Crystal structure of the Cenp-HIKHead-TW sub-module of the inner kinetochore CCAN complex

Abstract Kinetochores are large multi-subunit complexes that attach centromeric chromatin to microtubules of the mitotic spindle, enabling sister chromatid segregation in mitosis. The inner kinetochore constitutive centromere associated network (CCAN) complex assembles onto the centromere-specific Cenp-A nucleosome (Cenp-ANuc), thereby coupling the centromere to the microtubule-binding outer kinetochore. CCAN is a conserved 14–16 subunit complex composed of discrete modules. Here, we determined the crystal structure of the Saccharomyces cerevisiae Cenp-HIKHead-TW sub-module, revealing how Cenp-HIK and Cenp-TW interact at the conserved Cenp-HIKHead–Cenp-TW interface. A major interface is formed by the C-terminal anti-parallel α-helices of the histone fold extension (HFE) of the Cenp-T histone fold domain (HFD) combining with α-helix H3 of Cenp-K to create a compact three α-helical bundle. We fitted the Cenp-HIKHead-TW sub-module to the previously determined cryo-EM map of the S. cerevisiae CCAN–Cenp-ANuc complex. This showed that the HEAT repeat domain of Cenp-IHead and C-terminal HFD of Cenp-T of the Cenp-HIKHead-TW sub-module interact with the nucleosome DNA gyre at a site close to the Cenp-ANuc dyad axis. Our structure provides a framework for understanding how Cenp-T links centromeric Cenp-ANuc to the outer kinetochore through its HFD and N-terminal Ndc80-binding motif, respectively.


INTRODUCTION
Accurate chromosome segregation in mitosis and meiosis underlies the successful inheritance of genetic information by future generations. In eukaryotes, duplicated sister chromatids pairs are aligned at the metaphase plate of the mitotic spindle until separated at the onset of anaphase. Chromosomes are physically connected to the mitotic spindle by kinetochores, large protein complexes that both assemble onto centromeric chromatin and attach to and track the plus end of microtubules (1,2). The physical movement of chromosomes is powered by microtubule depolymerization, pulling chromosomes to centrosomes at opposite poles of the cell. As well as acting as load-bearing elements, kinetochores control chromosome segregation fidelity by monitoring tension and microtubule attachment to ensure biorientation, integrating this information to activate error correction mechanisms and the spindle assembly checkpoint (3,4).
Kinetochores, comprising over 50 different proteins, are delineated into the inner and outer kinetochore. The inner kinetochore specifically assembles onto the centromerespecific Cenp-A nucleosome (Cenp-A Nuc ) through the CCAN complex. CCAN then links the centromere to the KMN network that forms the outer kinetochore. The Ndc80 complex (Ndc80c) of the KMN network, together with the DASH/Dam1 complex of Saccharomyces cerevisiae and the Ska complex of vertebrates, attach to microtubules (5). The error-correction, tension sensing chromosome passenger complex (CPC) is located at the inner kinetochore by interacting with both CCAN and a proximal H3nucleosome, whereas proteins responsible for the spindle assembly checkpoint (SAC) assemble at the outer kinetochore through the KNL1 complex of the KMN network.

