorylation of the CIM motif at the onset of mitosis dissociates Ccp1 from CENP-T Cnp20 , allowing proper positioning of Ndc80C. Ccp1 associates with centromeres during mitosis in the phospho-null mutant of the CIM domain, leading to mislocalization of Ndc80C and severe chromosome segregation defects. Our study uncovers a previously unrecognized mechanism regulating kinetochore organization in regional centromeres.

Results
CENP-T Interacts with Ccp1. We previously showed that Ccp1 localizes to the centromeric regions in all cell cycle stages except mitosis (36). In order to study the molecular mechanism underlying cell cycle-dependent localization of Ccp1, a functional tandem affinity purification (TAP)-tagged Ccp1 was used (36). We purified Ccp1-TAP by tandem affinity purification and subsequently used mass spectrometry to identify its interacting proteins. Our results identified two centromere-specific proteins, CENP-A Cnp1 and CENP-T Cnp20 , associated with Ccp1 ( Fig. 1A and SI Appendix, Fig. S1 and Dataset S1). Our previous TAP purification of CENP-A Cnp1 also revealed that CENP-A Cnp1 interacts with Ccp1 (36). The interaction of CENP-T Cnp20 and Ccp1 was further confirmed by coimmunoprecipitation (co-IP) (Fig.  1B). The C terminus of CENP-T Cnp20 contains a histone fold domain known to interact with CENP-W New1 . Its N terminus is long, consisting of 383 amino acid (aa) residues. A conserved Ndc80 receptor motif from aa 70 to aa 87 was also identified in the N terminus (15).
CENP-T Is Required for Ccp1 Centromere Localization. Like higher eukaryotes, CENP-T Cnp20 in fission yeast is essential for cell growth (35). To investigate the role of CENP-T Cnp20 in centromeres, we created a temperature-sensitive (ts) mutant of CENP-T Cnp20 , cnp20-9, by random mutagenesis. The cnp20-9 mutant grows normally at room temperature, but does not survive at 36°C ( Fig. 2A). Through sequencing, we identified four amino acid mutations in cnp20-9, including S160C and T197S in the N-terminal domain, and L394W and A433D in the histone fold domain (Fig. 2B). We also found that GFP-tagged Cnp20-9 is delocalized from the centromeres at the restrictive temperature (SI Appendix, Fig. S2). Consistent with the key role of CENP-T Cnp20 in the assembly of the Ndc80 complex, we found that the association of Ndc80-GFP with centromeres is lost in cnp20-9 at the restrictive temperature ( Fig. 2 C and D).
To determine the effect of cnp20-9 on Ccp1 localization, we analyzed the cnp20-9 strain carrying GFP-Ccp1. We found that GFP-Ccp1 was delocalized from centromeres at the restrictive temperature in cnp20-9 at all stages of the cell cycle ( Fig. 2 E and F and SI Appendix, Fig. S3), whereas the localization of GFP-Ccp1 at centromeres only has a mild reduction in the CENP-A ts mutant, cnp1-1 (SI Appendix, Fig. S4). We next checked the distribution of CENP-T Cnp20 -GFP in ccp1Δ and found that its centromere localization was unaffected in the mutant (Fig. 2 G-I).
