Accurate chromosome segregation during mitosis is essential for the maintenance of genome integrity. Eukaryotic cells have evolved complex machinery to ensure the fidelity of chromosome segregation. Each chromosome directs the assembly of a kinetochore that mediates attachment to the mitotic spindle and is required for microtubule-dependent chromosome movement during mitosis. Kinetochores also act as signaling centers for the mitotic checkpoint, which delays the initiation of anaphase in response to improper chromosome attachment to the mitotic spindle (reviewed in 1
). The centromere is the region of the chromosome upon which the kinetochore is assembled. Human centromeric DNA is composed of repetitive α-satellite sequences but centromeres are thought to be epigenetically specified as centromeric DNA is not well conserved between species and no DNA sequences have been identified that are necessary or sufficient for kinetochore function in vertebrates (reviewed in 2
). Centromeric chromatin in all eukaryotes contains specialized nucleosomes in which histone H3 is replaced by a histone H3-variant called centromere protein A (CENP-A) (reviewed in 3
). CENP-A is currently the best candidate for the epigenetic mark that specifies centromere identity. CENP-A is essential for the recruitment to the centromere of most other proteins required for kinetochore function but the molecular basis for recognition of CENP-A-containing chromatin as the site of kinetochore assembly is poorly understood. Furthermore, the mechanisms that target newly synthesized CENP-A/H4 to established centromeric chromatin to maintain centromere identity have not been determined.
A fundamental limitation in understanding centromere assembly is the lack of well-defined biochemical assays for studying this process. In no case has a direct and specific interaction between CENP-A nucleosomes and any of the >75 proteins that make up a vertebrate mitotic kinetochore been demonstrated. In order to identify CENP-A-nucleosome interacting proteins we developed a simple and rapid binding assay using reconstituted mononucleosomes that contained α-satellite DNA derived from human centromeres and either histone H3 or CENP-A (). Potential CENP-A-nucleosome interacting proteins were selected for testing in this assay based on the analysis of kinetochore assembly in several organisms. We focused on the constitutive centromere associated network (CCAN) of proteins, which includes centromere protein (CENP)-C, H, I, and K through U, because the CCAN proteins are localized to centromeres during interphase and are required for the assembly of functional kinetochores in mitosis. Moreover, the localization by immuno-electron microscopy of CENP-C to the inner kinetochore plate and the recent demonstration that several of these proteins co-purify with CENP-A mononucleosomes suggests that CCAN proteins are likely to be the chromatin proximal elements of kinetochores4
. We also included CENP-B in our analysis because it has previously been shown to bind directly to a conserved 17-nucleotide motif called the CENP-B box present in α-satellite DNA 8
. We expressed and 35
S-labeled CENP-B, C, H, I, and K through U by coupled in vitro
transcription and translation (, S1
), incubated the labeled proteins with or without CENP-A mononucleosomes, and then resolved the mixtures by native gel electrophoresis. Both CENP-B and CENP-N displayed an increased relative migration in the presence of CENP-A nucleosomes assembled with α-satellite DNA suggesting that CENP-B and CENP-N bind directly to CENP-A nucleosomes (). We did not detect any change in the migration of CENP-C, H, I, K, L, M or O through U (, S1a
). We also tested hMis18α, hMis18β and M18BP1/hKNL-2 that have been implicated in CENP-A assembly but we did not detect any interaction with CENP-A nucleosomes in this assay (Figure S1b
We determined the contribution of DNA sequence and histone protein composition to CENP-B and CENP-N binding of CENP-A nucleosomes by alternately exchanging the nucleosomal DNA and histones. As expected, CENP-B bound to naked α-satellite DNA as well as both H3 and CENP-A nucleosomes that contained α-satellite DNA (Figure S1c
). CENP-B did not bind the synthetically derived 601 nucleosome positioning sequence or nucleosomes that contained the 601 DNA sequence. In contrast, CENP-N bound equally well to CENP-A nucleosomes that contained either the α-satellite DNA or the 601 DNA (). CENP-N did not bind the α-satellite or 601 DNA fragment or histone H3 nucleosomes assembled on either DNA fragment. CENP-N also bound to CENP-A/H4 tetrasomes reconstituted with α-satellite DNA that lacked histone H2A and H2B but did not specifically bind to soluble CENP-A/H4 tetramers (Figure S1d,e,f
). Thus, CENP-N binds specifically to CENP-A-associated chromatin.
