Temporal Characterization of Homology-Independent Centromere Coupling in Meiotic Prophase

Background Over the past thirty years several reports of the pairing or association of non-homologous centromeres during meiotic prophase have appeared in the literature. Recently, the homology-independent pairwise association of centromeres, termed centromere coupling, was also reported in budding yeast. It seems paradoxical that centromeres would pair with non-homologous partners during a process intended to align homologous chromosomes, yet the conservation of this phenomenon across a wide range of species suggests it may play an important role in meiosis. Principal Findings To better define the role of this phenomenon in budding yeast, experiments were preformed to place centromere coupling within the context of landmark meiotic events. Soon after the initiation of the meiotic program, centromeres were found to re-organize from a single cluster into non-homologous couples. Centromere coupling is detected as soon as chromosome replication is finished and persists while the recombination protein Dmc1 is loaded onto the chromosomes, suggesting that centromere coupling persists through the time of double strand break formation. In the absence of the synaptonemal complex component, Zip1, centromere coupling was undetectable, at all times examined, confirming the essential role of this protein on this process. Finally, the timely release of centromere coupling depends on the recombination-initiating enzyme, Spo11, suggesting a connection between events in homologous pairing/recombination and the regulation of centromere coupling. Conclusions Based on our results we propose a role for centromere coupling in blocking interactions between homologous centromeres as recombination initiation is taking place.


Introduction
In order to correctly segregate from one another at meiosis I, homologous chromosomes need first to become associated. In most organisms the selection of the segregation partners is based on sequence homology and recombination along the chromosome arms, and not the centromeric regions, as manipulation of homologous chromosomes so that they have non-homologous centromeres has little deleterious effect on their segregation behavior [1,2].
In many organisms, including budding yeast, crossing-over is repressed near the centromeres [3]. This crossover repression is beneficial, as crossovers in centromere-proximal regions of budding yeast, Drosophila and humans are associated with elevated levels of chromosome segregation errors in meiosis [4,5].
Over thirty years ago experiments in onion meiocytes revealed that the centromeres became organized in non-homologous pairs or small groups, prior to the period of homologous chromosome synapsis [6]. Subsequent similar observations in other species showed that the association of non-homologous centromeres during meiotic prophase is a widespread phenomenon (reviewed in [7]). Recently the phenomenon was reported in budding yeast where it was demonstrated that the centromeres of the thirty-two chromosomes form sixteen pairwise associations, usually between non-homologous partners, at a period prior to the alignment of homologous chromosomes [8]. To differentiate this homologyindependent centromeric association from the regular pairing between homologous chromosomes this phenomenon has been referred to as centromere coupling [8]. In budding yeast, centromere coupling has been shown to require the protein Zip1 [8]. Zip1 constitutes the central element of the synaptonemal complex (SC), and is necessary to bridge the lateral elements that assemble along the cores of the chromosomes, thus zippering the homologous partners together [9,10]. The exact role of Zip1 in centromere coupling is not known but it is tempting to think that it may perform a similar role as it plays for SC formation, directly bringing non-homologous centromeres together by its ability to self-associate.
Why, in many organisms, do centromeres become coupled with non-homologous partners when meiosis has evolved a series of elaborate mechanisms to ensure the pairing of homologous chromosomes? To help address this question we performed time course experiments to place the period of centromere coupling within the context of other meiotic landmark events, an approach that has not been taken by previous studies. The results favor a new model for the function of centromere coupling.