It has been shown that, although CENP-T deposition is not dependent on CENP-A, the association of CENP-T with centromeres is reduced when CENP-A is knocked out in several species (5,(38)(39)(40). We examined the localization of CENP-T Cnp20 -GFP in the S. pombe cnp1-1 mutant and showed that CENP-T Cnp20 at centromeres is also partially decreased at 36°C (SI Appendix, Fig.  S5). In addition, we found that CENP-A Cnp1 -GFP partially reduced its centromere localization in cnp20-9 at the restrictive temperature (SI Appendix, Fig. S6). We have previously shown that Ccp1 dissociates from centromeres in the mis16 mutant (36). We next examined CENP-T Cnp20 -GFP distribution in the mis16-53 mutant. We found that the association of CENP-T Cnp20 with centromeres is largely lost at the restrictive temperature in the mutant background (SI Appendix, Fig. S7). It is thus likely that Mis16 may regulate Ccp1 localization via CENP-T Cnp20 . It has been shown that Mis16 centromeric localization is not affected by the cnp1-1 mutant at the restrictive temperature (41), consistent with our observation that Ccp1 dissociates from centromeres in the mis16 mutant but still can stay at centromeres in cnp1-1. Fig. 2J shows a summary of localization dependency relationships between Ccp1, CENP-A Cnp1 , CENP-T Cnp20 , and Mis16. Together, our data indicate that CENP-T Cnp20 is required for Ccp1 centromere localization. To further confirm that Ccp1 interacts with CENP-T Cnp20 , the yeast two-hybrid system was performed in a heterologous system Saccharomyces cerevisiae. The CENP-W New1 known to have strong interaction with CENP-T Cnp20 was used as a control. Our yeast two-hybrid assays confirmed the interaction between Ccp1 and CENP-T Cnp20 (Fig. 3A). Our previous study indicated that Ccp1 forms a homodimer, which is necessary for the centromeric localization of Ccp1 (36). The yeast two-hybrid assays showed that the Ccp1 homodimer mutant, Ccp1-4A, was unable to interact with CENP-T Cnp20 (Fig. 3A and SI Appendix, Fig. S8), indicating that the dimeric structure is vital for its interaction with CENP-T Cnp20 .
In order to identify the interacting domain of CENP-T Cnp20 with Ccp1, we constructed the N-terminal and C-terminal domains of CENP-T Cnp20 : Cnp20-N (aa 1 to 368) and Cnp20-C (aa 369 to 479), respectively. Our yeast two-hybrid assays indicated that Ccp1 interacts with the N-terminal domain but not with the C terminus (SI Appendix, Fig. S9). To narrow down the Ccp1-interacting region of CENP-T Cnp20 , we constructed a series of N-terminal fragments of CENP-T Cnp20 , including aa 1 to 88, aa 89 to 196, and aa 197 to 368, and conducted yeast two-hybrid assays using Ccp1 as a bait. We found that Ccp1 associates with aa 1 to 88 (Fig. 3B). Further yeast two-hybrid analysis using partial deletion of the first 88 amino acids revealed that the first 55 amino acids interacts with Ccp1 at a similar level as the full-length CENP-T Cnp20 , whereas fragments aa 16 to 55 and aa 1 to 44 showed much weaker interaction ( Fig. 3C and SI Appendix, Fig. S10). Our results indicate that the first 55 amino acids of CENP-T Cnp20 are the minimal interaction domain with Ccp1, which we named the Ccp1interacting motif (CIM). The CIM domain lies close but does not overlap with the Ndc80 receptor motif (Fig. 3D). Our sequence analysis also identified multiple residues within the CIM domain that are conserved in CENP-T orthologs in Gallus gallus, Homo sapiens, and Mus musculus, as well as S. cerevisiae (SI Appendix, Fig. S11).
To further investigate how the CIM domain physically interacts with Ccp1, we performed in vitro pull-down assays using a peptide derived from Cnp20 1-56 and purified full-length Ccp1. We found that Cnp20 1-56 was able to pull down Ccp1 in vitro, indicating that the CIM domain indeed directly interacts with Ccp1 (Fig. 3E). We also performed in vitro pull-down assays using three peptides derived from Cnp20 1-25 , Cnp20 15-40 , and Cnp20 31-56 , respectively. We found that both Cnp20 15-40 and Cnp20 31-56 were able to pull down Ccp1 in vitro, and their interactions with Ccp1-4A are reduced (SI Appendix, Fig. S12).
