Recruitment of two Ndc80 complexes via the CENP-T pathway is sufficient for kinetochore functions

To form functional kinetochores, CENP-C and CENP-T independently recruit the KMN (Knl1C, Mis12C, and Ndc80C) network onto the kinetochores. To clarify the functions of the KMN network on CENP-T, we evaluated its roles in chicken DT40 cell lines lacking the CENP-C-KMN network interaction. By analyzing mutants lacking both CENP-T-Mis12C and CENP-C-Mis12C interactions, we demonstrated that Knl1C and Mis12C (KM) play critical roles in the cohesion of sister chromatids or the recruitment of spindle checkpoint proteins onto kinetochores. Two copies of Ndc80C (N-N) exist on CENP-T via Mis12C or direct binding. Analyses of cells specifically lacking the Mis12C-Ndc80C interaction revealed that N-N is needed for proper kinetochore-microtubule interactions. However, using artificial engineering to directly bind the two copies of Ndc80C to CENP-T, we demonstrated that N-N functions without direct Mis12C binding to Ndc80C in native kinetochores. This study demonstrated the mechanisms by which complicated networks play roles in native kinetochores.

combinations of components responsible for each function remain ambiguous. Therefore, it is critical to clarify these questions to understand kinetochore functions.
In this work, to address these questions, we examined the different roles of the KMN network on the CENP-T pathway in cells lacking the CENP-C-KMN network interaction, which exhibits some redundant functions to CENP-T. We constructed cell lines lacking the CENP-T-Mis12C interaction in the absence of the CENP-C-KMN network interaction and demonstrated defects in sister chromatid cohesion, reduction in the levels of Ndc80C, or loss of Knl1C and Bub1 in these cells. These results suggest that Mis12C-Knl1C (KM) interaction on the CENP-T pathway is critical for the recruitment of checkpoint proteins and Ndc80C or the cohesion of sister chromatids. While KM on CENP-T recruits Ndc80C, CENP-T directly binds to additional Ndc80C. We then constructed cell lines that specifically lacked the Ndc80C-binding site in Mis12C and demonstrated that one copy of Ndc80C on CENP-T was insufficient, suggesting that two copies of Ndc80C (N-N) are necessary for the functional kinetochore-microtubule interaction in DT40 cells. This suggests that the Mis12C-Ndc80C interaction is necessary for the microtubule-binding function of Ndc80C. However, using artificial engineering to directly bind the two copies of Ndc80C to CENP-T, we demonstrated that Ndc80C binding to Mis12C is not essential for microtubule binding. Our data indicate that the KMN network usually forms KMN-N (one copy of Ndc80C binds to KM as a KMN unit and the other directly binds CENP-T) on CENP-T for its functions, but KM-N-N (two copies of Ndc80C directly bind to CENP-T and KM independently binds to CENP-T) is functional in the network. We also demonstrated that Ndc80C binding to Mis12C is not essential for microtubule binding in human cells. Since the Ndc80C-Mis12C interaction is thought to be critical for the kinetochore functions 7,42,43 , our observations provide new insights for kinetochore studies and explain how the KMN network plays an essential role in kinetochore functions via the CENP-T pathway.

Results
The CENP-T-Mis12C interaction is essential on the CENP-T pathway. Mis12C binds to both CENP-C and CENP-T, and we have previously shown that the CENP-T-Mis12C interaction is much more critical than the CENP-C-Mis12C interaction in chicken DT40 cells 35 . However, deletion of the Mis12C-binding region (amino acids: aa 121-240 region) in chicken CENP-T did not lead to complete lethality in DT40 cells, although it caused a growth delay. This might be because Mis12C on CENP-C has redundant roles in kinetochore functions. We initially generated a cell line in which Mis12C-binding regions in both CENP-C and CENP-T were deleted (Fig. 1a-d). Since the Mis12C-binding site (N-terminal 73 aa region) in chicken CENP-C is dispensable in DT40 cells, we introduced mScarlet-fused CENP-C Δ73 into one endogenous β-actin allele with disrupting the endogenous CENP-C gene using CRISPR/Cas9-mediated genome editing ( Supplementary Fig. 1a-e). We chose the β-actin locus to obtain stable expression of mScarlet-fused CENP-C Δ73 , since the β-actin is ubiquitously expressed. Using this cell line, we introduced a cDNA for 3X FLAG-tagged CENP-T lacking the Mis12C-binding region (CENP-T Δ121-240 ) into one CENP-T allele ( Supplementary  Fig. 1f, g), and another allele was replaced with a drug resistance gene, and a wild-type (WT) CENP-T transgene was expressed under the control of a tetracycline (Tet)-responsive promoter (Tet-CENP-T, Fig. 1d). We called this line conditional knockout (cKO)-CT (Tet-Off) -CC Δ73 /CENP-T Δ121-240 , and CENP-T and CENP-C were completely replaced with 3X FLAG-CENP-T Δ121-240 and mScarlet-CENP-C Δ73 , respectively, after Tet addition in this ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28403-8 cell line (Fig. 1d). Using different combinations of CENP-T and CENP-C, we also generated various cell lines, such as cKO-CT-CC Δ73 /CENP-T Δ90 or cKO-CT-CC WT /CENP-T Δ121-240 cells ( Supplementary Fig. 1h). All cell lines generated in this study are summarized in Supplementary Table 1. We note that the CENP-T N-terminal region (aa 1-90) contains an Ndc80C-binding region (Fig. 1b), and CENP-T Δ90 does not directly bind to Ndc80C 35,39 .
