Homologous pairing and the role of pairing centers in meiosis

During meiosis, accurate segregation of homologous chromosomes relies on pairing of homologs to form so-called bivalents that interact with the meiotic spindle as a unit, enabling homologous centromeres to orient to opposite poles (Box 1; Fig. 1). In most eukaryotes, the formation of bivalents requires both homologous recombination and synapsis (Boxes 2 and 3). During the formation of bivalents, homologs usually enter meiosis unpaired and ‘search’ for homologous sequences during leptotene (Roeder, 1997; McKee, 2004). Chromosome synapsis initiates during zygotene and then extends from these initiation sites so that, by pachytene, homologous chromosome axes are fully aligned and synapsed (Page and Hawley, 2004). Once recombination is completed, the synaptonemal complex (SC) is disassembled, but homologs remain connected along their arms, through sister chromatid cohesion, and at discrete sites known as chiasmata until anaphase I (Carpenter, 1994). Chiasmata, in conjunction with sister chromatid cohesion, enable homologs to orient to opposite poles on the meiosis I spindle (Fig. 1).

Although synapsis and meiotic double-strand breaks (DSBs) are required for homologous pairing in most organisms, both DSB-independent and synapsis-independent meiotic segregation pathways have also been described (Zickler, 2006). In Bombyx mori (domesticated silkworm) females, synapsis occurs without crossovers and a modified form of the SC substitutes for chiasmata (von Wettstein et al., 1984). In Drosophila females, which utilize chiasmata to connect their three large chromosome pairs, pairing and segregation of the small fourth chromosomes proceeds without crossovers or chiasmata (Hawley and Theurkauf, 1993). By contrast, all four chromosome pairs in Drosophila males form stable bivalents in the absence of recombination, chiasmata or a SC (McKee, 1996).

Comprehensive and detailed studies in yeast and other model eukaryotes have revealed much detail on the mechanisms of synapsis and recombination, and these topics are summarized in many excellent reviews (Roeder, 1997; Page and Hawley, 2004; San Filippo et al., 2008; Inagaki et al., 2010). In this Commentary, we focus on ‘pairing’, the still largely mysterious process by which homologs find each other and form initial connections, and we will pay particular attention to the roles of pairing centers (PCs) in this process.

Before homologous chromosomes recombine and form a bivalent, they must find each other within the cell nucleus. In most organisms, the initiation of homologous pairing occurs at numerous sites along chromosomes by a mechanism that still remains unclear. These early interactions are then stabilized only at sites where there is good flanking homology between chromosomes. In many organisms, this sorting and stabilizing process appears to be promoted by a meiosis-specific organization of chromosomes called the ‘bouquet configuration’, which is initiated by a clustering of telomeres on the inner nuclear envelope. The bouquet appears to facilitate homologous recognition and alignment by concentrating chromosomes within a limited region of the nuclear volume, thus enabling chromosome movements that promote the identification of homologs, perhaps by the DNA DSB repair process (Box 3) (Hiraoka, 1998; Scherthan, 2001; Harper et al., 2004). These movements are facilitated by the attachment of telomeres to nuclear envelope proteins that contain Sad1 and Unc-84 (SUN) and Klarsicht, ANC-1 and Syne-1 homology (KASH) domains. The SUN–KASH bridge interacts with specific elements of the cytoskeleton,

such as dynein and kinesin, and provides a connection to cytoskeletal forces for moving chromosomes (Fridkin et al., 2009). An extreme example is observed in Schizosaccharomyces pombe in which a tight bouquet forms near the spindle pole body in early prophase I, which drags the whole nucleus back and forth several times within the cell, forming elongated horsetail nuclei (Chikashige et al., 1994; Scherthan et al., 1994). Studies in live yeast cells, in which specific loci on the chromosome arms are labeled or when GFP-tagged Rap1, a telomere-associated protein, is used to label telomeres, have shown that oscillatory chromosome movements promote alignment of homologous chromosomes in early meiotic prophase (Ding et al., 2004; Trelles-Stricken et al., 2005). By contrast, mutants that are defective in bouquet formation, such as taz1 (for telomere associated in Schizosaccharomyces) and bqt2 (for telomere bouquet protein 2) mutants in S. pombe (Cooper et al., 1998; Davis and Smith, 2006) and pam1 (for plural abnormalities of meiosis 1) mutants in maize (Golubovskaya et al., 2002), exhibit a reduction in homologous pairing. These findings support the notion that bouquet formation facilitates homologous recognition and pairing.