The CIM Domain of CENP-T Is Important for Ccp1 Localization. To determine the role of the CIM domain of CENP-T Cnp20 in Ccp1 localization, a CIM domain-deleted version of cnp20 + (cnp20-Δ1-55) was constructed under the control of its endogenous promoter and inserted at the ade6 + locus in the cnp20-9 mutant background. As a control, the full-length cnp20 + and Fig. 3. The N terminus of CENP-T contains a conserved CIM. (A) Yeast two-hybrid (Y2H) analysis confirmed the interaction between Ccp1 and CENP-T Cnp20 and also showed that the interaction is lost in the dimeric mutant of ccp1-4A. S. cerevisiae strains carrying the indicated plasmid combinations were spotted onto synthetic defined (SD) minimal medium plates lacking either Leu and Trp (SD ÀLT) or Leu, Trp, and His (SD ÀLTH) supplemented with either 60 mM 3-amino-1,2,4-triazole (3-AT) or 75 mM 3-AT, and grew at 30°C. AD, activation domain; BD, binding domain. (B) Y2H assays indicated that Ccp1 associates with the first 88 amino acids at the CENP-T Cnp20 . (C) The first 55 amino acids of CENP-T Cnp20 were identified by Y2H as a minimum Ccp1interacting motif. (D) A schematic diagram of the domain structure of CENP-T Cnp20 . Orange, the CIM domain; cyan, histone fold domain; purple, Ndc80 receptor motif. (E) In vitro pull-down assay demonstrates that the first 55 amino acids of CENP-T Cnp20 interact with Ccp1, but the interaction of the phosphomimetic 14D mutant of the region (14D) with Ccp1 is disrupted. Wild-type and phosphomimetic 14D mutant peptides derived from Cnp20 1-55 , and purified full-length Ccp1 were used. the first 88 amino acid-deleted cnp20 (cnp20-Δ1-88) were also inserted in ade6 + in the same manner. The full-length cnp20 + can completely rescue the growth defect of cnp20-9 at the restrictive temperature. On the other hand, cells carrying cnp20-Δ1-88 grow well at room temperature, but not at 36°C, the same as the cnp20-9 mutant alone (Fig. 4A). We further found that the deletion of the first 55 amino acids can also rescue cnp20-9, but has a slight growth defect (Fig. 4A). These results indicated that the CIM domain is not essential for cell viability, similar to the ccp1Δ mutant shown previously, whereas deletion of the Ndc80 receptor motif is lethal.
We next deleted the CIM domain of the endogenous cnp20 + (cnp20-ΔCIM) by CRISPR-Cas9. As a control, we also created a strain in which the first 30 amino acids in CENP-T Cnp20 (cnp20-Δ1-30) were deleted (Fig. 4B). To determine how the CIM domain affects Ccp1 localization, we examined the GFP-Ccp1 pattern in these strains. In the cnp20-Δ1-30 mutant, we found that GFP-Ccp1 is enriched in centromeres during interphase similar to wild type (WT), though the GFP signal is weaker (Fig. 4 C and D). But GFP-Ccp1 is completely disassociated from centromeric regions in cnp20-ΔCIM at all stages of the cell cycle ( Fig. 4 C and D and SI Appendix, Fig. S13). We concluded that the CIM domain is required for Ccp1 centromeric localization. Our data suggest that CENP-T regulates the recruitments of Ccp1 to centromeres via its CIM domain.
cnp20-ΔCIM Phenocopies the ccp1Δ Mutant. Our previous study had shown that the ccp1Δ mutant is sensitive to the microtubule destabilizing drug, thiabendazole (TBZ) due to chromosome segregation defects (36). To determine whether TBZ has a similar effect to cnp20-ΔCIM, we analyzed the growth of the mutant in rich media with TBZ as well as cnp20-Δ1-30. Our growth assays indicated that while the deletion of the first 30 amino acids has no significant growth defect, cnp20-ΔCIM displayed a TBZ-sensitive phenotype (Fig. 4B). Consistent with the results, our DAPI staining showed that more than 16% of cnp20-ΔCIM cells contain lagging chromosomes during mitosis ( Fig. 4 E and F).