Cloning, expression, purification and crystallization of Cenp-HIK Head -TW
Coding fragments of CTF3  (CENP-I N ), MCM16 137-182 (CENP-H C ), MCM22 130-239 (CENP-K C ), and full-length CNN1 (CENP-T) and WIP1 (CENP-W) previously amplified by PCR from S. cerevisiae genomic DNA (37) were cloned into the pU1 baculovirus expression vector (39). The gene expression cassettes for CTF3   ,   MCM16 137-182 , MCM22 130-239 , CNN1 and WIP1 were  further cloned into pF2 for generating a virus to express  the Cenp-HIK Head -TW complex (Head: denotes truncated  Cenp-H, Cenp-I and Cenp-K proteins representing the  Cenp-HIK Head domain as defined by (37)) using a modified MultiBac expression system (39). A double StrepII tag together with a TEV cleavage site were attached to the Cterminus of the Ctf3 1-245 protein. The same coding regions for the Cenp-HIK Head -TW complex were cloned into the pET28 plasmid for selenomethionine labelling using an E. coli expression system. For Cenp-TW expression, full length MCM16, MCM22, CNN1 and WIP1 were cloned into the MultiBac expression system, with a double StrepII tag attached to the C-terminus of Cenp-T.
All complexes except selenomethionine labeled Cenp-HIK Head -TW were expressed using the baculovirus-insect cell system as described (39). The cell pellet was lysed in a buffer of 50 mM Tris-HCl (pH 8.0), 200 mM NaCl, 1 mM DTT, 1 mM EDTA and loaded onto a Strep-Tactin column, and eluted in 5 mM desthiobiotin and the NaCl diluted to 100 mM for loading onto a Resource Q anion exchange column. The protein was then purified by size exclusion chromatography on a Superdex 200 size exclusion chromatography in buffer of 20 mM HEPES (pH 8.0), 150 mM NaCl, 1 mM DTT, 1 mM EDTA. Selenomethionine labeled Cenp-HIK Head -TW complex was produced in Escherichia coli using SeMet Medium Base Plus Nutrient Mix and Seleno Methionine Solution (Molecular Dimensions Ltd.). In brief, the cells were grown at 37 • C, shaken at 220 rpm to an OD 600 of 0.6. Protein expression was induced by addition of 0.3 mM IPTG, and the culture was incubated at 20 • C, 220 rpm for 16 h. The complex purification was performed as for the baculovirus/insect cell expression of native protein, except that 10 mM DTT and 5 mM EDTA were used in the buffers. For crystallization, the protein was concentrated to 5 mg/ml in a buffer containing 20 mM HEPES (pH 8.0), 150 mM NaCl, 1 mM EDTA and 1 mM DTT. Initial crystals were obtained by vapour diffusion in sitting drops in condition H11 of LMB 21 screening plate (sparse matrix screen, MD1-98) (40), containing 1.6 M NaK 2 PO 4 . The crystals were optimized in hanging drops with 1 M NaH 2 PO 4 and 0.38 M K 2 HPO 4 . Selenomethionine labelled Cenp-HIK Head -TW crystals grew in similar conditions with 10 mM DTT. Crystals were incubated in a cryoprotection buffer comprising 1 M NaH 2 PO 4 and 0.38 M K 2 HPO 4 and 25% glycerol prior to freezing in liquid nitrogen.

Crystallographic data collection and reduction
The high-resolution native dataset was collected at 100 K on beamline I24 at Diamond Light Source (DLS), Didcot, UK at a wavelength within the lead L-III absorption edge of 0.9465Å using a PILATUS3 6M detector (DEC-TRIS) with a crystal-to-detector distance to allow diffrac-tion to 3Å resolution at the detector edge. The eight collections were auto-processed using the XDS pipeline (41) in Xia2 (42) and merged together into a single dataset with Aimless (43). For the Se-SAD experiment, single datasets from 31 randomly orientated crystals were recorded at 100 K on beamline I03 at a wavelength of 0.9793Å using an Eiger2 XE 16M (DECTRIS) with a crystal-to-detector distance to cover diffraction to 2.9Å resolution at the detector edge. The optimal cluster of isomorphous datasets was obtained using the program BLEND (44). Selected datasets were auto-processed using the DIALS pipeline (45) in Xia2 and merged together into a single dataset with Aimless (43). Native data used in structure determination and refinement were anisotropically corrected using the STARANISO server (Tickle, I.J., Flensburg, C., Keller, P., Paciorek, W., Sharff, A., Vonrhein, C., Bricogne, G. (2018). STARANISO (http://staraniso.globalphasing. org/cgi-bin/staraniso.cgi). Cambridge, United Kingdom: Global Phasing Ltd.), however, isotropically processed intensities labelled IMEAN iso (and SIGIMEAN iso) have also been deposited to the Protein Data Bank.

