The Arabidopsis thaliana MND1 homologue plays a key role in meiotic homologous pairing, synapsis and recombination

Meiosis is the specialised cell division that sexually reproducing organisms undergo in order to reduce the chromosome number by half before gamete formation. Most cells of diploid organisms contain two corresponding - homologous - sets of chromosomes, one from the mother and one from the father. Both sets are replicated, recombined and subsequently distributed to four daughter cells in two successive rounds of cell division during the process of meiosis. During the first division (meiosis I), pairs of replicated homologous chromosomes are separated, whereas during the second division (meiosis II), it is the sister chromatids that are separated. Before homologous chromosomes are separated at meiosis I, they associate in pairs, synapse during prophase - culminating in the formation of a tripartite proteinaceous structure called the synaptonemal complex (SC) - and homologous recombination occurs, resulting in both crossover and non-crossover (or gene conversion) events. Homologous recombination results in the formation of chromosomes consisting of both paternal and maternal DNA sequences and consequently contributes to the genetic diversity of the four meiotic products (reviewed by Page and Hawley, 2003; Zickler and Kleckner, 1999).

Homologous recombination in meiosis proceeds by specialised repair of DNA DSBs, generated by the type-II topoisomerase-like enzyme Spo11 (Bergerat et al., 1997; Keeney et al., 1997). In many species, including Saccharomyces cerevisiae and Arabidopsis thaliana, the formation of the SC depends on meiotic recombination initiation by Spo11 (Grelon et al., 2001; Henderson and Keeney, 2004) and it is believed that the SC starts to form at these sites of broken DNA (Agarwal and Roeder, 2000; de Vries et al., 2005).

The DSB processing mechanisms that eventually lead to crossover and non-crossover events have been deciphered at the molecular level, primarily in yeast. DSB repair, which preferentially uses the homologous chromosome as a template, is mediated by two recombinases, Rad51 and Dmc1 (reviewed by Okada and Keeney, 2005). Following DNA DSB formation and resection, Rad51 is loaded onto single-stranded DNA (ssDNA). Rad51 has a role in both somatic and meiotic recombination, whereas the closely related strand-exchange factor Dmc1 is exclusively loaded onto ssDNA during meiosis (Pâques and Haber, 1999). Dmc1 physically interacts with Tid1/Rdh54 (Dresser et al., 1997), which mediates loading and is required, together with other factors, for inter-homologue recombination (but not inter-sister recombination) (Schwacha and Kleckner, 1997). One such factor is Hop2, which is indispensable for homologous recombination and inter-homologue bias during meiosis (Tsubouchi and Roeder, 2002). Hop2 requires the binding of Mnd1 to be functional (Chen et al., 2004). The budding yeast Mnd1 gene was identified in a screen for genes expressed early during meiosis (Rabitsch et al., 2001). Mutation of Mnd1 leads to meiotic arrest owing to hyper-resected DSBs and aberrant synapsis (Zierhut et al., 2004), a phenotype similar to that of hop2 mutants (Gerton and DeRisi, 2002; Leu et al., 1998; Tsubouchi and Roeder, 2002), leading to meiotic arrest owing to hyper-resected DSBs and aberrant synapsis (Zierhut et al., 2004). Furthermore, Mnd1 is a multicopy suppressor of a hop2ts mutation. Analyses of Hop2 in organisms other than budding yeast, such as fission yeast, mouse and Arabidopsis have underlined its importance for the regular processing of DSBs and synapsis between homologues (Nabeshima et al., 2001; Petukhova et al., 2003; Schommer et al., 2003). In vitro analysis of purified yeast Hop2 and Mnd1 proteins, has established that the heterodimer improves Dmc1 D-loop formation activity (Chen et al., 2004). In vitro data obtained from purified mouse proteins demonstrate that only Hop2 possesses strand invasion activity, that is inhibited by Mnd1, and that the Hop2-Mnd1 heterodimer stimulates the strand invasion activity of Dmc1 and Rad51 up to 35-fold (Petukhova et al., 2005). The current model of Hop2-Mnd1 function views the latter protein as a kind of match-maker, ensuring that the Hop2 strand invasion capacity is only elicited in the context of Dmc1 and/or Rad51. Moreover, it assumes that the Mnd1-Hop2 heterodimer and Dmc1 promote DSB repair specifically when homologous chromosomes are available as repair templates. If meiotic inter-homologue bias is disrupted by mutation of inter-homologue bias-determining factors such as Hop1 or Red1, mnd1 mutants can repair meiotic DSBs, by using the sister chromatid as a template (Zierhut et al., 2004).