The epigenetic state of S. pombe centromeres is manifested by position effect variegation (PEV). Reporter genes inserted within the centromeric core region are silenced stochastically. The silencing state correlates with the CENP-A nucleosome occupancy in centromeres (42,43). Deletion of ccp1 + results in increased silencing in centromere cores (37). To investigate the effect of cnp20-ΔCIM on PEV in centromeres, we crossed the mutant into the background with ade6 + in the centromere core. Silencing of ade6 + results in red colony color in the adeninelimiting media while its expression leads to white colonies. Wild-type cells carrying the reporter in the media thus display red/white color sectoring morphology due to PEV in centromeres; but colonies of ccp1Δ are mostly red due to increased silencing in the region (37). We found that single colonies of cnp20-ΔCIM also contain fewer white sectors than those of WT, and are predominantly red (Fig. 4 G and H). In addition, we found that similar to ccp1Δ (36), ∼20% of cnp20-ΔCIM mutant cells with a single fluorescence focus have disproportionately brighter CENP-A cnp1 -GFP signal (SI Appendix, Fig. S14), suggesting that the level of CENP-A cnp1 in centromeres substantially increase in these cells. Together, these results indicate that cnp20-ΔCIM phenocopies the ccp1Δ mutant, consistent with the idea that CENP-T Cnp20 recruits Ccp1 through its CIM domain to centromeres during interphase.
The Interaction between CENP-T and Ccp1 Is Disrupted in the cnp20-14D Mutant. We have demonstrated that CENP-T Cnp20 interacts with and recruits Ccp1 to centromeres through its CIM domain. CENP-T Cnp20 in S. pombe is an inner kinetochore protein known to associate with centromeric regions at all stages of the cell cycle (44), whereas Ccp1 is disassociated from centromeres during mitosis. The mechanism underlying the disassociation of Ccp1 from centromeres at the onset of mitosis is not known, nor is its biological significance.
Previous studies have shown that CENP-T was phosphorylated at multiple sites at its N terminus, including the conserved site on S74 within the Ndc80 receptor motif in S. cerevisiae, and T72 and S88 in G. gallus (16,21,22). Phosphorylation of CENP-T plays an important role in the recruitment of Ndc80C during mitosis (21)(22)(23)(24). We identified 14 serine and threonine residues in the CIM domain of CENP-T Cnp20 that can be potentially phosphorylated (Fig. 5A). We reasoned that the phosphorylation of the CIM domain may regulate the dissociation of Ccp1 from the centromeric region during mitosis. To test this idea, we mutated all serine and threonine residues in the CIM domain to phosphomimetic aspartic acid (cnp20-14D) at the endogenous sites by CRISPR-Cas9 (Fig. 5A). We found that, unlike in WT, the GFP-Ccp1 was completely delocalized from centromeres at all phases of the cell cycle (Fig. 5 B and C). These data suggest that phosphorylation of the CIM domain leads to disassociation of Ccp1 from centromeres.
In order to find the minimal phosphorylation sites that contribute to the centromeric positioning of Ccp1, we mutated a series of serine and threonine residues to aspartic acid, including cnp20-12D, cnp20-9D, cnp20-6D 9-14 , cnp20-6D 6-11 , cnp20-5D, cnp20-3D 6-8 , and cnp20-3D 9-11 at its endogenous locus (Fig. 5A). We examined the localization of GFP-Ccp1 using the same approach. We found that GFP-Ccp1 is still associated with centromeres in all these mutants, although at different degrees (Fig. 5 C-E). Ccp1 centromeric distribution is the least affected in cnp20-3D 6-8 and cnp20-3D 9-11 , whereas most cnp20-12D cells display delocalized GFP-Ccp1, although not as severely as in cnp20-14D. These results indicated that all predicted phosphorylation sites within the CIM domain of CENP-T Cnp20 are involved in the association of Ccp1 with centromeres.
To investigate how phosphorylation of the CIM domain affects the interaction between Ccp1 and CENP-T Cnp20 , we conducted yeast two-hybrid assays with Cnp20-14D and Ccp1. We found that the interaction between Cnp20-14D and Ccp1 was dramatically reduced (Fig. 5F). This is further confirmed by our in vitro binding assays (Fig. 3E). These data suggest that phosphorylation of the 14 serine and threonine residues within the CIM domain disrupts the interaction between CENP-T and Ccp1, which explains the Ccp1 delocalization phenotype in cnp20-14D. These results also suggest that a combination of phosphorylation events within the CIM domain leads to the disassociation of Ccp1 from CENP-T Cnp20 during mitosis.
