Eic1 links Mis18 with the CCAN/Mis6/Ctf19 complex to promote CENP-A assembly

CENP-A chromatin forms the foundation for kinetochore assembly. Replication-independent incorporation of CENP-A at centromeres depends on its chaperone HJURPScm3, and Mis18 in vertebrates and fission yeast. The recruitment of Mis18 and HJURPScm3 to centromeres is cell cycle regulated. Vertebrate Mis18 associates with Mis18BP1KNL2, which is critical for the recruitment of Mis18 and HJURPScm3. We identify two novel fission yeast Mis18-interacting proteins (Eic1 and Eic2), components of the Mis18 complex. Eic1 is essential to maintain Cnp1CENP-A at centromeres and is crucial for kinetochore integrity; Eic2 is dispensable. Eic1 also associates with Fta7CENP-Q/Okp1, Cnl2Nkp2 and Mal2CENP-O/Mcm21, components of the constitutive CCAN/Mis6/Ctf19 complex. No Mis18BP1KNL2 orthologue has been identified in fission yeast, consequently it remains unknown how the key Cnp1CENP-A loading factor Mis18 is recruited. Our findings suggest that Eic1 serves a function analogous to that of Mis18BP1KNL2, thus representing the functional counterpart of Mis18BP1KNL2 in fission yeast that connects with a module within the CCAN/Mis6/Ctf19 complex to allow the temporally regulated recruitment of the Mis18/Scm3HJURP Cnp1CENP-A loading factors. The novel interactions identified between CENP-A loading factors and the CCAN/Mis6/Ctf19 complex are likely to also contribute to CENP-A maintenance in other organisms.


Summary
CENP-A chromatin forms the foundation for kinetochore assembly. Replicationindependent incorporation of CENP-A at centromeres depends on its chaperone HJURP Scm3 , and Mis18 in vertebrates and fission yeast. The recruitment of Mis18 and HJURP Scm3 to centromeres is cell cycle regulated. Vertebrate Mis18 associates with Mis18BP1 KNL2 , which is critical for the recruitment of Mis18 and HJURP Scm3 . We identify two novel fission yeast Mis18-interacting proteins (Eic1 and Eic2), components of the Mis18 complex. Eic1 is essential to maintain Cnp1 CENP-A at centromeres and is crucial for kinetochore integrity; Eic2 is dispensable. Eic1 also associates with Fta7 CENP-Q/Okp1 , Cnl2 Nkp2 and Mal2 CENP-O/Mcm21 , components of the constitutive CCAN/Mis6/Ctf19 complex. No Mis18BP1 KNL2 orthologue has been identified in fission yeast, consequently it remains unknown how the key Cnp1 CENP-A loading factor Mis18 is recruited. Our findings suggest that Eic1 serves a function analogous to that of Mis18BP1 KNL2 , thus representing the functional counterpart of Mis18BP1 KNL2 in fission yeast that connects with a module within the CCAN/Mis6/Ctf19 complex to allow the temporally regulated recruitment of the Mis18/Scm3 HJURP Cnp1 CENP-A loading factors. The novel interactions identified between CENP-A loading factors and the CCAN/Mis6/Ctf19 complex are likely to also contribute to CENP-A maintenance in other organisms.

