AtREC8 and AtSCC3 are essential to the monopolar orientation of the kinetochores during meiosis

Accurate chromosome segregation during mitosis and meiosis is based on processes that are particularly well conserved among eukaryotes. One of these processes, sister chromatid cohesion, holds newly replicated chromatids together until anaphase. Mitotic cohesion is laid down during S phase and released at the metaphase-anaphase transition to allow chromatid segregation. Meiosis is a modified cell division in which a single S phase takes place before two rounds of chromosomal segregation. The homologous chromosomes are pulled to opposite poles of the cell at the first meiotic division (meiosis I), whereas the sister chromatids segregate only at meiosis II. A successful first meiotic division requires the association of homologous chromosomes as bivalents during prophase I, the monopolar attachment of sister kinetochores at metaphase I, and the preservation of centromere cohesion at anaphase I, when arm cohesion is released (reviewed by Watanabe, 2004).

Yeast mitotic cohesion depends on a multisubunit protein complex (the cohesin complex), comprising four essential proteins: Scc1, Scc3 and two members of the ubiquitous structural maintenance of chromosome (Smc) family: Smc1 and Smc3 (reviewed by Nasmyth, 2002). The cohesin complex may be organised in a ring in which the Smc1/3 heterodimer forms a large V closed by Scc1 (Anderson et al., 2002; Gruber et al., 2003; Haering et al., 2002), probably with the aid of Scc3, which is thought to bind Scc1 (Nasmyth, 2002). At anaphase, Scc1 cleavage by an endopeptidase (separase) releases cohesion, allowing chromatids to segregate (Haering and Nasmyth, 2003). During meiosis, cohesion is released in two steps. First, at anaphase I, arm cohesion is released but centromere cohesion retained, facilitating the release of chiasmata and the segregation of homologues. Centromere cohesion is then released at anaphase II, allowing the sister chromatids to separate. The mechanisms controlling these processes are only now becoming clear. In all organisms studied, Scc1 is mostly replaced at meiosis by a meiotic orthologue (Rec8). And as far as we know, this replacement is absolutely necessary for the maintenance of centromeric cohesion at meiosis I (Klein et al., 1999; Watanabe and Nurse, 1999). The protection of centromeric Rec8 at anaphase I appears to be mediated by the Sgo1 protein, recently identified in yeasts (Katis et al., 2004; Kitajima et al., 2004), and widely conserved across different species (Watanabe, 2004). Unlike the clear role Rec8 plays in centromeric cohesion at meiosis I, the role of Rec8 in kinetochore orientation at meiosis I remains ambiguous. Although it has been shown in Schizosaccharomyces pombe that Rec8 is required for kinetochores to be oriented to the same pole (Watanabe and Nurse, 1999; Watanabe et al., 2001; Yokobayashi et al., 2003), replacement of Rec8 by Scc1 in Saccharomyces cerevisiae preserves the monopolar orientation, showing that at least in this organism Rec8 is not obligatory for monopolar attachment of kinetochores (Toth et al., 2000). In other organisms, it has been difficult to assess whether Rec8 was involved or not in kinetochore orientations. So far, in all other organisms studied, Rec8 depletion either leads to an early release of cohesion (before metaphase I) or induces strong chromosome fragmentation (Bai et al., 1999; Bannister et al., 2004; Bhatt et al., 1999; Klein et al., 1999; Pasierbek et al., 2001).

Very few data are currently available for the other cohesin subunits, including Scc3. This protein was first identified as a member of the yeast mitotic cohesin complex (Toth et al., 1999) and was subsequently identified in Caenorhabditis elegans, the only organism in which this protein has been shown to play a role in both mitotic and meiotic cohesion (Pasierbek et al., 2003; Wang et al., 2003). In contrast to the sole Scc3 found in S. cerevisiae and in C. elegans, other organisms (Drosophila melanogaster, S. pombe, mammals, Xenopus laevis) possess several Scc3-like proteins. S. pombe has two Scc3-like proteins that play specific roles: Rec11 is involved in arm cohesion, whereas Psc3 is required for centromere cohesion during meiosis (Kitajima et al., 2003); only Psc3 has been shown to also be involved in mitotic cohesion (Tomonaga et al., 2000). The mammalian STAG3 gene, which encodes one of three Scc3 homologues, has been shown to be specific to meiosis (Pezzi et al., 2000), where it is probably involved in meiosis I sister arm cohesion (Prieto et al., 2001) and may therefore be the homologue of Rec11.

In this study, we investigated the function of the sole Arabidopsis Scc3 homologue (AtSCC3) and demonstrated its involvement in both meiotic and mitotic divisions. Our results suggest that both cohesins, AtREC8 and AtSCC3, are necessary for the monopolar orientation of the kinetochores at meiosis I and for the maintenance of centromeric cohesion at anaphase I. Further, our study highlights possible differences between AtREC8 and AtSCC3 during recombination repair and axis building. Finally, we demonstrate a role for AtSPO11-1 in stabilisation of the cohesin complex.

