Alleles of afd1 dissect REC8 functions during meiotic prophase I

Meiotic prophase is initiated by the formation of a unique and cytologically characteristic leptotene chromosome. The establishment of leptotene chromosome structure relies, at least partly, on the installation of proteinaceous axial elements (AEs) onto the chromosomes. Once leptotene chromosome structure is established, several unique meiotic events occur, including telomeres attaching to the inner nuclear envelope, telomere clustering into a bouquet and recruitment of the recombination machinery. This is followed by homologous pairing and synapsis. As the chromosomes zip up during synapsis, a central element is installed between the AEs (now called lateral elements) of the two homologous chromosomes, completing the formation of the synaptonemal complex (SC). Recombination is probably initiated in leptotene, and some recombination events mature in crossovers during synapsis, yielding chiasmata that, in conjunction with sister chromatid cohesion (SCC), hold maternal and paternal chromosomes together until the transition from metaphase I to anaphase I. Sister chromatids segregate to daughter nuclei during meiosis II. This two-step chromosome segregation relies on a sequential dissociation of SCC.

In meiosis, sister chromatids are held together through the first meiotic division, and rely on meiosis-specific SCC components. In yeast, REC8, a member of the α-kleisin superfamily of SMC protein partners (Schleiffer et al., 2003), is a key component of the cohesion complex and is absolutely required for chromosome segregation during meiosis (Watanabe and Nurse, 1999). Mutations in rec8 have been identified and characterized in a number of organisms including Saccharomyces cerevisiae, Schizosaccharomyces pombe, Caenorhabditis elegans, Arabidopsis thaliana and mouse, and their analysis has shown that the role of REC8 in cohesion is conserved across kingdoms (Uhlmann, 2003; Lee and Orr-Weaver, 2001; Nasmyth, 1999; Kitajima et al., 2003; Stoop-Myer and Amon, 1999; Page and Hawley, 2003). Furthermore, the analysis of rec8 mutant phenotypes revealed additional REC8 functions in leptotene chromosome structure, AE formation, homologous pairing, synapsis and recombination. However depending on the organism, the rec8 phenotypes are variable and the exact role of REC8 in all these events is controversial (Lin et al., 1992; Molnar et al., 1995; Bai et al., 1999; Bhatt et al., 1999; Klein et al., 1999; Watanabe and Nurse, 1999; Pasierbek et al., 2001; Chelysheva et al., 2005; Xu et al., 2005). In particular, the link between REC8 and synapsis is under debate. Although AE formation is impaired in rec8 mutants from most organisms studied, cytologically wild-type synaptonemal complexes (SCs) form in mouse rec8 mutants, suggesting that REC8 is not required for AE formation in mammals (Xu et al., 2005). The role of REC8 in homology recognition is also poorly understood. Characterization of rec8 mutants has shown that, depending on the organism, homologous recognition either does not occur at all (yeast, Arabidopsis), is initiated but is not completed (C. elegans) or occurs between sister chromatids rather than homologs (mouse) (Bai et al., 1999; Cai et al., 2003; Pasierbek et al., 2001; Xu et al., 2005).

One way to deepen our knowledge of REC8 functions is to study the phenotypic effects of an allelic series of this gene. Very few rec8 mutant alleles have been isolated, and monitoring chromosome structure in yeast, the best-documented model organism for REC8 functions, remains difficult owing to the small size of its chromosomes. Here, we report the cloning of absence of first division (afd1; GenBank accession number AY788900) in maize, and show that it encodes an α-kleisin homolog. Based on its sequence similarity to the rice and Arabidopsis rec8 gene, and its meiosis-specific function, we call afd1 a rec8 homolog. This is the first report of the cloning of a rec8/α-kleisin homolog in maize. In this study, we focused on the role of AFD1 during early prophase I. Based on the phenotype of the null-alleles, we show that AFD1 is required for establishing leptotene chromosome structure, bouquet formation, pairing and synapsis as already shown in other rec8 mutants. Our unique series of weaker afd1 alleles shows a range of defects in prophase I events, allowing us to investigate for the first time the role of AFD1/REC8 in the coordination of these events. We showed that AFD1 is not required for AEs initial recruitment but instead is controlling the extent of AEs elongation and their maturation into lateral elements during synapsis. Surprisingly, partial AEs elongation was sufficient for leptotene chromosome structure establishment and bouquet formation. By contrast, homologous pairing, synapsis and proper distribution of RAD51 depended on full AEs elongation, providing the basis for a model in which AFD1 controls homologous pairing through its role in AEs elongation and the subsequent distribution of the recombination machinery independent of bouquet formation.