Introduction
Genome integrity and stability are key to the propagation of genetic information through generations. Centromeres are the specialized sites on eukaryotic chromosomes where kinetochores, which govern spindle microtubule attachment, assemble during cell division. Centromere dysfunction, and consequent kinetochore instability, can therefore cause mis-segregation of chromosomes resulting in aneuploidy and genome instability. A plethora of evidence suggests that DNA sequence is neither necessary nor sufficient for the establishment of a functional centromere [1,2], and it is generally accepted that centromere specification is epigenetically regulated in many organisms [3,4].
CENP-A, a histone H3 variant, replaces canonical H3 in specialized nucleosomes that mark active centromeres. CENP-A is both necessary and sufficient to mediate the establishment and maintenance of functional kinetochores. CENP-A associates with a distinct set of proteins that allow it to promote its own propagation to preserve centromere location [5][6][7]. In the fission yeast Schizosaccharomyces pombe, Cnp1 CENP-A is restricted to the non-repetitive central domain of centromeres (approx. 7 that is flanked by heterochromatic outer repeats. Cnp1 CENP-A recruitment to the central domain depends on the conserved kinetochore proteins Mis18, Mis6 CENP-I/Ctf3 , Sim4 CENP-K/ Mcm22 and Scm3 HJURP [8][9][10][11][12]. Mis18 was originally identified as being required to maintain Cnp1 CENP-A at fission yeast centromeres. Subsequently, Mis18 function was shown to be conserved at vertebrate centromeres [11,13]. Mis18 associates with the nuclear protein Mis16, which is orthologous to Saccharomyces cerevisiae Hat2 and human RbAp46/48, which are known to act as histone chaperones in many distinct histone transactions [14,15]. Mis6 CENP-I/Ctf3 and Sim4 CENP-K/Mcm22 reside in a larger complex consisting of approximately [16][17][18] centromeric proteins [16][17][18], most of which are conserved and collectively form the complex known as CCAN, Ctf19 and Mis6 in vertebrates, budding yeast and fission yeast, respectively [16,19,20]. The modular CCAN/Mis6/Ctf19 complex forms the main scaffold of centromere proteins that is required to recruit outer kinetochore components in order to mediate and regulate kinetochore-microtubule attachments [21]. Scm3 HJURP is orthologous to the vertebrate CENP-Aspecific chaperone HJURP [22]. Scm3 HJURP directly associates with Cnp1 CENP-A and is required for Cnp1 CENP-A incorporation at centromeres [9,12].
In vertebrate cells, the assembly of new CENP-A into centromeric chromatin is uncoupled from replication and occurs in early G1, prior to S phase [23]. Consequently, following centromere replication, the level of resident CENP-A at centromeres is halved as a result of its distribution to the resulting two sister-centromeres [24]. The CENP-A loading factors Mis18a/b and Mis18BP1 KNL2 and the CENP-A chaperone HJURP Scm3 are transiently recruited to human centromeres in telophase and are required to replenish CENP-A in G1 [13,25,26]. Mis18 is recruited to human centromeres by Mis18BP1 KNL2 , which associates with centromeres via the C-terminus of the constitutive kinetochore component, CENP-C Cnp3/Mif2 [27,28]. Both Mis18a and Mis18BP1 KNL2 are required for the cell-cycle-regulated recruitment of HJURP Scm3 prior to G1 [25,28,29]. Interestingly, in fission yeast, the CENP-C orthologue Cnp3 is not essential for viability [30] and no Mis18BP1 KNL2 protein can be identified. It therefore remains unclear how Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP are recruited to centromeres to maintain Cnp1 CENP-A at fission yeast centromeres. Nevertheless, as in vertebrate cells, the Cnp1 CENP-A assembly factors Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP all display dynamic association with fission yeast centromeres during the cell cycle, but they are released from centromeres in mitotic prophase and re-associate in mid-anaphase [9,[11][12][13]. This suggests that fission yeast relies on an alternative to the CENP-C Cnp3/Mif2 -Mis18BP1 KNL2 interaction for the regulated recruitment of the Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP Cnp1 CENP-A assembly factors to centromeres.
Although many kinetochore proteins have clearly been conserved between different experimental organisms over evolutionary time, it is also evident that particular proteins have been lost in some organisms and replaced by alternative pathways. For example, both Drosophila and S. cerevisiae lack the Mis18 and Mis18BP1 KNL2 proteins. However, S. cerevisiae retains an Scm3 HJURP orthologue, whereas Scm3 HJURP function is replaced by the distinct Cal1 protein in Drosophila [31,32]. Moreover, in Arabidopsis the association dynamics of Mis18BP1 KNL2 with centromeres is more similar to that of S. pombe Mis18/Scm3 HJURP than human Mis18BP1 KNL2 / Mis18/HJURP Scm3 in that it departs only briefly from centromeres during mitosis, re-associating in mid-anaphase prior to Arabidopsis CenH3 CENP-A deposition in G2 [33,34]. It is well recognized that alternative solutions can evolve to mediate the same conserved process, and such alternatives can inform on parallel or other underlying pathways and mechanisms that may be more prevalent in one system relative to another.

Hat1 does not contribute to the maintenance of Cnp1 CENP-A chromatin
In S. cerevisiae, Hat2 is the orthologue of Mis16 and it has previously been shown to associate with the histone H4 acetyl-transferase Hat1 [36]. Related to this, levels of acetylated histones have been shown to increase at centromeres in mis16 and mis18 mutants [11]. It is possible that altered histone acetylation at centromeres relates to the function of His-Eic1

GST-Mis16
His-Mis16  rsob.royalsocietypublishing.org Open Biol. 4: 140043 Mis16 RbAp46/48/Hat2 and Hat1 at centromeres. We first confirmed that Mis16-Myc and Hat1-HA co-immunoprecipitate (electronic supplementary material, figure S1a) and then, to investigate the possibility that Hat1 possesses centromerespecific functions, we subjected Hat1-HA immunoprecipitates to LC/MS-MS analyses. However, we were unable to detect any Hat1-associated proteins apart from Mis16 RbAp46/48/Hat2 , confirming the analyses of others [35]. Moreover, Hat1-HA was not found to be enriched in the central kinetochore domain of centromeres (electronic supplementary material, figure S1b) and no loss of Cnp1 CENP-A from centromeres could be detected in cells lacking Hat1 (hat1D; electronic supplementary material, figure S1c