The Atscc3-1 mutant (EDT1 line) was obtained from the Versailles collection of Arabidopsis T-DNA transformants (Ws accession) (Bechtold et al., 1993). Its FST was obtained from https://genoplante.infobiogen.fr/flagdb/info. The Atscc3-2 mutant, line SALK_021769, was obtained from the collection of T-DNA mutants of the Salk Institute Genomic Analysis Laboratory (Col-0 accession) (SIGnAL, http://signal.salk.edu/cgi-bin/tdnaexpress) (Alonso et al., 2003) and provided by NASC (http://nasc.nott.ac.uk/). The Atrec8 is dif1-1 allele (Ler accession) is described elsewhere (Bhatt et al., 1999). The Atspo11-1-1 mutant (Ws accession) is also described (Grelon et al., 2001).

Arabidopsis plants were cultivated in a greenhouse or growth chamber under a 16-hour day/8-hour night photoperiod, at 20°C with 70% humidity. For culture in vitro, sterilised seeds were plated on Arabidopsis medium (Estelle and Somerville, 1987) diluted 1:2. Plates were incubated for 48 hours at 4°C in the dark and were then transferred to the growth chamber. For primary root measurement, plates were stood on end.

Protein sequence similarity searches were performed at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/BLAST/) and at the Arabidopsis Information Resource (TAIR, http://www.arabidopsis.org/Blast), using BLOSUM45 matrix and default parameters. Sequence analyses were performed with DNAssist software (http://www.dnassist.org).

The right border of the Atscc3-1 T-DNA was amplified by PCR with primers Stag1 (5′-GCAAGTTGGTTGCTAGTGATGTGG-3′) and TAG3 (5′-CTGATACCAGACGTTGCCCGCATAA-3′); the left border was amplified with Stag2 (5′-CTTATCTTCTCTGTCTGACCCGCC-3′) and LbBar1 (5′-CAACCCTCAACTGGAAACGGGCCGGA-3′). Wild-type AtSCC3 was amplified with primers Stag1 and Stag2. For Atscc3-2, oligonucleotides ON521769-1 (5′-CAAAATCCAAATGCCCAGAGAC-3′) and ON521769-2 (5′-TCCAGAAAGAAGGAACTCAAGAAC-3′) were used for the wild-type allele and ON521769-2 with LbSALK1 (5′-CATCAAACAGGATTTTCGCC-3′) for the mutant allele. AtREC8 and AtSPO11-1 wild-type and mutant alleles were amplified as previously described (Bhatt et al., 1999; Grelon et al., 2001).

We tested for allelism between the two Atscc3 mutations by crossing Atscc3-1^–/+ (female) and Atscc3-2^–/+ (male). Of the 117 F1 plants, 23 were Atscc3-1^+/–, 45 Atscc3-2^+/– and 49 were wild type for both loci. Thus, the two mutations are allelic and the heterozygous Atscc3-1/Atscc3-2 is lethal.

Double mutants for Atspo11-1-1 and Atscc3-1 or Atrec8 were obtained by crossing a plant heterozygous for the Atspo11-1-1 mutation with plants heterozygous either for Atscc3-1 or Atrec8 mutations. The resulting hybrids were self-pollinated. We selected plants from the F2 progeny, segregating 15:1 for kanamycin resistance in vitro. We used PCR screening to select the sterile plants in the F2 progeny homozygous for both mutations.

The anti-ASY1 polyclonal antibody has been described elsewhere (Armstrong et al., 2002). It was used at a dilution of 1:500. The anti-REC8 polyclonal antibody (Cai et al., 2003) was used at a dilution of 1:250. The anti-RAD51 antibody (Anderson et al., 1997) was used at a dilution of 1:10.

The anti-SCC3 antibody was raised as follows: a 1442-bp DNA fragment was amplified from AY091270 corresponding clone (U11370) with the T3 promoter and LC1 (5′-GTGCCTCGAGGATCAAGTGGTCATACA-3′) primers; it was inserted into pTOPO2-1 (Invitrogen) and sequenced. An NcoI-XhoI fragment containing the DNA sequence encoding the first 417 amino acids of AtSCC3 was then subcloned in-frame into pET29a digested with NcoI and XhoI (Novagen). The resulting construct was transferred to E. coli BL21 cells (Novagen). Upon induction, the 452 amino acid recombinant protein accumulated in the insoluble fraction and was resolubilised in a binding buffer containing 5 M urea, 50 mM Tris-HCl, pH 8, 300 mM NaCl and 5 mM imidazol. The resulting suspension was subjected to ultracentrifugation at 100,000 g for 1 hour. The supernatant was incubated for 90 minutes with 1 ml Ni-NTA Agarose (Qiagen) equilibrated in binding buffer. The resin was washed with 50 ml binding buffer and packed into an HR5/5 column connected to the Äkta Prime system (Amersham Biosciences). The protein was eluted using a gradient of imidazol (20-500 mM) and 0.3 ml fractions were collected at a rate of 0.3 ml/minute. Fractions containing AtSCC3 were pooled and the total amount of recombinant protein was determined with the Biorad Protein assay kit according to the manufacturer's instructions. Rabbit polyclonal antibodies were raised against the recombinant protein (Biogenes). The AtSCC3 antibody was used at 1:1000 dilution.

We observed the development of pollen mother cells by DIC microscopy, after clearing fresh buds of various sizes in Herr's buffer (phenol:chloral hydrate:85% lactic acid:xylene:clove oil; 1:1:1:0.5:1; v:v:v:v:v). We stained meiotic chromosomes with DAPI, as previously described (Ross et al., 1996).

