Molecular basis for Cdk1‐regulated timing of Mis18 complex assembly and CENP‐A deposition

Abstract The centromere, a chromosomal locus that acts as a microtubule attachment site, is epigenetically specified by the enrichment of CENP‐A nucleosomes. Centromere maintenance during the cell cycle requires HJURP‐mediated CENP‐A deposition, a process regulated by the Mis18 complex (Mis18α/Mis18β/Mis18BP1). Spatial and temporal regulation of Mis18 complex assembly is crucial for its centromere association and function. Here, we provide the molecular basis for the assembly and regulation of the Mis18 complex. We show that the N‐terminal region of Mis18BP1 spanning amino acid residues 20–130 directly interacts with Mis18α/β to form the Mis18 complex. Within Mis18α/β, the Mis18α MeDiY domain can directly interact with Mis18BP1. Mis18α/β forms a hetero‐hexamer with 4 Mis18α and 2 Mis18β. However, only two copies of Mis18BP1 interact with Mis18α/β to form a hetero‐octameric assembly, highlighting the role of Mis18 oligomerization in limiting the number of Mis18BP1 within the Mis18 complex. Furthermore, we demonstrate the involvement of consensus Cdk1 phosphorylation sites on Mis18 complex assembly and thus provide a rationale for cell cycle‐regulated timing of Mis18 assembly and CENP‐A deposition.


Introduction
Equal and identical distribution of chromosomes to each daughter cell during cell division is essential for maintaining genome integrity. A central player regulating this process is the kinetochore, a large proteinaceous structure assembled at a specialized region of the chromosome called the centromere [1][2][3]. Kinetochores physically couple chromosomes with spindle microtubules to facilitate chromosome segregation. Consequently, correct kinetochore assembly and function depends on centromeres being maintained at the right place on chromosomes [3].
In most eukaryotes, the centromeric chromatin is epigenetically defined by the enrichment of nucleosomes containing the histone H3 variant CENP-A [4][5][6][7]. To maintain centromere identity, new CENP-A must be deposited in each cell cycle. The timing of CENP-A deposition varies among species (G1-humans and G2-Schizosaccharomyces pombe); however, the molecular mechanisms by which it is achieved share considerable similarity [3,5,[8][9][10][11][12][13]. This process is initiated by the centromere targeting of the Mis18 complex (composed of Mis18a, Mis18b, and Mis18BP1) along with canonical histone chaperones, RbAp46/48 [14][15][16]. Centromere association of the Mis18 complex directly or indirectly makes the underlying chromatin permissive for CENP-A deposition. Mis18 proteins have also been shown to affect histone acetylation and DNA methylation at centromeres [14,17]. Mis18 centromere association subsequently allows HJURP, a CENP-A-specific chaperone, to associate with centromeres resulting in CENP-A loading [18,19]. Finally, chromatin remodelers like MgcRacGAP, RSF, Ect2, and Cdc42 have been suggested to fully stabilize the centromeric chromatin by a poorly understood maturation process [20,21]. Timely CENP-A deposition critical for centromere maintenance and function is determined by the kinase activities of Cdk1 and Plk1, which influence the centromere association of the Mis18 complex through negative regulation and positive regulation, respectively [22,23].
The Mis18 proteins contain two structurally distinct domains (Yippee/MeDiY and a C-terminal a-helix), both of which can self-oligomerize. Previously, we and others have shown that oligomerization of Mis18 proteins is required for their centromere association and function both in S. pombe and humans [24,25]. The human Mis18 paralogs, Mis18a and Mis18b, directly interact with Mis18BP1 and CENP-C, respectively, facilitating Mis18 complex formation and centromere association [14,26]. However, the molecular basis for the cell cycle-dependent regulation of Mis18 complex formation, and thus, CENP-A deposition has yet to be defined.
