Fission yeast (Saccharomyces pombe) centromere DNA is organized in a central core region flanked on either side by a region of outer repeat (otr) sequences. The otr region is known to be heterochromatic and bound by the Swi6 protein whereas the central core region contains an unusual chromatin structure involving the histone H3 variant Cnp1 (S. pombe CENP-A). The central core is the base for formation of the kinetochore structure whereas the flanking region is important for sister centromere cohesion. We have previously shown that the ultrastructural domain structure of S. pombe centromeres in interphase is similar to that of human centromeres. Here we demonstrate that S. pombe centromeres are organized in cytologically distinct domains even in mitosis. Fluorescence in situ hybridization of fixed metaphase cells revealed that the otr regions of the centromere were still held together by cohesion even after the sister kinetochores had separated. In live cells, the central cores and kinetochores of sister chromosomes could be distinguished from one another when they were subjected to mitotic tension. The function of the different centromeric domains was addressed. Transacting mutations affecting the kinetochore (nuf2) central core domain (mis6) and the heterochromatin domain (rik1) were analyzed in live cells. In interphase, both nuf2 and mis6 caused declustering of centromeres from the spindle pole body whereas centromere clustering was normal in rik1 despite an apparent decondensation defect. The declustering of centromeres in mis6 cells correlated with loss the Ndc80 kinetochore marker protein from the centromeres. Interestingly the declustered centromeres were still restricted to the nuclear periphery thus revealing a kinetochore-independent peripheral localization mechanism for heterochromatin. Time-lapse microscopy of live mis6 and nuf2-1 mutant cells in mitosis showed similar severe misaggregation phenotypes whereas the rik1 mutants showed a mild cohesion defect. Thus, S. pombe centromeres have two distinguishable domains even during mitosis, and our functional analyses support the previous observations that the kinetochore/central core and the heterochromatin domains have distinct functions both in interphase and mitosis.
The centromere in most species can be observed as a localized primary constriction of the metaphase chromosome. Centromeric DNA harbors a large protein complex, the kinetochore, which mediates the attachment of the chromosomes to the mitotic and meiotic spindle thereby ensuring proper chromosome segregation. Although centromeric DNA sequences have little homology between organisms, structurally the centromeres and kinetochores are similar between such distantly related species as Saccharomyces cerevisiae, Schizosaccharomyces pombe, Drosophila melanogaster and humans (reviewed by Pidoux and Allshire, 2000; Bjerling and Ekwall, 2002; Blower et al., 2002). Centromeric DNA in fission yeast occupies 40-100 kb on each chromosome, organized with a unique central core sequence (CC/cnt) (Chikashige et al., 1989; Clarke and Baum, 1990) flanked by inner (imr/B) and outer (otr/K+L) repeat arrays (Fig. 1A). The use of chromatin immunoprecipitation (ChIP) has established that the central core region is bound by proteins Cnp1 (fission yeast homologue of human CENP-A) Mis6 and Mis12 bind (Takahashi et al., 2000; Saitoh et al., 1997) and XMAP215 homologues, Dis1 and Mtc1/Alp14 (Garcia et al., 2001; Nakaseko et al., 2001). In contrast the outer repeated region is occupied by the chromodomain proteins Swi6 and Chp1 that are structural components of centromeric heterochromatin over the imr and otr repeats (Partridge et al., 2000). The passenger protein Bir1/Cut17 transiently associates with the imr and otr repeats and not with the central core region in mitosis (Morishita et al., 2001). Thus, the DNA sequence analysis that first revealed the symmetrical organization of S. pombe centromeres (Clarke et al., 1986; Chikashige et al., 1989) and the pattern of protein binding (established by ChIP analysis) to the central core and otr/imr regions suggested that the fission yeast centromeres are composed of two different domains. The contrasting phenotypes of mutants such as rik1 and swi6, with defects in heterochromatin (Ekwall et al., 1995; Ekwall et al., 1996), as compared to mutants such as mis6, with defects in the central core (Saitoh et al., 1997), indicated that these two domains are functionally distinct. Interphase centromeres in S. pombe are discretely clustered at the nuclear periphery, adjacent to the spindle pole body (SPB) that is located on the cytoplasmic side of the nuclear envelope (Funabiki et al., 1993). This clustering facilitated the description of cytological structural domains in fission yeast, and revealed a structural conservation between fission yeast and humans (Kniola et al., 2001). These results have both paved the way for the current study in which we present analyses of the structure of fission yeast centromeres during mitosis and of the structure of fission yeast centromeres in live cells, and also for further investigation into the putative specific functions of the centromere otr/imr and kinetochore central core domains in interphase and mitosis.