Preparation of prophase stage spreads was performed as described (Armstrong et al., 2002) with the following modifications. For one slide, ten flower buds of the appropriate meiotic stage were used. Anthers were isolated from buds directly in 10 μl enzyme digestion mixture (Armstrong et al., 2002). Then after adding another 10 μl digestion mixture, anthers were incubated for 2 minutes at 37°C. After this time, anthers were tapped out using a hook, made with a cytological needle to release pollen mother cells (PMCs). To prepare the spreads, 10 μl Lipsol spreading medium (1% Lipsol detergent in water buffered to pH 9.0 with borate buffer) was added to the droplet and incubated for 2 minutes at room temperature. During this time, the droplet was stirred with a hook. Then 20 μl of 4% paraformaldehyde, pH 8.0 (Sigma) was added and the slide was allowed to dry. For immunolocalization, slides were washed in distilled water and immersed in PBS with 0.1% Triton X-100. For metaphase stages, spreads were performed as described (Ross et al., 1996), then treated in the microwave according to the method of Leong and Sormunen (Leong and Sormunen, 1998) before immunostaining. Fluorescence immunolocalization was performed according to published methods (Armstrong et al., 2002).

All observations were made using a Leica DM RXA2 microscope; photographs were taken using a CoolSNAP HQ (Roper) camera driven by Open LAB 3.1.5 software; all images were further processed with Open LAB 3.1.5 or Adobe Photoshop 7.0.

The mitotic index was determined as described (Hartung et al., 2002) on roots isolated from five wild-type plants and ten Atscc3-1 plants grown for 21 days in the greenhouse.

Blast searches (Altschul et al., 1990) with several Scc3 proteins (S. cerevisiae, M. musculus Stag3 and S. pombe Rec11) revealed a single putative homologue among the Arabidopsis AGI proteins (http://www.arabidopsis.org/Blast/): At2g47980 (hereafter called AtSCC3). A cDNA clone corresponding to AtSCC3 was identified in the database (accession number AY091270). The protein encoded by AtSCC3 is 1098 amino acids long and is 21% identical and 40% similar to the ScScc3 protein (Blast2 sequence, Matrix Blosum 45) (Tatusova and Madden, 1999) (supplementary material Fig. S1B).

Using RT-PCR on mRNA from various tissues, we found that AtSCC3 was expressed equally strongly in roots, mature leaves, buds and plantlets (data not shown).

AtSCC3 is required for a normal vegetative development We searched for Atscc3 mutants in T-DNA insertion line collections. One mutant allele, Atscc3-1, was found to carry an insertion in exon 19 of AtSCC3 (supplementary material Fig. S1A,B). Sequencing of the DNA flanking the T-DNA in Atscc3-1 showed that T-DNA insertion was accompanied by a deletion of 30 bp of exon 19 and the insertion of 14 bp of foreign DNA. This allele encodes a putative truncated protein consisting of the first 927 amino acids of AtSCC3 plus an additional seven amino acids at the C-terminus. A second insertion allele, Atscc3-2, has a T-DNA insertion at the boundary between intron 5 and exon 6 of AtSCC3 (supplementary material Fig. S1A,B). The Atscc3-2 allele encodes a putative truncated protein containing only the first 193 amino acids encoded by AtSCC3. These two mutants were confirmed to be allelic by genetic tests (see Materials and Methods).

We investigated the effect of AtSCC3 disruption in plants by examining the progeny (self-fertilisation) of a heterozygous Atscc3-2^+/– plant. PCR genotyping of 115 plants from this progeny, using primers specific to the mutant or the wild-type allele, demonstrated the absence of homozygous mutant plants from this progeny (77 plants were heterozygous and 38 wild type), suggesting a defect in transmission of the mutant allele. Selfed Atscc3-2^+/– heterozygotes produced fewer seeds (22.9±1.9 seeds per silique, Fig. 1B) than did wild-type plants (54.4±3.6 seeds per silique, Fig. 1A) owing to the early abortion of some of the progeny. By analysing the transmission of the Atscc3-2 allele to the progeny in backcrosses, we proved that Atscc3-2 mutation was embryo-lethal (data not shown).

Sterile, dwarf plants were observed among the progeny of a heterozygous Atscc3-1^+/– plant (Fig. 1C,D,E). This phenotype segregated in a 3:1 ratio, indicating that the Atscc3-1 mutation was monogenic and recessive. We genotyped 142 mutant plants and showed that all were homozygous for the mutant allele Atscc3-1, demonstrating strong linkage between AtSCC3 disruption and the mutant phenotype. We investigated the phenotype of Atscc3-1 mutants on synthetic medium and in soil. Mutant organs were smaller than those of wild-type plants by a factor of two to five (Table 1) under all conditions tested. We also analysed the proportion of dividing cells in root tips. Mutant roots contained only one-third as many dividing cells as wild-type roots (Table 1).

These data indicate that AtSCC3 disruption is lethal and that a leaky mutation (Atscc3-1) greatly disturbs plant development and may be correlated with mitotic defects.

The fertility of Atscc3-1 plants was extremely low, with only very small numbers of seeds produced (Table 1). We therefore examined the reproductive development of this mutant, and found that the sterility of Atscc3-1 mutants was correlated with abortion of the male and female gametophytes (data not shown). With the aim of identifying the stages of sporogenesis and/or gametogenesis impaired in Atscc3-1, we examined developing male gametophytes by DIC microscopy of cleared buds.