After confirming the expression of each transgene using immunoblot analysis (Fig. 1d), we analyzed the growth of each cell line in the absence or presence of Tet. In the absence of Tet (expression of Tet-CENP-T), all cell lines grew well. After Tet addition, cKO-CT-CC WT /CENP-T Δ90 cells died by~96 h (Supplementary Fig. 1i). However, cKO-CT-CC WT /CENP-T Δ121-240 cells were still viable in the presence of Tet (Fig. 1e), although  they showed a growth delay, which is consistent with the results of our previous study 35 . We further tested the growth of cKO-CT-CC Δ73 /CENP-T Δ121-240 cells in the presence of Tet and found that these cells completely died by~120 h after Tet addition (Fig. 1e). These results indicate that Mis12C on CENP-C has redundant roles in the kinetochore functions, although it is not sufficient to completely compensate for the lack of CENP-T-Mis12C interaction. Therefore, to precisely evaluate the Mis12C function on the CENP-T pathway in native kinetochores, we must analyze it in the absence of CENP-C-KMN network interaction (expression of CENP-C Δ73 instead of CENP-C WT ). Although we defined the aa 121-240 region as the Mis12Cbinding region of chicken CENP-T, we attempted to define a more precise region of CENP-T required for Mis12C binding, using cKO-CT-CC Δ73 cells (Fig. 1f). In addition to centromeric signals, some mScarlet-fused CENP-C Δ73 localized to the noncentromeric chromosome region in these cells (Fig. 1g). This may be due to overexpression of mScarlet-fused CENP-C Δ73 under the control of β-actin promoter. Since we confirmed that cells expressing mScarlet-fused CENP-C Δ73 are viable, mScarlet-fused CENP-C Δ73 should be functional. We also introduced GFP-Dsn1 (a component of Mis12C) into endogenous Dsn1 alleles in these cells ( Supplementary Fig. 1j-l). We note that cKO-CT-CC Δ73 cells (None in Fig. 1g) died before the complete depletion of CENP-T. This explains why GFP-Dsn1 is still visible in cKO-CT-CC Δ73 cells, and we also presented normalization data for GFP-Dsn1 based on data from cKO-CT-CC Δ73 cells. As shown in Fig. 1g, GFP-Dsn1 levels were significantly reduced in either cKO-CT-CC Δ73 /CENP-T Δ161-200 or cKO-CT-CC Δ73 /CENP-T Δ201-216 cells, suggesting that either aa 161-200 or 201-216 regions is essential for Mis12C recruitment to the CENP-T pathway. Consistent with the Dsn1 levels in these cell lines, both cKO-CT-CC Δ73 /CENP-T Δ161-200 and cKO-CT-CC Δ73 /CENP-T Δ201-216 cells died by~72 h after Tet addition (Fig. 1h). These results suggest that Mis12C binds to the region of aa 161-216 of chicken CENP-T.
cKO-CT-CC Δ73 /CENP-T Δ121-240 cells died in the presence of Tet (Fig. 1e), which might be due to mitotic defects. To test this prediction, we examined the cell cycle distribution (Fig. 1i), rate of chromosome alignment (Fig. 1j, Supplementary Fig. 1m-o), and the number of spindle poles (Fig. 1k) in these cells in the presence of Tet. Strong accumulation of G2/M fractions was observed (Fig. 1i), and chromosomes were not properly aligned (Fig. 1j, Supplementary Fig. 1o) in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells. Abnormal numbers of spindle poles were observed (Fig. 1k). These phenotypes were similar to those observed in cKO-CT-CC Δ73 cells.
Ndc80C binding to CENP-T facilitates the Mis12C recruitment to CENP-T. To examine the functional roles of the KMN network on the CENP-T pathway, it is also important to know how each component of the network is recruited to CENP-T. Previous studies have suggested that CDK1-mediated phosphorylation of CENP-T facilitates recruitment of Ndc80C or Mis12C onto the kinetochore 37,39,41 . We further evaluated the effect of CENP-T phosphorylation on the recruitment of Ndc80C or Mis12C onto CENP-T in cKO-CT-CC Δ73 cells. We mutated potential CDK1 sites in CENP-T (CENP-T T72A-S88A ) responsible for Ndc80C binding 39 and introduced a mutant into cKO-CT-CC Δ73 cells (Fig. 2a, Supplementary Fig. 1f). To evaluate the levels of Ndc80C at kinetochores in these cell lines, we introduced Nuf2-GFP (a component of Ndc80C) into endogenous Nuf2 alleles ( Supplementary Fig. 2a-c). We measured Nuf2-GFP levels at kinetochores. The data indicated that Nuf2 levels were significantly reduced in cKO-CT-CC Δ73 /CENP-T T72A-S88A cells, compared to that in cKO-CT-CC Δ73 /CENP-T WT cells (Fig. 2b). These results support those of our previous study 39 .
We also examined the effects of CENP-T phosphorylation on Mis12C recruitment. We further observed eight serine/threonine residues in the aa 161-216 region of chicken CENP-T, and a mutant CENP-T cDNA, in which all serine/threonine residues were replaced with alanine (CENP-T 8A ), was introduced into cKO-CT-CC Δ73 cells (Fig. 2c, Supplementary Fig. 2d). GFP-Dsn1 was also introduced into the endogenous Dsn1 alleles in these cells to measure Mis12C levels at the kinetochores (Supplementary Fig. 2d). As shown in Fig. 2d, Mis12C levels in cKO-CT-CC Δ73 /CENP-T 8A cells were not significantly different from those in cKO-CT-CC Δ73 /CENP-T WT cells. Consistent with the data of Dsn1 levels, the growth of cKO-CT-CC Δ73 /CENP-T 8A cells in the presence of Tet was comparable to that of these cells in the absence of Tet (Fig. 2e). Considering these results, we conclude that phosphorylation of the aa 161-216 region of CENP-T is not essential for the recruitment of Mis12C onto the CENP-T pathway.
Previous CENP-T tethering experiments into a noncentromere locus suggested that the extreme N-terminal region of CENP-T required for Ndc80C binding might be involved in the CENP-T-Mis12C interaction 34,37 . However, it is unclear whether the same regulation occurs in native kinetochores. To examine whether the CENP-T N-terminal region facilitates Mis12C recruitment to CENP-T in native kinetochores, we Fig. 1 Disruption of the CENP-T-Mis12C interaction in cells lacking the CENP-C-Mis12C interaction leads to severe growth defects. a Recruiting pathway of KMN network via CENP-C or CENP-T. b Schematic representation of chicken CENP-T. Ndc80C-and Mis12C-binding domains were deleted in CENP-T Δ90 and CENP-T Δ121-240 , respectively. c Schematic representation of chicken CENP-C. The Mis12C-binding domain was deleted in CENP-C Δ73 . d Immunoblot analyses of various CENP-C or CENP-T in cKO-CENP-T (Tet-Off) cells. Expression of the CENP-T transgene (Tet-CENP-T) was terminated by 24 h after Tet addition. mScarlet-CENP-C Δ73 was expressed in CENP-C knockout cells. The 3X FLAG-fused CENP-T mutants were stably expressed. CENP-C migration was slightly slow, due to CENP-C phosphorylation in mitosis. Alpha-tubulin (tubulin) was used as the loading control. The asterisks indicate nonspecific bands. None depicts parental cKO-CENP-T cells. e The growth of cKO-CENP-T cells expressing 3X FLAG-CENP-T Δ121-240 in the absence or presence of Tet (-Tet or +Tet) with CENP-C WT (left) or mScarlet-CENP-C Δ73 (right). f Schematic representation of each CENP-T mutant.   (Fig. 2f, Supplementary Fig. 2e) and observed that Dsn1 levels in these cells were significantly lower than those in cKO-CT-CC Δ73 /CENP-T WT cells (Fig. 2f).