Structure determination, refinement and validation
The high-resolution dataset of the Cenp-HIK Head -TW complex referred to as 'native' in this study was collected at the lead L-III absorption edge from a crystal that had been soaked with 20 mM trimethyl lead acetate (TMLA) for 72 h prior to cryo-freezing. The resulting data set extended anisotropically to 2.9Å (Table 1). Pb-SAD phasing attempts failed despite detection of weak anomalous signal during data reduction. Molecular replacement using the Cenp-HIK Head heterotrimer (PDB: 5Z08) (37) and Cenp-TW heterodimer (PDB: 3B0C) (15) as search models, also failed to give correct solutions.
Se-SAD experiments were therefore pursued by merging together only the first 200 • wedge from a cluster of eleven isomorphous datasets out of 31 collections from randomly orientated SeMet crystals to obtain a dataset that extended to 3.8Å (Table 1), and that showed significant anomalous signal up to 6Å. An overall signal-to-noise ratio of 13 was deemed sufficient to phase a pentameric complex of 860 residues containing 10 selenomethionines resulting in a Bijvoet ratio of 4.2%. The selenium substructure was determined using the HKL2MAP graphical interface (46) with SHELXC, SHELXD and SHELXE (47). Multiple searches for the correct selenium substructure were performed over the 3.2-8Å resolution range, estimating one heteropentamer in the asymmetric unit using the program MATTHEWS COEF (48). A solution was obtained at 6.5 A showing at least five heavy atom sites. Successful phasing was achieved with phenix.autosol (49) generating maps at 6.5Å with clear ␣-helical features. The search models 5Z08 and 3B0C for heterotrimeric Cenp-HIK, and dimeric Cenp-TW, respectively, succeeded in a MOLREP (50) run using the phased translation function. This newly assembled heteropentamer model was successfully utilized in a MOLREP run against the high-resolution native structure factors, providing a solution that could be refined initially with REFMAC5 (51) and at later stages with phenix.refine (49). Manual building was carried out with Coot (52).
Crystallographic statistics are listed in Table 1. The difference anomalous maps calculated for the dataset from the TMLA-soaked crystal using the phases of the refined structure failed to localize any lead site confirming that these data are indeed native. An example of a 2Fo-Fc density map is shown in Supplementary Figure S1A, B. MolProbity (53) was used for model validation.
Structural conservation was analysed using Consurf (54,55). The Cenp-HIK Head -TW crystal structure was fit to the cryo-EM density maps of CCAN -Cenp-A Nuc and apo dimeric CCAN (37) using Chimera (56). This was possible because a 3D cryo-EM class of the CCAN-Cenp-A complex (EMD-11626) revealed EM density for the Cenp-HIK Head -TW sub-module. The remaining subunits of CCAN are as reported previously (37), except that we used the recent cryo-EM model of S. cerevisiae Ctf3c-Cnn1-Wip1 (Cenp-HIK-TW) (PDB 6WUC) (57) to fit the 'flexible 'joint' region that connects Cenp-HIK Body to CenpHIK Head : residues 139-142 of Cenp-H, residues 285-353 of Cenp-I, and residues 141-148 of Cenp-K as a rigid body fit in Coot (52), and additionally, the ␣-helix (residues 242-268) to Cenp-I Head . The fit of Cenp-HIK Head -TW into the CCAN -Cenp-A Nuc cryo-EM map (EMD-11626) was assessed using the Chimera-derived correlation coefficient between the calculated EM density of the fitted models and the cryo-EM map. The correlation coefficient for the fit of Cenp-HIK Head -TW to the assigned density is 0.926, slightly lower than that for the fit of Cenp-HIK Body (0.953). The higher correlation coefficient for Cenp-HIK Head alone Nucleic Acids Research, 2020, Vol. 48, No. 19 11175 (0.955), likely results from the weaker density for Cenp-TW compared with Cenp-HIK Head . The correlation coefficient for randomly placing Cenp-HIK Head -TW into Cenp-HIK Body density was 0.880, clearly distinct from the fits into the assigned HIK Head -TW density. Figures were generated using PyMOL (Molecular Graphics Systems, 2.03, Schrodinger) and Consurf (54,55).