Most organisms appear to use the type of pairing pathway described above, in which the telomere-led bouquet configuration facilitates presynaptic alignment, with the alignment stabilized by a combination of DSB repair and synapsis (Fig. 1). However, an alternative method for initiating chromosome pairing, which involves specialized pairing sites, has been described in both Drosophila and C. elegans.

The existence of specialized pairing sites in C. elegans was initially inferred from the effects of reciprocal translocations – chromosome rearrangements involving the exchange of chromosome segments between two non-homologous chromosomes (Fig. 2A) – on the frequency of recombination. In individuals that are heterozygous for such a translocation, recombination is severely suppressed to one side of each translocation breakpoint but is elevated on the other side of the breakpoint (Rosenbluth and Baillie, 1981; McKim et al., 1988; McKim et al., 1993). Similar behavior has also been reported for other types of rearrangements, such as deletions and duplications (Herman and Kari, 1989; McKim et al., 1993; Villeneuve, 1994). For example, duplications of the right end of the X chromosome rarely recombine with the homologous region of the normal X chromosome, whereas duplications of the left end of the X chromosome engage in recombination frequently (Herman and Kari, 1989). These findings suggest that the homologous pairing capacity (i.e. information enabling homologous chromosomes to pair and recombine) is restricted to one end of each chromosome, and this observation has led to the mapping of homolog recognition regions (HRRs) or PCs (the term we will use in this Commentary) near one end of each chromosome. Recent findings have verified the notion that autonomous homologous pairing

capacity is restricted to one end of each chromosome. In translocation heterozygotes, all chromosomes synapse as bivalents even though two of the pairs are therefore homologously synapsed only over part of their lengths. In these mismatched pairs, the ends of chromosomes that contain the PC synapse homologously, whereas the non-PC ends synapse non-homologously (Fig. 2B) (MacQueen et al., 2005). PCs have crucial roles in the homolog-pairing pathway, which is supported by the observation that deletion of both copies of a PC from homologs severely disrupts their recombination and segregation (Villeneuve, 1994; MacQueen et al., 2005).

Detailed analyses have demonstrated two roles for PCs. First, they act locally to stabilize homolog alignment in a synapsis-independent manner (MacQueen et al., 2002; MacQueen et al., 2005). In the absence of synapsis (i.e. in animals depleted of essential SC components) transient pairing occurs at all tested chromosome sites during leptotene and zygotene. The ends of all chromosomes that contain the PC remain paired throughout prophase I, whereas sites that are distant from PCs are largely unpaired by mid-pachytene. This suggests that PCs function to locally stabilize an earlier chromosome-wide pairing process and, indeed, deleting PCs does eliminate this preferential stabilization. A second role of PCs is to initiate synapsis, a process that, once initiated, is largely homology independent. These roles are apparently independent of each other as synapsis occurs even in PC-deletion heterozygotes, in which the PC lacks a pairing partner (MacQueen et al., 2005).

Each PC is bound by one of four zinc-finger proteins, HIM-8, ZIM-1, ZIM-2 and ZIM-3, which are encoded in a single gene cluster. Two of these proteins bind in a chromosome-specific manner; HIM-8 binds to the PC on the X chromosome and

ZIM-2 binds to the PC on chromosome V. The other two proteins bind to PCs on two different chromosomes – ZIM-1 to both the chromosome II and III PCs, and ZIM-3 to the PCs on chromosomes I and IV. Mutations in the genes that encode these proteins result in the expected chromosome-specific phenotypes. For example, him-8 mutations disrupt X chromosome pairing, recombination and segregation but do not affect the meiotic behavior of autosomes. Interestingly, the phenotypes of him-8 mutations are subtly, but consistently, more severe than the phenotypes that result from the deletion of the X chromosome PC, suggesting that HIM-8 acts at other sites in addition to the PC (Phillips et al., 2005; Phillips and Dernburg, 2006).