Here, we show that Mis18BP1, through its highly conserved Nterminal region comprising amino acids 20-130 (Mis18BP1 20-130 ), binds Mis18a/b hetero-oligomer via the Mis18a/b MeDiY heterodimer to form the Mis18 complex. Characterization of the oligomeric structures of the Mis18 subcomplexes revealed that the Mis18a/b complex is a hetero-hexamer with four copies of Mis18a and two copies of Mis18b, while the Mis18 holo-complex is a hetero-octamer with two additional copies of Mis18BP1. We identify two conserved consensus Cdk1 phosphorylation sites (T40 and S110) within Mis18BP1  and show that phospho-mimicking mutations of these residues disrupt Mis18 complex formation. This explains the molecular basis for Cdk1-mediated timing of Mis18 assembly and CENP-A deposition.

Results and Discussion
The N-terminal region of Mis18BP1 comprising amino acids 20-130 directly interacts with Mis18a/b to form the Mis18 complex Mis18BP1 contains two evolutionary conserved domains, the SANTA (residues 385-472) and SANT (residues 878-927) domains (Figs 1A and EV1A). It has previously been shown that a region overlapping the SANTA domain, Mis18BP1 475-878 , interacts with CENP-C while Mis18BP1 1-376 can interact with Mis18a [15,22,26]. The amino acid sequence analysis revealed the presence of a highly conserved 110 amino acid stretch within the first 130 amino acids of Mis18BP1 (residues 20-130; Mis18BP1 20-130 ; Fig 1A). We hypothesized that Mis18BP1 20-130 might be the minimal region that interacts with Mis18a/b. To test this, we expressed TetR-eYPF-Mis18a in HeLa 3-8 cells containing a synthetic alphoid DNA array with tetracycline operator sequences (alphoid tetO array) integrated in a chromosome arm [27] and analyzed the ability of Mis18a to recruit full-length Mis18BP1 (mCherry-Mis18BP1 fl ) as well as two different N-terminal fragments: mCherry-Mis18BP1 20-130 and mCherry-Mis18BP1 336-483 (covering the SANTA domain) to the tethering site. Mis18a tethered to the ectopic alphoid tetO array recruited Mis18BP1 fl and Mis18BP1 20-130 more robustly compared to the Mis18BP1 336-483 ( Fig 1B). As expected, Mis18a tethered to the synthetic array recruited Mis18b along with Mis18BP1 20-130 ( Fig EV1B). These data suggest that the Mis18BP1 20-130 is sufficient to interact with Mis18a/b to form the Mis18 complex in vivo.
To confirm that Mis18BP1 20-130 can directly interact with Mis18a/b in vitro, individually purified recombinant proteins (Mis18a/b and Mis18BP1 20-130 ) were analyzed using size-exclusion chromatography (SEC) before and after complex formation. While Mis18a/b and Mis18BP1 20-130 eluted at 11.0 and 15.7 ml, respectively, the complex containing Mis18a/b and Mis18BP1 20-130 eluted at 10.9 ml (Fig 1C). The elution volume of the Mis18a/Mis18b/ Mis18BP1 20-130 complex is not very different from Mis18a/b, suggesting that the binding of Mis18BP1 20-130 does not significantly alter the hydrodynamic radius of the Mis18a/b. These data, together with the in vivo tethering assays, confirm that Mis18BP1 20-130 is sufficient to make a stable and direct interaction with Mis18a/b to form the Mis18 complex.
The Mis18a MeDiY domain can directly interact with Mis18BP1 20-130 in vitro Mis18 proteins possess a globular MeDiY domain (Mis18a MeDiY : 77-187 and Mis18b MeDiY : 56-183) and a C-terminal a-helical domain (Fig 2A). Our previous structure-function analysis of S. pombe Mis18 revealed that its centromere association and function requires Mis18 MeDiY dimerization [24]. Stellfox et al [26] showed that mutations within the conserved C-X-X-C motif of the Mis18a MeDiY domain perturbed its ability to interact with Mis18BP1, suggesting a direct role of Mis18a for Mis18 complex formation. However, as the C-X-X-C motif mutation is expected to affect Zn binding required to stabilize the structure, it was not clear whether the inability of this mutant to interact with Mis18BP1 is a primary or secondary consequence.