In a directed Mutator (Mu)-tagging experiment for a very closely linked gene, tdy1, we obtained a new afd1 allele, afd1-2, that had a large deletion encompassing part of the afd1 gene. Sequencing analysis of one of the regions flanking the deletion in afd1-2 uncovered a truncated open reading frame (ORF) with close identity to the 5′ sequence of rec8. We obtained two additional insertion lines in this rec8 homolog, named afd1-3 and afd1-4, by reverse genetics in the trait utility system for corn (TUSC) transposon mutagenesis program at Pioneer Hi-Bred (Meeley and Briggs, 1995; Chuck et al., 1998). Sequencing the alleles revealed that afd1-1 contains a G to A alteration at the 5′ splice site of intron 16, introducing a premature stop codon and leading to a truncated protein lacking 141 C-terminal amino acids (Fig. 1A). A Mu insertion occurred 40 bp downstream of the donor site of intron 3 in afd1-3, and 80 bp downstream of the donor site of intron 1 in afd1-4 (Fig. 1B). All three new afd1 alleles were sterile and displayed comparable phenotypes to afd1-1 after anaphase I (Golubovskaya and Mashnenkov, 1975) (see below). Based on these results and an allelism test between all the afd1 alleles (see Materials and Methods), we concluded that we cloned the afd1 gene.

By using reverse transcriptase (RT) and rapid amplification of cDNA ends (RACE)-PCR we isolated a 2300 bp long full-length afd1 cDNA encoding a predicted protein of 602 amino acids. The afd1 gene is complex: it contains 20 exons and 19 introns. Although AFD1 displays a stronger sequence identity with mammalian RAD21 than mammalian REC8, AFD1 exhibits the highest sequence identity (39%) with the Arabidopsis REC8 ortholog SYN1, and a weaker sequence identity with the three Arabidopsis RAD21 (Bai et al., 1999; Bhatt et al., 1999; Cai et al., 2003). In yeast, RAD21 functions in mitotic cohesion, and REC8 functions specifically in meiotic cohesion. In other organisms with multiple α-kleisin/rad21/rec8 homologs, genes have been called rec8 based on sequence and meiosis-specific function. AFD1 has the typical REC8/RAD21 N-terminal and C-terminal domains and possesses four putative separase cleavage sites, based on the (D/E)xxR consensus sequence described by Cai et al. (Cai et al., 2003) as well as several other putative protein binding sites (data not shown).

By RT-PCR, we showed that the afd1 gene is expressed in leaves, tassel and ear, and to a lesser extent in roots and very young leaves of maize (Fig. 1C). afd1 expression is thus not meiosis-specific, although the mutant phenotype is limited to reproductive organs. By RT-PCR, we did not detect any afd1 transcripts in deletion allele afd1-2, as expected (Fig. 1D). Sequencing of afd1 cDNA in afd1-3 and afd1-4 did not reveal any alteration in the afd1 sequence. However, we detected lower levels of afd1 expression in these two mutants as estimated by semi-quantitative RT-PCR (Fig. 1D,E).

To check whether the AFD1 protein is expressed in afd1-3 and afd1-4, we generated an AFD1-specific antibody and performed immunocytochemistry during pachytene. In the wild type, a bright signal was detected between homologous chromosomes and this pattern was very similar to that of SYN1/AtREC8 in Arabidopsis (Cai et al., 2003) (Fig. 1F). In the deletion allele afd1-2, no staining was detected in the nucleus (Fig. 1F). In afd1-1, no staining was detected in the nucleus, suggesting that the protein is either degraded or not properly localized because of the truncation in the protein sequence (Fig. 1F). By contrast, we detected a signal in the nuclei of afd1-3 and afd1-4 mutants. In afd1-3, based on the distribution of antibody staining only, we could not recognize individual chromosomes, although we detected 2 to 5 patches of staining per nucleus (Fig. 1F, arrowheads). In afd1-4, the AFD1 signal colocalized with the chromosomes, but was diffuse and faint compared to the wild type (Fig. 1F). When using an antibody raised against SYN1, the Arabidopsis homolog of AFD1, we obtained a similar but less distinct staining pattern in the four afd1 alleles, confirming that the AFD1 staining truly correlates with the presence of AFD1 in the nucleus (supplementary material Fig. S1) (Cai et al., 2003). Altogether, these results showed that afd1-2, like most probably afd1-1, is a null-allele, whereas afd1-3 and afd1-4 are two weaker alleles of afd1.

In wild-type zygotene, based on the AFD1 immunostaining only, we could recognize synapsed and unsynapsed chromosome regions (Fig. 2A). To further investigate the localization of AFD1 during prophase I in the wild type, we performed a double immunostaining with the AFD1 antibody and an antibody raised against ASY1/HOP1 (Armstrong et al., 2002). HOP1 is a structural component of the axial element (Hollingsworth et al., 1990; Smith and Roeder, 1997). ASY1/HOP1 and AFD1 signals colocalized at leptotene and zygotene, demonstrating that AFD1 contributes to the AEs. In synapsed regions, however, the ASY1/HOP1 signal became much dimmer and the AFD1 signal was brighter than it was in unsynapsed regions (Fig. 2B; supplementary material Fig. S2). Interestingly, in meiocytes fixed with 2% (instead of 4%) formaldehyde ASY1/HOP1 as well as AFD1 staining persisted through pachytene and appeared to be double-stranded with both proteins present on each strand (Fig. 2C). Visualization of two separate lateral elements and the persistence of ASY1/HOP1 staining might be due to a less compact chromatin structure and improved epitope accessibility of material fixed in 2% formaldehyde. The ASY1/HOP1 signal observed in meiocytes fixed in 4% formaldehyde is similar to that of PAIR2, the recently characterized rice homolog of HOP1, which was not clearly detectable between synapsed chromosomes (Nonomura et al., 2006). However, in meiocytes fixed in 2% formaldehyde, the ASY1/HOP1 signal is similar to that observed in Arabidopsis (Armstrong et al., 2002). To conclude, kinetics and pattern of the AFD1 signal as well as its colocalization with ASY1/HOP1 strongly suggest that AFD1 and ASY/HOP1 are components of both axial and lateral elements of SCs.