Specific, but similar, target sequences for each of the zinc-finger proteins have recently been identified and found to be enriched in the PC regions of the appropriate chromosomes. These sequences are repeats of varying length and spacing that all have similar 12-bp core sequences, which have been shown to recruit the cognate zinc-finger proteins to their specific chromosomal target sites (Phillips et al., 2009). For example, deletion of the X chromosome PC abrogates recruitment of HIM-8 to the X chromosome. Insertion of arrays of target sequences onto a PC-deficient X chromosome restores HIM-8 recruitment and PC function (Phillips et al., 2009). It remains unclear whether the target sequences have any function other than recruitment of the zinc-finger proteins.

Once the zinc-finger proteins are recruited to PCs, the resulting protein–PC complexes attach to the nuclear envelope by interacting with the SUN-domain-containing protein SUN-1 and the KASH-domain-containing protein zygote defective protein 12 (ZYG-12) to form a bridge spanning the nuclear envelope. This mechanism is very similar to that mediated by the telomere bouquet and is generally considered to be a variant of it (Penkner et al., 2009; Sato et al., 2009). SUN-1 is required for movement of chromosome ends and forms dynamic aggregates at the sites of PC attachment to the nuclear envelope. ZYG-12 is necessary to localize dynein, a cytoskeletal motor protein, to the nuclear envelope. Subsequently, the PC–SUN-1–ZYG-12 complex moves chromosomes along the nuclear envelope using dynein-dependent microtubule forces (Fig. 2C). This movement is thought to facilitate homologous recognition and synapsis during early prophase. Besides mediating chromosome movements, SUN-1 and ZYG-12 also cooperate to inhibit initiation of synapsis between transiently associated non-homologous chromosomes (Sato et al., 2009). However, how homology is assessed is still an open question. It has been proposed that dynein is required for SC polymerization. When dynein exerts forces that oppose the association of homologous PCs, the resulting tension might induce a mechanochemical signal through SUN-1 and ZYG-12 that leads to the initiation of synapsis (Sato et al., 2009). Baudrimont and colleagues (Baudrimont et al., 2010) have characterized the dynamic movements of SUN-1–GFP aggregates – the equivalent of chromosomal attachment plaques – and demonstrated that multiple chromosome ends are brought together as SUN-1 foci fuse into SUN-1 patches. When homologous chromosomes encounter each other, sufficient affinity between them is generated to resist the cytoplasmic forces, so that synapsis can then be initiated, whereas cytoplasmic forces rapidly separate non-homologous chromosomes (Fig. 2C). The movement of chromosome ends through these patches continues until all of the homologous chromosomes have paired (Baudrimont et al., 2010).

In Drosophila the X and Y chromosomes share homology for the ribosomal RNA genes (the genes encoding 18S, 5.8S, 2S and 28S rRNAs; also known as rDNA), but are otherwise non-homologous. The rRNA genes are present in tandem arrays of 200–250 copies in the heterochromatin (genetically inactive chromatin) of the X chromosome and near the base of the short arm of the Y chromosome. In male meiosis, deletion of most of the proximal X chromosome heterochromatin, including the rRNA genes, results in a failure of X–Y pairing and high levels of X–Y nondisjunction (McKee, 1996). Insertions of transgenes that contain single complete rRNA genes on such X chromosomes substantially restore X–Y pairing and segregation, indicating that the rDNA functions as the X–Y pairing site (McKee and Karpen, 1990). Mapping studies have revealed that the pairing activity resides in 240-bp sequences that are found in tandemly repeated arrays of six to ten copies upstream of each rDNA transcription unit (Fig. 3A) (McKee et al., 1992). rDNA transgenes that include arrays of these 240-bp repeats restore pairing of rDNA-deficient X chromosomes, whereas rDNA transgenes lacking these repeats do not (McKee, 1996). Thus, the X–Y pairing site comprises the 240-bp rDNA repeats.