Materials and Methods
Strains and media
Media were prepared according to standard methods (Moreno et al., 1991). Strains used are listed in Table 1. Strains carrying genes fused to green and cyan fluorescent proteins (GFP and CFP) were constructed using the methods described previously (Bahler et al., 1998) with the following modifications. Plasmids containing CFP were constructed by replacing the entire coding region of GFP with that of CFP using standard techniques. Briefly, CFP was amplified by PCR using the cameleon pYC2 as template (Miyawaki et al., 1997) and subsequently ligated into pFA6 in which GFP had been released. Expression of N-terminally tagged genes was controlled by the thiamine repressed nmt1 promoter.
We performed GFP and CFP fluorescence microscopy on live cells using Openlab software version 3.1 and a ZEISS Axioplan 2 imaging microscope equipped with a Hamamatsu C4792-95 camera. Cells were prepared as described previously (Facanha et al., 2002; Pidoux et al., 2000). Cells were analyzed at 30°C with the exception of mis6-302 and nuf2-1 cells, which were analyzed at 36°C. The temperature during cultivation of mis6-302 and nuf2-1 cells was raised from 25°C to 36°C 6 hours and 4 hours, respectively, prior to analysis. Electron microscopy was carried out according to Kniola et al. (Kniola et al., 2001) and fluorescence in situ hybridization (FISH) was performed as described by Ekwall et al. (Ekwall et al., 1995) using the pRS140 probe to detect otr (Chikashige et al., 1989).
Frames were collected every 25 seconds and are displayed at a rate of 2 frames/second. Movies are available at http://jcs.biologists.org/supplemental and selected images are presented in Fig. 1B,C.
Results and Discussion
Distinct centromere domains during live mitosis
To obtain a more comprehensive understanding of centromere function and dynamics we first tested if mitotic fission yeast centromeres were organized into cytologically separable domains. FISH was used to detect centromeric outer repeat (otr) sequences in S. pombe cells immunolabeled for the kinetochore protein Ndc80. Ndc80 is a kinetochore marker protein that forms an evolutionarily conserved protein complex with Nuf2 and Spc24 (Wigge and Kilmartin, 2001). Digital deconvolution analysis (z=0.2 μm) revealed that cells in metaphase had a single unseparated cen (otr)-FISH signal on paired sister chromatids that were flanked by two Ndc80-GFP signals facing opposite cellular poles (Fig. 1B). The distance measured (0.48-0.53 μm) between the paired kinetochore signals correspond to the published kinetochore to kinetochore distances of 3-dimensional metaphase spindle models, based on EM serial sectioning (Ding et al., 1993). We therefore concluded that the heterochromatin (otr) domain was clearly distinct from the kinetochore domain in metaphase.
Using green and cyan fluorescent proteins (GFP and CFP) we were then able to explore the centromere/kinetochore dynamics and multilayered structure in real time. We were able to simultaneously detect two tagged proteins by using strains expressing chromosomal copies of a combination of both green-tagged and cyan-tagged proteins. As a reference point for spindle movement, the spindle pole body (SPB) marker Cut12-GFP (Bridge et al., 1998) or Cut12-CFP was used. We also CFP tagged the S. pombe central core binding protein Cnp1 (Takahashi et al., 2000). Cnp1 is the S. pombe homologue of human CENP-A (Warburton et al., 1997; Vafa and Sullivan, 1997), Drosophila CID (Blower and Karpen, 2001), budding yeast Cse4 (Meluh et al., 1998) and C. elegans HCP3 (Buchwitz et al., 1999). CENP-A is a histone H3 variant that creates a specialized chromatin structure, which determines the site of kinetochore assembly.