To determine whether any of the defects present in nuclei from afd1 null-alleles are partly rescued in the weaker afd1 alleles, we first examined DAPI-stained nuclei during early prophase I in the four afd1 alleles using 3D deconvolution light microscopy. Previously described criteria (Dawe et al., 1994; Bass et al., 1997; Golubovskaya et al., 2002) were used for accurate staging of the prophase I wild-type and afd1-mutant nuclei. From leptotene to zygotene in wild-type nuclei, the nucleolus moves to an off-center position and chromosomes form a crescent surrounding it (Fig. 3). Similar nucleolus behavior was observed in the four afd1 alleles, thus facilitating the staging in these nuclei (Fig. 3). In wild type, leptotene chromosomes are organized as thin threads (Fig. 3A). At zygotene, whole-length chromosome threads are observed and telomeres cluster on the nuclear envelope (Fig. 3E). At pachytene, homologous chromosomes are aligned (Fig. 3I) and the installation of the SC is complete. Structural defects of meiotic chromosomes in afd1-2 were similar to the ones observed in afd1-1: although leptotene, zygotene and pachytene stages could be identified based on nucleolus behavior and cell-shape and -size, chromosome morphologies could not be distinguished from each other because chromosomes formed a variable and fuzzy mass (Fig. 3B,F,J). By contrast, leptotene chromosome morphology was identified in both afd1-3 and afd1-4 mutants (Fig. 3C,D). In afd1-4 but not in afd1-3 the zygotene chromosome structure looked similar to that of wild type (Fig. 3G,H). As nuclei entered the stage corresponding to pachytene, chromosome morphology deteriorated in both weak alleles and was similar to that observed in afd1-1 and afd1-2 nuclei (Fig. 3K,L). Based on these data, we concluded that the absence of leptotene chromosome structure in the afd1 null-allele could be rescued in weaker afd1 alleles, strongly suggesting that AE elongation can occur in the weak afd1 alleles.

To investigate the role of AFD1 in AE recruitment and elongation, we used transmission electron microscopy (TEM) of prophase I nuclei spreads (see Materials and Methods). This technique allowed us to identify proteinaceous structures and in particular AEs, based on their width and linear shape. To confirm the TEM data, we also performed immunostaining experiments with ASY1/HOP1, a component of the AEs (Hollingsworth et al., 1990; Smith and Roeder, 1997; Armstrong et al., 2002).

In the wild type, at pachytene, TEM micrographs show ten distinct SCs, each of them exhibiting two lateral elements that surround a central element (Fig. 4A,F). Although we never detected a normal SC in afd1-1 and afd1-2, some AEs aggregates were observed in afd1-1 and afd1-2 (13.6±4.7 and 11.8±3.8 AEs aggregates per nucleus, respectively), showing that although AFD1 is required for normal AEs, it is not required for the initial recruitment of AEs (Fig. 4B,C). Although some linear shapes were observed in the AEs aggregates, they were shorter, often multilayered and the central element was never detected (Fig. 4G,H). This phenotype suggests that AFD1 is required for AE elongation and central element recruitment. In afd1-3, the number of synaptic structures slightly increased (18.9±4.5) and we also observed long stretches of `shaggy' AEs forming a `cobweb' around them, suggesting that AE elongation is partly rescued in afd1-3 (Fig. 4D,I). In afd1-4, long AEs were recruited on chromosomes (Fig. 4E,J). The total length of AEs per nucleus ranged from 20-52% of that of wild type; and six of the 30 nuclei examined, were closer to the 52% upper limit. This is consistent with the fact that, in light-microscopy images, zygotene chromosome structure in this allele resembles that of the wild type. This further confirmed that AE elongation was partially rescued in a weak afd1 allele. In addition to AEs agregates, we also observed parts of AEs that appeared synapsed - based on their characteristic distance from each other, leading to an increase of the total number of synaptic structures in afd1-4 (41.1±8.6). However, we never observed a central element or cytologically wild-type SCs in afd1-4 (Fig. 4J). The maximum length of synapsed regions in afd1-4 reached 24% of those in wild type. Furthermore, although more synapsis was detected in afd1-4, most of the AEs were paired non-homologous `self-synapsis', manifested by fold-backs of AEs on the same chromosomes (Fig. 4K,L) as well as exchanges of synaptic partners, manifested as chromosomes that pair with one partner in one region along its length and with a different partner in another region (Fig. 4M,N). The partial rescue of AE elongation in afd1-4 is thus necessary but not sufficient for homologous synapsis. At late pachytene, in all the afd1 alleles, unsynapsed AEs disappeared and only the AEs aggregates remained (Fig. 4S,U).

