In vitro centromere and kinetochore assembly on defined chromatin templates.

During cell division, chromosomes are segregated to nascent daughter cells by attaching to the microtubules of the mitotic spindle through the kinetochore. Kinetochores are assembled on a specialized chromatin domain called the centromere, which is characterized by the replacement of nucleosomal histone H3 with the histone H3 variant centromere protein A (CENP-A). CENP-A is essential for centromere and kinetochore formation in all eukaryotes but it is unknown how CENP-A chromatin directs centromere and kinetochore assembly. Here we generate synthetic CENP-A chromatin that recapitulates essential steps of centromere and kinetochore assembly in vitro. We show that reconstituted CENP-A chromatin when added to cell-free extracts is sufficient for the assembly of centromere and kinetochore proteins, microtubule binding and stabilization, and mitotic checkpoint function. Using chromatin assembled from histone H3/CENP-A chimaeras, we demonstrate that the conserved carboxy terminus of CENP-A is necessary and sufficient for centromere and kinetochore protein recruitment and function but that the CENP-A targeting domain--required for new CENP-A histone assembly--is not. These data show that two of the primary requirements for accurate chromosome segregation, the assembly of the kinetochore and the propagation of CENP-A chromatin, are specified by different elements in the CENP-A histone. Our unique cell-free system enables complete control and manipulation of the chromatin substrate and thus presents a powerful tool to study centromere and kinetochore assembly.

Metazoan centromeres are specified epigenetically by the presence of CENP-A nucleosomes[3]. Structural differences between CENP-A and histone H3 nucleosomes [2,4] and/or specific protein recognition elements in CENP-A appear to provide the information that specifies centromere identity and directs kinetochore assembly in a DNA sequence independent manner [5-10] ENREF 4. Moreover, many metazoan centromeres are complex in their organization with interspersed blocks of CENP-A nucleosomes and histone H3 nucleosomes assembled on long arrays of repetitive DNA [11-13]. The difficulty in purifying and manipulating complex centromeres has limited our understanding of how centromeric chromatin promotes centromere and kinetochore formation and chromosome segregation.

To mimic the arrays of CENP-A nucleosomes present in complex vertebrate centromeres, we reconstituted human CENP-A chromatin from recombinant components (Figure 1a). We generated saturated chromatin arrays by salt dialysis of purified histone proteins H2A, H2B, H4, and either CENP-A or H3 with a biotinylated DNA template containing 19 repeats of a 147 bp high affinity nucleosome positioning sequence (19×601) (Figure S1a,b) [14,15]. We bound the biotinylated arrays to streptavidin coated magnetic beads, thereby immobilizing the arrays so that they can be easily added to and recovered from cell extracts (Figure 1a, Figure S1c, d, e).

Figure 1

Reconstituted CENP-A chromatin supports centromere assembly in Xenopus egg extracts

(a) A schematic showing the reconstitution of CENP-A and H3 chromatin arrays and the attachment of the chromatin to magnetic beads via biotin end-labeled DNA. (b) Representative images comparing CENP-C binding to CENP-A and H3 chromatin arrays in CSF and interphase Xenopus extract. The left column shows the separate histone H4 staining used for normalization of the quantification, followed by staining for DNA, HsCENP-A and CENP-C. A merge image of the DNA (red) and CENP-C (green) channels is shown in the right column. Scale bar, 5 μm (c) Quantification of the array associated centromeric proteins CENP-C, CENP-N and CENP-K in CSF and interphase extracts, normalized to histone H4 levels. The levels are rescaled so that CENP-A arrays in CSF are set at 1. Error bars represent the standard error of the mean (SEM), n = 3 (p < 0.05 between CENP-A and H3 chromatin arrays for CENP-C, CENP-N and CENP-K).