Since the CENP-T N-terminal region is critical for Ndc80C binding, we attempted to demonstrate whether direct binding of Ndc80C to the CENP-T N-terminal region facilitates Mis12C recruitment. We generated cKO-Spc25 (a direct CENP-T-binding protein in Ndc80C) (Tet-Off) cells in the absence of the CENP-C-KMN interaction (cKO-Spc25-CC Δ73 cells) ( Supplementary   Fig. 2f-h) in which GFP-Dsn1 was introduced into endogenous Dsn1 alleles. As shown in Fig. 2g, Dsn1 levels in kinetochores were significantly reduced in cKO-Spc25-CC Δ73 cells after Tet addition, compared with those in these cells before Tet addition. Since these cells contained intact CENP-T, including the Mis12Cbinding region, data indicate that direct binding of Ndc80C to CENP-T facilitates Mis12C recruitment onto CENP-T. As Mis12C on CENP-T binds to Knl1C and additional Ndc80C to form a Knl1C-Mis12C-Ndc80C unit (KMN unit), we concluded that direct binding of Ndc80C to CENP-T facilitates the Since Ndc80C-CENP-T direct binding facilitates the Mis12C-CENP-T interaction to form KMN-N, it is possible that the tension caused by kinetochore-microtubule interactions might facilitate Mis12C recruitment onto CENP-T. Using nocodazole (for the absence of tension) or MG132 (for the presence of tension), we prepared chromosome spreads after confirming that chromosomes were aligned in the presence of MG132 (Fig. 2h, i, Supplementary Fig. 2i-m) in CENP-C-deficient cells expressing CENP-C Δ73 (CENP-C Δ73 cells). Further, we compared the levels of Ndc80C and Mis12C at kinetochores in the presence or absence of tension. Both Ndc80C and Mis12C levels did not change in CENP-C Δ73 cells in either the presence or absence of tension ( Fig. 2h, i, Supplementary Fig. 2i), suggesting that the tension from microtubules might not have facilitated Mis12C recruitment onto CENP-T. It would be an important further challenge to clarify the molecular mechanisms of how direct binding of Ndc80C to CENP-T facilitates Mis12C recruitment onto CENP-T in cells.
Knl1C on the CENP-T pathway is essential for sister chromatid cohesion. Since cKO-CT-CC Δ73 /CENP-T Δ121-240 cells led to severe mitotic defects (Fig. 1), we examined the levels of Mis12C-interacting proteins in these cells. We measured Knl1C levels in cells lacking CENP-T-Mis12C and CENP-C-Mis12C interactions (Fig. 3a, Supplementary Fig. 3a). As shown in Fig. 3a, Knl1C levels were significantly reduced in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells, and the reduction levels in these cells were similar to those in cKO-CT-CC Δ73 cells.
Additionally, Bub1 recruitment leads to the proper localization of the chromosome passenger complex (CPC) and Shugoshin. These complexes together with phosphorylated threonine 3 of histone H3 (H3T3ph) contribute to the maintenance of sister chromatid cohesion in centromeres 10,34,44 . We then examined the distribution of H3T3ph, which is governed by Haspin kinase, in cKO-CT-CC Δ73 /CENP-T Δ121-240 or cKO-CT-CC Δ73 /CENP-T WT cells. In cKO-CT-CC Δ73 /CENP-T WT cells, H3T3ph signals were concentrated in the centromeric region, but these signals were diffused along the entire chromosome arm in cKO-CT-CC Δ73 / CENP-T Δ121-240 cells (Fig. 3c). Consistent with this localization profile of H3T3ph in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells, Aurora B (a component of CPC) signals were also distributed in the non-centromeric region in these cells (Fig. 3d). We interpret that the reduction of Knl1C and Bub1 leads to a reduction in the phosphorylation levels of threonine 120 of histone H2A in centromeres governed by Bub1 44,45 , which causes the mislocalization of H3T3ph and Aurora B 46,47 . Centromere localization of H3T3ph and Aurora B is related to each other, and the reduction of upstream components affects their centromeric localization (Fig. 3c, d). In addition to these localization profiles, we observed that the distance between sister kinetochores increased in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells, compared to that in cKO-CT-CC Δ73 /CENP-T WT cells (Fig. 3e). These data suggest that Mis12C and Knl1C (KM) on the CENP-T pathway facilitate proper localization of CPC and H3T3ph to centromeric regions through Bub1 recruitment to maintain the cohesion of sister chromatids (Fig. 3f). KM is known to function on CENP-C 20,48 , and we emphasize that our data demonstrated KM functions on CENP-T in cells lacking CENP-C-KMN interaction.
Two copies of Ndc80C on CENP-T are essential for their proper functions. While KM (Knl1C and Mis12C) on the CENP-T pathway plays a critical role in the maintenance of sister chromatids cohesion and the recruitment of checkpoint proteins, the roles of Ndc80C must be clarified in the KMN network on CENP-T. CENP-T has two copies of Ndc80C (N-N): one copy directly binds to the CENP-T N-terminus and the second copy is recruited via Mis12C (Fig. 3f). Therefore, the network forms KMN-N. Therefore, it is important to address whether N-N is essential for its microtubule-binding function on the CENP-T pathway independent of KM functions or whether Ndc80C must bind to Mis12C for its function. To test whether N-N is necessary for chromosome segregation, we first examined Ndc80C levels in cells lacking CENP-T-Mis12C and CENP-C-Mis12C interactions. To evaluate Ndc80C levels, we introduced Nuf2-GFP into the endogenous Nuf2 alleles in cKO-CT-CC Δ73 /CENP-T Δ121-240 , cKO-CT-CC Δ73 /CENP-T Δ161-200 , or cKO-CT-CC Δ73 /CENP-T Δ201-216 cells ( Supplementary Fig. 2a, 3b). As shown in Fig. 4a, Ndc80C levels in these cells were significantly reduced, albeit these levels were higher than those in cKO-CT-CC Δ73 cells (None in Fig. 4a), suggesting that one copy of Ndc80C still binds to the Fig. 2 Direct binding of Ndc80C to CENP-T facilitates the recruitment of Mis12C into the kinetochore. a Amino acids alignment of CENP-T between various species. Potential phosphorylation sites (T72 and S88) are marked. b Nuf2-GFP levels at kinetochores in cKO-CENP-T cells expressing mScarlet-CENP-C Δ73 and either CENP-T WT , CENP-T Δ90 , or CENP-T T72A-S88A in the presence of Tet for 30 h. None indicates parental cKO-CENP-T/CENP-C Δ73 / Nuf2-GFP cells. Subtract values are presented using signal intensities of None. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated by one-way ANOVA followed by Tukey's test. c Amino acids alignment of CENP-T between various species. Mutated S/T residues are marked. d GFP-Dsn1 levels at kinetochores in cKO-CENP-T cells expressing mScarlet-CENP-C Δ73 and CENP-T 8A in the presence of Tet for 48 h. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated by two-tailed Welch's t-test. e The growth of cKO-CENP-T cells expressing mScarlet-CENP-C Δ73 and either CENP-T 8A or CENP-T WT in the absence or presence of Tet. f GFP-Dsn1 levels at kinetochores in cKO-CENP-T cells expressing mScarlet-CENP-C Δ73 and either CENP-T WT , CENP-T Δ121-240 , or CENP-T Δ90 in the presence of Tet for 30 h. None indicates parental cKO-CENP-T/CENP-C Δ73 /GFP-Dsn1 cells. Subtract values with None are presented. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated as in b. g GFP-Dsn1 levels at kinetochores in cKO-Spc25 cells expressing mScarlet-CENP-C Δ73 in the absence or presence of Tet for 18 h. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated as in d. h GFP-Dsn1 levels at kinetochores in CENP-C KO cells expressing HA-CENP-C Δ73 (CENP-C Δ73 cells) in the presence of nocodazole or MG132. CENP-T was immuno-stained. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated as in d. i Nuf2-GFP levels at kinetochores in CENP-C Δ73 cells in the presence of nocodazole or MG132. CENP-T was immuno-stained. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated as in d.   (Fig. 4a), which is consistent with our previous observation 35 .