To confirm the presence of AE revealed by the TEM data, we performed immunostaining experiments with the antibody against ASY1/HOP1 (Armstrong et al., 2002). ASY1 is the Arabidopsis homolog of HOP1 and a component of the AEs (Hollingsworth et al., 1990; Smith and Roeder, 1997; Armstrong et al., 2002). ASY1/HOP1 staining closely resembled the TEM data. Although we detected an ASY1/HOP1 signal in all the afd1 alleles, we only observed long-stretched staining of ASY1/HOP1 protein on afd1-4 chromosomes (Fig. 4O-R). In afd1-1 and afd1-2, only short protein stretches were detected (Fig. 4O,P). In afd1-3, we observed larger patches and lines of ASY1/HOP1 staining (Fig. 4Q). In afd1-4, the ASY1/HOP1 stained area was significantly more elongated than in the null-alleles (compare Fig. 4R with Fig. 4P). In late pachytene ASY1/HOP1 was present in all the afd1 alleles as stretches and short lines (Fig. 4T,V).

Based on these data, we conclude first that AFD1 is not absolutely required for the initial recruitment of AEs and acts downstream of ASY1/HOP1. Second, AFD1 is required for AE elongation, and this function depends on the strength of the afd1 alleles. Third, partial AE elongation alone is not sufficient to allow central element recruitment and homologous synapsis.

To further address the role of AFD1 in homologous pairing, we used a 5S rRNA locus-specific FISH probe to monitor the extent of homologous pairing in the afd1 alleles. In the wild type, at pachytene, the 5S rRNA loci are paired and appear as a single large focus per nucleus (Fig. 5A). In the four afd1 alleles, the 5S rRNA foci were seen as two distinct spots in 100% of the afd1-1 and afd1-2 cells, 97% of the afd1-3 cells, and 96% of the afd1-4 cells and, most often, each spot was doubled because the lack of arm cohesion increased the distance between sister chromatids (Fig. 5B-D). Therefore, as previously shown for homologous synapsis, full AFD1 activity is necessary for homologous pairing. The absence of homologous pairing in all the alleles did not correlate with the variable extent of AE elongation in the afd1 alleles. In afd1-4 in particular, even though we observed a threefold increase in the number of synaptic structures and a dramatic increase in AE elongation, the frequency of homologous pairing was comparable to that of the null-alleles. This further confirms that the AFD1-dependent homologous pairing does not rely solely on the presence of elongated AEs. To investigate further the absence of homologous pairing in afd1, we analyzed two events possibly involved in the AFD1-dependent homologous pairing: telomere clustering and RAD51 localization on chromosomes.

In the wild type, telomeres attach to the nuclear envelope at the leptotene-zygotene transition and form a tight bouquet at zygotene as described previously (Golubovskaya et al., 2002; Harper et al., 2004) (Fig. 5E). In afd1-1 and afd1-2, telomeres remained scattered throughout the nucleus and never formed a bouquet (Fig. 5F). In afd1-3, telomeres could attach to the nuclear envelope, but no bouquet was visible (Fig. 5G). In afd1-4, telomeres attached to the nuclear envelope and the bouquet formed in 20% of the nuclei (Fig. 5H). To check whether bouquet formation requires normal AEs at chromosome ends, we performed an immunoFISH experiment in the wild-type and in the afd1-4 allele using a telomere FISH probe and both the anti-AFD1 and anti-ASY1/HOP1 antibodies. We did not detect a stronger ASY1/HOP1 or AFD1 signal at the telomeres, showing that bouquet formation does not depend on a preferential recruitment of either AFD1 or ASY1/HOP1 at the chromosome ends, but instead depends on the recruitment of AEs on the entire chromosomes (supplementary material Fig. S3). On the basis of these data, we concluded that the rescued AE elongation in afd1-4 could partly rescue the formation of a bouquet, but the formation of the bouquet in afd1-4 was not able to significantly increase the frequency of homologous pairing in this allele. This suggests that bouquet formation is not a major contributor to the AFD1-dependent homologous pairing. Alternatively, it is possible that the presence of short AEs in afd1 prevents the maintainence of the bouquet for a long enough time to facilitate pairing.