Eic1 and Eic2 associate with centromeres dynamically through the cell cycle
CENP-A and all known fission yeast kinetochore proteins localize specifically at centromeres and are enriched over the central domain region of centromeres. Both Eic1 and Eic2 associate with Mis18, which is known to localize to fission yeast centromeres for most of the cell cycle, apart from early prophase to mid-anaphase of mitosis [13]. To examine the localization of Eic1 and Eic2, the endogenous genes expressing Eic1 and Eic2 were fused to GFP. Quantitative chromatin immunoprecipitation (qChIP) analyses demonstrated that, as with Mis18 and other kinetochore proteins, both Eic1-GFP and Eic2-GFP are enriched over the central kinetochore domain (central core plus imr repeats) of fission yeast centromeres (figure 2a). Genome-wide ChIP-seq analyses confirmed this finding and also demonstrated that Eic1-GFP and Eic2-GFP, much like Mis18-GFP and Scm3-GFP, are undetectable at other regions of the genome including the heterochromatic outer repeats of centromeres (figure 2b and electronic supplementary material, figure S2). Immunostaining revealed that both Eic1 and Eic2 GFP-tagged proteins colocalize with Cnp1 CENP-A at clustered centromeres in interphase cells (figure 2c(i),(ii)). However, as with Mis16 RbAp46/48/Hat2 and Mis18 [11,13], Eic1-GFP and Eic2-GFP dissociate from centromeres in early mitosis (prometaphase), just as the centromeres (Cnp1) begin to form two separate clusters as they biorient on the spindle (figure 2c(iii),(iv)). Moreover, both Eic1-GFP and Eic2-GFP reassociate with centromeres in mid-anaphase when centromeres (Cnp1) and chromosomes (DAPI) have clearly segregated to opposite spindle poles (figure 2c(viii),(ix)). Apart from Mis16 RbAp46/48/Hat2 and Mis18, the CENP-A chaperone Scm3 HJURP exhibits similar temporal dissociation-reassociation at centromeres during mitosis [9,12]. Thus, our analyses identify Eic1 and Eic2 as two additional proteins that are released from centromeres in early mitosis and reloaded on centromeres in mid-anaphase. The finding that Eic1 and Eic2 exhibit similar association dynamics as Mis18 through the cell cycle is consistent with our identification of Eic1 and Eic2 being tightly associated with Mis18.

Eic1 is required to maintain Cnp1 CENP-A at centromeres
Mis18 and associated proteins have been shown to be required to maintain CENP-A at centromeres in fission yeast and vertebrates [11,13,37]. If Eic1 and Eic2 are critical for Mis18 function then they should also be required to maintain normal Cnp1 CENP-A levels at centromeres. As the eic1 þ gene is essential for cell viability, we generated a conditional temperature-sensitive (ts) mutant of eic1 (eic1-1:hygR, hereafter referred to as eic1-1) in which a single amino acid substitution (Phe102Ser) at a conserved residue rendered cells inviable at 368C but with retained viability at 258C (figure 3a). We also generated a corresponding wild-type allele of eic1 (eic1 þ :hyg R ) as a control in which, as with eic1-1, a hygromycin resistance marker was inserted within the 3 0 UTR of eic1 at its endogenous locus for ease of genetic manipulation ( figure 3a). qChIP analyses demonstrated that Cnp1 CENP-A levels were significantly diminished at centromeres in eic1-1 cells at restrictive temperature (figure 3b; electronic supplementary material, figure S3a). Cells harbouring a second ts allele of eic1 (eic1-2:hyg R ; two substituted residues, Lys37Glu and Tyr55His, hereafter referred to as eic1-2) also demonstrated significant loss of Cnp1 CENP-A from centromeres at 368C (figure 3a,b; electronic supplementary material, figure S3a). Furthermore, eic1-1 and eic1-2 cells displayed sensitivity to the microtubule depolymerizing drug thiabendazole (TBZ), whereas eic1 þ :hyg R and the mis18-262 ts mutant did not confer TBZ sensitivity (figure 3c). TBZ sensitivity suggests that kinetochore-microtubule interactions are defective in eic1 mutants. Aberrant kinetochore function in eic1-1 cells was supported by cytological analyses, which revealed severe chromosome segregation defects (electronic supplementary material, figure S3b). qChIP analyses demonstrated that GFP-tagged Eic1-1 and Eic1-2 mutant proteins remained associated with centromeres even at 368C; thus, the observed phenotypes of eic1-1 and eic1-2 cells are not a consequence of the complete absence of Eic1 protein at centromeres (electronic supplementary material, figure S3c). In contrast to eic1 mutants, cells lacking the eic2 þ gene (eic2D) displayed no loss of Cnp1 CENP-A from centromeres (figure 3d). One possibility is that Eic2 is not required to maintain Cnp1 CENP-A at centromeres but is required to establish Cnp1 CENP-A on naive centromeric DNA. To test this, a plasmid bearing centromeric DNA that efficiently establishes Cnp1 CENP-A chromatin and functional kinetochores in wildtype cells (pH-cc2) [38] was transformed into eic2D cells. No defect in the establishment of Cnp1 CENP-A chromatin or functional kinetochores was observed (electronic supplementary material, figure S4). Furthermore, eic2D cells were not sensitive to TBZ (figure 3c). We conclude that Eic1 is required for Cnp1 CENP-A maintenance and kinetochore integrity, whereas Eic2 is dispensable. chaperone HJURP Scm3 to centromeres, and consequently the incorporation of CENP-A at centromeres [13,25]. We find that the levels of GFP-tagged Mis16 RbAp46/48/Hat2 and Mis18 associated with fission yeast centromeres are greatly reduced in the eic1-1 mutant even at 258C (figure 4a,b), consistent with the observed synthetic genetic interaction between eic1-1 and the GFP-tagged alleles of mis16 and mis18 (electronic supplementary material, figure S5). Additionally, we find that Scm3 HJURP association with centromeres is also dependent on Eic1 function (figure 4c). Interestingly, we find Mis6 CENP-I/Ctf3 association with centromeres to be only partially dependent on Eic1 (figure 4d), as intermediate levels of Mis6 CENP-I/Ctf3 are retained at centromeres even when Eic1 function is compromised. Thus Eic1, in conjunction with its interacting partners Mis16 RbAp46/48/Hat2 and Mis18, is essential to maintain normal levels of Cnp1 CENP-A as well as Cnp1 CENP-A loading factors on centromeres. Surprisingly, eic2D cells exhibited reduced levels of Mis18, Mis16 RbAp46/48/Hat2 and Mis6 CENP-I/Ctf3 at centromeres, whereas Scm3 HJURP remained unaffected (figure 4e-h). As normal levels of Cnp1 CENP-A are retained at centromeres and no defect in chromosome segregation was detectable in eic2D cells