We recently demonstrated that the essential centromere protein CENP-C directly recognizes the C-terminus of CENP-A in mononucleosomes but not in isolated CENP-A2/H42 tetramers [5] (our unpublished observations). Therefore, we tested in vitro translated human and Xenopus laevis CENP-C (Hs- and XlCENP-C respectively) for binding to reconstituted H3 and CENP-A chromatin. Human and Xenopus CENP-A are >50% identical (Figure S2a) and we find that both HsCENP-C and XlCENP-C bind specifically to HsCENP-A chromatin arrays in vitro, when compared to H3 chromatin arrays (Figure S2b).

Xenopus laevis egg extract is a widely used cell free system to study chromosome segregation[16]. Egg extracts are arrested in metaphase II of meiosis by the activity of cytostatic factor (CSF) and the cell cycle state of the extract can be transitioned into interphase by adding calcium. We developed a quantitative immunofluorescence assay to determine whether centromere proteins bound to CENP-A chromatin arrays when arrays were added to Xenopus egg extracts. CENP-N and CENP-K are centromere proteins that are required for proper centromere and kinetochore assembly in somatic cells, and we have previously shown that CENP-N, similar to CENP-C, directly binds to the CENP-A nucleosome[6]. We found that CENP-C, CENP-N and CENP-K specifically associated with CENP-A arrays independent of the cell cycle stage of the extract (Figure 1b,c and Figure S2c-f). The centromere protein CENP-T that binds to either H3 nucleosomes or DNA at centromeres did not selectively bind CENP-A chromatin arrays (Figure S3a,b)[17]. Similarly, the inner centromere protein Incenp and Polo like kinase 1 (Plk1) associated with both types of chromatin arrays (Figure S3c). Xenopus incenp is targeted to chromatin through phosphorylation of both H2A and H3 and thus may have affinity for both CENP-A and H3 chromatin[18-20] and Plk1 associates with chromatin in Xenopus egg extract independent of the kinetochore[21]. Furthermore, reconstituted chromatin segments are unlikely to generate paired sister chromatids with inner centromeres because naked DNA and linear DNA replicates inefficiently in these egg extracts[22]. The specific recruitment of the centromere proteins CENP-C, CENP-N and CENP-K, however, indicates that reconstituted CENP-A chromatin arrays can support essential steps in the centromere assembly process in vitro.

Functional kinetochores assemble on sperm chromatin in metaphase Xenopus egg extract. At high sperm concentration, microtubule depolymerization causes mitotic checkpoint activation, resulting in the increased association of checkpoint proteins with kinetochores, and cell cycle arrest[23]. We tested whether reconstituted CENP-A chromatin arrays support kinetochore assembly and checkpoint protein binding after microtubule depolymerization. We added CENP-A or H3 arrays to CSF arrested egg extracts and then cycled the extracts through interphase and back into mitosis, in the presence or absence of nocodazole, as outlined in Figure 2a and demonstrated in Figure S4a. The constitutive centromere protein CENP-C and the microtubule-binding kinetochore protein Ndc80 bound to CENP-A arrays in the presence or absence of nocodazole (Figure 2b,c and S4b). The spindle assembly checkpoint proteins CENP-E, Mad2, Rod and ZW10 associated with CENP-A chromatin at intermediate levels in the absence of nocodazole but upon microtubule depolymerization their binding increased 2-4 fold (Figure 2b). Western blot analysis showed that CENP-C and Ndc80 are precipitated with CENP-A arrays independent of microtubule depolymerization. ZW10 and Rod are enriched on CENP-A arrays upon nocodazole treatment in metaphase, regardless of whether the extract has been cycled through interphase (Figure 2c). These results indicate that CENP-A chromatin arrays respond to microtubule depolymerization by recruiting mitotic checkpoint proteins (Figure 2b,c and S4b).

Figure 2

CENP-A chromatin specifically recruits kinetochore proteins as a response to a mimic of kinetochore detachment from microtubules

(a) A schematic showing the experimental procedure. (b) Quantification of immunofluorescence analysis of CENP-C, Ndc80, CENP-E, Mad2, Rod or ZW10 recruitment to chromatin arrays with (+) and without (−) nocodazole (NOC). The levels are rescaled so that CENP-A arrays with (+) nocodazole are set at 1. Error bars represent SEM, n = 3 (p < 0.05 between − and + nocodazole for CENP-E, Mad2, Rod and ZW10 binding to CENP-A chromatin arrays). (c) Western blot analysis of CENP-C, Ndc80, Rod and ZW10 recruitment to CENP-A and H3 chromatin arrays with and without (+/−) nocodazole (NOC) in CSF and cycled egg extracts. H4 levels are shown as a loading control.