Mnd1 is present in the genomes of several organisms (Gerton and DeRisi, 2002; Zierhut et al., 2004) but so far has only been functionally characterised in budding and fission yeasts (Saito et al., 2004; Tsubouchi and Roeder, 2002). Purified Mnd1 protein from mouse has been biochemically studied in vitro (Petukhova et al., 2005), but no mutant has been generated. We present the first functional characterisation of an Mnd1 homologue in a higher eukaryote, AtMND1 from A. thaliana. We demonstrate that AtMND1 is essential for male and female meiosis, and that mutation of this gene leads to sterility. Although axial elements are formed normally, sister chromatid cohesion is established and recombination initiation appears to be unaffected (according to the formation of RAD51 foci) in mutant plants, chromosomes do not synapse. A mass of entangled chromosomes, interconnected by chromatin bridges and exhibiting severe chromosome fragmentation were observed during meiotic progression. This fragmentation depends on the presence of the SPO11-1 protein. These data demonstrate that AtMND1 is involved in meiotic recombination and repair, and that its absence leads to unprocessed or abnormally repaired meiotic DSBs. Using as yeast two-hybrid (Y2H) assay, we show that AtMND1 interacts with AHP2, the A. thaliana homologue of Hop2. We conclude that AtMND1 associates with AHP2 and that together they act as key proteins in meiotic recombination, a feature apparently conserved from yeasts to higher eukaryotes.

Zierhut et al. (Zierhut et al., 2004) present predicted homologues of the Saccharomyces cerevisiae Mnd1 gene found in BLAST searches (Altschul et al., 1997) against GenBank. Genes were found in Encephalozoon cuniculi, Schizosaccharomyces pombe, A. thaliana, Mus musculus and Homo sapiens. An Mnd1 gene homologue has not been found in either Drosophila melanogaster or Caenorhabditis elegans (Gerton and DeRisi, 2002), both of which differ from other organisms with respect to the initial stages of meiosis. The A. thaliana gene homologue of the yeast Mnd1 gene has been named AtMND1 (At4g29170). It shares 43% identity to the human, 42% to the mouse, 38% to the S. pombe and 26% to the S. cerevisiae gene homologues. AtMND1 has ten exons; its mRNA, including 5′ and 3′ UTRs, is 994 bp in length as determined by RACE (GenBank accession number: DQ248000), and its ORF is 693 bp in length (Fig. 1A). We obtained the AtMND1 cDNA sequence from the Arabidopsis Biological Resource Center (Yamada et al., 2003) and confirmed its identity by sequencing.

An insertion mutant line (SALK_110052), carrying a T-DNA insertion in the AtMND1 gene region, was identified in the Salk Institute Genomic Analysis Laboratory T-DNA collection (Alonso et al., 2003). In Fig. 1A, the position of the T-DNA insertion within the seventh intron of the AtMND1 gene is shown. PCR amplification of T-DNA border regions was accomplished by combining primers from the left border of the T-DNA with gene-specific primers. Sequencing of these PCR products revealed, that the T-DNA insert has two left borders and that a 85-bp fragment of the seventh intron of the AtMND1 sequence is deleted in the mutant plant (Fig. 1B). DNA gel blot analysis was performed to characterise the nature of the T-DNA insertion. DNA from individual plants was extracted, digested with the endonuclease HindIII, separated, and blotted. Detection of the AtMND1 sequence was performed with a probe corresponding to the entire genomic region of AtMND1. The blot shows the expected bands for wild-type (+/+) (0.8 kb band; 3.2 kb band), heterozygous (+/-) (0.8 kb band; 2.2/2.3 kb bands; 3.2 kb band) and homozygous mutant plants (-/-) (2.2/2.3 kb bands; 3.2 kb band) (Fig. 1A,C). Re-probing the blot with a probe derived from the T-DNA, indicated that there is only one T-DNA integration locus, which consists of multiple T-DNA repeats (data not shown). Plant genotypes were routinely determined by multiplex PCR, combining primers M2, M3 and LBa1 in one reaction (Fig. 1A,D). The expected DNA fragment lengths for wild-type and mutant alleles are ∼0.9 kb and ∼1.3 kb, respectively. Wild-type and homozygous mutant plants, selected by PCR, were analysed for expression of the AtMND1 gene by RT-PCR. The cDNA of AtMND1 (693 bp) was only detected in wild-type samples. By contrast, the ACTIN cDNA control (390 bp) was detected in mutant and wild-type samples (Fig. 1E). Expression of AtMND1 was found in buds, leaves and shoots (Fig. 1E). This finding is consistent with publicly available microarray data (http://www.genevestigator.ethz.ch).

Plants heterozygous for the Atmnd1 mutation were indistinguishable from wild-type plants and self-fertilisation produced homozygous mutants in the expected 3:1 ratio. Homozygous Atmnd1 mutants do not have any obvious growth aberrations during vegetative development. The mutants germinate at the same time as wild-type plants and develop at the same rate. Rosette leaves looked normal and bolting was not delayed. Inflorescences also looked normal but none of the siliques of Atmnd1 plants elongated. Whereas senescent shoots of wild-type plants had long siliques and a terminal inflorescence that ceased to develop further, Atmnd1 mutant plants possess only very short siliques and an inflorescence that produces flowers for a much longer time period than wild-type plants. The short siliques of Atmnd1 plants produced only 0.033 seeds/silique (n=4930) (Fig. 2A). About half of these seeds germinated (0.019% seeds/silique), but very few formed a plant body. None of these plants had a normal morphology and none produced seeds (data not shown).