Comparison of the early stages of microsporogenesis revealed no difference between wild-type and mutant plants (Fig. 2A,C): round pollen mother cells (PMCs) were distinguished within the anther locules. In wild-type anthers, these cells underwent two meiotic divisions to produce a characteristic tetrad of microspores enclosed in a callose wall (Fig. 2B). Meiosis products were also detected in mutant plants, but lacked the regular tetrahedral structure, and were either asymmetric tetrads or `polyads' containing more than four products (Fig. 2D), suggesting disturbance of the meiotic program in Atscc3-1.

We therefore investigated male meiosis in Atscc3-1 plants by staining chromosomes with 4′,6-diamidino-2-phenylindole (DAPI). Wild-type Arabidopsis meiosis has been described in detail (Ross et al., 1996), and the major stages are summarised here (Fig. 2E-I). During prophase I, meiotic chromosomes condense, recombine and undergo synapsis, resulting in the formation of five bivalents, each consisting of two homologous chromosomes attached to each other by sister chromatid cohesion and chiasmata, which become visible at diakinesis (Fig. 2F). Synapsis, the close association of two chromosomes via a synaptonemal complex (SC), begins at zygotene and is complete by pachytene, by which point the SC has polymerised along the length of the bivalents (Fig. 2E). At metaphase I, the five bivalents are easily distinguishable (Fig. 2G). During anaphase I, each chromosome separates from its homologue (Fig. 2H), leading to the formation of dyads corresponding to two pools of five chromosomes (not shown). The second meiotic division then separates the sister chromatids, generating four pools of five chromosomes (Fig. 2I), which give rise to tetrads of microspores (Fig. 2B). Chromosome condensation and synapsis occurred in Atscc3-1 PMCs (Fig. 2J,O). Nevertheless, fully synapsed chromosomes were rarely observed: among 118 cells at zygotene or pachytene stages in Atscc3-1, only 11 displayed full synapsis (9%), whereas among 91 wild-type PMCs at the same stages, 43 were found to be fully synapsed pachytene (47%). Further examination of pachytene-like stages in the mutant revealed an abnormal (fluffy) appearance of the chromatin. Condensation abnormalities were observed in pericentromeric heterochromatin, which was abnormally rounded in the mutant (compare Fig. 2E,J,O, asterisks). Nevertheless, condensation and progression through the subsequent stages of meiosis occurred in Atscc3-1 PMCs. At diakinesis and metaphase I (Fig. 2K,P,L,Q), a complex combination of bivalents and univalents was observed, with these structures often tangled. At anaphase I, we observed 15 to 20 chromosomes segregating toward the poles (Fig. 2M,R), suggesting the premature separation of sister chromatids at the first meiotic division. In some cells (9 of 85), we also detected chromosome fragmentation (Fig. 2R, arrows). Occasionally we also observed chromosomes that failed to separate their chromatids leading to the formation of chromatin bridges (Fig. 2M, arrows). At the second meiotic division, we observed random segregation of the separated chromatids, giving rise to variable numbers of daughter cells with different numbers of chromosomes. In some cases, we observed chromosome bridges that could correspond either to chromatids subjected to bipolar tension or to univalents that fail to separate their sister chromatids (Fig. 2N,S). An analysis of female meiosis in Atscc3-1 identified defects similar to those seen during male meiosis (data not shown). Thus, AtSCC3 is involved in both male and female meiosis and its disruption provokes an early release of cohesion at anaphase I.

We investigated the function of AtSCC3 during meiosis with the aid of antibodies against the N-terminal sequence of the protein. We co-immunolocalised AtSCC3 and ASY1, a protein associated with the axial element of meiotic chromosomes (Armstrong et al., 2002), and compared the two signals.

AtSCC3 immunolocalisation revealed a strong signal in the nuclei of wild-type PMC from meiotic interphase up to and including metaphase I (Fig. 3). During interphase, AtSCC3 was detected as foci in the nucleus (Fig. 3B,E). As the chromosomes condensed during leptotene, the AtSCC3 signal appeared on chromatin, delineating the chromosome axis. At this stage, staining with anti-SCC3 antibody was consistently more punctate than the linear staining of the chromosome axis observed with anti-ASY1 antibody (Fig. 3J,K). By zygotene and pachytene, AtSCC3 and ASY1 signals overlapped showing that at this stage chromosome axes were stained with AtSCC3 (Fig. 3M,N). At diplotene and diakinesis the signal weakened, but could still be observed on the chromosome axis (Fig. 3Q). At metaphase I, only the arms of each bivalent were stained with anti-SCC3 serum (Fig. 3T,W). Although the signal was faint and often punctate at this stage, the anti-SCC3 serum clearly delineated the two homologous arms of the bivalent (Fig. 3W,X, arrows). No signal corresponding to AtSCC3 was detected in centromeric regions during metaphase I, anaphase I or metaphase II (data not shown). ASY1 and AtSCC3 were colocalised at most meiotic stages, except metaphase, during which no ASY1 was detected, and in very young meiocytes, in which ASY1 was barely detectable, whereas AtSCC3 was detected (Fig. 3A,B). This suggested that AtSCC3 was present in meiotic nuclei before ASY1, probably in G1.