The X–Y pairing site is bound by the two proteins Stromalin in Meiosis [SNM; also known as Stromalin-2 (SA-2)] and Modifier of Mdg4 in Meiosis (MNM), which are required for stable homolog pairing and segregation in male, but not female, meiosis (Thomas et al., 2005). Both proteins localize to chromosomes throughout meiosis I until they suddenly disappear at anaphase I, coincident with homolog segregation (Thomas et al., 2005). Thus, these proteins appear to substitute for chiasmata in supporting the association between homologs.

Throughout meiosis I, SNM and MNM colocalize with each other and with the 240-bp repeats on the X–Y chromosome pair (Fig. 3B) (Thomas et al., 2005). Moreover, localization of SNM and MNM to the X chromosome is lost when the rDNA genes are deleted, but restored when transgenic 240-bp repeat arrays are inserted (Thomas et al., 2005; Thomas and McKee, 2007). These findings indicate that the 240-bp repeats function to recruit the SNM–MNM complex to the sex chromosomes. SNM and MNM also localize to autosomes, where they have a role in maintaining pairing of autosomal homologs (Thomas et al., 2005), which is discussed further below.

The role of the C. elegans PC sequences in pairing is not entirely clear. On the one hand, heterologous pairing or synapsis between native chromosomes that share the same PC protein and target sequences, such as chromosomes II and III or chromosomes I and IV, is never observed. On the other hand, multi-copy transgenes comprising large blocks of protein recruitment sequences, where the density of these sequences is much higher than in the wild type, can induce heterologous pairing when located on non-homologous chromosomes (Phillips et al., 2009). Thus, these sites can function as direct pairing sites when artificially concentrated, but they probably do not function in this way on native chromosomes. In the native situation, the PC sequences are interspersed with other unrelated sequences; therefore, these non-PC sequences might function to test for homology. On the basis of this interpretation, PC sequences might function indirectly to promote pairing of nearby sequences, but not to provide the main sites for stable homolog connections (Phillips and Dernburg, 2006). This is consistent with the view that the C. elegans PCs function in a manner similar to telomeres in the bouquet mechanism. As discussed above, telomeres are thought to promote pairing of nearby sequences rather than to provide homolog recognition sites directly.

By contrast, the 240-bp repeats in Drosophila probably function directly as pairing sites, rather than by merely stimulating the pairing of linked non-PC sequences. The best evidence for this comes from studies in which transgenic arrays of the 240-bp repeat were inserted at random sites in the euchromatin (the part of the chromosome most active in gene expression) of an X chromosome deficient for native rDNA. All such insertions were effective in partially restoring X–Y pairing (McKee, 1996). However, because the Y chromosome lacks homology to the X euchromatin it is hard to see how X chromosome euchromatic sequences near the PC insertions could contribute to pairing. A role for nearby non-PC sequences in pairing of normal X and Y chromosomes cannot be ruled out. The 240-bp repeats in these chromosomes are interspersed with longer rDNA transcription unit sequences that are shared between the X and Y chromosomes. These regions could serve as additional sites for pairing interactions, even though those sequences lack autonomous pairing capacity as isolated transgenes (McKee et al., 1992; McKee, 1996).

The PCs in both C. elegans and Drosophila have been shown to function in the stabilization or maintenance of pairing (McKim, 2005). However, the term ‘stabilization’ has different meanings in the two systems. As described above, in C. elegans the PCs act in early meiotic prophase (zygotene) to stabilize initial pairing interactions and promote synapsis (MacQueen et al., 2005). HIM-8 and the ZIM proteins also act early in meiosis, as shown by the timing of the him-8 and zim mutant phenotypes, which appear as early as zygotene, and by the fact that these proteins are removed from chromosomes by the end of pachytene (Phillips et al., 2005; Phillips and Dernburg, 2006). These observations suggest that PC proteins are needed only for pairing and synapsis and are dispensable for later steps in the homolog segregation pathway.