The recruitment of Cnp1 CENP-
Scm3 0 Epistasis analyses (double mutant combinations) frequently reveal positive and negative genetic interactions between mutants and consequently inform on the functional niche of specific proteins. The eic1-1 mutation exhibited significantly reduced growth when combined with ts mutations in mis18, scm3, cnp1 and mis6 (figure 5a and electronic supplementary material, figure S6a; summarized in figure 5c). We failed to generate eic1-1 mis16-53 and eic1-1 scm3-139 double mutant strains, suggesting that eic1-1 has a synthetic lethal interaction with both mis16-53 and scm3-139. Surprisingly, such synthetic interactions could also be observed for the eic1 þ :hyg R allele, which must compromise eic1 function in sensitized genetic backgrounds and therefore be considered a hypomorphic allele of eic1 (figure 5 and electronic supplementary material, figure S6a). These results, in conjunction with our ChIP analyses of Cnp1 CENP-A assembly factors (figure 4a-d), clearly demonstrate that Eic1 makes important contributions to the proper localization of Cnp1 CENP-A assembly factors and Cnp1 CENP-A itself.
eic2D showed no combinatorial effects when combined with cnp1, mis6 or mis16 ts mutants (electronic supplementary material, figure S6b; summarized in figure 5c). However, eic2D displayed a negative interaction with mis18-262, notably reducing growth at 368C relative to mis18-262 alone (figure 5b). By contrast, eic2D partially suppressed the temperature sensitivity of the scm3-139 (Leu73Phe) but not the scm3-15 (Ser281Leu) mutant (figure 5b; summarized in figure 5c). The negative interaction of Eic2 with Mis18 confirms that Eic2 contributes to the function of the Mis18 complex. As the protein levels as well as association of Scm3 HJURP with centromeres remain unaffected in eic2D cells (figure 4c; electronic supplementary material, figure S6c), the mechanism of partial suppression of temperature sensitivity observed in eic2D scm3-139 cells remains unclear. eic2D showed no combinatorial effects when combined with eic1-1 (figure 5b; summarized in figure 5c), suggesting that Eic2 may not influence Eic1 function even though Eic1 and Eic2 physically associate.     in strains expressing various ts kinetochore proteins. Eic1 association with centromeres was found to be entirely dependent on functional Mis18, as indicated by undetectable levels of Eic1 at centromeres in mis18-262 cells grown at 368C (figure 6a and electronic supplementary material, figure S7a). Lower Eic1 levels were also detected on centromeric central cores in mis16-53 and scm3-139 mutants at restrictive temperature; however, they were largely unchanged in cells expressing mutant Cnp1 CENP-A , Mis6 CENP-I/Ctf3 and Mis12 proteins (figure 6a and electronic supplementary material, figure S7a). Centromeric Eic1 levels also remained unaffected in eic2D cells ( figure 6b). Thus, the localization of Eic1 at kinetochores is mainly dependent on its partner proteins Mis16 RbAp46/48/Hat2 and Mis18, but not Eic2.

Eic1 and
The level of Eic2 at centromeres was found to be greatly dependent on functional Mis16 RbAp46/48/Hat2 , Mis18 and Eic1, and partially dependent on functional Scm3 HJURP , Cnp1 CENP-A and Mis6 CENP-I/Ctf3 (figure 6c,d and electronic supplementary material, figure S7b). Unexpectedly, Eic2 levels at centromeres were found to be reduced in mis12-537 (a mutation in an outer kinetochore component) cells at restrictive temperature (figure 6c and electronic supplementary material, figure S7b). These analyses suggest that Eic2 is also recruited to centromeres by other mechanisms that are independent of its association with Mis18.