To confirm that the Atmnd1 mutation caused the sterility phenotype, heterozygous Atmnd1 mutant plants were transformed with a T-DNA construct containing a genomic DNA fragment, harbouring the AtMND1 gene and its putative promoter region. Transformants were selected and simultaneously screened with PCR for the presence of the Atmnd1 mutation. Four individual transformants, containing the complementing T-DNA and harbouring the Atmnd1 mutant allele either in heterozygous or homozygous configuration were analysed for the production of sterile offspring plants. All offspring that contained the complementing T-DNA (n=108) were fertile, indicating complete reversion of the sterility phenotype in the presence of a genomic copy of AtMND1 (Fig. 2A).

Both the sterility observed in our Atmnd1 mutant plants as well the results of Mnd1 analyses in other organisms suggested that there is a meiotic defect in A. thaliana Atmnd1 mutants. The viability of pollen grains, the products of male meiosis, can easily be visualised by Alexander staining (Alexander, 1969). Wild-type anthers contain many mature and viable pollen grains, as visualised by the red staining of the cytoplasm (Fig. 2B). By contrast, Atmnd1 mutant plants develop anthers that are devoid of regular pollen grains. The aborted pollen grains, visualised by the green counter-stain of the pollen wall, are generally smaller and variable in size (Fig. 2B). 93% of all Atmnd1 anthers contained no viable pollen grains, 6% contained one and 1% contained two viable pollen grains (n=81). All of these viable pollen grains looked bigger and non-regular in shape (supplementary material Fig. S1G-J).

We investigated the behaviour of meiotic chromosomes in Atmnd1 mutants and in wild-type plants, by means of a meiocyte spreading technique (Fig. 3). In wild-type meiocytes, the ten A. thaliana chromosomes appear as thread-like structures in leptotene (Fig. 3A), undergo synapsis in zygotene (Fig. 3B) and are fully synapsed along their entire lengths, at pachytene (Fig. 3C). After the disappearance of the SC in diplotene (Fig. 3D), the resulting five bivalents condensed, revealing the presence of chiasmata in diakinesis (Fig. 3E,F). The bivalents assembled at the metaphase I plate (Fig. 3G) and segregated homologues at anaphase I (Fig. 3H). During meiosis II, the second round of segregation leads to the formation of four sets of five chromatids (Fig. 3I).

In Atmnd1 meiocytes, the first step of prophase I, leptotene, appeared similar to that in wild-type meiocytes (compare Fig. 3J with 3A). The first observable defect in Atmnd1 meiocytes was the absence of typical zygotene and pachytene stages, during which synapsis normally occurs. Instead, we observed zygotene-like (compare Fig. 3K with 3B) and pachytene-like stages (compare Fig. 3L to 3C), with a clear absence of extended synapsis. Some alignment seemed to occur at this stage but the short stretches of associated chromosome axes observed, could have been formed by chance during the spreading procedure. In any case, it became clear from subsequent stages that synapsis, even if initiated, did not extend to completion in the Atmnd1 mutant. Subsequently, chromosomes condensed progressively and altered chromosome structures during early diakinesis-like stage (Fig. 3M) and diakinesis-like stage (compare Fig. 3N with 3F). DNA masses erratically linked by chromatin bridges became visible. Instead of five separated bivalents, maintained by chiasmata, an entangled mass of chromosomes was observed at the metaphase-I-like stage (compare Fig. 3O with 3G). At anaphase I, the chromosome mass separated and chromosome fragments appeared (Fig. 3P). During the second meiotic division, chromosome fragmentation was visible at the metaphase-II-like stage (Fig. 3Q) and became more pronounced at the telophase-II-like stage (Fig. 3R).

As meiotic defects can have different consequences for male and female meiosis in Arabidopsis (Mercier et al., 2001), we investigated female meiosis and gametophyte development in Atmnd1 plants. We found that 2.4% of fully grown ovules in Atmnd1^-/- mutants contained apparently normal embryo sacs, 2.8% contained an embryo sac blocked during mitotic divisions and 94.8% contained a degenerated or a single nucleus embryo sac (n=611) (supplementary material Fig. S1A-F). We pollinated Atmnd1^-/- mutants with wild-type pollen and found 2.2% seed formation (1.08 seeds/silique, n=25) when compared with the wild type (50 seeds/silique). Only 0.68 seeds/silique were viable, leading to an overall fertility of 1.4%. This indicates that female meiosis is less affected than male meiosis. The cytological defects observed were similar to those seen in male meiosis (Fig. 4). Typical pachytene stages were not observed (compare Fig. 4D with 4A), an entangled chromosome mass instead of five bivalents was formed at metaphase I-like stages (compare Fig. 4E with 4B) and chromosome fragmentation appeared at anaphase I-like stages (compare Fig. 4F with 4C).