Although we observed that one copy of Ndc80C per one CENP-T still exists in cells lacking Mis12C, we analyzed the impact of one copy of Ndc80C without disrupting other Mis12C functions. Previous studies have suggested that the Dsn1 C-terminal region in Mis12C is critical for Ndc80C binding (Fig. 4b, c) 35,48 . Although another study suggested that the PVIHL motif in human Nsl1 in Mis12C might be also involved in Ndc80C binding 43 , this region in Nsl1 was not essential in DT40 cells, based on analyses using cKO-Nsl1 (Auxin-Inducible-Degron: AID) cells ( Supplementary Fig. 4a-e). Therefore, in this study, we focused on the Dsn1 C-terminal region. We generated cKO-Dsn1 (Tet-Off) cells, in which GFP-Dsn1 WT (cKO-Dsn1/ Dsn1 WT ) or GFP-Dsn1 Δ326-349 lacking the Ndc80C-binding region (cKO-Dsn1/Dsn1 Δ326-349 ) were expressed under the control of the endogenous Dsn1 promoter ( Supplementary  Fig. 5a). We confirmed that Dsn1 Δ326-349 formed Mis12C and bound to Knl1C but not to Ndc80C based on immunoprecipitation experiments using an anti-GFP nanobody (Fig. 4d).  In contrast, Ndc80C (Nuf2-mScarlet) levels in cKO-Dsn1/ Dsn1 Δ326-349 cells were significantly reduced, albeit Nuf2 signals were still visible (Fig. 4e, Supplementary Fig. 5h), suggesting that this mutation does not affect the CENP-T-Ndc80C direct interaction. We further investigated the growth rate of cKO-Dsn1/Dsn1 Δ326-349 cells and observed that these cells died by~96 h after Tet addition (Fig. 4f). Although cKO-Dsn1/Dsn1 Δ326-349 cells died, kinetochores in these cells may still be bound to microtubules, since some of the Ndc80C are still bound to CENP-T. We then stained microtubules and observed chromosome alignments in cKO-Dsn1/Dsn1 Δ326-349 cells. As shown in Fig. 4g, unaligned chromosomes and abnormal spindles were observed in these cells, and microtubules did not appear to bind to kinetochores. These data indicate that proper kinetochore-microtubule attachment does not occur in cKO-Dsn1/Dsn1 Δ326-349 cells. Consistent with microtubule staining, tension from microtubules was not applied to CENP-T in these cells, based on analyses using the tension sensor probe (Fig. 4h, Supplementary Fig. 4i, j). We used the talin rod (TR) tension sensor system 49,50 , similar to that used in our previous study 35 . We expressed vinculin head domain (VH)-mScarlet and CENP-T-TR, and VH-mScarlet localized to kinetochores via binding to CENP-T-TR. In cells with wild-type Dsn1, VH-mScarlet signals decreased upon nocodazole treatment. However, in cells expressing Dsn1 Δ326-349 VH-mScarlet signals did not decrease upon nocodazole treatment (Fig. 4h). Considering these results, one copy of Ndc80C directly binding to CENP-T is not sufficient for its microtubule-binding function, suggesting that two copies of Ndc80C (N-N) on CENP-T are essential for the establishment of appropriate kinetochore-microtubule interaction.
KM-Ndc80C interaction is not essential, and they are separable. Since some KM roles were ensured without interaction with Ndc80C (Fig. 5a), KM and Ndc80C might not be necessary to interact with each other and are separable (Fig. 5a). Alternatively, it is still possible that a second copy of Ndc80C might be necessary to interact with KM for its functions. To distinguish between these possibilities, we engineered the CENP-T N-terminus. If we introduced an additional Ndc80C-binding domain (aa 1-90) in the CENP-T N-terminal region, extra Ndc80C should be recruited without Mis12C binding (Fig. 5a, b). Initially, we introduced CENP-T 2x(1-90) into the endogenous CENP-T allele in cKO-CT-CC Δ73 cells and confirmed that CENP-T 2x(1-90) functions as wild-type CENP-T by examining the growth of these cells in the presence of Tet ( Supplementary Fig. 6a, b). We also observed an increase in Ndc80C (Nuf2-GFP) levels in cKO-CT-CC Δ73 /CENP-T 2x(1-90) cells owing to an additional Ndc80Cbinding site on CENP-T (Fig. 5c).