In wild-type maize, RAD51 foci form on the chromosomes at the beginning of zygotene and reach a maximum of about 500 per nucleus by mid-zygotene when chromosomes are pairing and synapsing. In zygotene, wild-type cells contain either single RAD51 foci or doubly contiguous RAD51 foci on paired chromosomes (Franklin et al., 1999). In pachytene, the RAD51 foci are brighter, often seen as two paired foci, and their number is reduced (Fig. 5I). In diplotene, no RAD51 foci are detected (Franklin et al., 1999). Mutants with defects in homologous pairing exhibit a significant decrease in the number of RAD51 foci at zygotene, corresponding to the degree of their pairing defects, suggesting that RAD51 might have a role in the homology search in addition to its known role in meiotic recombination (Franklin et al., 1999; Franklin et al., 2003; Pawlowski et al., 2003). As previously described, the number of RAD51 foci is reduced in afd1-1 compared with that of the wild type, even at pachytene (Pawlowski et al., 2003) (Fig. 5J). Additionally, we found that the shape of the RAD51 foci in afd1 is very different than that of the wild type. In particular, in all the afd1 alleles, most of the RAD51 foci aggregated in bright and long patches (Fig. 5J-L). The RAD51 aggregates in afd1-3 and afd1-4 were particularly long, reaching a maximun of 2 μm (Fig. 5L; supplementary material Fig. S4). In comparison, in the wild type, the brightest and longest foci, which are found at pachytene, are never longer than 0.5 μm (Fig. 5I). The number of these bright RAD51 aggregates was highest in afd1-4 (29±5), in comparison with afd1-1, afd1-2 and afd1-3 (7±2.6, 6.8±1 and 8.8±1.6, respectively), suggesting that rescuing AE elongation in afd1-4 facilitates the recruitment of more RAD51, but is not sufficient to form normal RAD51-foci morphology and distribution. In this respect, the number of cytologically wild-type, round RAD51 foci that were excluded from the RAD51 aggregates was extremely low (<10) in all the afd1 alleles (Fig. 5J-L; supplementary material Fig. S4), showing that AFD1 is required for the proper morphology and distribution of RAD51 on the chromosomes.

We have cloned afd1, the first rec8 homolog in maize. A new series of alleles of afd1 allowed us to develop for the first time a unified model for chromosome behavior during meiotic prophase I in maize. By analyzing the phenotype of weak and strong alleles, we showed that, (1) AFD1 is required for AEs elongation but not AEs recruitment, (2) bouquet formation requires elongated AEs but not the installation of the recombination machinery, (3) homologous pairing relies on the formation of complete AEs and, (4) complete AEs installation is required for the appropriate deposition of the recombination machinery independently of bouquet formation. This model is summarized in a cartoon as Fig. 6.

In the afd1 null-mutant, we see a series of significant defects. The first detectable defect is the absence of leptotene chromosome structure. By comparing null-alleles and weak afd1 alleles, we found a correlation between the presence of leptotene structure, the extent of AEs elongation and the extent of AFD1 localization on chromosomes, which strongly suggests that AFD1's earliest meiotic function is to control and contribute to AEs elongation. In meiocytes fixed in 2% formaldehyde, AFD1 and ASY/HOP1 colocalized not only on axial elements but also on lateral elements, strongly suggesting that AFD1 is actually a component of axial/lateral elements. These data fit the current model that REC8 is a structural component of the synaptonemal complex (Molnar et al., 2003; Chelysheva et al., 2005).

However, we show that AFD1 has more than a structural role in AEs formation. Based on the presence of the ASY1/HOP1 signal and the AE-like structure in afd1 null-alleles, we propose that AFD1 acts downstream of both ASY1/HOP1 and AE recruitment. The observation of longer AEs in the weaker afd1 allele and the lack of normal SC structure in all afd1 alleles suggests that instead of AEs recruitment, the main function of AFD1/REC8 is to control and/or contribute to AEs elongation and synapsis.

The absence of AEs elongation in the afd1 null-alleles has a strong impact on the subsequent steps of meiotic prophase and demonstrated that AEs elongation is crucial for bouquet formation, RAD51 distribution, homologous pairing and homologous synapsis. However, using weaker afd1 alleles, we are able for the first time to dissociate the effect of AFD1/REC8 on AEs formation from its role in bouquet formation, RAD51 distribution, homologous pairing and synapsis.

The mechanism by which the bouquet is formed is not known because very few bouquet mutants have been identified. The role of the cytoskeleton in bouquet formation is under debate, the proteins involved in forming the attachment to the nuclear envelope are not known and the role of AEs in bouquet formation has not been defined (Harper et al., 2004; Trelles-Sticken et al., 2005). In afd1-1, afd1-2 and afd1-3, the presence of short axial elements did not lead to bouquet formation. In afd1-4 alleles, 20% of the meiocytes had partially or completely formed bouquets, consistent with the increased levels of axial elements observed in these mutant plants. This is, to our knowledge, the first demonstration that bouquet formation depends on a threshold of AEs elongation. However, AEs elongation is most probably not sufficient for bouquet formation because in mouse rec8 mutants bouquet formation does not occur, even though the extent of AEs elongation is comparable to that of the wild type (Xu et al., 2005).