Eic1 associates with CCAN/Mis6/Ctf19 complex components that influence its recruitment to centromeres
To further investigate the functional niche of Eic1 and Eic2, immunoprecipitates of both GFP-tagged proteins were subjected to LC-MS/MS analyses to identify associated factors. Only Eic1 and Mis18 were detected in Eic2-GFP immunoprecipitates ( figure 7a)   , as these proteins are detectable to a significant degree at centromeres in eic1-1 cells even at non-permissive temperature. This is consistent with the finding that centromeric association of the CCAN component Mis6 CENP-I/Ctf3 also only partly depends on Eic1 function ( figure 4d).
Interestingly, Eic1-GFP association with centromeres was found to greatly depend on functional Mal2 CENP-O/Mcm21 (figure 8d), although it appeared to be largely independent of functional Mis6 CENP-I/Ctf3 (figure 6a and electronic supplementary material, figure S7a). Additionally, the mal2-1 ts mutant exhibited a more severe growth defect than mis6-302 when combined with eic1-1 (figures 5a and 8e). Together, these results suggest that the CCAN components Fta7 CENP-Q/Okp1 , Cnl2 Nkp2 and Mal2 CENP-O/Mcm21 may act in concert to recruit Eic1 to centromeres via their physical association with Eic1.

Discussion
In this study, we have identified and characterized two novel Mis18-interacting proteins, Eic1 and Eic2, in fission yeast. Eic1 is a small protein of approximately 13.4 kDa that is well conserved within Schizosaccharomyces species (figure 1b). The absence of an Eic2 orthologue in S. japonicus suggests that the primary sequence of Eic2 may have rapidly diverged over evolutionary time so that it is now undetectable within the S. japonicus genome; alternatively, other proteins may undertake the function of Eic2 in S. japonicus. The lack of any specific domains within Eic1 and Eic2 hinders the detection of similar or related proteins in other organisms, and thus orthologues of Eic1 and Eic2 remain unidentified outside of the Schizosaccharomyces clade.
The human Mis18 complex (consisting of hMis18a/b and Mis18BP1 KNL2 ) primes CENP-A assembly by associating with centromeres specifically in telophase, subsequently recruiting the CENP-A-specific chaperone HJURP Scm3 to mediate CENP-A deposition in G1 [13,25,26,28,29]. The phosphorylation of Mis18BP1 KNL2 by CDK1/2 has been shown to regulate the recruitment of Mis18BP1 KNL2 and timing of CENP-A assembly during the cell cycle [39]. In fission yeast, the known Cnp1 CENP-A loading factors, Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP , also exhibit a very specific cell-cycle-regulated localization to centromeres in that they dissociate before metaphase and re-associate in mid-anaphase [9,[11][12][13]. Our analyses demonstrate that both Eic1 and Eic2 are centromere-specific proteins with cell-cycle dynamics that are very similar to that of Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP ( figure 2 and electronic supplementary  material, figure S2). The re-association of these five factors in  anaphase-telophase is presumably required to allow the subsequent replication-independent deposition of Cnp1 CENP-A in G2 phase of the cell cycle [40,41]. Recent analysis of Arabidopsis Mis18BP1 KNL2 suggests that it conforms to the S. pombe, rather than the vertebrate, pattern of CENP-A loading factor recruitment to centromeres during the cell cycle [33]. Our analyses show that functional Eic1 is required for the association of its interacting partners Mis16 RbAp46/48/Hat2 and Mis18, as well as Scm3 HJURP with centromeres ( figure 4a -c). The eic1 þ :hyg R and eic1-1 alleles also display strong negative interactions when combined with mutant Cnp1 CENP-A or mutant Cnp1 CENP-A assembly factors, Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP (figure 5a; summarized in figure 5c). Such interactions are consistent with Eic1 being required to promote Cnp1 CENP-A assembly by mediating the association of the Cnp1 CENP-A chaperone Scm3 HJURP as well as Mis16 RbAp46/48/Hat2 and Mis18 with centromeres. Eic1 is required to recruit Mis18 to centromeres, and functional Mis18 has been shown to affect the localization of the CCAN/Mis6/Ctf19 complex component Mis6 CENP-I/Ctf3 to centromeres [11]. It was thus surprising that our analyses only detected a partial reduction in Mis6 CENP-I/Ctf3 levels at centromeres in eic1 mutant cells (figure 4d). Nevertheless, consistent with this finding, our analyses show that the association of other CCAN components with centromeres is only partly dependent on Eic1 function (Fta7 CENP-Q/Okp1 , Cnl2 Nkp2 and Mal2 CENP-O/Mcm21 ; figure 8a-c). The CCAN proteins Mis6 CENP-I/Ctf3 , Sim4 CENP-K/Mcm22 , Mis15 CENP-N/Chl4 and Mis17 CENP-U/Ame1 have been shown to facilitate Cnp1 CENP-A assembly [8,10,11]. Thus, although functional Eic1 is essential for the recruitment of the key Cnp1 CENP-A assembly factors Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP to centromeres, it is partly dispensable for maintaining constitutive CCAN components at centromeres.
Eic2, the second Mis18-interacting protein that we identified, contributes to Mis18 function (figure 5b) as it promotes normal levels of Mis18, Mis16 RbAp46/48/Hat2 and Mis6 CENP-I/Ctf3 association with centromeres ( figure 4e,f,h). However, Eic2 is   rsob.royalsocietypublishing.org Open Biol. 4: 140043 dispensable for maintaining Cnp1 CENP-A and Scm3 HJURP association with centromeres (figures 3d and 4g). It was therefore unexpected that loss of Eic2 (eic2D) resulted in the partial rescue of scm3-139 temperature sensitivity (figure 5b; summarized in figure 5c). Given that the levels of Scm3 HJURP protein and its association with centromeres remain unaffected in eic2D cells (figure 4g; electronic supplementary material, figure S6c), we speculate that Scm3 HJURP -Mis18 interactions are perhaps stabilized in the absence of Eic2. Regardless of these changes, the levels of loading factors remaining at centromeres in eic2D cells are