By contrast, analysis of tagged autosomal loci in mnn and snm mutants in Drosophila reveals that SNM and MNM are dispensable for pairing in early prophase; there is no diminution in pairing frequencies in these mutants relative to the wild type (Thomas et al., 2005). Instead, mnm and snm mutations disrupt chromosome behavior from mid-prophase when chromosome territories – chromosome-specific nuclear domains that contain both homologs of a bivalent – appear more diffuse when compared with that in the wild type. Subsequently, when chromosomes condense at prometaphase I, they do so as univalents in snm and mnm mutants (Thomas et al., 2005). Moreover, SNM and MNM are retained on chromosomes until anaphase I, indicating a much later role for these proteins in pairing maintenance than that of the HIM-8 and ZIM proteins. Thus, both SNM and MNM and the HIM-8 and ZIM proteins function to stabilize pairing, yet they do so at distinct stages of the homolog segregation process.

The findings described above show that PCs can perform essential roles in meiotic chromosome pairing. An interesting question is whether such roles are confined to specific isolated cases or whether PCs are general phenomena. In light of the compelling evidence for a PC on the X–Y chromosome pair in Drosophila, an obvious question is whether PCs contribute to pairing of other Drosophila chromosomes, in either male or female meiosis.

In light of the evidence for PC-directed X–Y chromosomal pairing, it has also been suggested that pairing of autosomes in Drosophila male meiosis involves specific sites (Vazquez et al., 2002). However, numerous studies involving diverse techniques have failed to provide any convincing evidence for such sites on autosomal chromosomes 2 and 3, which together account for ~80% of the Drosophila genome. These studies, which have been extensively reviewed elsewhere (McKee, 1998; McKee, 2004), demonstrate that the euchromatic regions of chromosomes 2 and 3 pair at multiple interstitial sites in early prophase and that heterochromatic regions lack autonomous pairing capacity. However, they do not rule out the possibility of specific non-autonomous pairing sites (i.e. sites at which connections depend upon prior alignment of homologs in linked euchromatic regions) in centric heterochromatin. A recent fluorescent in situ hybridization (FISH) analysis, using probes to chromosome-specific repeated sequences, failed to detect any such stable pairing sites (i.e. sites that remain paired throughout meiosis I) in the heterochromatic regions of the major autosomes (Tsai et al., 2011), thus providing further evidence against PC-based pairing of the major autosomes. However, the small fourth chromosomes did remain paired at a specific heterochromatic site in >90% of spermatocytes throughout prophase I, suggesting that the fourth chromosome contains a PC. It remains to be determined whether fourth chromosome pairing will map to a specific site or whether it is a chromosome-wide phenomenon.

The failure to detect PCs on the major autosomes during Drosophila male meiosis leaves unanswered the important question of how these homologs remain connected after the loss of intimate allelic pairing at the mid-G2 transition. The autosomal homologs share a common territory throughout mid- and late-prophase I and condense into well-aligned bivalents at prometaphase I; thus, this indicates that an unknown factor could keep them together. SNM and MNM are involved in this process, as mutations in both lead to a loss of territory definition (Thomas et al., 2005). SNM and MNM are also observed within autosomal chromatin (Thomas et al., 2005) (J.-H.T., unpublished results), but the binding sites of SNM and MNM on autosomes remain undefined. Furthermore, recruitment of MNM, and perhaps SNM, to autosomes depends upon the Teflon (TEF) protein, which is required for segregation of the autosomes but not the sex chromosomes (Tomkiel et al., 2001; Thomas et al., 2005). We have suggested, by analogy to chiasmata in recombinational meiosis, that stable connections between autosomal homologs exist at different sites in different meiotic cells (Tsai et al., 2011). Time-lapse analyses in living spermatocytes using GFP-tagged chromosomal sites could be useful for revealing such stable connections; a stable connection site that happened to lie sufficiently close to a tagged chromosomal site should restrict the relative mobility of the tagged homologous alleles, perhaps dramatically so in favorable cases.