Domain transfer experiments have previously suggested that a contiguous portion of the loop I and helix II region within the histone fold of CENP-A, called the CENP-A targeting domain (CATD), is sufficient for CENP-A function in vivo11
. We tested whether the CATD was sufficient for CENP-N binding using reconstituted nucleosomes that contained the histone H3CATD
chimera. CENP-N bound efficiently to H3CATD
-containing nucleosomes (), and dose-response experiments revealed that the affinity of CENP-N for H3CATD
nucleosomes was indistinguishable from the affinity of CENP-N for wildtype CENP-A nucleosomes (apparent Kd
=163 nM ±60 vs. 169 nM ±70, respectively) (, Table S1
). The CATD imparts structural differences to CENP-A nucleosomes, in comparison to H3 nucleosomes, that have been suggested to act as an epigenetic mark within chromatin to specify centromere identity13
. Our data suggest that CENP-N recognizes the unique structural information encoded by the CATD within CENP-A nucleosomes.
Depletion of CENP-N with siRNA causes defects in kinetochore assembly and chromosome congression during metaphase, and results in the loss from the centromere of most other CCAN components5
. However, the depletion of other CCAN subunits, including CENP-H, CENP-I, and CENP-K, results in mitotic phenotypes similar to those caused by CENP-N depletion4
. Furthermore, CENP-N co-purifies with CENP-H, CENP-I, CENP-M, CENP-K, CENP-L and CENP-T, which are all likely to be interdependent for centromere localization based on pairwise dependency relationships4
. Accordingly, a specific function for CENP-N in recognizing centromeric chromatin in vivo
cannot be inferred from siRNA-based studies.
In order to directly determine the role of CENP-A nucleosome binding by CENP-N in centromere assembly we generated mutants in CENP-N that specifically reduced CENP-A nucleosome binding affinity. Conserved charged and polar amino acids within CENP-N were changed to alanine based on sequence alignments of CENP-N orthologs from several species (Figure S2a
). Following an initial characterization of the mutants in vitro
and in vivo
and Table S1
), two of the point mutants, R11A and R196A, were selected for detailed analysis. Both the R11A and R196A mutants displayed reduced CENP-A nucleosome binding when compared to wildtype CENP-N (). Dose-response experiments indicated that the CENP-N R11A and R196A mutants had a 6-fold and a 2-fold reduction in CENP-A nucleosome binding, respectively (, Table S1
), suggesting that both R11 and R196 in CENP-N contribute to the recognition of CENP-A nucleosomes. In addition, the C-terminus of CENP-N (amino acids 289–339 in human CENP-N) is more highly conserved among vertebrates than the rest of the protein (Figure S2b
). We therefore constructed a truncation mutant of CENP-N lacking the C-terminus (called CENP-NΔC). The CENP-NΔC mutant did not affect nucleosome binding when compared to wildtype CENP-N ().
Identification and characterization CENP-N mutants defective in CENPA-nucleosome binding
We generated stable HEK293 cell lines that expressed green-fluorescent protein (GFP) fusions to either wildtype CENP-N or the CENP-N mutants to determine how changing the affinity of CENP-N for the CENP-A nucleosome affected CENP-N localization and centromere assembly in vivo
. Western blotting indicated that each cell line expressed comparable levels of the respective GFP-CENP-N protein, with the exception of the CENP-NΔC mutant, which was reduced ~7-fold compared to wildtype GFP-CENP-N ( and S4a
). The HEK293 cell lines express CENP-N from a single genomic locus under the same promoter (Methods) suggesting that the decreased protein level is a general property of the CENP-NΔC mutant. Interestingly, while the levels of CENP-H and CENP-K were not affected in any of the stable cells lines, the expression of wildtype GFP-CENP-N or either of the two GFP-CENP-N point mutants caused a significant reduction in endogenous CENP-N protein (). A reduction of endogenous CENP-N protein has also been described in cells depleted of CENP-H suggesting that CENP-N is unstable when not associated with other CCAN components14
. We conclude that wildtype GFP-CENP-N and the GFP-CENP-N R11A and R196A mutant proteins effectively replaced endogenous CENP-N in these cell lines.