We began an investigation of centromere domain movements in live cells by examining the spindle growth using the Cut12-CFP marker (see Movie 1). As seen in earlier studies (Nabeshima et al., 1998), we found that the SPBs moved in three distinct phases. In the first phase, the SPB was duplicated. In the second phase the spindle length remained nearly constant (2.5-3.0 μm) while the sister kinetochores made oscillatory movements back and forth. In the third phase, which corresponds to anaphase B, the spindle rapidly extended (up to 10 μm). By taking advantage of the nearly constant spindle length of the second stage of spindle growth we were able to establish the relative position on the metaphase spindle of our marker proteins. The central core domain of the centromere detected with CFPCnp1, and the kinetochore detected with Ndc80-GFP can be seen occupying slightly different positions along the spindle axis when compared to the localization of Cut12-CFP/GFP in cells with equal axis lengths (2.7-2.9 μm distance between SPBs) (Fig. 1C; Movies 1 and 2). The slightly different localizations of Cnp1 and Ndc80 during phase two were directly visualized in triple-tagged live cells expressing Ndc80-GFP, CFP-Cnp1 and Cut12-CFP (Fig. 1C, top; see Movie 3). The perfect colocalization of the dual color of the control strain expression of Cut12 (Fig. 1C, bottom) shows that the pixel differences between Cnp1 and Ndc80 (Fig. 1C, top) are significant. This finding reinforced the evidence for a multilayered domain structure in mitotic fission yeast centromeres. A schematic model illustrating our observations on S. pombe centromere structures at metaphase is presented in the bottom panel of Fig. 1A.
Different centromere domains and different centromere subfunctions
Kinetochore assembly, binding of kinetochore microtubules, orientation of sister kinetochores to opposite poles and their poleward movements are separable centromere functions which may be organized by different structural domains of the centromere. Evidence from Drosophila shows that cohesion function is independent from kinetochore assembly (Lopez et al., 2000) and the results of RNAi inhibition of CID (Drosophila CENP-A) also argues that the centromere and flanking heterochromatin are physically and functionally separable protein domains that are required for different centromere functions (Blower and Karpen, 2001). In fission yeast, mutations in cohesin, which is normally enriched in the flanking heterochromatin, do not seem to affect Cnp1 loading in the central core region (Toyoda et al., 2002). To further test if the different domains we characterized had different centromere functions, we analyzed trans-acting mutations that affect each of the domains. The rik1-304 mutation specifically disrupts centromeric heterochromatin (Allshire et al., 1995; Ekwall et al., 1996). The mis6 mutation abolishes Cnp1 loading to the central core region (Saitoh et al., 1997; Takahashi et al., 2000). The nuf2-1 mutation disrupts the kinetochore (Nabetani et al., 2001).
We began by analyzing the effect of mis6 mutation on interphase centromere clustering and kinetochore behavior using otr FISH combined with immunofluorescence microscopy (IF) of Ndc80-GFP. It has previously been shown that the first mitosis is normal when synchronized mis6 mutant cells are shifted to the restrictive temperature but the second mitosis is severely defective and lethal (Saitoh et al., 1997). Furthermore it was shown that a high proportion of mis6 interphase cells in a unsynchronized culture, 6 hours after being shifted up to the restrictive temperature, had scattered centromere FISH signal, not clustered at the SPB (Saitoh et al., 1997). These results were confirmed in our study where the control wild-type interphase cells invariably showed one large centromere FISH signal overlapping with the Ndc80-GFP signal (Fig. 2A,B) whereas 63% of mis6 interphase cells had multiple dispersed after 6 hours at the restrictive temperature. One of the cen (otr) FISH signals usually overlapped with the Ndc80 signal. At the permissive temperature, 13% of mis6 cells gave multiple cen (otr) spots. Our interpretation of the loss of the Ndc80 kinetochore marker that correlates with scattered cen (otr) signals is that the central core region affected by mis6 was required to connect the kinetochore component Ndc80 to the centromere, and that failure in this process led to declustering of the centromeres.