A weak signal was obtained with the anti-SCC3 antibody in Atscc3-1 pollen mother cells, confirming the leaky expression of AtSCC3 in Atscc3-1 mutants. However, when AtSCC3 signal was detected (7 out of 18 cells), it was either very faint or associated with the nucleolus (4 out of 7 cells) or forming patches in the nucleus but not on the chromatin of Atscc3-1 plants, suggesting that Atscc3-1 plants produce a truncated version of AtSCC3 that accumulates at low levels and displays an aberrant distribution (Fig. 3Z,AC). However, ASY1 staining was perfectly normal in the Atscc3-1 mutant background (Fig. 3Y,AB, but also Fig. 4I,M).

The staining of vegetative nuclei with anti-SCC3 serum (Fig. 3AF and supplementary material Fig. S2) confirmed the involvement of AtSCC3 in both meiosis and mitosis, in contrast to that observed for ASY1 (Fig. 3AE). AtSCC3 was detected throughout the mitotic cell cycle during interphase (G1, S, G2, supplementary material Fig. S2C-L) as well as on the chromosome axis as the chromosomes condense (supplementary material Fig. S2A,B).

As the Atscc3-1 mutant displayed a mix of univalents and bivalents at metaphase I, we wondered whether recombination was occurring at normal levels in the mutant background. We analysed the nuclear distribution of the protein RAD51, which is an essential component of the recombination machinery. Its appearance on meiotic chromosomes during prophase is thought to reflect the progression of recombination repair (Masson and West, 2001). RAD51 staining of meiotic chromosomes has been described in several plant species: maize (Franklin et al., 1999), Lilium (Terasawa et al., 1995) and Arabidopsis (Mercier et al., 2003). We used an antiserum directed against tomato RAD51 and obtained results similar to those for other species. RAD51 foci appeared at leptotene, were most abundant at zygotene (Fig. 4A-H), and tended to disappear during pachytene (not shown). A similar pattern of RAD51 staining was observed in Atscc3-1 pollen mother cells (Fig. 4I-P). Quantification of this staining at zygotene stages showed that approximately the same number of foci were present in both genotypes (222±59, n=10 for Atscc3-1 cells; 210±23, n=7 in wild-type cells). Therefore, the number of recombination initiation sites is unlikely to be much lower in Atscc3-1 cells than in the wild type.

DIF1/SYN1 (hereafter AtREC8), another member of the cohesin complex in Arabidopsis, is thought to be the Arabidopsis Rec8 homologue (Bai et al., 1999; Bhatt et al., 1999; Cai et al., 2003). In Atrec8 mutants, PMCs display multiple meiotic defects and univalents and chromosome fragmentation is observed at metaphase I (Bai et al., 1999; Bhatt et al., 1999) (Fig. 5E). Chromosome fragmentation and chromatin bridges were evident at anaphase I (Fig. 5F; observed in 28 cells out of 36; 77%), leading previous investigators to suggest that Atrec8 was defective in recombination repair (Bai et al., 1999; Bhatt et al., 1999). In order to check this hypothesis and to compare AtREC8 with AtSCC3, we introduced the Atspo11-1-1 mutation (Grelon et al., 2001) into the Atrec8 and Atscc3-1 mutant backgrounds. In Atspo11-1-1 mutants, the dramatic decrease in the formation of double-strand breaks (DSB) prevents recombination, resulting in univalents being observed at metaphase I, rather than bivalents (Grelon et al., 2001) (Fig. 5C). These univalents segregate randomly at meiosis I (Fig. 5D), and sister chromatid segregation does not occur before meiosis II (not shown). When the Atspo11-1-1 mutation was introduced into Atrec8 or Atscc3-1 mutants, PMCs at meiosis I also had ten univalents (Fig. 5I,K), and no more chromosome fragmentation was observed. At anaphase I, the univalents of both double mutants, Atspo11-1-1Atscc3-1 and Atspo11-1-1Atrec8, underwent a mitosis-like division, with separation of the sister chromatids of all univalents (Fig. 5J,L) but reductional division was never observed (n=44 for Atspo11-1-1Atscc3-1 and n=19 for Atspo11-1-1Atrec8). In both cases this first equational division was followed by random segregation of the chromatids at meiosis II (not shown). Thus, in both Atscc3-1 and Atrec8 mutants, centromere cohesion is maintained through prophase but lost at anaphase I, and sister kinetochores have a bipolar rather than a monopolar orientation, transforming the first meiotic division into a mitotic one.

We investigated the effect of disrupting one cohesin on the location of its putative partner. The antibodies against AtREC8 used here have been described elsewhere, and AtREC8 has been detected on chromosome arms from meiotic interphase to anaphase I (Cai et al., 2003) (Fig. 6B). The distribution of AtREC8 was similar in Atscc3-1 and wild-type PMCs (Fig. 6E,H). By contrast, when we incubated anti-SCC3 antibody with Atrec8 meiocytes, we detected AtSCC3 in young meiocytes (interphase, Fig. 6K) but not at subsequent meiotic stages (Fig. 6N,Q). This suggests that AtREC8 is required for the correct binding of AtSCC3 to meiotic chromosome axes or for stabilisation of that binding.