sufficient to maintain normal Cnp1 CENP-A levels. The fact that the localization and function of Mis18, but not Eic1 or Cnp1 CENP-A (figures 3d, 4e, 5b and 6b), is compromised in eic2D cells suggests that Eic1 and Eic2 function independently. Eic2 may act to bolster Mis18 function under particular conditions of stress. Eic1 association with centromeres is entirely dependent on functional Mis18, but only partly dependent on functional Mis16 RbAp46/48/Hat2 and Scm3 HJURP . Moreover, Eic1 levels at centromeres remain largely unaffected when Cnp1 CENP-A , Mis6 CENP-I/Ctf3 or Mis12 function is compromised (figure 6a and electronic supplementary material, figure S7a). These and other observations (figure 4a-d) suggest that Eic1 acts in concert with Mis16 RbAp46/48/Hat2 and Mis18 to promote recruitment of the Cnp1 CENP-A -specific chaperone Scm3 HJURP to centromeres and thereby facilitates Cnp1 CENP-A assembly. Thus, although no obvious protein similarity is evident between Eic1 and Mis18BP1 KNL2 , we propose that Eic1 serves as the functional counterpart of Mis18BP1 KNL2 [13,33,37].
The maintenance of epigenetically specified centromeres on chromosomes requires a feedback mechanism where, once established, constitutive centromere proteins themselves recruit the CENP-A assembly factors and thereby ensure their preservation. In vertebrates, the CENP-A assembly factors HJURP Scm3 /Mis18/Mis18BP1 KNL2 are recruited by the interaction of Mis18BP1 KNL2 with CENP-C Cnp3/Mif2 [27,28]. Our analyses suggest that Eic1 performs the equivalent function to Mis18BP1 KNL2 by association with the Fta7 CENP-Q/Okp1 , Cnl2 Nkp2 and Mal2 CENP-O/Mcm21 subunits of the constitutive CCAN/Mis6/Ctf19 complex (figure 7). In support of this view, we find that Eic1 recruitment to centromeres is particularly dependent on functional Mal2 CENP-O/Mcm21 , and eic1 mal2 double mutants are severely compromised (figure 8d,e). By contrast, Eic1 association with centromeres is only partly dependent on the CCAN component Mis6 CENP-I/Ctf3 (figure 6a and electronic supplementary material, figure S7a). Mcm21 Mal2/CENP-O and Okp1 Fta7/CENP-Q along with Ctf19 Fta2/CENP-P and Ame1 Mis17/CENP-U are known to form COMA, a biochemically distinct subcomplex within the larger Ctf19 complex in S. cerevisiae [19,42], much like the stable CENP-O/P/Q/U subcomplex described in vertebrate cells [43]. We propose that Mal2 CENP-O/Mcm21 , Fta7 CENP-Q/Okp1 and Cnl2 Nkp2 form an analogous module within the CCAN/Mis6/Ctf19 complex that ensures the propagation of Cnp1 CENP-A chromatin and kinetochores by recruiting Eic1 and consequently the Cnp1 CENP-A assembly factors Mis16 RbAp46/48/Hat2 , Mis18 and Scm3 HJURP to fission yeast centromeres.
Undoubtedly, the analyses of centromere-kinetochore architecture in distinct model systems have provided insights into the function of particular components. The lack of specific components in one organism reveals its reliance on alternative pathways. For example, Drosophila lack the kinetochore protein CENP-T Cnp20/Cnn1 , and consequently their kinetochores are completely reliant on the KMN (Knl1-Mis12-Ndc80) pathway for attachment to microtubules [4,18]. In vertebrates, Mis18BP1 KNL2 directly associates with CENP-C Cnp3/Mif2 , allowing the maintenance and propagation of centromeric chromatin through the recruitment of Mis18, HJURP Scm3 and thus CENP-A [27,28]. Remarkably, the CENP-A CID assembly pathway appears to be completely rewired during evolution of Drosophila because the main CENP-A loading factors HJURP Scm3 , Mis18 and Mis18BP1 KNL2 have been lost and replaced by the Cal1 protein, which directly associates with CENP-C Cnp3/Mif2 [31,32]. In S. cerevisiae, where centromere specification is driven by the recognition of specific DNA elements by DNA-binding proteins [15], a self-propagation mechanism is no longer necessary and consequently proteins equivalent to Mis18/ Mis18BP1 KNL2 , Cal1 or Eic1 are absent. Interestingly, fission yeast Cnp3 CENP-C/Mif2 is not essential indicating that it is not required for the propagation of Cnp1 CENP-A chromatin. Eic1 appears to perform an equivalent role to Mis18BP1 KNL2 but associates with the essential constitutive CCAN components Mal2 CENP-O/Mcm21 , Fta7 CENP-Q/Okp1 and Cnl2 Nkp2 , rather than Cnp3 CENP-C/Mif2 . Thus, it is possible that the highly conserved CCAN components CENP-O Mal2/Mcm21 and CENP-Q Fta7/Okp1 also function in other organisms to recruit CENP-A assembly factors; they may even have an equivalent underlying function that is redundant with CENP-C Cnp3/Mif2 at vertebrate centromeres.
In conclusion, our analysis has identified Eic1 as a factor that connects Mis16 RbAp46/48/Hat2 and Mis18, which are required for Cnp1 CENP-A incorporation, with constitutive kinetochore components within the CCAN/Mis6/Ctf19 complex. We propose that Eic1 serves two major functions: (i) priming Cnp1 CENP-A assembly in concert with Mis16 RbAp46/48/Hat2 and Mis18 and (ii) promoting kinetochore integrity in conjunction with the CCAN/Mis6/Ctf19 complex. The interdependency relationships among Eic1, Mis16 RbAp46/48/Hat2 and Mis18, along with their similar dynamic localization to centromeres, indicate that Eic1, Mis16 RbAp46/48/Hat2 and Mis18 form a complex that is temporarily released from centromeres during mitosis and can associate with centromeres independently of Cnp1 CENP-A . The novel physical interaction via Eic1 that we have uncovered between constitutively bound CCAN components and Mis18 is likely to be fundamental for the temporal regulation of Eic1-Mis16 RbAp46/48/Hat2 -Mis18 association with centromeres, and thereby the cell-cycle-dependent assembly of Cnp1 CENP-A chromatin. The identification of Eic1 and Eic2 adds to the repertoire of known cell-cycle-regulated centromeric proteins involved in maintaining CENP-A at centromeres and provides a good example of how distinctly different proteins can contribute to a conserved cellular process in diverse organisms.