Until recently, Drosophila females were also thought to utilize specific sites to pair their chromosomes, but recent findings have discredited this idea. As in C. elegans, the ‘pairing sites’ in Drosophila females were identified in flies that were heterozygous for reciprocal translocations. Females that were heterozygous for X; 4 translocations, carrying one normal X chromosome and one that is broken into two pieces each attached to a portion of the tiny fourth chromosome, were utilized. In each genotype in the original study, X recombination was found to be suppressed in a distinct interval around the translocation breakpoint but occurred at normal frequencies elsewhere on the X chromosome (Hawley, 1980). Analysis of several such translocations suggested that the X chromosome was subdivided into three discrete autonomously recombining regions defined by four widely distributed ‘boundary sites’. As the translocation breakpoints disrupted recombination only within the region they interrupt, it was thought that these sites functioned as alignment sites and that adjacent pairs of boundary sites are required to be in cis in order to mediate alignment of the intervening region (Hawley, 1980). A recent study found similar behavior for another group of translocations, which led to the mapping of two boundary sites on chromosome arm 3R (Sherizen et al., 2005). However, molecular analysis of pairing in females that were heterozygous for these 3R translocations revealed that both pairing and synapsis in the recombinationally suppressed regions still occurred at normal frequencies (Sherizen et al., 2005). Normal levels of pairing and synapsis were also observed in females that were heterozygous for a normal sequence X chromosome and a multiply rearranged balancer X chromosome, a genotype in which recombination is completely suppressed (Gong et al., 2005). Thus, if the boundary sites identified in the previous studies do have roles in pairing or synapsis, then this role appears to be more subtle than originally hypothesized. At least within euchromatic intervals, alignment and synapsis in Drosophila female meiosis apparently does not rely on specific defined sites but rather on multiple interactions throughout homologous regions.

As described above, in Drosophila, analyses of pairing have failed to identify any additional PCs and have instead indicated that general homology pairing is the predominant pairing mechanism. Moreover, nothing similar to the PCs of C. elegans or the X–Y chromosome pair in Drosophila has been described in other organisms. Nevertheless, as summarized below, phenomena suggestive of PC-like properties have been described for specialized chromosomal sites, including nucleolus organizer regions (NORs), telomeres and centromeres. As noted above, telomere meiotic function above has similarity to the functioning of C. elegans PCs. However, the data on centromere pairing are particularly intriguing and will be analyzed in some depth below.

In general NORs are not thought to have prominent roles in pairing or synapsis. In organisms in which preferential synapsis initiation sites have been mapped, these sites do not coincide with NORs (Page and Hawley, 2004). Indeed in budding yeast, NORs are apparently excluded from synapsis (Tsubouchi et al., 2008). However, PC-like behavior has been reported for NORs in ahp2 mutants of Arabidopsis thaliana. AHP2 is a homolog of the homologous-pairing protein 2 (HOP2), which is conserved among yeast, animals and plants and has been shown, in several organisms, to be required for proper homolog partner choice. In Arabidopsis, the ahp2 mutation was found to severely disrupt meiotic pairing and synapsis at most genomic sites. However, the short arms of chromosomes 2 and 4, where the two NORs are located, exhibit normal pairing frequencies and normal SC formation. These findings indicate that the NORs act as cis-acting pairing and synapsis initiation sites in ahp2 mutants (Stronghill et al., 2010), in a manner reminiscent of C. elegans PCs. The extent to which NORs contribute to pairing of chromosomes 2 and 4 in wild-type plants, and whether NORs exhibit similar behavior in other organisms, still remains to be determined.