Furthermore, the analysis of bouquet formation and RAD51 distribution in the afd1 mutant alleles allowed us to investigate the relation between bouquet formation and recombination machinery. In afd1-4, distribution and morphology of the RAD51 foci were impaired even though bouquet was present, showing that bouquet formation occurs independently of RAD51 distribution. This is consistent with the observation that spo11 mutants in yeast and Sordaria, which lack recombination, make normal bouquets (Storlazzi et al., 2003; Trelles-Sticken et al., 1999). Although REC8 is required for bouquet formation in maize and mouse (Xu et al., 2005), the role of REC8 in bouquet formation is not conserved in yeast. A rec8 deletion induced a persistent telomere clustering in prophase-I-arrested cells in budding yeast, suggesting that REC8 is required for the resolution of the bouquet - rather than its formation - during zygotene in this organism (Trelles-Sticken et al., 2005). Knowing that recombination mutants also have persistent bouquets, this could be an indirect effect mediated by the requirement of REC8 to recruit the recombination machinery (Harper et al., 2004).

The relation between bouquet formation and homologous pairing is under debate. The maize mutant pam1 is specifically impaired in the formation of a full cluster of telomeres (Golubovskaya et al., 2002). In pam1, homologous pairing is dramatically reduced, suggesting that bouquet formation facilitates the recognition of homologous chromosomes. Mechanistically, clustering telomeres may facilitate homologous pairing by initiating the pairing of a few loci that could allow the rest of the chromosome to zip-up mechanically (Bass et al., 1997; Harper et al., 2004). Here, we showed on the contrary that, the presence of a bouquet in afd1-4 is not sufficient to promote homologous pairing of the 5S rRNA locus, a sequence near the telomeres. Absence of pairing is therefore most probably due to the greatly reduced number of recombination inititation events in the afd1 mutants, rather than a direct effect of the bouquet. In contrast to wild-type nuclei, afd1-1 nuclei are not stained by the TUNEL assay suggesting that meiosis-specific DNA damage, i.e. double-strand breaks, are not made in the afd1 null-allele (data not shown). Altogether these data suggest that the role of REC8 in bouquet formation is independent from its role in homologous pairing.

Although we could uncouple AEs elongation and bouquet formation from homologous pairing, we were not able to dissociate defects in homologous pairing from defects in RAD51 distribution and polymerization in any of the afd1 alleles. It is therefore possible that the role of AFD1 in homologous pairing relies on the recombination machinery in addition to the AEs. Based on previous work in maize, RAD51 is thought to participate in recognizing homologous chromosomes prior to homologous synapsis (Franklin et al., 1999; Franklin et al., 2003; Pawlowski et al., 2003; Pawlowski et al., 2004). Thanks to the large meiotic chromosomes in maize, we were able to observe the morphology of RAD51 foci with high resolution in the afd1 alleles. Morphology of RAD51 foci in afd1 was similar, although the foci number was higher compared with that in maize phs1, dsyCS and segII mutants, which are also impaired in recognizing homologous chromosomes and homologous synapsis (Pawlowski et al., 2003; Pawlowski et al., 2004) (I.N.G., unpublished data). The RAD51 clusters in these mutants have been proposed to be a manifestation of failure in the search for chromosomal homology (Pawlowski et al., 2003; Franklin et al., 2003; Pawlowski et al., 2004). Based on these observations, the role of AFD1 in recruiting and controlling the distribution of the recombination machinery might contribute, directly or indirectly, to the recognition of homologous chromosomes.

The absence of first division 1 mutant originally received its name because of the presence of an equational first division due to a complete lack of sister chromatid cohesion (SCC) (Golubovskaya and Mashnenkov, 1975). This mutant has been used in further studies to explore the link between SCC and histone phosphorylation (Kaszas and Cande, 2000), and RAD51 and ZmSGO1 recruitment (Pawlowski et al., 2003; Hamant et al., 2005). This work is the first description of AFD1 as a REC8 homolog in maize, and confirms the role of REC8 in SCC and kinetochore orientation in maize. We have shown previously that ZmSGO1 maintains centromeric cohesion at the end of MI, and that ZmSGO1 recruitment is impaired in afd1-1 (Hamant et al., 2005). Defects in centromeric cohesion in the afd1 mutants are thus probably the result of the loss of both AFD1 and ZmSGO1 at the centromere. Interestingly, fission yeast rec8 mutants also display an equational first division. The characterization of the MOA1 protein has recently allowed the dissection of the functions of REC8 at the centromeres (Yokobayashi and Watanabe, 2005), providing a complementary analysis to our study focused on events in prophase I.

In summary, the characterization of a series of afd1 alleles has unraveled the role of AFD1/REC8 in controlling AEs elongation and the formation of lateral elements. It allowed us to monitor the impact of the extent of AEs elongation on bouquet formation and has demonstrated that, AFD1-dependent homologous pairing requires full elongation of AEs and is associated with the proper distribution of the recombination machinery but does not rely on bouquet. We believe that allelic series of rec8 in other organisms would provide new information on the role of REC8 in the coordination of prophase I events and might help us to understand the differences between yeast, C. elegans, Arabidopsis and mammalian rec8 mutant phenotypes, and to dissect the functions of REC8 that are conserved throughout species.