Yeast strains and standard techniques
Standard methods were used for fission yeast growth, genetics and manipulation [44]. Five-fold serial dilutions of the indicated strains were spotted onto YES media containing Phloxine B for growth assays, or DMSO or 12.5 mg ml 21 TBZ for TBZ sensitivity assays. Gene deletions, tagging and centromeric plasmid transformations were carried out by either the lithium acetate transformation method or electroporation. The eic1 þ :hyg R allele was generated by integrating a hygromycin resistance marker within the 3 0 UTR of eic1 þ rsob.royalsocietypublishing.org Open Biol. 4: 140043 at its endogenous locus. The eic1-1 and eic1-2 alleles were generated likewise, but derived by error-prone PCR using the GeneMorph II random mutagenesis kit (Agilent Technologies). The epitope-tagged alleles of Eic1, Eic2, Hat1, Fta7 and Cnl2 were generated at their respective endogenous loci by integrating an in-frame GFP, 3HA or 13myc cassette at their respective C termini [45].

Immunoaffinity purification and mass spectrometry
For Mis16 and Mis18 pulldowns, 5 g of pulverized S. pombe cells expressing Myc-tagged Mis16 or Mis18 were used for immunoprecipitation with anti-Myc antibody 9E10 (Covance) coupled to Protein G Dynabeads (Life Technologies), alongside an untagged control. For Eic1 and Eic2 pulldowns, 5 g of pulverized S. pombe cells expressing GFP-tagged Eic1 or Eic2 were used for immunoprecipitation with anti-GFP antibody A11122 (Life Technologies) coupled to Protein G Dynabeads, alongside an untagged control. After washes, Dynabeads with immunoprecipitated material were subjected to on-bead tryptic digestion, following which the samples were treated as described [9]. The average number of unique peptides corresponding to proteins that were reproducibly enriched in the epitope-tagged samples but consistently absent in the untagged controls, over three independent biological replicates, is presented.

Multiple sequence alignment
Orthologues of Eic1 and Eic2 among the Schizosaccharomyces species were identified by BLAST, PSI-BLAST or synteny searches against the Schizosaccharomyces group database available at the Broad Institute Schizosaccharomyces Comparative Genome Project [46]. Primary sequences of Eic1 or Eic2 orthologues were aligned using CLUSTALW and T-COFFEE.

Co-immunoprecipitations and western analyses
For co-immunoprecipitation experiments, 2 g of pulverized S. pombe cells expressing the indicated epitope-tagged proteins were used for immunoprecipitation using anti-GFP antibody A11122 (Life Technologies), anti-Myc antibody 9E10 (Covance), anti-Myc antibody 9B11 (Cell Signaling) or anti-HA antibody 12CA5 (Roche) coupled to Protein G Dynabeads (Life Technologies). The same antibodies were also used for western analyses as indicated. Anti-Bip1 was used as a loading control where indicated.