Centromere coupling dynamics in WT cells
To characterize centromere coupling we compared the kinetics of the formation of non-homologous and homologous centromere pairs in budding yeast cells harvested at intervals from cultures induced to enter meiosis. The kinetochore protein Mtw1 was epitope-tagged, allowing us to score the number of kinetochore foci in spread nuclei. Diploid yeast cells have sixteen homologous chromosome pairs. If the centromeres of the chromosomes are arranged in pairs, each nucleus will exhibit sixteen Mtw1 foci. As described previously [11,12] we observed that the centromeres are tightly clustered early in meiosis. Chromosome spreads typically exhibited fewer than ten Mtw1 foci at the time of meiotic induction (Fig. 1A, T = 0), about two hours later a population appeared that exhibited about sixteen foci (Fig. 1A, T = 2-T = 5), consistent with pairing of the centromeres. In order to quantify the emergence and disappearance of different types of centromere organization we categorized nuclei with fewer than twelve foci as ''clustered'', nuclei with twelve to twenty foci as ''paired'' and nuclei with more than twenty as ''dispersed'' (Fig. 1A).
Cells with about sixteen Mtw1 foci could be engaged a period of centromere coupling or homologous alignment of the chromosomes (or a transition between the two). To distinguish between these possibilities both copies of chromosome I were GFP-tagged at their centromeres (CEN1) [13]. When homologous chromosomes become synapsed, the GFP-tags of the two CEN1's will be juxtaposed; yielding one single or two closely paired fluorescent dots that co-localize with a single Mtw1 focus. Conversely, centromere coupling has been shown to occur largely between non-homologous partners [8] so during a period of centromere coupling, the two CEN1 GFP-tags will usually be associated with different Mtw1 foci.  non homologous coupling, to homologous alignment as cells proceed through meiotic prophase, a progression similar to that described for centromeres in wheat (reviewed in [7]). Throughout the time course, spreads with more than twenty Mtw1 foci (dispersed) were rare (Figs. 1B and 1C). In our experiments the exact timing of meiosis varied somewhat between isogenic strains and also in independent repetitions using same strain ( Fig 1C). Thus, we measured SC formation/disassembly during the multiple time courses (Fig 1C, linear Zip1). Despite small variations from one repetition to the next, the relative order of centromere coupling with respect to pairing and the SC formation is always identical (Fig 1C). In some time courses the proportion of clustered cells increased at late time points (Fig. 1C, red lines). This occurred in samples that entered meiosis more quickly after the switch to meiosis inducing medium ( Fig 1C). The increase in clustered cells, is probably attributable in part to the recently described reclustering of the centromeres at the side-by-side spindle pole bodies right after pachytene [14], and also to the fact that once cells became bi-nucleate in these experiments, they were excluded from the analysis. One population of mononucleate cells are those cells that do not progress through the meiotic program and never form asci (typically about 20% of the cells with the strains used here). These cells remain as mononucleate cells with clustered centromeres and no Zip1 staining, and represent an ever-growing proportion of the mononucleate cell population as cells that have entered the meiotic program become multinucleate.