Microtubule binding is a hallmark of kinetochore function and decondensed sperm chromatin efficiently supports spindle formation in egg extracts (Figure 3a, left panel)[24]. However, chromatin assembled on naked DNA induces spindle formation in Xenopus egg extracts independent of kinetochores[25]. When we added CENP-A and H3 chromatin beads into mitotic egg extract we observed microtubule polymerization around the majority of CENP-A arrays but only around a subset of H3 arrays (Figure 3a, left panel). We quantified the amount of microtubule polymer associated with each type of array and found significantly more microtubules associated with CENP-A chromatin beads (Figure 3b and Figure S5a). This indicates that CENP-A chromatin preferentially stabilizes microtubules or promotes their polymerization. We observed heterogeneous microtubule structures around the CENP-A chromatin beads ranging from bipolar spindles to stabilized microtubules or microtubule bundles (Figure 3a, Figure S5a,b). A second property of functional kinetochores is that kinetochore associated microtubule bundles (k-fibers) are stable to cold treatment which depolymerizes non-kinetochore microtubules. We asked whether kinetochores assembled on CENP-A chromatin could stabilize microtubules to cold shock by incubating the microtubule assembly reactions for 10 min at 4°C. We found that kinetochores assembled on CENP-A chromatin arrays stabilized microtubules to cold shock similar to kinetochores assembled on native sperm chromatin while H3 chromatin arrays did not (Figure 3a,c and Figure S5c). When we completely depolymerized microtubules with nocodazole we observed Mad2 recruitment to native sperm centromeres and CENP-A chromatin beads but not H3 chromatin beads (Figure 3a,c and Figure S5c). These results indicate that CENP-A chromatin arrays, similar to native sperm chromatin, assemble functional kinetochores that promote microtubule binding, k-fiber stabilization and spindle checkpoint function (Figure 3a).

Figure 3

Kinetochores assembled on reconstituted CENP-A chromatin bind microtubules and generate a mitotic checkpoint signal

(a) Representative images of microtubule polymerization induced by sperm or reconstituted CENP-A and H3 chromatin. Microtubules (green) and Mad2 (magenta) levels are shown. Scale bar, 10μm (b) Quantification of tubulin and DNA associated with CENP-A and H3 chromatin beads. Error bars represent SEM, n = 5 (c) Quantification of tubulin and Mad2 levels associated with CENP-A and H3 chromatin beads after cold shock (4°C) and nocodazole (NOC) treatment. Error bars represent SEM, n = 5 (d) Western blot showing phospho-Wee1 (P-Wee1) levels as an indicator of the cell cycle stage and tubulin levels as a loading control. Samples from different time points after release from mitotic arrest (t 0′, t 10′, t 20′, t 30′, t 40′) are shown for CENP-A and H3 chromatin arrays, each incubated with nocodazole (+) or with DMSO (−) as a control. (e) Quantification of four independent experiments showing the phospho-Wee1 signal intensity (P-Wee1 signal) over time (min). Error bars represent SEM, n = 4.