Chromosomes undergo pairing during meiosis in wild-type plants, with telomeres clustering during leptotene at the nucleolus and associating in pairs before recombination-mediated alignment of chromosomes (Armstrong et al., 2001). In zygotene, FISH signals corresponding to regions adjacent to telomeres are paired or are in close proximity (Fig. 5A, red signal corresponds to a sub-telomeric region of chromosome 2), whereas interstitial chromosomal regions are paired or not, depending on the progression of zygotene (Fig. 5A, green signal corresponds to an interstitial region of chromosome 1). In pachytene, chromosome pairing is completed and only one signal can be seen for each FISH probe (Fig. 5B). Homologous chromosomes separate during anaphase I, a process that requires sister-chromatid cohesion to be lost in arm regions (reviewed by Watanabe, 2004). Therefore, a pair of FISH signals, corresponding to the two sister-chromatids, are frequently seen on each homologue (Fig. 5C). In Atmnd1 mutants, no pairing of homologues is observed. In zygotene, the sub-telomeric regions are sometimes paired or in close proximity to each other (red signal, Fig. 5D), but no pairing was ever detected during meiotic progression for the two FISH probes analyzed (pachytene-like stage shown in Fig. 5E). At the anaphase-I-like stage (Fig. 5F), chromosome fragmentation and chromosome bridges were visible and the FISH signals indicate that homologous chromosomes 1 and 2 do not regularly segregate.

To determine whether SPO11 function is involved in the chromosome fragmentation observed in prophase I in Atmnd1 mutant plants, we generated Atmnd1 spo11-1 double mutants. Although A. thaliana possesses three SPO11 homologues, SPO11-1 is probably the only meiotic protein of the three SPO11 proteins (Grelon et al., 2001; Yin et al., 2002) (M. Grelon, INRA IJPB, Versailles, France, personal communication). Meiotic progression of spo11-1 mutants was compared with Atmnd1 spo11-1 double mutants (Fig. 6). Leptotene stages (not shown) were comparable with wild-type meiocytes. However, neither single nor double mutants displayed paired chromosomes; thus, only failed zygotenes (Fig. 6A,G) and only unsynapsed chromosomes of a pachytene-like stage (Fig. 6B,H) were observed. During diakinesis (Fig. 6C,I), chromosomes condensed and ten univalents were visible which then, after alignment at metaphase I, were randomly distributed at anaphase I (Fig. 6D,J). Randomly distributed chromosomes aligned in metaphase II (Fig. 6E,K) and chromatids, after separation at anaphase II, were incorporated into polyads (Fig. 6F,L). The Atmnd spo11-1 double mutant displayed the same meiotic defect as the spo11-1 mutant, indicating that the function of SPO11 is epistatic to AtMND1. In other words, the DNA fragmentation and chromosome entanglements seen in Atmnd1 mutants stem from the activity of the Spo11-1 enzyme.

ASY1 is a meiosis-specific protein intimately associated with the chromosome axes during prophase I (Armstrong et al., 2002). SCC3 is a member of the cohesin complex and localizes along chromosome axes (Chelysheva et al., 2005). Both proteins behaved similarly in Atmnd1 mutants compared with wild-type plants during the leptotene stage (data not shown) indicating that axial elements are formed normally and that sister-chromatid cohesion is established in Atmnd1 mutants. In Atmnd1 mutants, chromosome axes visualised with antibodies against ASY1 and SCC3, respectively, are not incorporated into an SC structure. We could not identify a pachytene stage in Atmnd1 mutants as we did in the wild type (Fig. 7).

The chromosome fragmentation defect observed during meiosis in Atmnd1 is compatible with a defect in the processing of DSBs that initiate recombination. We thus investigated the behaviour of RAD51, a key protein involved in mediating strand invasion during DSB repair. We immunolocalised the RAD51 protein on chromosome spreads and obtained results similar to those previously reported (Higgins et al., 2004). Numerous foci appeared at leptotene (Fig. 8A), probably corresponding to sites at which recombination had been initiated, and were observable during zygotene (Fig. 8B). The Atmnd1 leptotene and zygotene stages were indistinguishable from wild-type cells, with respect to RAD51 focus formation (Fig. 8C,D), indicating that DSBs are made at a normal level.

AHP2 is the Arabidopsis homologue of Hop2 (Schommer et al., 2003), a yeast protein that interacts with Mnd1. We used a Y2H assay to test for a potential interaction between AtMND1 and AHP2. For this, we cloned the AtMND1 cDNA into a yeast expression vector in-frame with a GAL4 DNA-binding domain, and the AHP2 cDNA in-frame with a GAL4 activator domain, and vice versa. Plasmids encoding DNA-binding domain and activator-domain fusion proteins were transformed together into yeast strain AH109 (Clontech). Control experiments were performed, by transforming AH109 with combinations of fusion-protein-containing vectors and empty vectors (Fig. 9). Whereas all plasmid combinations enabled yeast strain AH109 to grow on synthetic dropout (SD) medium selecting for the presence of the plasmids (SD -Leu/-Trp), only simultaneous transformation of plasmids encoding AtMND1 and AHP2 fusion proteins led to growth on selective plates (SD -Leu/-Trp/-His). We therefore conclude that AtMND1 interacts with AHP2.