As CENP-T Δ121-240 lacks the binding of Mis12C, which binds to Knl1C and Ndc80C, it loses KM and one Ndc80C copy. Therefore, in addition to defects due to a lack of KM roles (Fig. 3), microtubule-binding defects owing to Ndc80C reduction could occur in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells (Fig. 4a). We further stained microtubules and observed chromosome alignments in cKO-CT-CC Δ73 /CENP-T Δ121-240 , cKO-CT-CC Δ73 / CENP-T WT , or cKO-CT-CC Δ73 cells (Fig. 5d). We frequently observed unaligned chromosomes, and the spindle shape was not normal in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells, as observed in cKO-CT-CC Δ73 cells (Fig. 5d). We also observed abnormal numbers of spindle poles (Fig. 5e), a higher rate of misalignment of chromosomes (Fig. 5f), and an increase in G2/M populations (Fig. 5g) in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells. These data suggest that proper kinetochore-microtubule attachment and spindle integrity were disrupted in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells. However, the addition of the Ndc80C-binding site to CENP-T Δ121-240 (CENP-T 2x(1-90)_Δ121-240 ) might rescue defects in proper kinetochore-microtubule attachments. We further generated cKO-CT-CC Δ73 /CENP-T 2x(1-90)_Δ121-240 cells (Supplementary Fig. 6c) and analyzed the phenotype of these cells. Abnormal spindle shape, improper kinetochore-microtubule attachments, number of spindle poles, and defects in chromosome misalignment were largely suppressed in cKO-CT-CC Δ73 /CENP-T 2x(1-90)_Δ121-240 cells, although recovery of chromosome misalignment was not strong, compared with cKO-CT-CC Δ73 /CENP-T WT cells (Fig. 5d-f, Supplementary Fig. 6d). These data suggest that some Ndc80C-related defects in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells were partially suppressed by the addition of extra Ndc80C-binding sites. However, significant cohesion defects were still observed in cKO-CT-CC Δ73 / CENP-T 2x(1-90)_Δ121-240 cells (Fig. 5h). In addition, cKO-CT-CC Δ73 / CENP-T 2x(1-90)_Δ121-240 cells still revealed G2/M accumulation (Fig. 5g) and ultimately died by~96 h after Tet addition (Fig. 5i), implying that the addition of extra Ndc80C to CENP-T did not completely suppress mitotic defects in cKO-CT-CC Δ73 /CENP-T Δ121-240 cells. One possible explanation is that KM mitotic functions are essential for cell growth. Another possible explanation is that the Fig. 4 Two copies of Ndc80C (N-N) are essential for proper kinetochore-microtubules interaction. a Nuf2-GFP levels at kinetochores in cKO-CENP-T cells expressing mScarlet-CENP-C Δ73 and various CENP-T mutants in the presence of Tet for 30 h. None indicates parental cKO-CENP-T/CENP-C Δ73 / Nuf2-GFP cells. Subtract values were presented using signal intensities of None. DNA was stained with DAPI. Error bars show the mean and standard deviation. p values were calculated by one-way ANOVA followed by Tukey's test. b Schematic representation of chicken Dsn1 Δ326-349 . c Schematic representation explaining that chicken Dsn1 Δ326-349 does not bind to Ndc80C but binds to Knl1C via the formation of Mis12C. d Immunoprecipitation with anti-GFP in cKO-Dsn1 cells expressing either GFP-Dsn1 WT or GFP-Dsn1 Δ326-349 in the presence of Tet. Samples were subjected to immunoblot analyses using anti-Dsn1, -Mis12, -Knl1, and -Ndc80 antibodies. e Nuf2-mScarlet levels at kinetochores in cKO-Dsn1 cells expressing either GFP-Dsn1 WT or GFP-Dsn1 Δ326-349 in the presence of Tet for 30 h. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated by two-tailed Welch's t-test. f The growth of cKO-Dsn1 cells expressing either GFP-Dsn1 WT or GFP-Dsn1 Δ326-349 in the absence or presence of Tet (-Tet or +Tet). None indicates parental cKO-Dsn1 cells. g Microtubule staining using an anti-tubulin antibody (red) for cKO-Dsn1 cells expressing either GFP-Dsn1 WT or GFP-Dsn1 Δ326-349 in the presence of Tet for 30 h. The cells were incubated in ice-cold medium for 10 min prior to fixation. DNA was stained with DAPI. h mScarlet-fused vinculin head domain (VH-mScarlet) localization to kinetochore in cKO-Dsn1 cells expressing CENP-T-TR reporter and either GFP-Dsn1 WT or GFP-Dsn1 Δ326-349 . Disordered region of CENP-T (aa 241-529) was exchanged with the chicken TR domain (aa 482-889) (CENP-T TR), was expressed from the CENP-T locus. Cells were treated with or without nocodazole (Noc + or -), and mScarlet signals were measured. DNA was stained with DAPI. Error bars indicate the mean and standard deviation. p values were calculated as in a. To distinguish between these two possibilities, we expressed CENP-T 2x  in cKO-Dsn1/Dsn1 Δ326-349 cells ( Supplementary  Fig. 6e), which lost one copy of Ndc80C and retained KM functions (KM-N). Although the KMN network forms KMN-N (one copy of Ndc80C binds to Mis12C, and another Ndc80C directly binds CENP-T) in wild-type chicken cells, KM-N-N (two copies of Ndc80C directly bind to CENP-T distinct from KM) may be functional (Fig. 5a). Strikingly, expression of GFP-CENP-T 2x  in cKO-Dsn1/Dsn1 Δ326-349 cells completely compensated for mitotic defects in the cKO-Dsn1/Dsn1 Δ326-349 cells in the presence of Tet (Fig. 5j-m, Supplementary Fig. 6f). Abnormal  (Fig. 5n). These data indicate that two copies of Ndc80C can be functional without direct binding to Mis12C. This result indicates that direct binding of Ndc80C to Mis12C is not required for the microtubule-binding function of Ndc80C. Comparing the phenotype of cKO-Dsn1/Dsn1 Δ326-349 cells expressing CENP-T 2x(1-90) (Fig. 5j-n) with that of cKO-CT-CC Δ73 /CENP-T 2x(1-90)_Δ121-240 cells (Fig. 5d-i), we demonstrated that KM has essential mitotic roles, distinct from the microtubule-binding functions of N-N. We conclude that KM and N-N (two copies of Ndc80C) are separable and have distinct essential functions for cell growth (Fig. 5a).

Discussion
We previously revealed that the CENP-T pathway is the main pathway for the recruitment of the KMN network onto kinetochores, compared to the CENP-C pathway, in chicken DT40 cells 35 . However, the specific roles of the KMN network on the CENP-T pathway remain unclear. The KMN network binds to both CENP-C and CENP-T. Therefore, to precisely evaluate the roles of the KMN network on the CENP-T pathway, we must remove the CENP-C-KMN interaction to eliminate some redundant functions of CENP-C. Moreover, since some proteins have multiple roles, we must use specific mutants of each protein, which do not disrupt their multiple functions to avoid misinterpretation of its protein function. Based on our analyses, we concluded that the KMN network forms KMN-N in chicken cells and that KM and N-N have distinct roles: Proper kinetochoremicrotubule interaction via N-N (two copies of Ndc80C) and the maintenance of sister chromatid cohesion, recruitment of one copy of Ndc80C or checkpoint function via KM (Knl1C and Mis12C) (Fig. 7a, b). Importantly, we observed that the Mis12C-Ndc80C interaction was not necessary, and KM and N-N are separable (Fig. 7a, b). Since the Ndc80C-Mis12C interaction is thought to be critical for the kinetochore functions 7,42,43 , our observations provide new insights for kinetochore studies. There are multiple recruitment pathways of the KMN network onto kinetochores in other organisms, in addition to the CENP-C and CENP-T-pathways 52,53 , and the pathway choice is variable among different species 54 . Although many species have more than two pathways, Drosophila melanogaster and Caenorhabditis elegans have only the CENP-C pathway [55][56][57] . In contrast, several Lepidoptera and Mucor have only the CENP-T pathway 58,59 . These observations suggest that each pathway can recruit the KMN network onto the kinetochore, and the CENP-T and CENP-C pathways might be redundant in organisms possessing more than two pathways. However, our previous study indicated that the CENP-T-KMN network interaction is essential, but the CENP-C-KMN network interaction is dispensable in chicken DT40 cells. Nevertheless, we hypothesized that the CENP-C-KMN network interaction plays a role, because cells lacking Mis12C binding to CENP-T in the presence of the CENP-C-KMN network interaction do not completely die. This study demonstrated that cells completely died in the absence of Mis12C interaction with both CENP-C and CENP-T, suggesting that Mis12C on CENP-C plays a role, which might contribute to sister chromatid cohesion or checkpoint function via KM or robust kinetochore-microtubule attachment via additional Ndc80C since Mis12C binds to Knl1C and Ndc80C. Therefore, to analyze the functional roles of KMN in each pathway, we must characterize it with the depletion of some redundancies. This strategy is essential for evaluating the function of each complex in different organisms with various recruitment pathways of the KMN network.