Ccp1 Associates with Centromeres during Mitosis in the cnp20-14A Mutant. We next investigated how the dephosphorylation of the CIM domain in CENP-T Cnp20 affects Ccp1 localization. We replaced all of the 14 serine and threonine residues with phospho-null alanine in the CIM domain of CENP-T Cnp20 at its endogenous site (cnp20-14A) and examined the distribution pattern of GFP-Ccp1 throughout the cell cycle in the mutant background. As a control, we also created a mutant in which 6 serine and threonine residues within CIM were replaced by alanine (cnp20-6A) (Fig. 6A). GFP-Ccp1 in WT is delocalized from centromeres during mitosis. In contrast, we found that GFP-Ccp1 in all cnp20-14A mutant cells remains associated with centromeres during all the stages of the cell cycle ( Fig. 6 B and C). We also observed centromeric localization of GFP-Ccp1 in cnp20-6A cells, although the GFP signal is very weak (SI Appendix, Fig. S15). These data indicate that dephosphorylation of the CIM of CENP-T Cnp20 results in a tight association of Ccp1 with centromeres. Consistent with this idea, our yeast two-hybrid assays demonstrated that, unlike Cnp20-14D, Cnp20-14A showed a strong interaction with Ccp1 (Fig. 5F). Together, our results suggest that the CIM domain of CENP-T Cnp20 is phosphorylated at the onset of mitosis leading to dissociation of Ccp1 from centromeres, while the dephosphorylation of the CIM domain at the end of mitosis allows Ccp1 to reassociate with centromeres and remain stably bound throughout interphase.
The cnp20-14A Mutant Disrupts the Positioning of the Ndc80 Complex during Mitosis. To determine the effect of cnp20-14A on chromosome segregation, we analyzed the mutant using TBZ growth assays. We found that cnp20-14A is highly sensitive to TBZ, even more strongly than the cnp20-ΔCIM mutant ( Fig. 6D and  SI Appendix, Fig. S16), indicating that cnp20-14A leads to severe chromosome segregation defects. We further utilized time-lapse microscopy to examine chromosome segregation in cnp20-14A. The mutant cells expressing the GFP-tagged Spc25, a subunit of Ndc80C and the Sad1-mCherry (spindle pole body [SPB] marker), were analyzed. After the onset of mitosis, kinetochores oscillate between the two SPBs and segregate toward the SPBs during anaphase. Wild-type cells complete the segregation within 14 min (Fig. 6E). But most cnp20-14A cells display mitotic delay, and more than 12% of mutant cells failed to complete chromosome segregation within 30 min (Fig. 6 E  and F).
Given that the CIM domain and the Ndc80 receptor motif in CENP-T Cnp20 are adjacent to each other, we reasoned that the association of Ccp1 with Cnp20-14A during mitosis might interfere with the recruitment of Ndc80C to the Ndc80 receptor motif. To test whether this is the case, we examined closely the effect of cnp20-14A on the positioning of Ndc80C using cells carrying the GFP-tagged Spc25 and the mCherry-tagged tubulin. In wild type, Ndc80C oscillates along the microtubules between the two SPBs before moving to the two poles (Fig. 7A). However, we found that Spc25-GFP appeared not to attach to microtubules in ∼20% of cnp20-14A mitotic cells, indicating that dephosphorylation of the CIM domain leads to mislocalization of Ndc80C during mitosis (Fig. 7A). Importantly, our co-IP results indicated that the interaction between Cnp20-14A and Ndc80 is significantly reduced during mitosis (Fig. 7B). Furthermore, since Ccp1 is not present at centromeres in mitosis, we performed co-IP experiments using either cells synchronized in mitosis or unsynchronized cells, which is predominantly at G2. These co-IP experiments showed that the interaction between Ndc80 and CENP-T Cnp20 is drastically increased during mitosis (Fig. 7C), supporting the idea that the presence of Ccp1 interferes with the interaction between Ndc80 and CENP-T Cnp20 .