Centromere coupling and the appearance of Zip1
Centromere coupling has been shown to require the SC protein Zip1 [8]. We therefore compared the timing of centromere coupling with the timing of Zip1 loading onto the chromosomes. 90% (n = 726) of chromosome spreads with clustered centromeres showed no Zip1 staining ( Fig. 1B and Fig. 2). The detection of centromere coupling and localization of Zip1 to the chromosomes appeared simultaneously (Fig 1C, compare ''coupled'' and ''punct. Zip1''). About 75% of cells undergoing centromere coupling exhibited numerous Zip1 punctate foci ( Fig. 1B and 2). The initial study of centromere coupling reported that in chromosome spreads exhibiting centromere coupling, Zip1 foci and kinetochore pairs occur in approximately equal numbers (about sixteen) and exhibit a high level of co-localization [8]. In contrast, in spreads that exhibited punctate Zip1 staining and were undergoing centromere coupling, we observed considerable variation in the total number of Zip1 foci (average 22.4 +/2 S.D. 7.2, n = 20 spreads). This result is consistent with previous studies of Zip1 association with chromosomes in early meiosis [15,16]. Some of the Zip1 foci we detect co-localize with kinetochores (Mtw1-13XMYC) but most do not (29.0% co-localization of Zip1 foci to Mtw1 foci, +/2 S.D. 12.0, n = 20 spreads) ( Fig. 1 B), a result similar to that previously observed by Bardhan et al [17]. The punctate Zip1 signals detected at centromeres in these experiments were not of uniform intensity suggesting that different amounts of Zip1 are localized to different centromeres. The modest Zip1/ Mtw1 co-localization reported here does not necessarily mean that Zip1 is not located at all centromere couples and could be explained if small numbers of Zip1 molecules (not detectable by our staining) are sufficient for centromere coupling. The result also shows that Zip1 localizes to other chromosomal loci besides centromeres in early meiosis.
Finally, about 10% of the cells that were engaged in centromere coupling (CEN1's associated with separate Mtw1 foci) exhibited linear SC (Fig. 2) demonstrating that, in some cells, considerable synapsis occurs before the homologous CEN1's become alignedconsistent with the model that the alignment of chromosome arms generally precedes the alignment of homologous centromeres. As expected, almost all cells with paired homologous CEN1's showed positive Zip1 staining: 50% linear staining, typical of synapsed chromosomes [9], and 42% punctate or patchy staining, corresponding to SC assembly and SC disassembly stages.
Centromere coupling dynamics in spo11 and zip1 mutants Centromere coupling was originally noted in spo11D mutants [8]. The ease of obtaining chromosome spreads in the time point used in those experiments suggests that cells might persist in the centromere coupling stage in spo11D mutants. To test this, we evaluated early centromere behavior in spo11D mutants, which exhibit severe defects in the pairing and synapsis of homologous chromosome arms [18,19]. Whereas in the wild-type strain, the frequency of cells exhibiting centromere coupling begins to diminish by three/four hours in meiosis ( Fig. 1 C), in spo11D mutants, the proportion of cells with non-homologous coupled centromeres continues to rise and persists at all times evaluated (Fig. 3A, spo11D). Thus, the timely release from centromere coupling depends on the presence of Spo11p.
Previous work (single time point experiments) revealed that zip1 mutant cells in late meiotic prophase do not exhibit centromere coupling, suggesting that Zip1 is necessary either for the initiation of centromere coupling or its persistence [8]. To distinguish between these possibilities we evaluated centromere coupling in cells harvested from multiple time points following meiotic induction (Fig. 3A, zip1D). A requirement for Zip1 in initiation of coupling would predict that coupling would not be observed at any time point, whereas a role in maintenance could result in the observation of centromere coupling at early time points that does not persist until late meiotic prophase. We did not detect centromere coupling in zip1D cells at any time point evaluated. Instead, concomitant with the exit from centromere clustering we observed the appearance of chromosome spreads with dispersed centromeres (more than 20 Mtw1 foci) (Fig. 3A, zip1D and 3C). Some of the dispersed spreads we observe in zip1D mutants reflect a role for Zip1 in pairing the centromeres of homologous chromosomes in late prophase [14,19,20,21]. In the absence of ZIP1, homologs become aligned, but because Zip1 is necessary for tight pairing of homologous centromeres in late meiotic prophase,  Figure 1 were classified according to their pattern of Zip1 staining. Zip1 categories: absent (as in the clustered example , Fig 1 B), punctate/ patchy (an example of punctate staining in the Dispersed example, Fig. 1 Fig. 1 B) or linear (an example is the bottom row of Homologous in Fig. 1 B). Clustered (n = 726), dispersed (n = 30), non-homologous paired (n = 186) and homologous paired (n = 139). Error bars: 95% confidence intervals. doi:10.1371/journal.pone.0010336.g002 their centromeres are often slightly separated and thus visualized as side-by-side Mtw1 foci [14,20,21]. Because side-by-side foci are scored as two separate foci in our assay, this yields higher counts of Mtw1 foci. Consistent with this, in the zip1D strain the ''dispersed'' spreads from the four and five hour time points often feature sideby-side CEN1 GFP dots associated with separate but adjacent Mtw1 foci (37% at T = 4 hr, 67% at T = 5 hr) (Fig. 3B). The model that coupling requires Zip1 also predicts that in a spo11D zip1D double mutant, cells exhibiting dispersed, individual, centromeres will accumulate, since the loss of Zip1 should eliminate the centromere coupling and loss of Spo11 should eliminate homologous alignment. This was in fact observed (Fig. 3A, spo11D zip1D); spo11D zip1D double mutants transition from the clustered stage to a dispersed organization of centromeres.