Two copies of Ndc80C (N-N) are necessary and sufficient for proper kinetochore-microtubule attachment in DT40 cells, suggesting that binding of a single copy of Ndc80C to spindle microtubules is not sufficient for full kinetochore functions. Single-molecule experiments demonstrated that the multivalency of Ndc80C efficiently tracked depolymerizing microtubules, whereas a single copy of Ndc80C did not 51 , which supports our observations. Other biophysics experiments suggest that microtubule pulling forces from kinetochores vary (0.4-8 pN) 60,61 , suggesting that the double increase of microtubulebinding proteins might not be sufficient for stable chromosome segregation. However, it is unclear how much the kinetochore with one copy of Ndc80C is unstable, and such kinetochores might not generate sufficient force for chromosome segregation. Thus, the extra copy of Ndc80C might have synergistic effects on kinetochore function in vivo. It is important to know how extra copies of Ndc80C contribute to kinetochore function.
The CENP-C-KMN network interaction is not sufficient in DT40 cells since Mis12C is easily dissociated from CENP-C 35 . However, forced binding of Mis12C to CENP-C stabilizes the CENP-C pathway. In this situation, there is only one copy of Ndc80C on the kinetochore, which possesses only the CENP-C pathway. However, kinetochores with one copy of the Ndc80C were functional in this case. This mechanism might be explained by the amount of protein at the kinetochore 38 . Suzuki et al. 38 proposed~215 molecules of CENP-C and~72 molecules of CENP-T in a human kinetochore. Since only~70 molecules of Mis12C bind to CENP-C, two-thirds of CENP-C are usually not used for the KMN network interaction. However, CENP-C has the capacity to accept more KMN components, and forced binding of Mis12C to CENP-C would cause a stable CENP-C pathway, even if the kinetochore with the CENP-C pathway has only one copy of Ndc80C. Additionally, this measurement may explain why two copies of Ndc80C are necessary for CENP-T. Compared with CENP-C amounts, CENP-T amounts are small at each kinetochore 38 . To maintain sufficient amounts of Ndc80C, two copies of Ndc80C per CENP-T may be critical for chromosome segregation. Independent of the microtubule-binding function of two copies of Ndc80C (N-N), Knl1C and Mis12C (KM) have some essential roles, since even when cells had two copies of Ndc80C for proper kinetochore-microtubule interaction in cKO-CT-CC Δ73 /CENP-T 2x(1-90)_Δ121-240 cells, the cells died. Although it is still unclear which KM functions are critical for cell viability, KM has various roles, including the maintenance of sister chromatid cohesion or the recruitment of checkpoint proteins. KM binds to Bub1, which further recruits various factors required for the maintenance of sister chromatid cohesion, including the cohesin complex. The cohesin complex, H3T3ph, and CPC localize on entire chromosomes until prophase, but they localize to centromeric regions in metaphase cells. KM appears to play an essential role in proper centromere localization of H3T3ph or CPC in metaphase (Fig. 3f), since these localizations are dispersed into entire chromosomes in mutant cells lacking the Mis12C-CENP-T interaction. KM either directly or indirectly facilitates the removal of H3T3ph or CPC from chromosome arms and protects their removal from centromeres in metaphase for sister chromatid cohesion.
The functions of KM and N-N on the CENP-T pathway are relatively predictable based on previous observations 2,6 , however, it is surprising to find that KM and N-N are separable, which means that Mis12C is not necessary for binding to Ndc80C. This striking result contrasts with a previous yeast study 62 , which proposed that the microtubule-binding activity of yeast Ndc80C is regulated by its interaction with the MIND complex (yeast Mis12 complex).
For the CENP-C-KMN network interaction, Mis12C is the sole platform for Ndc80C; therefore, this interaction is critical. On the CENP-T-KMN network interaction, as one copy of Ndc80C is not sufficient, the Mis12C-Ndc80C interaction is critical in chicken cells. However, if Ndc80C is artificially supplied to CENP-T, this Ndc80C-recruitment function for Mis12C becomes dispensable (Fig. 5). Although this was observed in an artificial chicken situation, the Ndc80C-Mis12C interaction is dispensable in human cells, since humans have three copies of Ndc80C on CENP-T (KMN-N-N) (Fig. 7a) 35,41 . To bind to Ndc80C, CENP-T phosphorylation of the threonine residue of the TPR motif in its N-terminus is critical (Fig. 7c) 39,41 . Two TPR motifs exist in primate CENP-T, albeit a TPR motif in the extreme N-terminus of CENP-T is not clearly detected in CENP-Ts other than the primate CENP-T, while Chinese hamsters have two TPR motifs (Fig. 7c). This suggests that the TPR motif was acquired after the primate lineage to ensure robust kinetochore-microtubule interactions. Although we prefer to interpret that the copy number of Ndc80C binding sites on CENP-T is a critical factor for the increase in Ndc80C amounts in most organisms, it is also possible that the amount of Ndc80C would be increased in different ways. Therefore, this point is important and needs to be addressed.
While KM and N-N are separable, one copy of Ndc80C is usually supplied via Mis12C binding; however, the advantage of this system is unknown. This may be related to the Mis12C recruitment mechanism on the CENP-T pathway. Direct binding of Ndc80C to CENP-T facilitates the recruitment of Mis12C (Fig. 2), which further recruits additional copy of Ndc80C. This is an efficient way to recruit essential factors to the CENP-T pathway, and it is not necessary to have two Ndc80-binding sites in CENP-T. In in vitro experiments, Mis12C can bind to CENP-T, even if CENP-T does not directly bind to Ndc80C 41 . Therefore, the efficient Mis12C recruitment mechanism via direct Ndc80C binding to CENP-T cannot be explained only by performing in vitro studies. It would be an important further challenge to clarify the regulatory mechanisms of Mis12C recruitment onto the CENP-T pathway.