Centromeres have been reported to pair during meiosis in a wide variety of organisms (reviewed by Stewart and Dawson, 2008). In addition to clustering of both homologous and non-homologous centromeres, which is a common feature of pre-meiotic and early meiotic nuclei, the pairwise association of centromeres before the general onset of pairing and synapsis has also been observed in budding yeast and wheat (Martinez-Perez et al., 1999; Tsubouchi and Roeder, 2005). In both of these cases, however, early pairwise associations are not homologous but instead involve apparently random ‘couplings’ of centromeres, with the pairings becoming homologous as cells proceed through meiotic prophase. However, this transition appears to be driven by homologous interactions initiated in other chromosomal regions rather than by any homologous interactions of the centromeres themselves. FISH analyses in wheat show that telomeric and sub-telomeric regions pair earlier than centromeres in meiosis and that the transition from non-homologous to homologous centromere associations is driven by the progression of synapsis from the telomere towards the center of the chromosome. This suggests that the homology at specific sequences near telomeres, rather than at centromeres, is involved in the correct recognition and selection of partners (Corredor et al., 2007). Moreover, when the wheat centromeres are replaced with those from the corresponding rice chromosomes there is no effect on the pairing patterns in wheat nuclei, indicating that centromeres have no role in the sorting of homologous from non-homologous chromosomes. Similarly, interchromosomal ‘centromere swaps’ have no effect on meiotic chromosome pairing and segregation in budding yeast (Clarke and Carbon, 1983).

Remarkably, however, in budding yeast, these non-homologous centromere couplings seem to have an important role in synapsis. During zygotene (by which time non-homologous couplings have been largely replaced by homologous associations), a majority of the segments of the polymerized SC central element proteins (including the molecular zipper ZIP1) either have one end at a centromere or incorporate a centromere within them. This is consistent with the idea that synapsis frequently initiates at centromeres and can propagate either unidirectionally or bidirectionally (Tsubouchi et al., 2008). Moreover, ZIP1 localizes to centromeres before the onset of general synapsis, and the early non-homologous centromere couplings are completely dependent on ZIP1 (Tsubouchi and Roeder, 2005). Thus, centromeres in yeast share some similarity to PCs in C. elegans and initiate synapsis, but, unlike PCs, they do not appear to have a main role for pairing partner identification.

Centromere pairing has also been reported later in meiotic prophase, after the disassembly of the SC in a number of organisms (Stewart and Dawson, 2008). In budding and fission yeast and Drosophila females, these interactions are important for segregation of achiasmate chromosomes (i.e. those that lack chiasmata). In budding yeast, the centromeres of achiasmate chromosomes pair with each other during late prophase, irrespective of whether the chromosomes are homologs or non-homologs, and these chromosomes segregate preferentially to opposite poles with moderate efficiency (Kemp et al., 2004). Recent evidence shows that these late-prophase centromere couplings, like the early-prophase couplings described above, require ZIP1. Moreover, loss of ZIP1 randomizes segregation of achiasmate chromosomes (Gladstone et al., 2009). In the same study, it was also found that centromeres of homologous chromosomes pair in late prophase and that this pairing promotes the orientation of homologous centromeres to opposite poles. Thus, the budding yeast centromere provides an example of a chromosome pairing site with important roles in both synapsis and segregation, but which does not contribute directly to homologous partner choice.

In Drosophila females, centromeric heterochromatin regions pair throughout meiotic prophase and this pairing serves to promote the segregation of achiasmate homolog pairs (Dernburg et al., 1996; Karpen et al., 1996). Unlike in yeast, achiasmate chromosome segregation in Drosophila is at least partly homology driven [i.e. non-exchange X chromosomes segregate preferentially from other non-exchange X chromosomes, rather than from a non-exchange non-homolog (Hawley et al., 1992)]; this is also more efficient as it yields segregation frequencies of ~100% in many cases. However, achiasmate centric pairing in Drosophila is not limited to centromeres. Mapping studies have shown that the pairing ability is diffusely distributed throughout large tracts of centric heterochromatin and that pairing frequency depends on the length of heterochromatic homology (Hawley et al., 1992; Karpen et al., 1996). The involvement of such extensive regions probably explains the homology dependence of achiasmate segregation in Drosophila and could also contribute to its high efficiency. Thus, although the Drosophila case provides the only compelling example in which centric pairing drives chromosome assortment and segregation on a homologous basis, it is unclear what role the centromeres themselves have in this process. An interesting possibility is that centromere associations do occur, perhaps non-homologously, as in budding yeast, but that these serve to promote homology testing and, eventually, enable the formation of stable connections within flanking heterochromatic domains.