We observed that the ASY1 signal in Atscc3-1 PMC (Fig. 6D,G and Fig. 3Y,AB) was the same as that of the wild type (Fig. 3A,D,G,J,M and Fig. 6A). However, the distribution of ASY1 was strongly abnormal in Atrec8 PMC (Fig. 6M,P), with most of these cells showing only short stretches of ASY1 staining in Atrec8. These results suggest that a normal chromosome axis develops in Atscc3-1, but not in Atrec8 mutants. We investigated whether this axis abnormality was a consequence of the chromosome fragmentation observed in the Atrec8 mutant by analysing the distribution of ASY1 in the Atspo11-1-1Atrec8 double mutant, which lacks SPO11-induced DSBs and displays no chromosome fragmentation (see above). We found that even in the absence of chromosome fragmentation, ASY1 was distributed unevenly, in short stretches, in Atspo11-1-1Atrec8 PMCs (Fig. 7A,D), indicating that the axis defect of the Atrec8 mutant was independent of DSBs.

When we analysed the distribution of ASY1 and AtREC8 in the Atspo11-1-1Atscc3-1 double mutant, young meiocyte nuclei were labelled by both ASY1- and AtREC8-specific antibodies (Fig. 7G,H), but the AtREC8 signal was difficult to detect once axis formation was complete and at subsequent meiotic stages (Fig. 7K,N,Q). When we analysed the ASY1, AtSCC3 and AtREC8 staining in the Atspo11-1-1 single mutant, we found that their distributions were more similar to that seen in the wild type (Fig. 7S-Z′), suggesting that AtREC8 stabilisation or correct association with chromosomes needs the presence of both AtSCC3 and AtSPO11-1.

Of all the subunits of the cohesin complex, we know the least about the role of Scc3 during mitosis and meiosis. In both S. cerevisiae and C. elegans scc3 mutants, cohesion is lost before anaphase I, leading to extensive missegration (Toth et al., 1999; Pasierbek et al., 2003; Wang et al., 2003). In S. pombe, which has two Scc3 homologues, the ubiquitous Psc3 and the meiosis-specific Rec11; Rec11 is located primarily on chromosome arms, whereas Psc3 is localised at the centromere, defining distinct chromosomal regions (Kitajima et al., 2003).

Here we report the identification and functional analysis of the sole Scc3 gene in the Arabidopsis genome. Several of our results suggest that AtSCC3 is a mitotic cohesin: first, AtSCC3 is ubiquitously expressed in all tissues undergoing cell division; second, the embryo lethality of the Atscc3-2 null allele and the mitotic defects of the leaky Atscc3-1 allele, all indicate involvement in mitosis. We also found that AtSCC3 was involved in meiosis, as we detected AtSCC3 in meiotic nuclei as early as interphase, bound to the chromosome axis from early leptotene through to anaphase I. The AtSCC3 localisation pattern during meiosis is very similar to that for AtREC8 (Cai et al., 2003), which, at least in yeasts, has been shown to be the meiotic partner of Scc3 (Toth et al., 1999; Kitajima et al., 2003). In addition, the meiotic phenotype of the Atscc3-1 mutant confirmed the involvement of AtSCC3 in meiotic sister chromatid cohesion.

In all models used in the study of meiosis, the cohesin Scc1 is mostly replaced by its meiotic equivalent, Rec8, allowing the cohesin complexes to fulfil their meiotic functions. During the anaphase of the first meiotic division, the dissociation of Rec8 from the chromosome arms releases arm cohesion, allowing homologous chromosomes to segregate, whereas centromeric cohesion is maintained until anaphase II owing to the protection of Rec8 from cleavage by Sgo1, recently isolated in S. cerevisiae and S. pombe (Katis et al., 2004; Kitajima et al., 2004; Marston et al., 2004; Rabitsch et al., 2003). In this report, we show that chromosomes in Atscc3-1, Atspo11-1-1Atrec8 and Atspo11-1-1Atscc3-1 mutants separate their sister chromatids at anaphase I. This indicates that in Arabidopsis, both AtREC8 and AtSCC3 are absolutely necessary for the maintenance of cohesion at centromeres in anaphase I. Interestingly, centromere cohesion was not lost until anaphase I, either in the Atscc3-1 or in Atrec8 mutants, suggesting that neither of these proteins is absolutely necessary for cohesion until this stage. This is in contrast to that found for these cohesins in C. elegans (Pasierbek et al., 2001; Pasierbek et al., 2003; Wang et al., 2003), S. cerevisiae (Klein et al., 1999) but similar to S. pombe rec8Δ phenotype or mouse rec8 mutant (Watanabe and Nurse, 1999; Xu et al., 2005). In S. pombe rec8 mutants, Rad21 (the Scc1 homologue) was found to relocate to centromeres and consequently maintained centromeric cohesion until separase action at anaphase I (Yokobayashi et al., 2003). Moreover, the localization of Rad21 in meiotic prophase cells has also been shown in mouse (Parra et al., 2004), suggesting its involvement in meiotic cohesion in other organisms as in fission yeast. As the Arabidopsis genome contains three other Scc1-like sequences (Dong et al., 2001), we hypothesize that one of these Scc1 homologues probably behaves in a similar manner in the Atrec8 background, and maintains cohesion until anaphase I. The situation of Atscc3-1 is more striking as there are no additional Scc3 homologues identifiable in the Arabidopsis genome that could partially compensate for the loss of AtSCC3. We cannot exclude the possibility that the small amount of the truncated protein still present in the Atscc3-1 mutant may function with Scc1 to maintain cohesion during these early stages but is not protected at anaphase I. Alternatively, a non-cohesin mechanism may function in sister chromatid cohesion during the early stages of meiosis. In this regard, it should be noted that Scc1-dependent cohesion is not protected during anaphase I, as has been demonstrated in both budding and fission yeast (Watanabe, 2004). In plants, a specific mutation in Switch1 (encoding a protein involved in the establishment of sister chromatid cohesion during S phase) leads to a loss of cohesion only during late prophase (Mercier et al., 2003; Mercier et al., 2001). Specific antibodies raised against AtREC8 and AtSCC3 do not stain the centromeres after prometaphase, despite their involvement in centromere function (Cai et al., 2003) (this study). Indeed, many studies have reported a failure of antibodies against cohesins in general (Cai et al., 2003; Losada et al., 1998; Waizenegger et al., 2000) and against SCC3 in particular (Pasierbek et al., 2003) to detect their antigens on metaphase chromosomes despite the clear involvement of these proteins in centromeric cohesion. Thus the inability to detect AtREC8 and AtSCC3 on meiotic centromeres after prometaphase is neither unprecedented nor is it unusual, suggesting that the structure of the centromere at this stage prevents the antibodies from accessing their target.