Cenp-HIK-Cenp-TW interaction analysis
To test the Cenp-HIK-Cenp-TW interaction and consequence of disrupting this interface on their mutual interaction, wild type and mutant forms of Cenp-HIK and Cenp-TW were mixed and incubated at equal molar ratios of 10 M at 4 • C for 1 h in a buffer of 10 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 1 mM DTT, and then resolved on a Micro-S200 size exclusion chromatography column.

Cenp-HIK Head -TW architecture
The structure of the S. cerevisiae Cenp-HIK Head submodule at 2.9Å resolution is essentially identical to its counterpart in the thermophilic yeast (38) (Figure 1A and Supplementary Figure S2A). Cenp-HIK Head is assembled from the N-terminus of Cenp-I Head (residues 1-241) and the C-termini of Cenp-H (residues 147-181, Cenp-H Head ) and Cenp-K (residues 136-237, Cenp-K Head ). Cenp-I Head is composed of five HEAT repeat-like motifs that generate a double layer of ␣-helices, capped at its C-terminus by a single ␣-helix, which then connects to a linker region running along the Cenp-I Head HEAT-domain to form a 13th ␣-helix located at the N-terminus of the domain ( Figure 1A). The single ␣-helix of Cenp-H Head inserts between Cenp-I Head and Cenp-K Head . Cenp-K Head comprises three ␣-helices followed by a short two-stranded ␤-sheet. Three loops that connect the pair of ␣-helices of three contiguous HEAT repeats (L3, L4 and L5), contribute to the interface of Cenp-HIK Head with Cenp-TW ( Figure 1A).
S. cerevisiae Cenp-TW resembles that of chicken Cenp-TW (15) ( Figure 1A and Supplementary Figure S2B), which are in turn related structurally to the canonical histone dimers of H2A-H2B and H3-H4. Differing from chicken Cenp-W, a disordered 20-residue loop connects ␣-helices H1 and H2 in S. cerevisiae Cenp-W. Cenp-W comprises a canonical histone fold domain (HFD) of three ␣-helices, whereas Cenp-T includes a HFD followed by two solvent exposed ␣-helices (H4 and H5) termed the histone fold extension (HFE) (13)(14)(15). Although we crystallized a complex of Cenp-HIK Head -Cenp-TW using the full-length Cenp-T subunit, in our structure the N-terminal 268 residues are disordered, consistent with disorder predictions, with only the HFD and HFE being visible.
The crystal structure of the Cenp-HIK Head -TW submodule reveals that the major contacts between Cenp-HIK Head and Cenp-TW involve Cenp-I Head and Cenp-K Head of Cenp-HIK and Cenp-T of Cenp-TW, with Cenp-H Head and Cenp-W making minor contributions ( Figure  1A). Consistent with our structure, it was previously shown that the interaction of Cenp-T with Cenp-HIK Head is not dependent on Cenp-W (14). Together, interactions between Cenp-TW and Cenp-HIK Head comprise two contiguous interfaces. In interface 1, the C-terminal anti-parallel ␣helices (H4 and H5) of the HFE of Cenp-T combine with H3 of Cenp-K to create a compact three helix bundle (Figure 1A). This interface is augmented by the Cenp-I Head HEAT repeat domain that contributes a long loop (L3, residues 89-104). This loop in turn is stabilized through contacts with Cenp-H and Cenp-K (Thr91, Val94 and Arg97 of Cenp-I Head ). Interface 1 is dominated by a network of electrostatic interactions involving Arg206 and Arg210 of Cenp-K and Asp338, Glu357 and Ser354 of Cenp-T ( Figure  1B). Phe341 of Cenp-T buttresses the Cenp-K Arg206 and Arg210 side-chains. Other electrostatic interactions include Glu342 of Cenp-T with Tyr205 of Cenp-K ( Figure 1B).
In interface 2, the conserved L5 loop (residues 177-182) of Cenp-I Head inserts into a cavity within the Cenp-TW dimer formed by the interface of H1 of the Cenp-T HFD, turn connecting H4-H5 of the Cenp-T HFE, and N-terminus of Cenp-W ( Figure 1A). His177 Cenp-I in the L5 loop, a highly conserved residue, makes an important contribution by providing a C-terminal helix cap to H4 of Cenp-T ( Figure 2C, D and Supplementary Figure S3). The nearby Arg173 of Cenp-I forms an electrostatic interaction with Glu346 of Cenp-T. Leu350 of the Cenp-T H4-H5 turn bridges interfaces 1 and 2, forming non-polar contacts with all three chains of Cenp-HIK, specifically Phe142 of Cenp-I ( Figure 1B). Additionally, the Cenp-I L4 loop contacts the H4-H5 turn of Cenp-T in which the highly conserved Glu351 accepts a hydrogen bond from Ser143 of Cenp-I ( Figure 1B).
The HFE of Cenp-T (Cenp-T HFE ) was previously implicated in mediating Cenp-TW interactions with Cenp-HIK from in vitro reconstitution studies (14). In this study, mutating the three residues of Cenp-T HFE (Glu346, Leu350, Glu351), that our structure shows mediates contacts at the Cenp-TW -Cenp-HIK Head interface ( Figure 1B), revealed that Cenp-TW failed to associate with Cenp-HIK, whereas contacts with Cenp-W were retained (14). To further assess our structure, we mutated His345, Leu350 and Ser354 (L350R, S354Y and H345R) of Cenp-T and Thr91 of Cenp-I (T91Y) (full-length proteins). As assessed by size exclusion chromatography (SEC), Cenp-HIK Mut and Cenp-TW Mut assembled correctly. Mutating Cenp-T alone, but not Cenp-I alone, abrogated the Cenp-HIK and Cenp-TW association ( Figure 3). We assume that in the Cenp-I T91Y mutant, conformational changes of the solvent exposed Tyr91 side chain compensate for its substitution for Thr91.