Centromere coupling is co-incident with early events in homologous recombination
Determining the placement of centromere coupling among the landmark events of early meiosis might provide clues to the function of non-homologous centromere associations. Thus we compared the timing of centromere coupling to pre-meiotic DNA replication and the early stages of homologous meiotic recombination.
Flow cytometry was used to monitor DNA replication in time course experiments. The emergence of 4n cells coincided with the release of the centromere clustering and the emergence of cells exhibiting centromere coupling (Fig. 4A). The percentage of 4n cells at the final time point roughly equals the number of cells with their centromeres released from the cluster, indicative of entry into the meiotic program ( Fig 4A).
Formally, the experiments here do not reveal whether centromeres within the cluster are already organized in couples, though the apparent requirement for Zip1 for coupling and the absence of Zip1 from cells in the clustering stage make this model less appealing. The data do demonstrate that when replication has finished clusters are giving way to dispersed couples. To compare the relative timing of centromere coupling and the early steps of homologous recombination, we asked whether cells involved in coupling have their chromosomes decorated with Dmc1 ( Fig. 4B  and 4C). Dmc1 loads onto the chromosomes after meiotic DNA double strand breaks (DSBs) are created and prior to chromosomal synapsis, playing a critical role in mediating the formation of early recombination intermediates between homologous chromosomes [22,23]. Chromosome spreads were first classified into one of the four categories described above (clustered, dispersed, coupled and homologous centromeres paired), and then scored for staining with antibodies against Dmc1.
As expected, cells with clustered centromeres were nearly always negative for Dmc1 staining (Fig. 4B and 4C). Nearly all cells (96%) exhibiting homologous centromere alignment were positive for Dmc1 staining (Fig. 4B and 4C), suggesting that homologous alignment of the CEN1's rarely precedes initiation of homologous recombination. It makes intuitive sense that the alignment of homologous chromosomes would be led by a homology-based recombination process rather than by elements (centromeres) that interact in a homology-independent fashion. When Dmc1 staining was evaluated on cells undergoing coupling, 71% were found to exhibit clear positive staining (Fig. 4B and 4C) suggesting that centromeres are still coupled with non-homologous partners when double strand break repair initiates (Fig. 4D). The fact that 29% of the cells exhibiting centromere coupling were negative for Dmc1 suggests that coupling precedes the loading of Dmc1 onto chromosomes.