Somatic expression of MND1 has been observed in humans (Zierhut et al., 2004) and in A. thaliana (Fig. 1E) but not in yeast. Data from publicly accessible microarray databases (http://www.genevestigator.ethz.ch) show that AtMND1 expression is upregulated approximately fivefold in response to genotoxic stress (AHP2 is upregulated eightfold). These expression data point to a potential role of Mnd1 in somatic cells, presumably in DNA repair; this hypothesis has not yet been tested in mammals, as there is no mutant available. However, we investigated this issue in wild-type and Atmnd1 mutant plants exposed to different genotoxic treatments. Mutant plants were germinated on plates containing different concentrations of hydroxyurea (HU). HU depletes the pool of dNTPs by inhibiting ribonucleotide reductase (RNR) (Krakoff et al., 1968), leading to a replication block and subsequent DNA DSBs. The length of developing roots in Atmnd1 plants was monitored and compared with wild-type plants (Culligan et al., 2004). We could not detect any influence of the Atmnd1 mutation on root growth on media plates with or without HU (data not shown). In addition, we germinated seedlings on media plates and exposed them to different doses of gamma-radiation to induce DSBs (Garcia et al., 2003). Exposure to gamma-radiation interfered with development to a similar extent in mutant and wild-type plants (data not shown). Furthermore, Atmnd1 mutant plants had no growth defect when grown under normal conditions, contrary to the DNA-repair-deficient mutants mre11 (Bundock and Hooykaas, 2002) and rad50 (D. Vezon and M. Grelon, INRA, IJPB, Versailles, France, personal communication). Therefore, it was concluded that AtMND1 has no essential role in somatic DNA repair, at least under the conditions tested.

The model plant Arabidopsis thaliana is advantageous for studying meiosis, as it is amenable to molecular, cytological and genetic analysis. Furthermore, all stages of meiosis can be studied because mutations affecting meiotic DNA repair do not trigger a cell-cycle check point, in contrast to yeast and mammals. It is also emerging that model organisms such as S. cerevisiae do not reflect the full complexity of protein interactions in higher eukaryotes. Even though many proteins involved in meiotic DNA repair and recombination are conserved through evolution, their interaction partners, the chronology of their action and the impact of mutating them can be different. For example, the conserved protein Mre11 is necessary for DSB formation and processing in yeast, but in A. thaliana it is only necessary for the ensuing DNA processing after DSB formation (Puizina et al., 2004). Therefore, to fully understand the function of a meiotic protein such as Mnd1, and to reveal its interplay with other, potentially unknown protein factors, it is vitally important that it is characterised in different species.

In this study, we identified the Arabidopsis MND1 homologue, AtMND1, and analysed its function through the isolation and characterisation of an Atmnd1 mutant. This mutation leads to a strong male and female sterility phenotype that can be rescued by integrating a transgene carrying a wild-type copy of the AtMND1 gene, confirming that the observed phenotype is caused by the Atmnd1 mutation. In addition, we demonstrate that AtMND1 interacts with AHP2 (the Arabidopsis Hop2 homologue) in a Y2H assay, an interaction that has also been observed in yeasts and mammals (Petukhova et al., 2005; Saito et al., 2004; Tsubouchi and Roeder, 2002). The Mnd1-Hop2 complex is therefore conserved in higher eukaryotes.

As synapsis is severely disrupted, pachytene stages cannot be observed in either male or female Atmnd1 meiocytes. Although it is possible that some synapsis initiation occurs in Atmnd1, synapsis does not advance to a substantial extent. ASY1, a protein associated with the chromosome axes (Armstrong et al., 2002; Caryl et al., 2000) and SCC3 (Chelysheva et al., 2005), a cohesion complex component, both localise to unsynapsed chromosomes at the right time, but extensive juxtapositioning of chromosome axes has not been observed with either marker. Furthermore, FISH probes associated with chromosomes 1 and 2, were not found in a paired configuration in Atmnd1, apart from early telomere pairing, that seems unaffected in Atmnd1 mutants, but is not maintained during meiotic progression. Taken together, the existence of RAD51 foci, and the dependence of the observed meiotic aberrations on the SPO11-1 protein, demonstrate the existence of chromatin breaks, a prerequisite for synapsis in Arabidopsis. Normal RAD51 focus formation has also been observed in S. cerevisiae mnd1 and hop2 mutants (Gerton and DeRisi, 2002; Tsubouchi and Roeder, 2002; Zierhut et al., 2004) as well in mouse Hop2 mutants (Petukhova et al., 2003). The absence of homologous pairing and the subsequent observation of entangled multivalents are compatible with the existence of some kind of non-homologous interactions. This would be consistent with observations made in yeast mnd1 and mouse Hop2 mutants, which both have defects in SC formation.