Structural conservation
Residues at the Cenp-HIK Head -Cenp-TW interface are highly conserved across eukaryotes, as shown by multiple sequence alignments of the regions of Cenp-I, Cenp-K and Cenp-T responsible for mediating the Cenp-HIK Head -Cenp-TW interface (Supplementary Figure S4). Sequence conservation is mapped onto the surfaces of Cenp-HIK Head and Cenp-TW in Figure 2. The electrostatic interactions at the interface involving Arg206 and Arg210 of Cenp-K, His177 of Cenp-I with Glu342, Glu351 and Glu357 of Cenp-T HFE , are well conserved from yeast to vertebrates. Notably Glu351 of Cenp-T HFE is an invariant residue, and the neighbouring Leu350, located at the hub of the Cenp-T interface with Cenp-I and Cenp-K, is also highly conserved, and as mentioned earlier, mutation of Leu350 and Glu351 ablates Cenp-HIK -Cenp-TW interactions (this work and (14)). This supports the view that the Cenp-HIK Head -Cenp-TW architecture will be evolutionarily conserved.

Fitting Cenp-HIK Head -TW to CCAN-Cenp-A cryo-EM structure
We fitted the Cenp-HIK Head -TW domain into the cryo-EM density of the CCAN-Cenp-A complex (37) (Figure 4). This was possible because a 3D class of the CCAN -Cenp-A complex (EMD-11626) revealed EM density for Cenp-HIK Head -TW. This showed a good fit of the Cenp-HIK Head -TW crystal structure with the assigned density connected to the body of Cenp-HIK (Figure 4). The Cenp-HIK Head -TW sub-module contacts the DNA gyre of the Cenp-A nucleosome close to the dyad axis and SHL3, opposite the DNAbinding channel of Cenp-LN that engages the unwrapped DNA duplex at one of the Cenp-A Nuc DNA termini (Figures 4A and 5A) (37). Plotting the electrostatic potential of CCAN with the fitted Cenp-HIK Head -TW sub-module ( Figure 5C) reveals that the regions of CCAN that contact the DNA gyre of Cenp-A Nuc are positively charged. These include the Cenp-LN-DNA binding channel (37) and a positively-charged surface generated by the combination of Cenp-I Head and Cenp-T that is directed toward the Cenp-A DNA gyre. However, the exact basic residues that create this positively-charged surface are not well conserved in Cenp-I and Cenp-T homologs ( Figure 5B). We were unable to isolate stable Cenp-HIK Head -TW complexes with Cenp-A Nuc , preventing us from testing the functional roles of the basic residues of Cenp-I Head and Cenp-T that contact the DNA gyre.