Discussion
Like several other organisms, budding yeast has been shown to exhibit a period in which centromeres become grouped in a homology independent fashion [8]. Zip1 was shown to play a role in this phenomenon [8,20]. These first studies used a single time point analysis to evaluate centromere-coupling. The analysis revealed key features of the centromere coupling phenomenon in budding yeast, but left unanswered questions related to the timing of the phenomenon, an issue that has been difficult to address in other systems in which homology-independent meiotic centromere interactions have been reported. We have used this approach to directly position centromere coupling in relation to other meiotic events, study its dynamics and better understand the role of Zip1 in this phenomenon.
As demonstrated previously [8], we found that centromere coupling unequivocally takes place before SC formation. The tight centromere clusters that typify cells entering the meiotic program are largely maintained through S-phase and centromeres then transition quickly into non-homologous pairs. It is not clear whether centromeres emerge from the clusters as couples, disperse from the clusters and then become coupled, or both. Individual centromeres clearly do not linger long between clustering and coupling. Cells with these dispersed centromeres were observed, but at very low frequencies. Similarly we do not observe a stage of dispersed centromeres between coupling and homologous pairing. Thus the transition from non-homologous to homologous partners may be brief, or centromere coupling may be dynamic process, as suggested by Tsubouchi and Roeder [8] such that not all the centromeres disengage from their partners simultaneously.
We have studied the dynamics of centromere coupling in spo11D mutants. We have observed that the process is dramatically affected by the SPO11 deletion; centromere coupling persists at all times evaluated. Thus, SPO11 is required, at least indirectly, for the transition from non-homologous coupling to homologous pairing of the centromeres, suggesting a connection between events promoted by Spo11 activity, presumably through the formation of double strand breaks, and a triggered release from coupling after homologous recombination is initiated.
Zip1 is the structural element that holds the lateral elements of the SC together [10]. Zip1 could be imagined to perform a similar role during centromere coupling by bridging the centromeres of non-homologous chromosomes. Our results showing that centromere coupling is abolished at all time points in zip1 deletion mutants are consistent with this sort of direct structural role for Zip1. However, the formal possibility remains that centromeres still couple in zip1D cells, but the engagements are short-lived and not detectable with our experimental approach. If Zip1 is the structural element holding centromeres together, Zip1 should localize to the centromeres when they are coupled. We have observed more variable numbers and intensities of punctate Zip1 foci and a lower incidence of Zip1/kinetchore co-localization than previous studies. It seems from these observations and work of others [15,24] that 1) Zip1 may localize to more than just centromere regions in early meiosis, and 2) that relatively small (undetectable in our assays) amounts of Zip1 may be sufficient to promote coupling.
What is the role of centromere coupling? Centromere coupling has been proposed to promote homologous pairing by holding the centromeres of two chromosomes together while homology at the arms is being assessed [8]. However, ZIP1 deletion mutants do not have dramatic pairing defects along the chromosome arms [25] and some studies suggest that the bringing together of homologous centromeres is dependent upon interactions between the arms [1,2]. The results here suggest an alternative possibility; centromere coupling might be preventing the formation of deleterious crossing-over at the centromere. A recent analysis of the global distribution of meiotic crossovers in budding yeast suggested that Zip1 is necessary for the repression of centromeric crossing-over [26]. The authors concluded that Zip1, in some way, directs DSBs repair towards the sister chromatids. We have shown that centromere coupling initiates prior to the association of Dmc1 with the chromosomes and that Dmc1 loading occurs when centromeres are coupled with nonhomologous partners. This suggests that Zip1 may be manifesting its role in centromere crossover repression by coupling centromeres with non homologous partners. We propose two different mechanisms by which the coupling process might play this role. First, the sequestration of centromeres with nonhomologous partners (Fig. 4D, centromere sequestration) may spatially or topologically prevent interactions between homologous centromere regions, leaving sister chromatids as default partners for repairing DSBs (Fig. 4D, recombination complex marked with an asterisk). Alternatively, (Fig. 4D, crossover inhibition) Zip1 could promote inter-sister over inter-homolog repair of DSBs by blocking the recruitment, or affecting the function, of components of the recombination process that are required for interhomolog events (Fig. 4D, crossover inhibition, orange circles). It has been noted that in early prophase Dmc1 and Zip1 foci are non-overlapping, consistent with the notion that Zip1 deposition may exclude components of the recombination machinery [22]. By this model, centromere coupling might act to increase the polymerization or stability of Zip1 around the centromeres or could be an innocuous by-product of Zip1 assembly properties.

Yeast strains and culture conditions
All strains were obtained by matings of TSP50 and TSP52, or their isogenic derivatives. We used standard yeast culture methods [27]. To induce meiosis, cells were grown in YP-acetate to 3-4610 7 cells per ml, and then shifted to 1% potassium acetate at 10 8 cells per ml.

Strain construction
PCR-based methods were used to create complete deletions of ORFs and epitope-tags [28]. Some deletions were created by using PCR to amplify deletion-KANMX insertions from the gene deletion collection (Invitrogen) and these products were then used for transformations. The plasmid pJN2 targeted 256 lacO repeats to CEN1 (coordinates 153583-154854). Correct integration was confirmed genetically. The P CYC1 -lacI-GFP cassette was inserted as part of pAFS152, a gift from Aaron Straight. Strain genotypes are reported in Table S1.

FACS analysis of DNA replication
FACS analysis and determination of the proportion of 4n cells were performed according to published protocols [30,31]. Time points during the first five hours of meiosis were analyzed. At the five hour time point, 50% of cells had not entered meiosis (50% of the cells retained clustered centromeres and were negative for Zip1 staining). This strain typically exhibits about 80% asci by 48 hours after the switch to meiosis-inducing medium.