As Mis18a MeDiY and Mis18b MeDiY share 44% sequence similarity, we tested whether the Mis18b MeDiY can also directly interact with Mis18BP1. Size-exclusion chromatography analysis of a sample containing Mis18b MeDiY and Mis18BP1 20-130 (using Superdex 75 10/ 300) showed that they elute at distinct elution volumes, 13.9 and 12.2 ml, respectively, demonstrating a lack of interaction ( Fig 2C). This, together with a published report [26], shows that Mis18a, rather than Mis18b, mediates Mis18BP1 interaction to form the Mis18 complex. We had previously shown that Mis18a MeDiY prefers to heterodimerize with Mis18b MeDiY over homo-dimerizing with itself [28]. Hence, we wanted to determine whether Mis18a MeDiY could still bind Mis18BP1     Mis18a/b forms a hetero-hexamer comprising four copies of Mis18a and two copies of Mis18b We and others have shown that homo-and hetero-oligomerization of Mis18 proteins are crucial for Mis18 function in S. pombe [24] and in humans [25], respectively. However, the precise subunit composition of human Mis18a/b hetero-oligomer and its consequence on Mis18 complex assembly and function remain elusive. First, we determined the absolute molecular mass of untagged Mis18a/b complex using SEC combined with multi-angle light scattering (SEC-MALS). The measured MW of the untagged Mis18a/b complex was 151.2 AE 2.9 kDa (calculated MWs of Mis18a and Mis18b are 25.9 kDa and 24.7 kDa, respectively) and correlated well with the calculated MW of a hetero-hexamer (151.8 kDa, using the average MW of Mis18 proteins 25.3 kDa; Figs 3A and EV3A). This contrasts with a previous report by Nardi et al [25], which used glycerol-based gradient experiments to show that Mis18a/b is a hetero-tetramer.
As Mis18a and Mis18b are very similar in size (~25 kDa), it is almost impossible to accurately determine the subunit stoichiometry within the Mis18 hetero-hexamer. Hence, we introduced a noticeable size variation by purifying His-GFP-Mis18a in a complex with His-Mis18b. Interestingly, the measured MW of His-GFP-Mis18a/His-Mis18b complex was 287.2 AE 5.5 kDa, matching the calculated MW of a hetero-hexamer containing four copies of His-GFP-Mis18a and two copies of His-Mis18b (276 kDa; Figs 3B and EV3B). The measured MW of Mis18a/b complex formed using His-Mis18a and His-GFP-Mis18b revealed same stoichiometry as above (measured MW = 232.2 AE 4.5 kDa; calculated MW of 4 His-Mis18a : 2 His-GFP-Mis18b = 221.4 kDa; Figs 3C and EV3C) and demonstrates that Mis18a/b is a hetero-hexamer with a 4:2 stoichiometry.

Mis18a/b hetero-hexamer is assembled from hetero-trimers of C-terminal a-helical domains and hetero-dimers of MeDiY domains
We had previously shown that the MeDiY domains of Mis18a and Mis18b form a homo-dimer and a monomer, respectively, but can form a hetero-dimer [28]. In addition, Nardi et al [25] have shown that the C-terminal a-helical domains also have the ability to oligomerize. These observations together with our data that the Mis18a/b complex is a hetero-hexamer (Fig 3A-C) prompted us to define the stoichiometry of the Mis18a/b C-terminal helical assembly. Using individually purified His-GFP-Mis18a C-term (Mis18a 188-end) and His-MBP-Mis18b C-term (Mis18b 184-end), we reconstituted the C-terminal helical assembly and analyzed their composition using SEC-MALS (Figs 2A and 3D). The measured MW of His-GFP-Mis18a C-term /His-MBP-Mis18b C-term complex was 115.5 AE 2.2 kDa, which matches a calculated MW of a hetero-trimeric assembly with 2 His-GFP-Mis18a C-term and 1 His-MBP-Mis18b C-term (119.6 kDa; Figs 3D and EV3D). This suggests that the formation of the full-length hetero-hexameric Mis18a/b assembly requires further oligomerization of Mis18a/b hetero-trimers mediated by Mis18a MeDiY /b MeDiY hetero-dimers (Fig 3E).  (Fig 4A).