These data suggest that dissociation of Ccp1 from centromeres by phosphorylation of CIM at the onset of mitosis promotes the interaction between CENP-T and Ndc80C, leading to the proper assembly of the outer kinetochores. In fact, we found that cnp20-14D has no obvious chromosome missegregation defects (SI Appendix, Fig. S17); actually unlike the cnp20-Δ1-55 mutant, cnp20-14D is even more resistant to TBZ than WT (Fig. 6D). It is likely that the phosphomimetic cnp20-14D mutant, but not cnp20-Δ1-55, can stabilize the interaction between CENP-T and Ndc80. Consistent with this, our co-IP assay showed that Ndc80 interacts more strongly with Cnp20-14D than CENP-T Cnp20 (SI Appendix, Fig. S18).
The CIM Domain Is Phosphorylated by CDK1. We suspected that CDK1 may be responsible for the phosphorylation of the CIM domain in CENP-T Cnp20 at the onset of mitosis. To test this, we performed the in vitro kinase assays using human recombinant CDK1-cyclin B and a synthesized peptide derived from Cnp20  . The highly conserved human CDK1 is known to be capable of functionally replacing the CDK1/Cdc2 in fission yeast (45). A peptide derived from the phospho-null version of the CIM domain, Cnp20 1-55-14A , was used as a control. We then utilized the mobility shift of phosphorylated proteins on Phos-tag polyacrylamide gel electrophoresis (PAGE) to detect changes in phosphorylation status. We observed migrating bands slower than their nonphosphorylated counterpart from the in vitro assay using Cnp20 1-55 . However, no slow migrating bands were identified in the assay with the Cnp20 1-55-14A peptide (Fig. 7D). These data demonstrate that CDK1 is capable of phosphorylating the CIM domain of CENP-T Cnp20 . Various slow migrating bands were observed in the assay with Cnp20 1-55 (Fig. 7D), indicating that the domain contains multiple phosphorylation sites, consistent with our point mutation analysis. CDK1/Cdc2 in fission yeast is essential for viability. The temperature-sensitive alleles of cdc2 are arrested at either G2/ M or G1/S phase at the restrictive temperature (46), preventing us from assessing its role in Ccp1 regulation in vivo. The mitotic cyclin B is expressed specifically in the G2/M phase and forms the complex with CDK1 to trigger mitosis. We reasoned that if we constitutively overexpressed cyclin B/Cdc13 in fission yeast, CDK1 might phosphorylate the CIM domain of CENP-T Cnp20 throughout different stages of the cell cycle, leading to dissociation of Ccp1 from centromeres. Indeed, we found that when Cdc13 is overexpressed in cells carrying GFP-Ccp1, most of the cells displayed high fluorescence signals in the nucleoplasm with weaker GFP-Ccp1 at centromeres, more than 25% in which GFP signals were totally diffuse through the nucleus (Fig. 7E and SI Appendix, Fig. S19). These data support the idea that the CIM domain is phosphorylated by CDK1. Furthermore, consistent with the functional studies described above, our in vitro binding assay showed that phosphorylation of the CIM domain by CDK1 results in reduced interaction between the CIM domain and Ccp1 (SI Appendix, Fig. S20).

Discussion
Assembly of the outer kinetochore during mitosis is tightly regulated. But the molecular mechanism for this regulation, especially in regional centromeres, remains poorly understood. CENP-T, a key inner kinetochore component, is known to promote the recruitment of Ndc80C, a complex essential for kinetochore-microtubule interactions during mitosis (5,(15)(16)(17)(18)(19)(20).