The chromodomain proteins Swi6 (fission yeast homologue of human HP1) localizes to centromeres mating type region and telomeres (Ekwall et al., 1995). Swi6 is required for proper formation of heterochromatin as well as sister chromatid cohesion, and the recruitment of Swi6 to the otr centromeric DNA is dependent on Clr4 and Rik1 (Ekwall et al., 1996; Bernard et al., 2001; Nonaka et al., 2002). To investigate if centromeric heterochromatin was affected by a mutation in mis6, we immunostained fixed mis6 cells with anti-Swi6 and found that Swi6 was still localized on the declustered centromeres (Fig. 2C). While centromeric Swi6 spots were declustered the telomere and mating type spots of Swi6 still appeared relatively normal in mis6 cells. To verify that heterochromatin was indeed separated from the previously identified anchor structures and plate like structures near the SPB (Kniola et al., 2001) we fixed wild-type and mis6 cells using high pressure freezing (HPF) and examined them by Swi6 immunoelectron microscopy (EM). Centromeric electron-dense centromeric heterochromatin was normal in wild-type controls cells but was detected at distances far away from the SPB and anchor structure in mis6 cells, suggesting that the loss of Cnp1 played a role in anchoring the centromere DNA to the SPB via the kinetochores (Fig. 3B,E). The structure of heterochromatin in rik1 cells was also analyzed by HPF-EM. Using this method, the heterochromatin domain was not readily detectable and appeared less electron dense than in wild-type control cells, whereas the plate-like `anchor' structures indeed appeared to be intact (Fig. 3C,F). This indicated that the kinetochore assembly mediated by the central core domain was functionally independent of the heterochromatin domain. To test this further, we analyzed the rik1 mutant by cen (otr) FISH, and as previously demonstrated (Ekwall et al., 1996) we found that the clustering of centromeres in interphase was intact. Furthermore, the colocalization with kinetochore component Ndc80 was normal in 100% of rik1 cells (data not shown). However, the area of the cen (otr) FISH signal in interphase was enlarged from 0.2±0.1 in wild-type cells to 0.8±0.2 μm2 in rik1 cells (n=50). This indicated that the centromeric heterochromatin was decondensed in cells with compromised rik1 (Fig. 2D). In contrast, the mis6 mutant cells showed a cen (otr) FISH area of 0.2±0.1 μm2 at 25°C and 0.4±0.2 μm2 at 36°C (n=50). The slight increase in the area of the otr signal in mis6 cells at 36°C can be accounted for by the scattering of centromeres rather than a decondensation effect. These findings substantiated the notion that heterochromatin and central core/kinetochore domains have separate functions in interphase cells.
Microtubules are not required for centromere clustering
To investigate centromere clustering in live interphase cells we first tested whether clustering was dependent on microtubules by treating cells expressing Cut12-GFP and CFP-Cnp1 with 10 μg/ml the microtubule poison thiabendazole (TBZ). The resulting live-analysis movies (see Movies 4 and 5) revealed that, although TBZ suppressed the SPB movements, there was no detectable declustering of centromere (CFP-Cnp1) signals in the TBZ-treated cells.
Centromere declustering effects of nuf2 and mis6
In order to visualize what happens to declustered centromeric DNA in live cells, an integrated lacI-GFP fusion protein that binds lacO repeats integrated in the vicinity of centromere 1 (Nabeshima et al., 1998) was used in combination with Cut12-CFP and the mutants rik1, nuf2-1 and mis6. 73% and 59% respectively of nuf2 and mis6 mutant cells showed an abnormal positioning of the cen1::lacO-LacI-GFP signals i.e. cen1 far away from SPB, whereas cen1 positioning was normal, at the SPB, in all wild-type and rik1 cells (Fig. 4A). To test if the declustered centromeres in the mis6 mutant maintained attachments to the nuclear periphery, we immunostained the Pom152-GFP nuclear envelope marker (Bjerling et al., 2002) and FISH stained cen (otr), which detects all three centromeres. This experiment showed that most of the declustered centromeres in mis6 cells were still preferentially found at the nuclear periphery, on average 0.30 μm from the nuclear rim (Fig. 4B,C; see also supplemental Fig. S1). Thus, since mis6 is required for association of Ndc80 to the centromere (Fig. 2A) we conclude that kinetochore components are normally required for anchoring centromeres to the SPB, but an unknown, independent mechanism may be responsible for peripheral positioning of the declustered centromeric heterochromatin domains, even in the absence of kinetochore components.