We directly determined the ability of the CENP-N point mutants to interact with other CCAN proteins and with CENP-A nucleosomes in vivo. Immunoprecipitation of wildtype GFP-CENP-N showed that GFP-CENP-N bound to CENP-H, CENP-K and CENP-A nucleosomes (, S4b
). The GFP-CENP-N R11A and R196A mutants were also associated with CENP-H and CENP-K, though the R196A mutant bound CENP-H less well when compared to wildtype CENP-N. Importantly, the R11A and R196A mutants both displayed defects in CENP-A nucleosome binding, the severity of which was consistent with the binding of each mutant to reconstituted CENP-A nucleosomes. Thus, residues R11 and R196 within CENP-N contribute to CENP-A nucleosome binding in vitro
and in vivo
. The comparatively low levels of the GFP-CENP-NΔC present in stable cells and the inability of this mutant to down-regulate endogenous CENP-N protein levels suggested that the GFP-CENP-NΔC mutant might not be stably associated with other CCAN proteins. Indeed, GFP-CENP-NΔC did not bind to CENP-H, CENP-K or CENP-A nucleosomes ().
In order to understand the mechanism by which the C-terminus of CENP-N mediates CENP-N association with CCAN proteins we expressed epitope-tagged centromere proteins and untagged CENP-N in reticulocyte extracts and performed pairwise binding experiments by anti-myc immunoprecipitation. Myc-tagged CENP-L, but not untagged CENP-L, efficiently co-precipitated with CENP-N indicating a direct interaction between CENP-N and CENP-L (). Myc-tagged CENP-L bound the CENP-N R11A and R196A mutants as efficiently as wildtype CENP-N (), consistent with the efficient association of these CENP-N mutants with CENP-H and CENP-K in vivo. However, the GFP-CENP-NΔC mutant did not bind to CENP-L. These data suggest that CENP-N associates with other CCAN components via a direct interaction with CENP-L that requires the highly conserved C-terminus of CENP-N.
We compared the localization of the CENP-N R11A, R196A and ΔC mutants to wildtype CENP-N. Wildtype GFP-CENP-N localized exclusively to centromeres as indicated by colocalization with endogenous CENP-A (). Both the R11A and R196A mutants also localized to centromeres but did so inefficiently. Quantification showed that the levels of the CENP-N R11A and R196A mutants at centromeres was reduced to 32 ± 4% and 39 ± 6%, respectively, of the levels of wildtype CENP-N (). The difference in localization efficiency between the GFP-CENP-N R11A and R196A mutants was not as great as would be predicted from their relative CENP-A nucleosome binding affinities (), suggesting that the other CCAN proteins likely contribute to CENP-N localization efficiency in vivo. Nevertheless, these data show that mutations that reduce CENP-A nucleosome binding by CENP-N result in quantitative defects in the centromere-specific localization of CENP-N.
CENP-N mutants exhibit centromere assembly defects
We did not detect the CENP-NΔC mutant at centromeres in our stable cell line suggesting that association with other CCAN subunits is an important step in the recruitment of CENP-N to centromeres (). However, transiently transfected cells expressing the GFP-CENP-NΔC mutant from a strong promoter occasionally contained detectable levels of the mutant protein at centromeres (). Thus, the C-terminus of CENP-N is not absolutely required for CENP-N centromere localization. Instead, our data indicate that association with the CCAN stabilizes CENP-N and likely increases the efficiency of CENP-N centromere localization.
Mutations in CENP-N that affected CENP-A nucleosome binding caused dominant defects in centromere assembly in our stable cell lines. Quantification of CENP-H, CENP-I and CENP-K in cells expressing the CENP-N R11A and R196A mutants indicated that the levels of each protein at centromeres in interphase cells was significantly reduced when compared to cells expressing wildtype CENP-N (). Thus, reducing the level of CENP-N at centromeres led to a reduction in the levels of a subset of other CCAN subunits. The defect in CENP-H, CENP-I and CENP-K localization in the mutant cells was not as severe as that of CENP-N itself suggesting that a small amount of remaining endogenous CENP-N may also contribute to centromere assembly in these cell lines (). The CENP-N R11A and R196A mutants did not alter the levels of CENP-A or CENP-C at centromeres.