During meiosis I, sister kinetochores are mono-orientated, allowing the attachment of the two sisters by microtubules emanating from the same pole. In S. cerevisiae, a protein complex (the monopolin) is necessary for establishing this monopolar attachment (Toth et al., 2000; Rabitsch et al., 2003), but Rec8 does not play such a critical role in this process, as expression of the mitotic cohesin Scc1 by the Rec8 promoter can also provide monopolar orientation at metaphase I (Toth et al., 2000). It is not clear yet if this situation is conserved among species because the monopolin complex has not yet been described in any other organism and because in S. pombe, in contrast to S. cerevisiae, Rec8 was shown to be necessary for monopolar orientation of the kinetochores (Watanabe and Nurse, 1999; Yokobayashi et al., 2003). The phenotype of rec8 mutants in other species was less informative because either cohesion was lost prematurely (before metaphase) as in S. cerevisiae (Klein et al., 1999), or rec8 depletion provokes apoptosis as in mammals (Bannister et al., 2004) or strong chromosome fragmentation as in Arabidopsis (Bai et al., 1999; Bhatt et al., 1999) (this study). We abolished this chromosomal fragmentation by introducing the Atspo11-1-1 mutation into the Atrec8 background, leading to the production of ten univalents. As these univalents align correctly on a metaphase plate and are submitted to a bipolar tension leading to a mitotic-like division, it demonstrates that in Arabidopsis, as in S. pombe, AtREC8 disruption leads to a bipolar orientation of the kinetochores. We observe that AtSCC3 depletion induces exactly the same bipolar kinetochores as AtREC8 depletion. This suggests that either AtREC8 per se is not sufficient for the monopolar orientation of the kinetochores, or that it may be rendered inactive in the absence of the other members of the cohesin complex as AtSCC3. Further investigation on the effect that AtSCC3 depletion has on other cohesins at the centromeres could reveal more about the mechanisms regulating kinetochore orientation.

Scc3 interacts directly with Scc1 in the cohesin complex in yeast mitosis (Toth et al., 1999), Xenopus oocytes and human cell lines (Losada et al., 2000). In S. pombe meiosis, the two Scc3 homologues, Psc3 and Rec11, interact with Rec8 (Kitajima et al., 2003) and in C. elegans, SCC-3 and REC-8 seem to control each other's location on the chromosome (Pasierbek et al., 2003). We provide evidence here that both AtREC8 and AtSCC3 are required for monopolar kinetochore orientation and for the maintenance of centromere cohesion at anaphase I, suggesting that they too are part of the same complex in meiosis. However, our data also suggest that these two cohesins display some specificity in their function, in particular they differ in their involvement in meiotic DSB repair.

In addition to its mechanistic role during cell division, cohesion between sister chromatids seems to be required for efficient DSB repair during the mitotic and meiotic cell cycles. Many cohesin mutants are sensitive to radiation and/or defective in DSB repair during the vegetative cell cycle (van Heemst and Heyting, 2000). Meiotic recombination is initiated by DSB formation, induced by Spo11 (for a review, see Keeney, 2001). The rec8 and smc3 mutants of S. cerevisiae, and C. elegans and Arabidopsis rec8 mutants accumulate broken chromosomes during meiosis which are thought to reflect defects in recombination repair (Bai et al., 1999; Bhatt et al., 1999; Klein et al., 1999; Bai et al., 1999; Bhatt et al., 1999; Klein et al., 1999; Pasierbek et al., 2001). It has also been shown that rec8 and rec11 mutations in S. pombe reduce meiotic recombination, whereas mutations in psc3 do not (DeVeaux and Smith, 1994; Kitajima et al., 2003; Krawchuk and Wahls, 1999). In mammals, a meiotic variant of Smc1 (SMC1β) is required for chiasma formation, but not for the early steps of meiotic recombination (Revenkova et al., 2004). Cohesion (or cohesins) is therefore necessary for meiotic recombination, probably for the repair of meiotic DSBs, but the mechanism by which these processes are linked remains largely unknown. We confirmed the role of AtREC8 in meiotic DSB repair showing that Atrec8 fragmentation is abolished in an Atspo11-1-1 mutant background. In the Atscc3-1 mutant, we detected only a low level of fragmentation (in only 10% of the cells compared to 77% of Atrec8 cells). This low level of fragmentation was probably not due to a large decrease in recombination initiation because RAD51 foci were formed normally; almost normal synapsis was observed at pachytene; and bivalents were formed in the Atscc3-1 mutant. Therefore, AtSCC3, unlike AtREC8, might not be a key player in meiotic DSB repair in Arabidopsis.