MeDiY dimerization interface of Mis18a is required for Mis18 oligomerization and Mis18BP1 binding
Our previous structural analysis of the S. pombe MeDiY domain identified a putative substrate-binding pocket and a dimerization interface, both required for Mis18 function [24]. To evaluate whether equivalent surfaces in human Mis18a are required for Mis18BP1 binding, we introduced mutations V82E/Y176D (Mis18a DimerM ) at the MeDiY dimeric interface of Mis18a/b and L136A/Y152A/C167A/S169K (Mis18a PocketM ) in the putative substrate-binding pocket using a S. pombe-based homologymodeled structure (Fig 4B). We then analyzed Mis18BP1 binding to His-GFP-Mis18a/His-Mis18b (Mis18a wt /b), His-GFP-Mis18a DimerM / His-Mis18b (Mis18a DimerM /b), and His-GFP-Mis18a PocketM /His-Mis18b (Mis18a PocketM /b) complexes using a Ni-NTA pull-down assay. While Mis18a wt /b and Mis18a PocketM /b interacted with Mis18BP1 20-130 , Mis18a DimerM /b failed to do so efficiently (Fig 4C). This suggests that the Mis18a MeDiY interface proposed to mediate dimerization with Mis18b MeDiY is also required for Mis18BP1 binding (in agreement with Fig 2D). Size-exclusion chromatography analysis of Mis18a DimerM /b, unlike Mis18a wt /b, showed the presence of several distinct populations including an aggregated and a smaller MW species (Fig EV4). The measured MW of the smaller MW species (the sample eluting at 11.7 ml) using SEC-MALS was 129.3 AE 2.5 kDa (Figs 4D and EV4). This correlates with the calculated MW of a complex containing 2 His-GFP-Mis18a DimerM and 1 His-Mis18b (137.8 kDa). This strengthens the notion that the Mis18a/b hetero-trimer is the core oligomeric unit formed through the interactions of the C-terminal helices which then assemble into a hetero-hexamer via the MeDiY-dimerization interface (

EMBO reports
Mis18 complex assembly and regulation Frances Spiller et al and 4A). We conclude that the Mis18a MeDiY dimerization interface is required both for Mis18BP1 binding and for the higher order oligomerization of the Mis18a/b hetero-trimer ( Fig 4A). We next evaluated the contribution of Mis18a MeDiY dimerization interface on Mis18BP1 binding in vivo by tethering TetR-eYFP-Mis18a DimerM to the alphoid tetO array and probing the recruitment of mCherry-Mis18BP1  . In agreement with the in vitro data, the TetR-eYFP-Mis18a DimerM failed to recruit mCherry-Mis18BP1 20-130 (Fig 4E). Furthermore, this mutant, unlike the TetR-eYFP-Mis18a wt , was unable to deposit CENP-A at the tethering site (Fig 4F). This confirms that the Mis18a MeDiY dimerization interface is required for Mis18BP1 binding and CENP-A deposition.