Cenp-A nucleosomes with a right-handed DNA configuration have been proposed (58), although in crystal and cryo-EM structures of yeast and vertebrate Cenp-A Nuc , the DNA is left-handed. We explored whether a Cenp-A nucleosome with a right-handed DNA chirality would be compatible with CCAN assembly. We assumed that a right-handed Cenp-A Nuc is an octasome with similar dimensions to the left-handed Cenp-A Nuc . Aligning the unwrapped DNA of the left-handed Cenp-A Nuc (defined by its interaction with the Cenp-LN DNA-binding groove) onto the equivalent segment of the modelled right-handed Cenp-A Nuc , indicates that the DNA gyre wrapping the histone octamer, located between the unwrapped DNA and Cenp-QU in the left-handed Cenp-A Nuc ( Figure 5A, left panel), would instead be positioned between the unwrapped DNA segment and Cenp-HIK for a right-handed Cenp-A Nuc . In this situation the DNA gyre would clash with Cenp-HIK Body , and less severely with Cenp-HIK Head (data not shown). However, it is possible that these steric clashes may be alleviated by rotating Cenp-A Nuc , and/or conformational changes of CCAN. Thus, this analysis suggests that the proposed right-handed Cenp-A Nuc might be less suited to bind CCAN.

Comparison of yeast Cenp-HIK Head -TW and chicken Cenp-TWSX
In vertebrates, Cenp-TW associates with the histone foldlike proteins Cenp-S and Cenp-X to form a Cenp-TWSX heterotetramer (15), reminiscent in architecture to the H3-H4 tetramer. Cenp-TWSX interacts with and supercoils DNA and has been proposed to form a nucleosome-like  particle to contribute to kinetochore attachment to chromatin (12,15). In budding yeast, ChiP-seq studies indicated that binding of Cenp-TW occurs at the core of the centromere (14), and that Cenp-TW does not form a separate nucleosome-like particle. This result is consistent with our cryo-EM structure of the CCAN-Cenp-A complex showing that a modelled Cenp-HIK Head -TW sub-module would contact the DNA gyre of the Cenp-A nucleosome (37). Budding yeast Cenp-TW closely resembles chicken Cenp-TW (15) (Supplementary Figure S2B). A notable difference however, anticipated in an earlier study (13), is that the central ␣-helix H2 of the Cenp-T HFD is three turns longer in chicken Cenp-T than in S. cerevisiae. In the chicken Cenp-TWSX complex, this region of Cenp-T interacts with Cenp-  S. We and others were unable to isolate a complex of Cenp-TW with the Cenp-SX homologs Mhf1 and Mhf2 using recombinant proteins (14), and Mhf1/Mhf2 did not associate with a kinetochore-Cenp-A nucleosome complex assembled de novo using yeast extracts (59). However, Mhf1 was previously reported to co-purify with kinetochore subunits in yeast extracts (14). Our structural study showing that the site on chicken Cenp-TW that binds to Cenp-SX is conserved in S. cerevisiae Cenp-TW ( Figure 2B), leaves open the possibility that the budding yeast Cenp-SX homologs interact with the yeast kinetochore in the context of centromeric chromatin. Furthermore, in the chicken Cenp-TWSX complex, Cenp-SX binds to a site on Cenp-TW that is opposite to the Cenp-HIK Head -binding site of budding yeast Cenp-TW, thus compatible with Cenp-HIK Head -Cenp-TW interactions ( Figure 2B).