We also showed that AtREC8 is a key player in chromosome axis structure during meiosis because ASY1 staining highlighted a fragmented and split axis in the Atrec8 mutant. This is consistent with REC8 providing the basis for AE assembly (Eijpe et al., 2003) and with the involvement of REC8 in axis structure, as has been shown in various organisms (Klein et al., 1999; Eijpe et al., 2000; Molnar et al., 2003). We also show that these axis defects are independent of Spo11-induced DSBs, because axis structure was not rescued in Arabidopsis Atspo11-1-1Atrec8 double mutants. The ASY1 axis appeared perfectly normal in Atscc3-1 mutants, suggesting that in contrast to AtREC8, AtSCC3 might not be involved in this process.

Of course we cannot exclude the fact that these differences between Atrec8 and Atscc3-1 can be explained by the leaky nature of Atscc3-1 mutation even if Atscc3-1 and the double mutant, Atspo11-1-1Atscc3-1, display a fully penetrant phenotype for centromeric cohesion and kinetochore orientation (see above), and for AtREC8 absence of labelling in Atspo11-1-1Atscc3-1 (Fig. 7). Further investigations with a specific meiotic RNAi extinction of AtSCC3 would help to answer this question. The functions of Scc1 and Scc3 have already been uncoupled in Drosophila mitotic cells, where depletion of Scc1 (Drad21) or Scc3 (SA1) homologues resulted in a different phenotype, even though there are no other homologues of these proteins in Drosophila (Vass et al., 2003). Therefore, although Rec8 and Scc3 are likely to be part of the same complex, these cohesins may also act independently to fulfil additional functions, through specific interaction with various proteins (recombination, axis building, etc.).

Cohesin behaviour was strongly modified by the introduction of the Atspo11-1-1 mutation into the Atscc3-1 mutant (compare Figs 6 and 7). Although AtREC8 was present in normal amounts during meiotic interphase in the double mutant, it was barely detectable at later stages (Fig. 7). The presence of AtREC8 at meiotic interphase shows that the protein is normally imported into the nuclei. Although it was not possible to determine at this stage whether AtREC8 was associated with chromatin, the Atspo11-1-1Atscc3-1 double mutant formed perfectly normal axial elements (AE) (according to ASY1 staining, Fig. 7), indicating that its disappearance at later stages resulted from dissociation from the chromatin rather than a loading defect. Our results clearly indicate that AtREC8 is required for axis building but is not involved in the stabilisation of axial elements because a normal axis may persist even if AtREC8 dissociates from chromosomes. Furthermore, these results clearly show that the stabilisation of AtREC8 on chromatin needs both proteins AtSCC3 and AtSPO11-1 to be present. However, it is impossible to determine whether this stabilisation involves SPO11-mediated DSBs or another role of SPO11. There must be some kind of redundancy between AtSPO11 and AtSCC3 in stabilisation of the cohesin complex as the destabilisation of AtREC8 was observed only in the Atspo11-1-1Atscc3-1 double mutant. In yeast, Spo11 has been shown to perform functions other than initiating DSB formation in meiotic prophase (normal S-phase length and early homologue pairing) (Cha et al., 2000). Spo11 has also been shown to be associated with the entire length of the chromosome axis in mouse (Romanienko and Camerini-Otero, 2000) and in Sordaria, Spo11-GFP staining gives a linear signal along chromosomes during the bouquet stage (Storlazzi et al., 2003), suggesting possible alternative roles of Spo11 in these organisms as well.

In conclusion, we have identified the Arabidopsis SCC3 cohesin and demonstrated that it is involved both in meiotic and mitotic divisions. In meiosis, both AtSCC3 and AtREC8 are necessary for kinetochore orientation and centromere cohesion at anaphase I but not for arm and centromere cohesion until metaphase I. It is striking to note that although some organisms possess a single copy of Scc3 (S. cerevisiae, C. elegans and A. thaliana) others have additional Scc3 homologues (S. pombe, mammals and D. melanogaster) and have diversified Scc3 functions, these differences are without any obvious link with either species proximity, genome size or chromosome number. It remains to be seen if organisms lacking an additional meiotic Scc3 homolog have selected another unrelated protein to provide some cohesin function in place of meiotic Scc3 or if the sole Scc3 protein deals with the specific meiotic Scc1 homolog (Rec8).

We are grateful to Georges Pelletier, Ivan Le Masson and Fabien Nogué for helpful discussions and critical reading of the manuscript. We thank H. Offenberg and C. Heyting for providing RAD51 antibody, C. Franklin for ASY1 antibody and C. Makaroff for SYN1 antibody. We also wish to thank all the Arabidopsis group members for their help in screening the T-DNA transformant collection of Versailles. We thank Halima Morin for technical advice.