The kinetochore is a multi-protein complex with complicated protein-protein networks that often exhibit redundant functions. Therefore, it is necessary to dissect the functional roles of each kinetochore complex. In this study, using precise information on the binding sites of each complex, we defined the distinct functions of KM and N-N on the CENP-T pathway and demonstrated that they are separable. It is difficult to distinguish these functions using only a simple knockout Fig. 6 Mis12C-Ndc80C interaction is dispensable in human cells. a Schematic representation of the KMN network on CENP-T in chicken and human cells. Chicken has KMN-N on CENP-T. While cKO-Dsn1/Dsn1 Δ326-349 cells (KMN-N) died, expression of CENP-T 2X  in these cells (KMN-N-N) rescued growth deficiency. In human cells, KMN-N-N is on CENP-T. The phenotype of cKO-Dsn1 lacking the Mis12C-Ndc80C interaction (cKO-hDsn1/hDsn1 Δ325-356 cell: KMN-N-N) will be tested. The KMN network on the CENP-C pathway in chicken and human cells is also presented. Both gDsn1 Δ326-349 and hDsn1 Δ325-356 do not bind to Ndc80C on the CENP-C pathway. b Immunoprecipitation with anti-RFP in AID-based cKO-hDsn1 cells expressing either mScarlet-hDsn1 WT or -hDsn1 Δ325-356 in the presence of nocodazole for 24 h and IAA for last 2 h. Samples were analyzed by anti-Dsn1, and -Ndc80 antibodies. c Ndc80 levels at kinetochores in cKO-hDsn1 cells expressing mScarlet-hDsn1 WT or -hDsn1 Δ325-356 in the presence of IAA for 2 h. DNA was stained with DAPI. Ndc80 was stained with an anti-Ndc80 antibody (red). CENP-T was stained with an anti-CENP-T antibody (green). None indicates parental cKO-hDsn1 (AID) cells. Error bars indicate the mean and standard deviation. p values were calculated by one-way ANOVA followed by Tukey's test. d Tubulin staining in cKO-hDsn1 cells expressing hDsn1 WT or hDsn1 Δ325-356 in the presence of IAA for 2 h. Kinetochores were stained with an anti-human CENP-A antibody (red). The insets present the kinetochore-microtubule bindings. DNA was stained with DAPI. e Images of mitotic chromosomes stained with anti-CENP-A in cKO-hDsn1 cells (None) expressing hDsn1 WT or hDsn1 Δ325-356 in the presence of IAA for 2 h. DNA was stained with DAPI. The graph summarizes the percentage of cells with misaligned chromosomes in each line. Error bars indicate the mean and standard deviation. p values were calculated as in c. f Images of anaphase for cKO-hDsn1 cells (None) expressing hDsn1 WT or hDsn1 Δ325-356 in the presence of IAA for 2 h. DNA was stained with DAPI. The graph summarizes the percentage of cells with abnormal anaphase in each line. Error bars indicate the mean and standard deviation. p values were calculated as in c. g The growth of cKO-hDsn1 cells expressing mScarlet-hDsn1 WT or -hDsn1 Δ325-356 in the absence or presence of IAA.
approach for the components of the KMN network. Furthermore, functional kinetochores could be created by providing an artificial Ndc80C-binding site on CENP-T to explain the distinct KMN functions in the CENP-T pathway. Therefore, to control or manipulate kinetochore functions in the future, we must understand the precise roles of each complex in vivo, including redundant functions. We believe that this study provides a model for analyzing the precise function of each kinetochore complex in vivo and provides critical insights into kinetochore functions.

Methods
Cell culture. Chicken DT40 cells 63 were cultured in Dulbecco's modified Eagle's medium (Nacalai tesque, 08459-64) containing 10% fetal bovine serum (Sigma, 172012-500 mL), 1% chicken serum (Thermo Fisher, 16110-082), 10 µM 2-Mercaptoethanol (Sigma, M3148), and penicillin (100 unit/mL)-streptomycin (100 µg/mL) (Thermo Fisher, 15140-122), at 38.5°C with 5% CO 2 . For gene inactivation in tetracycline-dependent conditional knockout cell lines (cKO-CENP-T, cKO-Dsn1, and cKO-Spc25 cells), 2 µg/ml tetracycline (Sigma, T-7660) was added. For degradation of GFP-AID-Nsl1 in cKO-Nsl1 (AID) cells, 500 µM 3-Insole acetic acid (IAA; Wako, 090-07123) was added. For examination of the cell growth for each cKO cell line, tetracycline or IAA was added to the culture medium at 0 h and the cell numbers were measured at each time point. Cell Fig. 7 The KMN network plays distinct roles on CENP-T. a Summary of numbers of K, M, and N in various mutants and their phenotypes in chicken and human cells. b A model of the essential roles of the KMN network on the CENP-T pathway. KM recruits Bub1, H3T3ph, and Aurora B to maintain sister chromatid cohesion and bind to checkpoint proteins. In addition, KM recruits one Ndc80C copy. However, two copies of Ndc80C (N-N) are critical for the establishment of kinetochore-microtubule interactions. As Ndc80C is not necessary to bind to Mis12C, KM and N-N are separable. In CENP-T 2X(1-90)_Δ121-240 cells (N-N lacking KM), cohesion defects were still observed, even if sufficient amounts of Ndc80C were bound to kinetochores. In Dsn1 Δ326-349 cells (KM-N lacking one N), only one copy of Ndc80 localized kinetochores, and proper chromosome alignments did not occur, while cohesion defects were not observed. c Amino acid sequence alignments of the CENP-T extreme N-terminal region, conserved second N-terminal region, and Dsn1 C-terminal end for Ndc80C binding between various species.
numbers were normalized to that before tetracycline or IAA addition (0 h) for each cell line.
Human RPE-1 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and penicillin (100 unit/mL)-streptomycin (100 µg/mL) at 37°C with 5% CO 2 . For degradation of GFP-AID-hDsn1 in cKO-hDsn1 (AID) cells, 250 µM IAA was added. For examination of the cell growth for each cKO cell line, IAA was added to the culture medium at 0 h and the cell numbers were measured at each time point. Cell numbers were normalized to that before IAA addition (0 h) for each cell line. To disrupt the expression of the endogenous CENP-T, the Neomycin resistance cassette (SV40 Neo R ) was cloned into the pBSKS_CT 2k vector (pBSKS_ CT KI_SV40 Neo R ). To integrate these constructs into the endogenous CENP-T locus by using CRISPR/Cas9-mediated homologous recombination, the pX335_ggCENP-T containing single-guide RNA (sgRNA) against a genomic sequence around the start codon of CENP-T and SpCas9 nickase (D10A) gene 35 was used.