In cells, unattached kinetochores activate the mitotic checkpoint and delay mitotic exit until all chromosomes are properly attached and aligned [26,27]. We tested whether kinetochores assembled on CENP-A chromatin arrays could generate a mitotic checkpoint response to microtubule depolymerization and delay the cell cycle. We mixed CENP-A and H3 chromatin with CSF extracts, cycled the reactions through interphase and then cycled them back into mitosis in the presence or absence of nocodazole (Figure 2a). We then released the extract from mitosis into interphase a second time and monitored the kinetics of this transition by measuring the mitosis specific phosphorylation of Wee1 (phospho-Wee1) (Figure 3d). Upon release from mitosis, phospho-Wee1 levels rapidly declined and were undetectable after 30 minutes in control extracts containing CENP-A chromatin or H3 chromatin, as well as in extracts containing H3 chromatin in the presence of nocodazole (Figure 3d and 3e). In extracts containing CENP-A chromatin and nocodazole, the phospho-Wee1 signal increased until 20 min after calcium addition and subsequently declined until 40 minutes after calcium addition to a level only slightly lower than the levels prior to release (Figure 3d and 3e). In the presence of CENP-A chromatin and nocodazole, cyclin B levels rapidly declined but then stabilized, similar to the response observed for native sperm chromatin[23]. However cyclin B was not stabilized in the presence of H3 chromatin and nocodazole (Figure S5d,e). We estimate that the number of CENP-A nucleosomes we are adding to the egg extract exceeds the CENP-A nucleosome concentration required to activate the checkpoint using sperm nuclei[23]. The lower efficiency of reconstituted arrays for checkpoint signaling may be due to the comparatively short length of our reconstituted CENP-A chromatin to native CENP-A chromatin or the lack of replicated sister chromatids and inner centromeres important for tension dependent checkpoint activation. Despite these differences, our synthetic CENP-A chromatin supports a mitotic checkpoint response that mimics the response of native kinetochores to microtubule depolymerization.

The reconstituted chromatin system we have developed provides a distinct experimental advantage over native metazoan centromeric chromatin because the chromatin template can be easily manipulated to dissect the roles of histone proteins in centromere function. A central question in centromere function is how CENP-A chromatin directs the assembly of the centromere and kinetochore. CENP-N recognizes the CATD region of the CENP-A nucleosome while CENP-C binds the C-terminal tail of CENP-A [5,6]. However, the relative importance of these two recognition mechanisms in centromere and kinetochore assembly is incompletely understood.

We generated chromatin arrays containing chimeric CENP-A/H3 proteins to ask how the CENP-A CATD domain and the CENP-A C-terminus influence centromere and kinetochore assembly (Figure 4a). We characterized the level of histone exchange and/or loss from the arrays during incubation in extracts and found that the majority of recombinant HsCENP-A nucleosomes were stable during the incubation indicating low exchange and/or loss rates (Figure S6a,b). We detected a low level of phosphorylated histone H3 on CENP-A chromatin arrays in CSF extract (11.7% ± 7% compared to H3 arrays) and in extract that had been cycled through interphase and back into mitosis (22% ± 13% compared to H3 arrays) (Figure S6c,d). The chimeric arrays containing CENP-A with the histone H3 tail (CENP-A+H3C) exhibited similar levels of exchange (Figure 6c,d). The Xenopus CENP-A present in the extract did not appreciably exchange onto any of the arrays (detection limit ~5-10% exchange) (Figure S6c). The absence of gross rearrangements or bulk histone exchange suggests that chromatin arrays can be used to dissect how individual domains of CENP-A influence kinetochore assembly.

Figure 4

The CENP-A C-terminus is required for centromere and kinetochore assembly in Xenopusegg extract