We next examined the localization dependence of CENP-H, CENP-I, and CENPK, on CENP-N by depleting CENP-N with siRNA in HeLa cells (, S5a
). CENP-N depletion led to a substantial reduction in the levels of CENP-H, CENP-I and CENP-K at centromeres () consistent with previous studies demonstrating CENP-H reduction in cells depleted of CENP-N 14
. The localization defects of CENP-H, CENP-I and CENP-K in CENP-N depleted cells were more severe than in cells depleted of CENP-A, indicating that the observed reduction in CENP-H, CENP-I and CENP-K levels at centromeres is not an indirect consequence of the reduced CENP-A levels in the CENP-N depeleted cells (, see below). CENP-C centromere localization was also reduced in CENP-N depleted cells but not to the extent that other CCAN proteins were (). Thus, the localization dependence of CCAN proteins on CENP-N function was similar between CENP-N-depleted cells and the stable cell lines expressing CENP-N mutants that have defects in CENP-A nucleosome binding.
Depletion of CENP-N affects centromere assembly
Depletion of CENP-N caused a reduction in total CENP-A protein () and CENP-A levels at centromeres (, S5b
). CENP-N is therefore required for the maintenance of centromeric chromatin. Similar defects have been described in S. pombe
cells with mutations in the CENP-N ortholog Mis15, suggesting that CENP-N function is evolutionarily conserved16
. CENP-H, CENP-I and CENP-K are required for the deposition of newly synthesized CENP-A at centromeres4
. We therefore asked whether the reduction in CENP-A levels within centromeric chromatin in CENP-N depleted cells was due to a failure to load new CENP-A at centromeres. To address this we employed the pulse labeling method based on the SNAP-tag to determine the fate of newly synthesized protein. Using this strategy, the assembly of nascent CENP-A was previously shown to be restricted to early G1 phase 17
. siRNA mediated reduction of CENP-N levels resulted in a significant reduction of newly synthesized CENP-A-SNAP recruitment to centromeres (, S5c
) indicating that the loss of steady state levels of centromeric CENP-A is caused, at least in part, by a defect in CENP-A assembly.
We have identified CENP-N as the first protein to bind specifically to CENP-A nucleosomes and shown that the direct binding of CENP-A nucleosomes by CENP-N is required for centromere assembly. The DNA sequence-independent binding of CENP-A nucleosomes by CENP-N suggests that CENP-N recognizes the epigenetic mark in chromatin that specifies centromere identity. Importantly, while CENP-N targets directly to CENP-A nucleosomes, we find that CENP-N itself is required for recruiting new CENP-A to the centromere, suggesting that CENP-N is part of a feedback loop responsible for propagating centromeric chromatin in dividing cells ().
CENP-N mutants with a two-fold or six-fold reduction in apparent binding affinity for CENP-A nucleosomes localized equally efficiently to centromeres in vivo, suggesting that CCAN proteins other than CENP-N may also provide direct interactions with chromatin that are important for centromere assembly. The interdependence of several CCAN proteins, including CENP-N and CENP-T, for centromere localization is consistent with this possibility5
. A complex of CENP-T and CENP-W was recently shown to bind directly to DNA in vitro
and the CENP-T/W complex associates with histone H3-containing nucleosomes18
. CENP-N and the CENP-T/W complex are therefore likely to cooperate in providing multiple distinct chromatin contacts that are required for the localization of a subset of other CCAN proteins, including CENP-H, CENP-I and CENP-K, to centromeres.
CENP-C assembly at centromeres is less sensitive to CENP-N depletion than the other CCAN proteins we examined, consistent with previous results suggesting that CENP-I and CENP-C are independently recruited to centromeric chromatin in human cells19
. Nevertheless, CENP-A is required for both CENP-N and CENP-C centromere localization indicating that multiple centromere localization pathways downstream of CENP-A exist5
. CENP-C has previously been shown to bind directly to DNA in vitro but it is unclear how such an activity could translate into the centromere-specific localization observed for CENP-C in vivo 21
. Identifying the molecular mechanisms by which CENP-C is recruited to CENP-A chromatin in the absence of CENP-N and understanding how these distinct centromere recognition pathways are integrated at the level of chromatin will provide important insights into centromere assembly and structure.