The mScarlet-fused CENP-C Δ73 expression vector was generated by cloning of CENP-C Δ73 cDNA into the pmScarlet-C1 (Addgene, 85042) vector. To express mScarlet-fused CENP-C Δ73 under control of the endogenous β-actin (ACTB) promoter in DT40 cells, mScarlet-fused CENP-C Δ73 and the Hygromycin B resistance gene expressed under control of the PGK promoter (PGK Hygro R ) were cloned into the pBSKS_ACTB 2k vector, in which about 2 kb genome fragment (5′ and 3′ homology arm regions (approximately 1 kb each)) around the ACTB start codon is cloned 35 , by In-Fusion ® HD Cloning Kit (pBSKS_ACTB KI_mScarlet-CENP-C Δ73 _PGK Hygro R ). To express the HA-fused CENP-C Δ73 under control of the endogenous ACTB promoter in DT40 cells, mScarlet region of the pBSKS_ACTB KI_mScarlet-CENP-C Δ73 _PGK Hygro R was replaced with the HAtag (pBSKS_ACTB KI_HA-CENP-C Δ73 _PGK Hygro R ). To integrate these constructs into the endogenous ACTB locus by using CRISPR/Cas9-mediated homologous recombination, pX335_ggACTB containing sgRNA against a genomic sequence around the start codon of ACTB and SpCas9 nickase (D10A) gene 35 was used.
To express the GFP-fused Nuf2 under control of the Nuf2 promoter in DT40 cells, GFP-fused Nuf2 and PGK Puro R or PGK EcoGPT were cloned into pBSKS_Nuf2 2k vector, in which about 2 kb genome fragment (5′ and 3′ homology arm regions (approximately 1 kb each)) around the end of the exon1 of Nuf2 genomic sequence is cloned, by In-Fusion ® HD Cloning Kit (pBSKS_Nuf2 KI_Nuf2-GFP_PGK Puro R and pBSKS_Nuf2 KI_Nuf2-GFP_PGK EcoGPT). To express the mScarlet-fused Nuf2 under control of the Nuf2 promoter in DT40 cells, mScarlet-fused Nuf2 and ACTB BSR or PGK EcoGPT were cloned into pBSKS_Nuf2 2k vector by In-Fusion ® HD Cloning Kit (pBSKS_Nuf2 KI_Nuf2-mScarlet_ACTB BSR and pBSKS_Nuf2 KI_Nuf2-mScarlet_PGK EcoGPT). To integrate these constructs into the endogenous Nuf2 locus by using CRISPR/Cas9-mediated homologous recombination, we designed sgRNA against a DNA sequence (GCCACCCTTAAGGTGTGTG) in the intron 1 of the Nuf2 genomic sequence using CRISPOR 64 , and cloned it into the BbsI site of the pX330 (pX330_ggNuf2).
To disrupt expression from the endogenous CENP-C gene by using CRISPR/ Cas9 system, sgRNA against a DNA sequence in the exon2 of the CENP-C genomic sequence (CGAGCAAGATTCTGCGGGCA) was designed using CRISPOR 64 , and cloned it into the BbsI site of the pX330 (pX330_ggCENP-C).
To express the GFP-fused Nsl1 WT , the pEGFP-C3_Nsl1 WT was used, and leucine 220 and leucine 222 residues of pEGFP-C3_Nsl1 WT were substituted with glutamic acid (pEGFP-C3_Nsl1 EE ) by PCR based mutagenesis for the expression of GFP-fused Nsl1 EE .
To express the VH-mScarlet under control of the endogenous ACTB promoter in DT40 cells for the tension sensor assay, GFP-sequence of VH-GFP (a gift from T. Maresca, University of Massachusetts Amherst, Amherst, MA, USA) was replaced with mScarlet (VH-mScarlet). Then, VH-mScarlet and PGK EcoGPT were cloned into pBSKS_ACTB 2k vector (pBSKS_ACTB 2k_VH-mScarlet).
Generation of cell lines. The cell lines established in this study are listed in Supplementary Table 1.
To generate cKO-Nsl1 (AID) cells, the pAID1.2-NEGFP-Nsl1 was transfected with the pX330_ggNsl1 and pX330_pAID1.2 using Neon Transfection System with 6 times of pulse (1400 V, 5 msec) in CL18 cells. The transfected cells were selected in the medium containing 25 µg/mL Blasticidin S hydrochloride in 96 well plates to isolate single-cell clones.
To express the GFP-fused Nsl1 WT and Nsl1 EE , the linearized pEGFP-C3_Nsl1 WT or pEGFP-C3_Nsl1 EE was transfected using Gene Pulser II electroporator (Bio-Rad, 165-2105) in cKO-Nsl1 (AID) cells. The transfected cells were selected in the medium containing 2 mg/mL G418 in 96 well plates to isolate single-cell clones.
For the genomic DNA extraction of RPE-1 cells, cells were collected after trypsinizing, washed with PBS, and suspended in 0.2 mg/mL Proteinase K in 0.1% PBST (PBS, 0.1% Tween 20). The suspended cells were incubated at 55°C for 90 min and heated at 96°C for 15 min.
To investigate the mutation sequence in the chicken CENP-C gene, the genomic regions flanking the sgRNA target site were amplified by PCR using the following primers: Fw: 5′-ATGCATCAACCAGGAGGCTGTC -3′, Rv: 5′-CTAAGGCACA CCATTAGTTTTGG -3′. The PCR amplicons were cloned into pBluescript II SK (-) and sequenced.
The target integration of several constructs was confirmed by PCR. The primers used for PCRs were listed in Supplementary Table 2.
The fluorescence signal intensities of H3T3ph and Aurora B in the centromere region and the noncentromere region were quantified using Fiji. The centromere region was selected by the 10 × 15 pixels (length × width) region along the chromosome axis (the region of interest: ROI-A). This region was enlarged with the 10 × 20 pixels (length × width) region (ROI-B) for background subtraction. To determine the background signal intensity, the sum of fluorescence signal intensity of ROI-B was subtracted with the sum of fluorescence signal intensity of ROI-A (ROI-B -ROI-A). The mean of fluorescence signal intensity on the centromere region (ROI-A) was subtracted with the mean of the background signal intensity. The noncentromere region was selected by the length size (pixel length size depends on length of chromosome arm) × 15 pixels (length × width) region along with chromosome axis. Background subtraction was performed in the same way as the centromere region. The fluorescence signals on 4 chromosomes in each cell ( Fig. 3c: n = 14 (WT) and 13 (Δ121-240) cells; Fig. 3d, Supplementary Fig. 5e, 5f: n = 13 cells) were quantified. p values were calculated by two-tailed Welch's t-test.
The percentage of cells with misaligned chromosomes (Fig. 6e) or cells with abnormal chromosome (Fig. 6f) were calculated. Each experiment was repeated three times ( Fig. 6e: n = 24 or 25 cells per experiment) or four times (Fig. 6f: n = 24 or 25 cells per experiment), and mean and standard deviation of three (Fig. 6e) or four (Fig. 6f) experiments are presented. p values were calculated by one-way ANOVA followed by Tukey's test.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Source data are provided with this paper. All data supporting the findings of this study are available within the paper and supplementary information. Source data are provided with this paper.