Here we found that CENP-T interacts with Ccp1, a conserved homodimer of the NAP family proteins in fission yeast, and demonstrated that the dimeric structure of Ccp1 is required for the interaction. We reported previously that Ccp1 associates with CENP-A and acts as a counteracting CENP-A loading factor to balance the CENP-A and histone H3 at centromeres (36). Interestingly, Ccp1 associates with centromeres at interphase but dissociates from centromeres during mitosis (36,37). We showed that Ccp1 centromere localization depends on Mis16 (36). In this study, we found that the CENP-T is also required for the Ccp1 centromere localization. In addition, we found that the association of CENP-T Cnp20 with centromere depends on Mis16 (SI Appendix, Fig. S7). It is possible that Mis16 recruits Ccp1 to centromeres through CENP-T Cnp20 .
We further identified the first 55 amino acids at the N terminus of CENP-T as the CIM. Notably, the CIM domain is situated next to the Ndc80 receptor motif. We found that the CIM domain can be phosphorylated. Phosphorylation of the CIM domain disrupts the interaction between CENP-T and Ccp1, leading to dissociation of Ccp1 from centromeres. On the other hand, dephosphorylation of the domain results in tight interaction of CENP-T with Ccp1 and consequently the association of Ccp1 with centromeres during mitosis. We further showed that dephosphorylation of CIM disrupts the interaction between CENP-T and Ndc80C. As a result, the phospho-null mutant of CIM exhibits mislocalization of the Ndc80 complex during mitosis and severe chromosome segregation defects. Our data suggest that occupancy of the Ccp1 homodimer at the CIM domain during interphase prevents the interaction of the adjacent Ndc80 receptor motif of CENP-T with Ndc80C. Removing Ccp1 from CENP-T makes the Ndc80 receptor motif accessible during mitosis, which may allow Ndc80C to interact with CENP-T to be properly assembled. Nevertheless, we found that only ∼20% of the cnp20-14A cells displayed mislocalized Scp25-GFP signal and the mutant cells are viable, suggesting that alternative mechanisms are also used in recruiting the Ndc80 complex during mitosis.
We found that multiple amino acids in the CIM domains are conserved across different species, especially in S. pombe, M. musculus, and H. sapiens (SI Appendix, Fig. S11). The Ndc80 receptor motif which is next to the CIM domain is considered as a conserved domain for the interaction between the Ndc80 complex and CENP-T by previous studies (15,16,(21)(22)(23)(24). Phosphorylation of the Ndc80 receptor motif is required for the interaction (16,(21)(22)(23)(24). Based on our sequence alignment analysis, there are also a few key residues conserved in the motif (SI Appendix, Fig. S11). It is likely that both the Ndc80 receptor motif and the Ndc80 complex coevolve during evolution to maintain the interaction. Similar to the Ndc80 receptor motif, we believe that the CIM domain is also functionally conserved.
We used both the yeast two-hybrid system and in vitro pulldown assays to confirm the interaction between the CIM domain and Ccp1. Nevertheless, we noted that our yeast two-hybrid results are not completely consistent with our in vitro pull-down data. We think it is likely due to the fact that two assays were performed in two very different systems. Unlike the yeast twohybrid system, the pull-down assays were performed in the nonstringent conditions with much higher protein levels. Also, the truncated BD-Cnp20-N in our yeast two-hybrid assays showed interactions with both AD (empty vector) and AD-Ccp1-4A (the homodimer mutant), but the full-length BD-Cnp20 did not show any interactions with AD or AD-Ccp1-4A (SI Appendix, Fig. S9). We believe that the truncated BD-Cnp20-N is likely selfactivating. On the other hand, the self-activation activity of fulllength BD-Cnp20 may be inhibited by the configuration of the full-length CENP-T Cnp20 protein.