Recombinant protein co-expression and binding assays
Codon-optimized ORFs of Eic1, Mis16 and Mis18 were synthesized by GeneArt (Life Technologies), for expression in E. coli. These were then sub-cloned into pGEX-6P-1 (GST) (GE Healthcare, gift from J. Welburn) or pEC(K)-3CHis (His) (gift from A. A. Jeyaprakash) expression vectors as indicated, and co-transformed into the BL21-Gold-pLysS E. coli strain (Agilent Technologies). Co-transformants were grown in SuperBroth supplemented with carbenicillin and kanamycin as appropriate, and the indicated proteins co-expressed by induction with 0.3 mM IPTG. GST pulldowns were done using glutathione agarose (Sigma). His pulldowns were done using Ni-NTA agarose beads (Qiagen). Samples were resolved on NuPAGE Bis-Tris gels (Life Technologies) and stained using InstantBlue (Expedeon).

Quantitative chromatin immunoprecipitation
The indicated S. pombe strains were grown in YES media at 328C. If appropriate, cells were shifted to restrictive temperature (368C) for 8 h, or continued to grow for the same length of time at permissive temperature (258C) before fixation. For ChIPs on centromeric plasmids, cells harbouring pH-cc2 [38,47] were grown in PMG media minus adenine, minus uracil, at 328C. To confirm that plasmids were behaving episomally and had not integrated, a plasmid stability test was performed at the time of fixation. Cells (100-1000) were plated onto YES supplemented with 1/10th adenine and allowed to form colonies. Samples exhibiting no integrations were used for ChIP. ChIP was performed as described [9], or with the following modifications. Cells were fixed with 1% formaldehyde (Sigma) for 20 min at room temperature and lysed using a bead beater (Biospec Products). Cell lysates were sonicated in a Bioruptor (Diagenode) (15 min, 30 s On and 30 s Off at 'High' (200 W) position). For Cnp1 CENP-A ChIPs, anti-Cnp1 CENP-A antiserum was used with Protein G agarose beads (Roche). For all other ChIPs, Protein G Dynabeads (Life Technologies) were used along with anti-GFP A11122 (Life Technologies), anti-HA 12CA5 (Roche) or anti-Myc 9B11 (Cell Signaling), as appropriate. Immunoprecipitated DNA was recovered using the Chelex-100 resin (BioRad) [48]. ChIPs were analysed by real-time PCR using Lightcycler 480 SYBR Green (Roche) with primers specific to the central cores of centromere 2 (cc2) or centromeres 1/3 (cc1/3), the innermost repeats of centromere 1 (imr) or act1. All ChIP enrichments were calculated as % DNA immunoprecipitated at the locus of interest relative to the corresponding input samples, and normalized to % DNA immunoprecipitated at the act1 locus. Histograms represent data averaged over at least three biological replicates. Error bars represent standard deviations from at least three biological replicates.

ChIP-seq analysis
Schizosaccharomyces pombe strains expressing the indicated GFP-tagged proteins were grown in YES media at 328C to 1.25 Â 10 9 cells at a density of 1 Â 10 7 cells ml 21 . Cells were fixed for 15 min in 1% formaldehyde (Sigma) and lysed with 0.4 mg ml 21 Zymolyase 100 T (AMS Biotechnology Europe) in PEMS for 1 h at 368C. Cell lysates were sonicated in a Bioruptor (Diagenode) (20 min, 30 s On and 30 s Off at 'High' (200 W) position) and immunoprecipitated overnight using anti-GFP antibody A11122 (Life Technologies) and Protein G Dynabeads (Life Technologies). The samples were washed and cross-links reversed using 1% SDS for 4 h at 658C. Immunoprecipitated DNA was recovered using a Qiagen PCR purification kit, and centromeric enrichment was verified by qPCR using Lightcycler 480 SYBR Green (Roche). Illumina libraries were prepared following the TruSeq Nano DNA kit (Illumina) guidelines using NEXTflex (Bio Scientific) adapters with internal barcodes. Multiplexed libraries were 100 bp paired-end sequenced on an Illumina HiSeq2000 (Ark Genomics, Edinburgh, UK). ChIP-seq data were mapped onto the S. pombe genome assembly EF2 rsob.royalsocietypublishing.org Open Biol. 4: 140043 (Ensemble) using BOWTIE2. Coverage calculations and peak calling were done using MACS and PEAKSPLITTER.

Centromeric plasmid selection system and stability
pH-cc2 (H denotes an otr heterochromatic element and cc denotes central domain DNA) carries ura4 þ and sup3-5 (suppressor of ade6-704) selection systems [38,47]. Cells without ura4 þ cannot grow on minus-uracil plates, while ade6-704 cells do not grow without adenine and form red colonies on 1/10th adenine plates. The sup3-5-tRNA gene suppresses a premature stop in ade6-704, allowing growth on minus-adenine plates. Cells containing pH-cc2 form a high percentage of white or sectored colonies on 1/10th adenine indicator plates, demonstrating their relative mitotic stability in wild-type cells. In cells lacking Clr4, however, their mitotic stability is lost due to a lack of heterochromatin-dependent centromeric cohesion. To confirm that plasmids were behaving episomally and had not integrated, a plasmid stability test was performed at the time of fixation for ChIP. Cells (100-1000) were plated onto YES supplemented with 1/10th adenine and allowed to form colonies. Samples exhibiting no integrations were used for ChIP.

Note added in proof
A concurrent study has identified Eic1 and Eic2 as Mis19 and Mis20, respectively, in association with Mis18 [49].