Conformational flexibility of the Cenp-HIK Head -TW submodule
We analysed the conformational variability of the Cenp-HIK Head -TW sub-module by comparing the conformations of Cenp-HIK-TW in the context of three states (i) the CCAN -Cenp-A Nuc complex (37), (ii) the CCAN dimer (37), and (iii) the recently reported Cenp-HIK-TW complex (Ctf19c-Cnn1-Wip1) (57). Superimposing all three structures onto the structurally invariant Cenp-HIK Body shows that Cenp-HIK Head -TW sub-module adopts a range of conformations facilitated by the flexible joint that connects Cenp-HIK Body and Cenp-HIK Head -TW ( Figure 4C and Supplementary Figure S5). In the CCAN-Cenp-A Nuc complex, the conformation of Cenp-HIK Head -TW has swung out relative to that in the isolated Cenp-HIK-TW complex and apo dimeric CCAN, to create a straightened configuration for Cenp-HIK-TW. This conformational flexibility provides space for Cenp-A Nuc in the CCAN-Cenp-A Nuc complex.

Model for (CCAN) 2 -Cenp-A Nuc
Based on the dyad symmetry operator of Cenp-A Nuc , we generated a model for two CCAN protomers assembled onto a single Cenp-A Nuc ( Figure 6A). This model suggests how a Cenp-A nucleosome would be supported by the CCAN dimer with the unwrapped DNA ends of Cenp-A Nuc interacting with the DNA-binding surface of CCAN. As reported previously, 2D classification of the cryo-EM dataset identified 2D projections consistent with such a model, and the molecular mass measurements determined using AUC and SEC-MALS (37). In the model there is some overlap of Cenp-TW with Cenp-U, Cenp-Q and Nkp2 of the symmetry-related CCAN protomer ( Figure 6B). The conformational flexibility of Cenp-HIK Head -TW would be expected to allow for relief of this steric overlap.

CONCLUDING REMARKS
The crystal structure of Cenp-HIK Head -TW reported here provides the missing structural data to complete the CCAN atomic model, and defines the interface between Cenp-HIK and Cenp-TW. Fitting the Cenp-HIK Head -TW atomic model to the cryo-EM map of CCAN -Cenp-A Nuc indicates that the positively charged surface of both Cenp-I Head and Cenp-T interact with the DNA gyre of Cenp-A Nuc . This agrees with previous data that Cenp-TW binds DNA (12). Although we think it unlikely that budding yeast Cenp-TWSX forms a nucleosome-like particle at a distinct locus from the Cen locus, our study leaves open the possibility that the budding yeast Cenp-SX orthologs interact with Cenp-TW in the context of chromatin. The model of CCAN-Cenp-A Nuc with the fitted Cenp-HIK Head -TW provides a framework for understanding how Cenp-T links centromeric Cenp-A Nuc to the outer kinetochore through its N-terminal Ndc80 complex-binding motif.
While this manuscript was in preparation, the cryo-EM structure of the S. cerevisiae Ctf3c-Cnn1-Wip1 (Cenp-HIK-TW) complex was deposited (PDB: 6WUC) (57). In this structure, the Cenp-HIK-Cenp-TW interface, involving the HFE of Cenp-T is well defined, with conclusions similar to ours. However, the Cenp-T E346 -Cenp-I R173 salt-bridge interaction is not formed, and in addition, the HFDs of Cenp-T and Cenp-W, which were not well resolved in the cryo-EM maps, were modelled on the chicken Cenp-TW coordinates.

DATA AVAILABILITY
Protein coordinates and MTZ file have been deposited with RCSB, ID: 6YPC. The cryo-EM map of CCAN -Cenp-A Nuc used to fit CCAN -Cenp-A Nuc including the Cenp-HIK Head -TW sub-module was deposited with EMD, ID EMD-11626.