(a) A schematic showing the different CENP-A/H3 chimeras used in this study. The numbers on top represent the amino acid (AA) within HsCENP-A. (b) Quantification of immunofluorescence analysis of CENP-C, CENP-K and CENP-N recruitment to wild type and chimeric arrays. The relative amounts of each centromere protein bound to the arrays are shown relative to CENP-A arrays set to 1. Error bars represent SEM, n = 3 (p ≤ 0.05 for all proteins binding to CENP-A arrays compared to chimeric arrays except for the H3 arrays containing the CENP-A C-terminus). (c) Quantification of immunofluorescence analysis of Ndc80, CENP-E, Mad2 recruitment to chimeric chromatin arrays with (+) and without (−) nocodazole (NOC). Values are displayed relative to CENP-A arrays in the presence of nocodazole set to 1. Error bars represent SEM, n = 4. The efficiencies of recruitment of kinetochore proteins to CENP-A and H3+CAC arrays in nocodazole were not statistically distinguishable (p ≥ 0.26 for Ndc80, CENP-E and Mad2). (d) Quantification of microtubule binding to CENP-A, H3, H3+HsCAC and H3+XlCAC chromatin arrays represented as percentage of beads associated with tubulin levels above threshold (dark grey bars, left y-axis). Average DNA levels on chromatin beads are shown representing the levels of chromatin arrays bound to beads (light grey bars, right y-axis). Error bars represent SEM, n = 4 for CENP-A and H3 arrays, n = 5 for H3+HsCAC arrays and n = 2 for H3+XlCAC arrays. (e) Western blot analysis shows phospho-Wee1 (P-Wee1) levels as an indicator of the cell cycle stage at t 0′ (0 min) and t 40′ (40 min) after mitotic exit. Tubulin levels are shown as a loading control. (f) Quantification of the phospho-Wee1 signal intensity over time (P-Wee1). Error bars represent SEM, n = 5.

Using our in vitro centromere and kinetochore assembly assay, we found that CENP-C bound with equal efficiency to chromatin arrays assembled with either wild type CENP-A or with chimeras of histone H3 with the CENP-A C-terminal 6 amino acids (H3+CAC) but not CENP-A+H3C (Figure 4b and S7a, left panel). This demonstrates that the CENP-A C-terminus is necessary and sufficient for recruiting CENP-C to CENP-A chromatin arrays in egg extracts as it is for CENP-A mononucleosome binding in vitro[5].

CENP-K depends on CENP-C for its association with Xenopus sperm centromeres [28] and CENP-K also associated with the wild type and H3+CAC arrays (Figure 4b and Figure S7a). Surprisingly, we found that H3+CAC arrays recruited CENP-N as efficiently as wild-type CENP-A arrays, even though these arrays lack the CATD recognition element for CENP-N[6]. CENP-N binding to either CENP-A+H3C or H3+CATD arrays was no better than its binding to H3 chromatin arrays, indicating that the CENP-A C-terminus is required for CENP-N association with CENP-A chromatin in Xenopus egg extract (Figure 4b and Figure S7a). The lack of Xenopus CENP-N binding to H3+CATD and CENP-A+H3C chromatin arrays is not due to species differences because XlCENP-N binds HsCENP-A mononucleosomes in vitro in the absence of CENP-C (Figure S7b). The association of CENP-N and -K with chromatin arrays was dependent on CENP-C as CENP-C depletion from the extract (Figure S8a) reduced the binding to background levels (Figure S8b,c). This was not due to depletion of CENP-N or CENP-K by CENP-C as we have previously shown that complementation of CENP-C depleted extracts restores CENP-K binding and CENP-K is known to depend on CENP-N for its centromere localization [6,29,30]. The dependence of CENP-N on CENP-C for its localization to CENP-A arrays may reflect a role for CENP-C in altering the geometry of centromeric chromatin to promote CENP-N’s access to CENP-A nucleosomes or may reflect the assembly of CENP-N into the larger CCAN complex recruited to the centromere via CENP-C. Our results demonstrate that CENP-C recognition of the CENP-A C-terminus is necessary and sufficient for CENP-N and CENP-K association with chromatin arrays in Xenopus egg extract.

We analyzed the chromatin requirements for mitotic kinetochore formation using the experimental strategy illustrated in Figure 2a. The kinetochore proteins Ndc80, CENP-E, Mad2, Rod and ZW10 are efficiently recruited to wild type and H3+CAC chromatin arrays, but not to CENP-A+H3C or H3+CATD chromatin arrays (Figure 4c, Figure S9a, Figure S10a). Similar to wild type CENP-A chromatin, only the checkpoint proteins CENP-E, Mad2, ZW10 and Rod increased in their association with H3+CAC after microtubule depolymerization (Figure 4c, Figure S9a, Figure S10a). As with wild type CENP-A arrays, the H3+CAC arrays showed increased associated microtubule polymer indicating that the C-terminus of CENP-A directs the formation of microtubule binding or stabilization activity (Figure 4d). Human and Xenopus CENP-A differ by two amino acids in their C-terminal tail (Figure S2a) and chimeric nucleosome arrays containing the Xenopus C-terminal tail of CENP-A fused to H3 (H3+XlCAC) were equally efficient in CENP-C recruitment and microtubule binding as HsH3+CAC arrays (Figure 4d and Figure S10b); indicating that the mode of interaction between CENP-C and CENP-A is conserved.