Cdk1 consensus sites of Mis18BP1 are directly involved in Mis18a/b binding
Mis18 complex assembly and its centromere localization have been suggested to be regulated by Cdk1 in a cell cycle-dependent manner [22,23]. Interestingly, the amino acid sequence analysis revealed two consensus Cdk1 phosphorylation sites, T40 and S110, within Mis18BP1   (Fig 5A), which we show here is able to interact with Mis18a/b (Fig 1). Notably, Mis18BP1 S110 has previously been shown to be phosphorylated in vivo [23]. We hypothesized that phosphorylation of Mis18BP1 T40 or/and S110 negatively regulates its association with Mis18a/b and hence Mis18 complex assembly.
To further study the cell cycle regulation of Mis18BP1, we tested the recruitment of mCherry-Mis18BP1 20-130 to the alphoid tetO array by TetR-eYFP-Mis18a during different stages of mitosis ( Fig 5D). Interestingly, in agreement with a previously reported suggestion [22], Mis18a was unable to recruit Mis18BP1 20-130 wt during early stages of mitosis when Cdk1 levels are high. On the contrary, a non-phosphorylatable mutant of Mis18BP1 (mCherry-Mis18BP1 20-130 T40A/S110A ) was recruited by TetR-eYFP-Mis18a to the array throughout the cell cycle. Consistent with this, recombinantly purified Mis18BP1 20-130 T40A/S110A interacted with Mis18a/b as analyzed by SEC (Fig EV5).
The temporal regulation of Mis18 complex formation defines the timing of HJURP-mediated CENP-A deposition essential for centromere inheritance and function [18,22,23]. Oligomerization of Mis18 proteins and Cdk1 activity are both key regulators of Mis18 complex assembly [22][23][24][25]. Our work presented here, in agreement with a parallel study by Pan et al [29], shows that the Mis18a/b complex is a hetero-hexamer made of 2 Mis18a/b hetero-trimers, each with 2 Mis18a and 1 Mis18b that are held together by the hetero-trimeric C-terminal a-helical assembly. One Mis18a MeDiY from each Mis18a/b trimer hetero-dimerizes with Mis18b MeDiY to form a Mis18a/b hetero-hexamer. This arrangement results in 2 Mis18a MeDiY /b MeDiY hetero-dimers that bind two copies of Mis18BP1 and 2 Mis18a MeDiY that possibly form a homo-dimer ( Fig 4A). However, our data show that Mis18a MeDiY can bind Mis18BP1, albeit less efficiently compared to Mis18a MeDiY /b MeDiY hetero-dimer contradicting the suggestion by Pan et al [29] that neither Mis18a MeDiY nor Mis18b MeDiY can interact with Mis18BP1 on their own. The apparent contradiction could be due to the higher ionic strength buffer used by Pan et al [29] in their binding assays. Whether Mis18a/b hetero-hexamer can bind more than two copies of Mis18BP1 mediated via the free Mis18a MeDiY under specific circumstances is yet to be determined. Finally, we identified two highly conserved Cdk1 consensus sites 70 amino acids apart within the Mis18 binding region of Mis18BP1 (T40 and S110). While mutating both these amino acid residues to phosphomimicking residues completely abolished the ability of Mis18BP1 to bind Mis18a/b in vitro and in vivo, individual phospho-mimic mutants (T40E or S110D) failed do so efficiently. This suggests that Mis18BP1 binds Mis18a/b possibly via a bipartite binding mode. Overall, our findings together with the recent independent study by Pan et al [29] provide key insights into the molecular basis for cell cycle-dependent Mis18 complex assembly and function.