Consequences in mitosis
Mis6 has previously been shown to influence spindle length, as shown by the increase of metaphase spindle length in mis6 mutants, which may be caused by an imbalance between the kinetochore forces that are pulling and the kinetochore-independent forces that are pushing on the centromeres (Goshima et al., 1999). To visualize mitotic defects in mis6 and nuf2-1 cells we monitored cen1::lacO-LacI-GFP and the SPB marker Cut12-CFP by time-lapse microscopy. We found that both mis6 and nuf2-1 mutant cells displayed severe defects in movement of sister-centromeres in relation to the SPB compared with wild-type control cells (Fig. 5). In contrast, the majority of rik1 mutant cells displayed similar movements of sister-centromeres to those of wild-type cells (data not shown). To characterize the behavior of centromeres in mis6 and nuf2-1 cells we measured the distance from one SPB to the other and the distance from one SPB to the centromeres of the two segregating cen1::lacO-LacI-GFP signals (cen1 and cen1*, respectively). Table 2 shows that on similar length, 4.0-5.0 μm, spindles the average distance between the one SPB and the closest centromere (cen1) was relatively similar in wild-type mis6 and nuf2-1 cells (0.5±0.31, 1.72±0.42 and 0.74±0.08 μm, respectively). However, on similar length 4.0-5.0 μm spindles the furthest-away sister centromere, cen1*, was 3.85±0.78 μm from the same SPB in wild-type cells, 1.93±0.46 μm in mis6 cells and 1.1±0.36 μm in nuf2-1 cells. Thus, whereas wild-type centromeres were segregated by at least 3.5 μm, centromeres had not significantly segregated from each other in mis6 and nuf2-1 cells. This severe mis-segregation phenotype is in accordance with the phenotypes reported earlier for mis6 and nuf2-1 (Saitoh et al., 1997; Nabetani et al., 2001). The fact that the two phenotypes are so similar indicates that loss of kinetochore components from centromeres in mis6 is functionally equivalent to the compromised kinetochore structures in nuf2-1. The consequence in both mutants is likely the occurrence of monotelic attachments in which only one functional sister kinetochore binds kinetochore microtubules or completely unattached sister kinetochores, resulting in spindle-kinetochore force imbalance and rapid spindle elongation as previously observed (Goshima et al., 1999). In contrast, the defects in the heterochromatin domain caused by rik1 lead to decondensation of the heterochromatin. This decondensation had no effect on clustering and caused no dramatic effect on sister kinetochore behavior (data not shown) except for the previously observed lagging chromosome phenotype (Ekwall et al., 1996). This phenotype is likely caused by a cohesin defect, since Rik1 is required to localize Swi6 to centromeres (Ekwall et al., 1996) and Swi6 is required in its turn for the recruitment of cohesins to the centromeric heterochromatin (Bernard et al., 2001; Nonaka et al., 2002). Interestingly, we also observed a decondensation of centromeric heterochromatin in rik1 cells (Fig. 2D). In budding yeast, cohesins have been shown to regulate condensin function after condensins are associated with chromatin (Lavoie et al., 2002). Therefore, we speculate that the decondensation observed in rik1 is explained by loss of Swi6 causing a loss of cohesins and hence misregulation of condensins. In our experiments decondensation or cohesion defects were not observed in transacting mutations that affect central core or kinetochore. In contrast, these mutants showed defects in clustering of centromeres at the SPB and severe mitotic defects. Thus, the kinetochore and the central core are functionally distinct from the flanking heterochromatin domain, both with respect to the centromere function in interphase and the mitotic behavior of sister centromeres.
Movies available online
K.E. is a Royal Swedish Academy of Sciences Research Fellow supported by a grant from the Knut and Alice Wallenberg Foundation. H.A. was the recipient of a VR-NT post doctoral grant. We thank R. Silverstein for critical reading of the manuscript. We thank M. Yanagida and Y. Hiraoka and I. Hagan for the generously providing us with strains. We thank K. Hultenby and S. Mengarelli (Huddinge Hospital, EM unit) for technical assistance with HPF and EM. This project was supported by Strategic (SSF-JIG) and Medical Research Council (VR-M 31X-12562) grants to K.E.