We assayed the ability of chimeric nucleosome arrays to promote mitotic checkpoint arrest after microtubule depolymerization and found that H3+CATD and CENP-A+H3C did not delay the exit from mitosis but that H3+CAC did (Figure 4e,f). The delay of mitotic exit caused by H3+CAC arrays was less effective than that of CENP-A chromatin arrays indicating that regions of CENP-A in addition to the C-terminus increase the effectiveness of checkpoint signaling, possibly by stabilizing CCAN and kinetochore protein interactions with chromatin (Figure 4e,f). Taken together, our data demonstrate that the primary chromatin determinant for functional centromere and kinetochore assembly is the C-terminus of CENP-A and its recognition by CENP-C.

Here we have shown that reconstituted CENP-A chromatin, in the absence of native centromeric DNA, is necessary and sufficient for centromere and kinetochore assembly. Our data imply that short domains of CENP-A chromatin are sufficient for assembling core components of the centromere and kinetochore in the absence of higher order organization of centromeric chromatin and interspersed domains of H3 chromatin. Using our in vitro system, we have directly assessed how domains of CENP-A participate in centromere and kinetochore assembly, even when the mutations we analyze would be expected to be lethal in vivo. We find that, the CENP-A C-terminus is both necessary and sufficient for the recruitment of centromere and kinetochore proteins, for microtubule binding and for a checkpoint response to microtubule depolymerization. We suggest that CENP-A performs two functions that can be separated molecularly: 1) the CENP-A CATD provides a recognition mechanism for targeting of CENP-A to centromeres to maintain centromeric chromatin [2,6-8] and 2) the CENP-A C-terminal tail domain recruits the conserved centromere protein CENP-C to promote centromere and kinetochore assembly[5]. We envision the use of more complex chromatin templates to understand the importance of higher order chromatin organization and regulatory modifications in centromere assembly and function.

Methods Summary

Histone proteins and chimeras were purified as described previously [5,6,15] and assembled onto a biotin end-labeled tandem array of 19 high affinity nucleosome positioning sequences (19×601) by salt dialysis[14]. Chromatin arrays were bound to streptavidin coated magnetic Dynabeads (Invitrogen). X. laevis extracts were prepared as previously described [16] and centromere protein binding to chromatin arrays was performed in freshly prepared CSF egg extract for 1 hour with or without calcium addition. Arrays were fixed in formaldehyde and stained for centromere proteins by indirect immunofluorescence. Kinetochore and checkpoint protein assembly was assayed by adding arrays to extracts released into interphase with calcium for 80 min followed by readdition of CSF extract in the presence or absence of nocodazole (10 μg/ml) for another 90 min. To analyze microtubule binding, chromatin arrays were incubated in CSF for 90 min. Reactions were sedimented through a glycerol cushion onto a coverslip followed by tubulin immunofluorescence. Chromatin array dependent inhibition of mitotic exit, was assayed as described for kinetochore protein binding, but calcium was added a second time to release extracts into interphase. The cell cycle state monitored by western blotting using the α-phospho-Wee1 antibody generously provided by Jim Ferrell.

Images were collected as 13 axial planes at 2 μm intervals on a Nikon Eclipse-80i microscope using a 60×, 1.4 NA PlanApo oil lens and a CoolSnapHQ CCD camera (Photometrics) with MetaMorph software (MDS Analytical Technologies). Axial stacks were maximum intensity projected and quantified using custom software. For normalization of each experiment, a separate Histone H4 staining was performed to quantify the exact array coupling efficiency.