The recessive afd1-1 reference allele was originally induced with N-nitroso-N-methylurea in the W23 inbred in 1974 (Golubovskaya and Mashnenkov, 1975). The afd1-2 came from a directed Mutator (Mu)-tagging experiment of the maize leaf gene tiedyed1 (tdy1), which is tightly linked with afd1-1. The TUSC procedure (Meeley and Briggs, 1995; Chuck et al., 1998) was used to isolate the afd1-3 and afd1-4 lines by reverse genetics.

Previously, two mutations (afd1-1 and tiedyed1) had been mapped at the distal region of the long arm of chromosome 6 with a basic set of B-A translocations. To estimate the linkage between the two genes, heterozygous plants for afd1-1 reference allele were crossed with homozygous tdy1/tdy1 plants. We only considered the F2 progeny that segregated the afd1 phenotype. Of 1000 F2 progeny, we never obtained an afd1-tdy1 double mutant but we did obtain three segregating phenotypes: green-fertile, green-sterile and tiedyed-fertile in a ratio 2:1:1, showing that the two genes are closely linked. A fine-mapping of the afd1 gene was conducted. For simple sequence repeats (SSRs) segregation analysis, the afd1-1 heterozygotes (W23) were out crossed onto the Mo17 and B73 inbred lines and self-pollinated. Polymorphisms between W23 and Mo17 inbred lines were identified for chromosome 6 SSR umc 2059 and bngl 1136 markers, both located on bin 6.08. A total of 48 afd1-1 mutants from segregated F2 families were used to fine-map the gene. The PCR samples were separated by electrophoresis on a 4.5% Apex agarose/TAE gel and stained with ethidium bromide to visualize bands. The chromosomal position of afd1 was also confirmed by in situ hybridization (Wang et al., 2006). To demonstrate allelism between the four mutant lines (afd1-1, afd1-2, afd1-3, afd1-4), heterozygotes for all alleles were crossed with each other, and F1 progenies were tested for segregation of sterility. F1 of all combination-segregated one-fourth of sterile plants and three-quarters of fertile plants. Cytological analysis showed that F1 afd1-2/afd1-1 displayed the same phenotype as afd1-1. Heterozygotes afd1-1/+ and hemizygotes for the deletion allele afd1-2/+ displayed a wild-type phenotype.

RT-PCR and RACE-PCR primers used to amplify afd1 were: OH-16 (5′-ATGGCGGCGACGCTCCACTCGAAGATC-3′), OH-27 (5′-TCTCGTACACGATCACCACGCCACC-3′), OH-57 (5′-GATCTTCGAGTGGAGCGTCGCCGCCAT-3′), OH-62 (5′-ACAGTGACCATAGTGTTACTCCTGGAAGT-3′), OH-61 (5′-GACTTCCAGGAGTAACACTATGGTCACTGT-3′), OH-80 (5′-TTCAGTAAGTGCTTTCAGGACAACTTAGTT-3′), OH-91 (5′-AAGAGACAAGTAGACAACCCAGCTCT-3′). RT-PCR primers used to amplify a partial smc3 cDNA were: OH-33 (5′-GAAGCAAAGGGGGTCACTTGAGAAAGC-3′), OH-36 (5′-TTCACGGAAGTGCCTTGCAACCCCTTT-3′).

Genomic DNA was isolated from secondary ears and leaves from the wild-type, and homozygotes of four afd1 alleles using a CTAB-based protocol (Doyle and Doyle, 1990). DNA (1 μg) was used as a template in a 25 μL reaction. PCR amplification was performed with the Taq polymerase Roche kit (Roche Diagnostics, IN). PCR cycles were as follows: ten cycles of 30 seconds at 94°C, 30 seconds at 65°C (-1°C/cycle), 1.5 minutes at 72°C, followed by 35 cycles of 30 seconds at 94°C, 30 seconds at 55°C, and 1.5 minutes at 72°C, followed by 5 minutes at 72°C and were performed in a GeneAmp PCR system 2400 (PerkinElmer, Wellesley, MA). Amplified fragments were cloned in pCR2.1 (Invitrogen, Carlsbad, CA) and sequenced in both directions.

Total RNA from A344 tissue-roots (1 cm from the tip), 7-day-old leaves, 3-week-old leaves, young tassel- and ear-containing meiocytes, and old tassel-containing pollen - was isolated following the Trizol reagent protocol (Gibco-BRL). RNA was also isolated from tassel-containing meiocytes from homozygotes of the four afd1 alleles following the same protocol. 2 μg of each RNA sample were run on a 1.2% agarose gel and the intensities of rRNA bands were compared to assess the amount of RNA template. These RNAs were subjected to reverse transcription using the MMLV reverse transcriptase kit (Promega, Madison, WI). PCR amplification was then performed with 5 μl of the reverse transcription sample. Primers OH-33 and OH-36, which are specific for the maize smc3 gene, were used as an internal control and confirmed that an equal amount of amplifiable cDNA was available in all the samples. Primers OH-16 and OH-61 were used to amplify part of the afd1 cDNA (exon 2 to exon 15). RACE samples were prepared with the RFLM-RACE Kit (Ambion, Austin, TX) according to the manufacturer's instructions. The 5′ RACE primers were OH-27 and OH-57 and the 3′ RACE primer was OH-62. Continuity of the afd1 cDNA was checked by PCR with primers OH-91 and OH-80. RT-PCR and RACE-PCR products were resolved, cloned and sequenced as described above.