Mouse Hop2 mutants perform limited synapsis, most of it non-homologous (Petukhova et al., 2003). More extended synapsis occurs in S. cerevisiae mnd1 mutants, but it is assumed to be non-homologous (Zierhut et al., 2004). This assumption is based on the similarity of SC extension in yeast mnd1 and hop2 mutants (Tsubouchi and Roeder, 2002) because synapsis in the latter mutant largely involves non-homologous chromosomes (Leu et al., 1998). Thus, it appears that yeast, mammals and plants share a common synapsis control pathway that depends on Mnd1-Hop2. It is interesting to note that the Mnd1-Hop2 complex is not conserved in all eukaryotes. C. elegans (Dernburg et al., 1998) and D. melanogaster (McKim et al., 1998), which differ in their initial steps of meiotic recombination (because Spo11 is not required for intact SC formation), lack the Mnd1-Hop2 complex (reviewed by Gerton and Hawley, 2005).

Mnd1 and Hop2 are required for meiotic recombination in yeast, but not for its initiation (Gerton and DeRisi, 2002; Tsubouchi and Roeder, 2002). Similarly, the mammalian HOP2 protein is not required for the initiation of DSBs, but only for their repair (Petukhova et al., 2003). Our results from Arabidopsis thaliana, are consistent with this view. Chromosome fragmentation and entanglements involving multiple chromosomes are observed during meiotic progression in Atmnd1 mutants. Both fragmentation and chromosomal interconnections are only seen in the presence of SPO11-1, the central protein in DSB formation (Grelon et al., 2001). We propose that non-homologous interactions of chromosomes in meiotic prophase I lead to interconnected DNA masses, which can be observed at metaphase-I- and anaphase-I-like stages. Chromatin links that can be seen in metaphase-I-like stages support this view. Nevertheless, this hypothesis has to be proved by identifying non-homologous interactions. The majority of DSBs in Atmnd1 mutants remain unrepaired, hence leading to chromosome fragmentation. This defect is comparable to yeast mnd1 mutants, in which hyper-resected DSBs accumulate (Gerton and DeRisi, 2002; Tsubouchi and Roeder, 2002).

It has been proposed that the Mnd1-Hop2 complex cooperates with Dmc1 to promote stable-strand invasion (Gerton and DeRisi, 2002). Consistent with this proposal, Rad51 and Dmc1 physically interact with Hop2-Mnd1, and a purified Mnd1-Hop2 heterodimer stimulates the in vitro strand-invasion activity of Dmc1 (Chen et al., 2004; Petukhova et al., 2005). Nevertheless, Mnd1-Hop2 does not co-localise with Rad51 on meiotic chromosomes, arguing against a simple model in which the Mnd1-Hop2 complex directly promotes Dmc1 activity at the DSB site (Zierhut et al., 2004). Intriguingly, Arabidopsis Atdmc1 and Atmnd1 (and ahp2) mutants have different phenotypes. The Atdmc1 mutation does not cause chromosome fragmentation, which would be reminiscent of a defect in DSB repair, but instead leads to the formation of univalents, which lack chiasmata (Couteau et al., 1999). It is thought that in the absence of AtDMC1, DSBs are repaired by the sister chromatid in Arabidopsis. In agreement with this conclusion, the additional depletion of RAD51 leads to DNA fragmentation in the Atdmc1 mutant background, indicating that repair of DSBs depends on the presence of RAD51 in Atdmc1 mutants (Siaud et al., 2004). Thus, it appears that in the absence of AtDMC1, repair occurs preferentially by the sister chromatid, whereas its presence promotes repair by the homologous chromosome. The mnd1 and ahp2/hop2 mutants in Arabidopsis and S. cerevisiae are unable to repair DSBs by using the sister chromatid. However, if the homologous chromosome is not present as a repair template, owing to disruption of SC lateral element proteins, such as Red1 or Hop1, S. cerevisiae mnd1 mutants can repair DSBs via the sister chromatid (Zierhut et al., 2004). Further experiments are needed to resolve this issue in Arabidopsis. Other proteins involved in the control of inter-homologue bias during meiotic recombination are probably to be found in the group of RecA-related proteins and amongst accessory proteins of RecA-like proteins, such as the Dmc1 loading factor Tid1/Rdh54. Future efforts are needed to identify additional regulators and to decipher the relationship between RecA-like proteins and the AtMND1-AHP2 complex, and to understand how together they promote inter-homologue bias during meiotic recombination in plants.

All plants were germinated on media containing MS salts and 1% sucrose, and were transferred to soil after 3 weeks. Plants were grown at 22°C with a 16:8 hour light:dark photoperiod. Seeds of the Atmnd1 A. thaliana SALK_110052 line were obtained from the Nottingham Arabidopsis Stock Center (Nottingham, UK). Since Atmnd1 mutants are sterile, the mutant allele was maintained by self-fertilization of heterozygous plants. Atmnd1/spo11-1 double mutants were generated by crossing plants heterozygous for the Atmnd1 mutation to plants heterozygous for an insertion in the SPO11-1 gene (Grelon et al., 2001). Double mutants were identified by PCR of the F2 population, obtained by self-fertilising F1 plants heterozygous for both genes. Transformation of plants heterozygous for the Atmnd1 mutation with pBIB-Hyg/gAtMND1, using Agrobacterium tumefaciens strain GV3101, was performed as described (Bechtold et al., 1993).