Proteins were expressed, either individually or together with their binding partners, using E. coli BL21 gold grown in LB media. Cultures were induced with 0.35 mM IPTG at 18°C overnight. Cell lysis was carried out by sonicating cells re-suspended in a lysis buffer containing 20 mM Tris (pH 8.0), 250 mM NaCl (or 500 mM for Mis18BP1 20-130 ), 35 mM imidazole, and 2 mM bME. Lysis buffer was supplemented with 10 lg/ml DNase, 1 mM PMSF, and cOmplete (EDTA-free, Roche). Proteins were purified from the clarified lysates by affinity chromatography using a 5 ml HisTrap HP column (GE Healthcare). The proteinbound resin was washed with lysis buffer, followed by a buffer containing 20 mM Tris (pH 8.0), 1 M NaCl, 50 mM KCl, 10 mM MgCl 2 , 2 mM ATP, 35 mM imidazole, 2 mM bME, and with a final lysis buffer wash. Proteins were eluted using a lysis buffer supplemented with 500 mM imidazole and dialyzed overnight into 20 mM Tris (pH 8.0), 75-100 mM NaCl, and 2 mM DTT. All proteins were subjected to anion exchange chromatography using the HiTrap Q column (GE Healthcare). Appropriate fractions were pooled, concentrated, and injected into a Superdex 200 increase 10/300 or Superdex 75 10/300 column (GE Healthcare) equilibrated with 20 mM Tris (pH 8.0), 100-250 mM NaCl, and 2 mM DTT. Fractions were analyzed on SDS-PAGE and stained with Coomassie blue.
Ni-NTA pull-down assay was performed in 20 mM Tris (pH 8.0), 75-500 mM NaCl, 10% glycerol, 0.5% NP-40, 35 mM imidazole, and 2 mM bME. Proteins were mixed with two times molar excess of Mis18BP1 20-130 and made up to 200 ll with buffer before incubated for 30 min at 4°C with 60-120 ll of slurry that had been equilibrated in buffer. Beads were then washed four times with 1 ml of buffer, and bound protein was eluted by boiling in SDS-PAGE loading dye before being analyzed by SDS-PAGE.

Cell culture and transfection
The HeLa 3-8 overexpressing CENP-A-SNAP integration cell line, containing a synthetic a-satellite (alphoid) DNA array integration with tetO sites (alphoid tetO array) integrated in a chromosome arm, was maintained in DMEM (Gibco) supplemented with 5% fetal bovine serum (Biowest) and penicillin/streptomycin (Gibco). Cells were grown at 37°C and 5% CO 2 . Transfections were performed in parallel with XtremeGene-9 (Roche) following manufacturer's instructions. Briefly, 24 h after plating, cells grown in 12-well plates were incubated with transfection complexes containing: 0.25 lg of each vector, 100 ll of Opti-MEM (Invitrogen), and 3 ll of Xtreme-Gene-9 reagent for 36 h.

Immunofluorescence and quantification
Thirty-six hours after transfection, cells growing on coverslips were fixed with 2.6% paraformaldehyde in PBS 1× buffer for 10 min at 37°C. Cells were then treated with permeabilization buffer (PBS containing 0.2% Triton X-100, Sigma) for 5 min at room temperature and blocked with permeabilization buffer containing 3% BSA for 1 h at 37°C. Mouse anti-CENP-A (AN1; 1/500 dilution) and Cy-5-conjugated donkey anti-mouse secondary antibody (Jackson Immunoresearch) were used for centromere staining.
Micrographs were acquired in the Centre Optical Instrumentation Laboratory on a DeltaVision Elite system (Applied Precision) using an inverted Olympus IX-71 stand, with an Olympus UPlanSApo 100× oil immersion objective (numerical aperture (NA) 1.4) and a Lumencor light source. Camera (Photometrics Cool Snap HQ), shutter and stage were controlled through Soft-Worx (Applied Precision). The 0.2-lm-spaced z-stacks were deconvolved using SoftWorx and analyzed using ImageJ software (NIH, Bethesda). For CENP-A signal quantification, a custommade macro in ImageJ modified from Bodor et al [30] was used. Briefly, the CENP-A signal (Cy-5) found at the alphoid TetO array (identified by the eYFP signal) was determined for every z-section within a 7-square pixel box. The mean signal intensity in the array was obtained and the background subtracted using the minimum intensity values within the section. The average intensity of the CENP-A signal from endogenous centromeres was used to normalize. At least three biological independent experiments were performed for each assay.
Expanded View for this article is available online.