A smear acetocarmine technique was used to investigate chromosome segregation in afd1 and to confirm the afd1 mutant phenotype in plants from afd1 families that segregated a male sterile phenotype. Immature tassels were fixed in Farmer's fixative (95% ethanol and glacial acetic acid at the ratio of 3:1) and stained with 2% acetocarmine, squashed, and observed with a light-microscope (Golubovskaya et al., 1993).

For FISH, immunostaining and immunoFISH, most anthers from pre-emerged tassels were fixed with 4% formaldehyde in buffer A and stored as described in Golubovskaya et al. (Golubovskaya et al., 2002); for immunostaining experiments, remaining anthers were fixed with 2% formaldehyde, as mentioned in the text. Meiocytes were embedded in polyacrylamide as described in Hamant et al. (Hamant et al., 2005). Staging criteria were as described previously (Dawe et al., 1994; Bass et al., 1997; Golubovskaya et al., 2002). FISH was performed as described in Golubovskaya et al. (Golubovskaya et al., 2002). Results on figures are representative of at least 25 meiocytes for each stage.

To generate anti-AFD1 antibody, we cloned a partial afd1 cDNA corresponding to amino acids 227-462 of the AFD1 protein (a region that was successfully used to raise the SYN1/AtREC8 antibody in Arabidopsis) into the pGEX-4T-3 plasmid in translational fusion with GST. The protein was expressed in E. coli BL21 and purified with the GST purification kit (Amersham Biosciences, Piscataway, NJ). The AFD1 antibody was then raised in rats (Strategic Biosolutions, Newark, DE). No staining on the chromosomes was obtained with the pre-immune antiserum in the wild-type or the AFD1 antibody in the afd1-2 null-allele.

Immunostaining was performed as described by Pawlowski et al. (Pawlowski et al., 2003) with the following modifications: 4% BSA (no donkey serum) was used to block the samples, antibodies were not pre-cleared (except anti-RAD51) and secondary antibodies were incubated for 1 hour and washed three times only. In addition to the anti-AFD1 antibody, we used the anti-ZmSGO1 antibody (1:50) (Hamant et al., 2005), anti-RAD51 antibody (1:200) (Franklin et al., 1999; Pawlowski et al., 2003; Franklin et al., 2003), an anti-ASY1 antibody (1:50), kindly provided by C. Franklin (University of Birmingham, UK) (Armstrong et al., 2002) and a SYN1 antibody (1:50), kindly provided by C. Makaroff (Miami University, Oxford, OH) (Cai et al., 2003). We used Cy3-conjugated F(ab')₂ donkey anti-rat IgG and FITC-conjugated F(ab')₂ donkey anti-rabbit IgG secondary antibodies (1:50; Jackson ImmunoResearch, West Grove, PA). 3D-stacks of images were collected and analyzed as described by Golubovskaya et al. (Golubovskaya et al., 2002). Results on figures are representative of at least 25 meiocytes for each stage.

The ImmunoFISH protocol was adapted from Kaszas and Cande, (Kaszas and Cande, 2000). On day 1, we followed the FISH protocol (Golubovskaya et al., 2002) and hybridized the probe overnight. On day 2, we washed the pads once with 1× SSC with 1× PBS (15 minutes); once with 1× PBS with 0.1% Tween 20 (15 minutes), followed by the immunostaining protocol (see above) with the following modifications: both the primary and secondary antibodies were incubated for 1 hour only and washed three times.

We used transmission electron microscopy to study the extent of synapsis in the afd1 alleles in chromosome spreads of male meiocyte nuclei. The spreads were treated with silver nitrate, which stains the AEs present at leptotene and early zygotene that later become the lateral elements of the SC, after the SC is formed between paired chromosomes at pachytene. SC spreads from 24 afd1-1, 29 afd1-2, 25 afd1-3, 30 afd1-4 mutant meiocytes were prepared and analyzed as described in detail by Golubovskaya et al. (Golubovskaya et al., 2002). Length of the synaptic structures was measured with softworx (Applied Precision, WA, USA).

New materials described in this publication are available for non-commercial research purposes upon acceptance and signing of a material-transfer agreement.

We thank Chris Makaroff for providing the SYN1 antiserum and Chris Franklin for providing the ASY1 antibody. We thank Meredith Sagolla for help with microscopes, and Lisa Harper, Wojciech Pawlowski, Ray Chan and David Mets for discussions and critical reading of the manuscript. We thank Barbara Rotz and her staff for excellent care of plants in the field and greenhouses. Part of this project was made possible through the use of technology developed by Pioneer Hi-Bred International, Inc. This research was supported by grants from the National Institutes of Health GM 048547, and the USDA 2003-35301-13152 to W.Z.C.