DNA from flowers of individual plants was extracted according to Cocciolone and Cone (Cocciolone and Cone, 1993). DNA gel blot analysis of the AtMND1 locus was performed with 1 μg genomic DNA digested with HindIII. A ³²P-labelled genomic fragment amplified by PCR with primers M1 and M4 was used as a probe for hybridisation. Plants were genotyped by PCR with two gene-specific primers and one primer specific for the left border of the T-DNA. Primers M2, M3 and LBa1 were used for the AtMND1 locus. A 990 bp product was generated from wild-type alleles and a 1342 bp product from the Atmnd1 mutant allele. Primers SPO1, SPO2 and LB-BAR2 were used for the SPO11-1 locus, resulting in a 943 bp band and a ∼500 bp product for wild-type and spo11-1 alleles, respectively. For amplification of border regions, primers N610052U and LBSALK1 were combined in one reaction and N610052L and LBSALK1 in another reaction. Sequences of both PCR products were obtained (Genoscreen) and analysed with DNASSIST software (http://www.dnassist.org).

For construction of the complementing construct pBIB-Hyg/gAtMND1, BAC clone ATF19B15, obtained from ABRC (Columbus, Ohio), was digested with XmnI and a 3806 bp genomic DNA fragment, harbouring the AtMND1 gene and its putative promoter region, was ligated to the SmaI-digested plant binary vector pBIB-Hyg (Becker, 1990) and used for transformation.

For construction of the AtMND1-GAL4DB fusion protein vector, the AtMND1 cDNA sequence, contained within plasmid U50561 (GenBank accession number BT005435), was amplified by PCR (Ex-Taq polymerase/TaKaRa, according to the manufacturer's instructions) with primers M1 and M4 and ligated into plasmid pCR2.1.TOPO (Invitrogen). The cDNA was released as a NdeI/BamHI fragment from this vector and ligated into vector pGBKT7 (Clontech) digested with the same enzymes, giving rise to the Y2H vector pGBKT7/AtMND1. For construction of the AHP2-GALAD fusion protein, vector pGAD10 (Clontech) was digested with EcoRI, followed by Klenow treatment and subsequently digested with BamHI. AHP2 cDNA (Schommer et al., 2003) was digested with XbaI, followed by Klenow treatment and subsequently digested with BamHI. Ligation of these DNA fragments resulted in Y2H vector pGAD10/AHP2. For construction of the AHP2-GAL4DB fusion protein vector, AHP2 cDNA, contained within plasmid pGAD10/AHP2, was amplified by PCR (Ex-Taq polymerase/TaKaRa) with primers AHP2_Ncodn and AHP2_PstIup. The PCR fragment was digested with restriction enzymes NcoI and PstI and ligated into vector pGBKT7 digested with the same enzymes, giving rise to Y2H vector pGBKT7/AHP2. For construction of the AtMND1-GAL4AD fusion protein vector, the AtMND1 cDNA, contained within plasmid U50561, was amplified by PCR (Ex-Taq polymerase/TaKaRa) with primers Mnd1_EcoRIdn and MndI_PstIup. The PCR fragment was digested with restriction enzymes EcoRI and PstI and ligated into vector pGAD424 (Clontech) digested with the same enzymes, giving rise to the Y2H vector pGAD424/AtMND1. All PCR generated AtMND1 and AHP2 fragments were sequenced to confirm error-free amplification. pGAD10 and pGAD424 are identical, and only differ in their multiple-cloning sites. Details of all primers used in this study can be found in supplementary material Table S1.

Total RNA was extracted from the indicated tissue using the TriReagent solution (Sigma, St Louis, MO), followed by DNA digestion (DNase, RNase free, 10 U/μl, Roche), phenol/chloroform extraction and RNA precipitation. 2 μg total RNA was used as a template for reverse transcription and generation of AtMND1 and ACTIN2/7 (At5g09810) cDNAs. Synthesis was primed by oligonucleotides M1 and Actin2.2 in one reaction and catalysed by AMV-RT as recommended by the manufacturer (Promega). One-tenth of the cDNA was used for PCR amplification with primers M1 and M4, and Actin2.1 and Actin2.2, respectively, applying standard PCR conditions. For rapid amplification of the 3′ end of the AtMND1 cDNA ends (RACE), 5 μg total RNA from buds of wild-type plants was reverse-transcribed with AMV-RT (Promega) using primer AnchordT, as described above. 5 μl of this reaction was used as a template in a standard PCR reaction with primers Anchor and Mnd1_3RACE1. The PCR reaction was diluted 1:100 and 1 μl was used as a template for a second PCR, with nested primer Mnd1_3RACE2 and primer Anchor. The amplification product was gel-purified, ligated into PCR cloning vector pCR2.1TOPO (Invitrogen) and transformed into E. coli. Plasmids were isolated from five individual transformants and inserted PCR fragments were sequenced. For the RACE of the 5′ end of AtMND1, 3 μg RNA was reverse-transcribed using primer Mnd1_5RACE1 and M-MLV RT as recommended by the manufacturer (Promega). A polyA⁺ tail was generated by incubating half of the cDNA reaction with TdT (MBI Fermentas), according to the manufacturer's manual. 5 μl of this reaction was used in a standard PCR reaction with primers AnchordT and Mnd1_5RACE1. The PCR reaction was diluted 1:100 and 1 μl was used for a second PCR using nested primers Anchor and Mnd1_5RACE2 in a standard PCR reaction. This PCR reaction was diluted 1:100 and 1 μl was used for the third PCR using primer Anchor and the nested primer Mnd1_5RACE3. The amplification product was analysed as described for the 3′ product.