Our in vitro kinase assays further demonstrated that the CIM domain is phosphorylated by CDK1. This was further supported by the fact that constitutive overexpression of Cdc13/ Cyclin B results in dissociation of Ccp1 from centromeres. Presumably, overexpression of Cdc13 causes CENP-T to be phosphorylated during interphase, which disrupts the interaction between the CIM domain of CENP-T and Ccp1, a phenomenon that is similarly observed in the Cnp20-14D mutant. Our in vitro kinase assays indicated that the domain contains multiple phosphorylation sites. Consistent with this, our point mutation analysis revealed that phosphorylation of all of the 14 potential phosphorylation sites within CIM is required for the dissociation of Ccp1 from centromeres during mitosis. It has been shown that CDK1 in multiple species specifically phosphorylates the N terminus of CENP-T, especially the Ndc80 receptor motif, during mitosis (16,(21)(22)(23)(24). Considering the conserved role of CENP-T and the close distance between the CIM domain and the Ndc80 receptor motif, it is likely that the CIM domain is also phosphorylated by CDK1 in the same cell cycle-regulated manner. Nevertheless, we found that only four of 14 phosphorylation sites in the CIM domain of CENP-T are located within known CDK1 consensus motifs. Overexpression of Cdc13 also only partially dissociates Ccp1 from centromeres. These data suggest that additional kinases are responsible for phosphorylation of the CIM domain. Consistent with this idea, our in vitro pull-down assay showed that phosphorylation of the CIM domain by CDK1 results in only partial loss of its interaction with Ccp1.
Our study uncovers a previously unrecognized mechanism behind the regulation of kinetochore organization in regional centromeres. This work also provides a molecular explanation for the observed Ccp1 dissociation from centromeres during mitosis. Based on our data, we propose the following model: The CIM domain of CENP-T is phosphorylated at the onset of mitosis by CDK1. The phosphorylation expels Ccp1 from CENP-T, which allows the Ndc80 receptor motif to interact with Ndc80C. At the end of mitosis, the CIM domain is dephosphorylated by an unknown phosphatase. This leads to the reassociation of Ccp1 with the CIM domain (Fig. 7F), which blocks the binding of Ndc80C to CENP-T. Similar to Ccp1, some other centromeric proteins, including HJURP/Scm3, Mis16, Mis18, Eic1/Mis19/Kis1, and Eic2/Mis20 (41,(47)(48)(49)(50)(51)(52) also dissociate from centromeres during mitosis. The mechanism of competitive exclusion between two proteins through phosphorylation identified here may be a general principle also used by these centromeric proteins to promote kinetochore assembly.
Whereas the point centromere in S. cerevisiae contains a single nucleosome (53,54), CENP-A-containing nucleosomes at regional centromeres are interspersed with canonical histone H3-containing nucleosomes. The histone H3 at centromeres is important for centromeric localization of CENP-A (55,56). The CENP-T complex has been shown to be associated with histone H3 and DNA in centromeres (5,25,27). CENP-T/W/S/X can form nucleosome-like structure in vitro (25,26). How the CENP-T particles and CENP-A nucleosome in centromeres are coordinated to establish homeostasis is unknown. Ccp1 counteracts the loading of CENP-A to ensure proper balance of CENP-A and H3 in centromeres as well as removing mislocalized CENP-A at ectopic sites (36). In this study, we found that CENP-T interacts with Ccp1 and is required for Ccp1 localization. Deletion of the CIM domain of CENP-T results in dissociation of Ccp1 from centromeres. Similar to the ccp1Δ mutant, the CIM-deleted mutant is sensitive to TBZ and also exhibited increased silencing in centromeres, indicative of an increased amount of CENP-A nucleosomes in the region. We propose that CENP-T recruits Ccp1 to centromeres at the end of mitosis to remove excessive CENP-A nucleosomes and allow room for the incorporation of sufficient CENP-T-containing particles to centromeres.

Materials and Methods
Standard media and genetic analysis for fission yeast were used (57). The temperature-sensitive allele of cnp20 was created by marker reconstitution mutagenesis (58). Details of the materials and methods used in this study, including mutant generation, immunoprecipitation, TAP-tag purification, yeast two-hybrid assays, in vitro pull-down assays, microscopy, in vitro kinase assays, and phos-tag PAGE are provided in SI Appendix, SI Materials and Methods. Fission yeast strains used in this study are listed in SI Appendix, Table S1.
Data Availability. All study data are included in the article and/or supporting information.