Interestingly, Drosophila also provides what is perhaps the only clear example of truly homologous centromere pairing (as opposed to centric heterochromatic pairing), but in male rather than female meiosis. Using a GFP-tagged centromere protein to visualize centromeres in live cells, centromeres have been found to cluster non-specifically in early prophase, when euchromatic sequences are tightly paired, then to sort into pairwise and strictly homologous associations shortly after the loss of homologous pairing in the chromosome arms (Vazquez et al., 2002; Yan et al., 2010). A recent FISH study demonstrated that this pairing is centromere-specific and does not extend even into very nearby pericentromeric heterochromatin (heterochromatin situated near to a centromere) (Tsai et al., 2011), and thus could be an example of PC-like behavior. However, homologous centromere pairing is short-lived; centromeres become unpaired by mid-prophase I and remain unpaired throughout the remainder of meiosis I. The functional significance of these transient pairings and the basis for the homology dependence is unknown. Centromeres of different Drosophila chromosomes appear not to share DNA sequence homology (Sun et al., 2003), so there could be a sequence basis for such specificity. Alternatively, the specificity could be entirely adventitious, driven by the homologous pairing of linked arms earlier in prophase. Centromere pairing occurs shortly after the homologous chromosome pairs have resolved into separate nuclear territories, so that, when they pair, it is probable that a centromere only has access to the centromere of its homolog. It will be of interest to determine how centromeres pair in experimental situations in which both homologous and non-homologous pairing partners are available.

Overall, the evidence indicates that centromeres pair actively and specifically with each other, but that such pairings are generally not homology driven. Centromere pairing can nevertheless play important roles in synapsis and homolog segregation.

We have described two prominent examples of the use of PCs to mediate homologous pairing. In both cases, the PCs function as recruitment sites for specialized pairing proteins. However, the pairing proteins and sites function somewhat differently in the two systems. In C. elegans PC-mediated homologous pairing is analogous to the telomere-led bouquet in mediating the pairing of linked sequences and the initiation of synapsis between homologs. Drosophila males, which lack recombination, synapsis and chiasmata, have evolved a specialized pairing site for the otherwise non-homologous X–Y chromosome pair and a unique protein complex containing SNM and MNM that substitutes for chiasmata, thereby providing stable inter-homolog connections. Surprisingly, however, Drosophila apparently does not utilize specific sites to pair any of their other chromosomes (with the possible exception of the tiny fourth chromosome pair) in either sex. Instead, general homology pairing appears to underlie homolog alignment and partner choice in both sexes. Important questions about the male achiasmate segregation system remain unanswered. One such question is how homolog alignment is translated into stable inter-homolog connections, particularly on the major autosomes, which appear to be unconnected throughout most of meiotic prophase despite occupying a common territory. Another is how SNM and MNM are recruited to autosomes. Finally, it remains to be determined how specific connections between the four different pairs of homologs are mediated by SNM and MNM. How chromosome specificity is achieved is also an important unanswered question in the C. elegans system. The precise role of the PC sequences and PC proteins in homolog pairing also remains to be established.

Overall, there is little evidence that PCs are widely used as a solution to the pairing problem. In most organisms, cytological and molecular evidence points to multiple sites along chromosomes that are able to initiate homologous interactions (Bozza and Pawlowski, 2008; Roeder, 1997). Nevertheless, most organisms do seem to rely on specific sites to help with various aspects of chromosome pairing, synapsis and segregation. Telomeres do not pair directly but do have a prominent role in homolog partner choice in many organisms, a role that seems remarkably similar, mechanistically, to that played by PCs in C. elegans. However, centromeres often do pair directly with each other, albeit non-homologously. Centromere pairing has been shown to contribute to synapsis initiation, centromere orientation and achiasmate segregation but, with the possible exception of Drosophila females, probably not to homologous partner choice. Further research in a variety of model organisms should shed light on the relationships among these diverse pairing systems.