DAPI staining of male meiotic chromosomes was performed as described (Ross et al., 1996). Female meiotic chromosomes were observed as described (Motamayor et al., 2000).

The BAC clones F1N21 (chromosome 1) and F11L15 (chromosome 2) were obtained from ABRC (Columbus, Ohio), and used as probes. BAC DNA was isolated using the Qiagen Mini or Midi Prep kit and labelled by nick translation, following the manufacturer's instructions (Roche Diagnostics), with SpectrumGreen-dUTP (Vysis) and Cy3-dUTP (Amersham Pharmacia Biotech), respectively.

Chromosome spreads were prepared according to Armstrong et al. (Armstrong et al., 2001), except that whole flower buds were digested in 0.33% (w/v) cellulase `Onozuka R-10' (Serva) and 0.33% (w/v) pectolyase (Sigma-Aldrich) for 90 minutes at 37°C, and meiocytes were left in 60% acetic acid at 45°C for a few seconds before re-fixation.

The FISH procedure was modified from Pedrosa et al. (Pedrosa et al., 2001) and Heslop-Harrison et al. (Heslop-Harrison et al., 1991) as follows. Preparations were pre-treated with RNAase for 30 minutes only and chromosomes were denatured in hybridization mix for 3 minutes at 73°C. Photographs were taken on a Zeiss Axioplan microscope (Carl Zeiss) equipped with a mono cool-view CCD camera (Photometrics, Tucson, AZ) and the IPLab spectrum software (IPLab, Fairfax, USA). Digital images were imported into Adobe Photoshop CS version 8 for final processing.

The immunolocalisation of the ASY1 and RAD51 proteins on chromosome spreads was carried out as described (Armstrong et al., 2002), except that whole buds, instead of isolated anthers, were used. No pre-blocking was applied. 10 μl anti-ASY1 (rabbit, 1:1000) and 10 μl anti-AtRAD51 (rat, 1:500), both diluted in blocking buffer (PBS + 0.1% Triton + 3% BSA), were used in combination. The secondary antibodies (20 μl anti-rat Cy3, Dianova/Szabo, 1:1000, and 20 μl anti-rabbit FITC, Sigma, 1:200 in blocking buffer) were applied consecutively and incubated at 37°C for 30 minutes. Slides were mounted in 2 μg/ml DAPI in Vectashield. Photographs were taken as described for FISH.

For ASY1 and SCC3 detection, chromosome spreads were prepared and immunofluorescence microscopy performed as described (Armstrong et al., 2002), using a Leica DMRXA2 fluorescence microscope. The ASY1 polyclonal antibody (Armstrong et al., 2002) was used at a working dilution of 1:500. The SCC3 antibody was described (Chelysheva et al., 2005) and used at a working dilution of 1:2000. Images were captured with a coolSNAP camera (Roper Scientific) driven by Openlab software (Improvision). These images were then processed with Adobe Photoshop 6.0 to improve image quality.

Indicated combinations of plasmids were co-transformed into yeast strain AH109 (Clonetech) following the protocol for high-efficiency transformation of yeast (Burke et al., 2000). Following transformation, colonies were selected for the presence of the plasmids, inoculated in liquid synthetic drop-out (SD) media (lacking the amino acids leucine and tryptophan, with the exception of untransformed strain AH109, which was grown in YPAD), grown to saturation and plated onto SD media plates, lacking the indicated amino acids. SD media plates lacking the amino acids leucine, tryptophan and histidine (SD -Leu/-Trp/-His) were supplemented with 1 mM 3-Amino-1,2,4-triazol (Sigma). Serial 1:5 dilutions were made in water and 3 μl of each dilution was used to yield one spot. Plates were incubated at 30°C for two (SD -Leu/-Trp) or three (SD -Leu/-Trp/-His) days before taking pictures. The pictures were taken with a Nikon Coolpix 4500 digital camera and processed in Adobe Photoshop CS version 8.

We are grateful to Chris Franklin for sharing antibodies directed against ASY1 and RAD51. We thank Carla Schommer for helping constructing pGAD10/AHP2 and Franz Klein, Josef Loidl and Karel Riha for critical reading of the manuscript and/or stimulating discussions. This work was supported by a grant from the Austrian Science Foundation (P18036) to P.S. and by the Austrian Academy of Sciences (GMIGmbH).