OsSpo11-4, a Rice Homologue of the Archaeal TopVIA Protein, Mediates Double-Strand DNA Cleavage and Interacts with OsTopVIB

DNA topoisomerase VI from Archaea, a heterotetrameric complex composed of two TopVIA and two TopVIB subunits, is involved in altering DNA topology during replication, transcription and chromosome segregation by catalyzing DNA strand transfer through transient double-strand breaks. The sequenced yeast and animal genomes encode only one homologue of the archaeal TopVIA subunit, namely Spo11, and no homologue of the archaeal TopVIB subunit. In yeast, Spo11 is essential for initiating meiotic recombination and this function appears conserved among other eukaryotes. In contrast to yeast and animals, studies in Arabidopsis and rice have identified three Spo11/TopVIA homologues and one TopVIB homologue in plants. Here, we further identified two novel Spo11/TopVIA homologues (named OsSpo11-4 and OsSpo11-5, respectively) that exist just in the monocot model plant Oryza sativa, indicating that at least five Spo11/TopVIA homologues are present in the rice genome. To reveal the biochemical function of the two novel Spo11/TopVIA homologues, we first examined the interactions among OsSpo11-1, OsSpo11-4, OsSpo11-5, and OsTopVIB by yeast two-hybrid assay. The results showed that OsSpo11-4 and OsTopVIB can self-interact strongly and among the 3 examined OsSpo11 proteins, only OsSpo11-4 interacted with OsTopVIB. Pull-down assay confirmed the interaction between OsSpo11-4 and OsTopVIB, which indicates that OsSpo11-4 may interact with OsTopVIB in vivo. Further in vitro enzymatic analysis revealed that among the above 4 proteins, only OsSpo11-4 exhibited double-strand DNA cleavage activity and its enzymatic activity appears dependent on Mg2+ and independent of OsTopVIB, despite its interaction with OsTopVIB. We further analyzed the biological function of OsSpo11-4 by RNA interference and found that down-regulated expression of OsSpo11-4 led to defects in male meiosis, indicating OsSpo11-4 is required for meiosis.


Introduction
Topoisomerase VI (TopVI), originally identified from the hyperthermophilic archaeon Sulfolobus shibatae, regulates DNA topology by catalyzing DNA strand transfer through transient double-strand breaks in the presence of Mg 2+ and ATP and is the only topoisomerase that can relax positively supercoiled DNA, thereby being essential to DNA replication, transcription and chromosome segregation [1,2]. TopVI consists of 2 distinct subunits, TopVIA and TopVIB. TopVIA is a catalytic subunit responsible for DNA binding and cleavage, and TopVIB is involved in ATP binding and hydrolysis. TopVIA is characterized by 5 conserved functional motifs, I to V, with 2 domains: CAP (catabolite activator protein) and toprim (topoisomerases as well as DNA primases). Motif I and II constitute a helix-turn-helix fold of the CAP domain, which exists widely in DNA-binding proteins. The conserved toprim domain spanning motifs III to V is involved in DNA binding and cleavage, with the DXD sequence responsible for coordinating metal ions [3,4,5]. The N-terminal region of S. shibatae TopVIB is characterized by a Bergerat fold, which is found in proteins of the GHKL family (DNA gyrase, Hsp90, bacterial histidine kinases and MutL families) [6]. The Bergerat fold, consisting of 3 motifs (B1 to B3), is responsible for ATP binding and hydrolysis [6,7]. The C-terminus of TopVIB is structurally homologous to that of GyrB, termed the transducer domain (motif B4), which is required for the transmission of structural signals and conformational changes [8].
In this study, we identified two novel Spo11/TopVIA homologues (named OsSpo11-4 and OsSpo11-5, respectively) from the model monocot plant rice (O. sativa L. ssp. japonica). We observed the interaction of OsSpo11-4 and OsTopVIB by both yeast two-hybrid and pull-down assays. We found that OsSpo11-4, rather than OsSpo11-1 or OsSpo11-5, was able to cleave doublestrand DNA (dsDNA) in vitro. This is the first in vitro enzymatic evidence that a plant Spo11/TopVIA homologue has doublestrand DNA cleavage activity. We further analyzed the function of OsSpo11-4 in growth and development using RNAi approach and showed that OsSpo11-4 was essential for meiosis.

Identification of SPO11 and TOPVIB homologues in rice
Previous studies have identified three TopVIA/Spo11 homologues in rice, namely OsTop6A1/OsSpo11-1, OsTop6A2/ OsSpo11-2, and OsTop6A3/OsSpo11-3 [23,35]. Here, using amino acid sequences of the yeast Spo11 protein as query to search the rice genome sequence in TIGR (http://rice.plantbiology. msu.edu/) with the TBLASTN program, we identified 3 proteins homologous to Spo11 in japonica rice: one is the identified OsSpo11-1 protein (NCBI accession number: GU170363); and the other two are novel proteins which were designated OsSpo11-4 (GU177866) and OsSpo11-5 (GU170364) respectively, as putative members of the Spo11/TopVIA family. The full-length cDNAs of OsSpo11-4 and OsSpo11-5 were isolated by RT-PCR and RACE, using gene-specific primers. The genomic sequences of the 2 novel genes were further examined by BLASTN searches of the TIGR rice genomic database with respective full-length cDNA used as queries. Each of the 2 genes is present in the rice genome as a single-copy gene (OsSPO11-4: Os12g0622500 and OsSPO11-5: Os11g0545300). Comparison of cDNAs with genomic DNA sequences showed that OsSPO11-5 consists of 11 introns and 12 exons, with the largest open reading frame (ORF) of 2145 bp encoding 714 putative amino acids, whereas OsSPO11-4 contains only 1 intron and encodes a predicted protein consisting of 487 amino acids ( Figure S1). Further searches of public database using the two genes as queries showed that OsSPO11-4 and OsSPO11-5 are also present in the genome of indica rice (EAY83148 and EEC68321, respectively) but absent in completely sequenced genomes of other plants that include dicots such as Arabidopsis and monocots such as maize, indicating that both genes exist just in rice genome.
In order to analyze the evolutional relationship of Spo11/ TopVIA homologues, members of this family were identified from plants, animals, fungi and representatives of archaea using Blast searches against public database. Phylogenetic analyses showed that plant Spo11 homologues fell into 4 distinct groups, which are respectively represented by Spo11-1, Spo11-2 and Spo11-3 in Arabidopsis and rice, and OsSpo11-5&OsSpo11-4 ( Figure 2). The five OsSpo11 proteins are more closely homologous to their corresponding Spo11/TopVIA members from other organisms than to each other except for OsSpo11-5 with OsSpo11-4. Moreover, it has been known that Spo11-1 and Spo11-2 are involved in meiosis, Spo11-3 in endoreduplication, while functions of OsSpo11-5 and OsSpo11-4 are unclear previously. Therefore, the five SPO11 genes in rice might not arise through recent duplication events, but represent different ancient paralogues.
The identified OsTOP6B of indica rice also exist in japonica rice (AY371050), which is named OsTOPVIB in our study. OsTopVIB contained the 4 motifs (B1 to B4) conserved in TopVIB proteins from other organisms [23] ( Figure S2). In archaeal TopVIB, Asn in motif B1 (Asn 42 in SsTopVIB) is identified as Mg 2+ binding residue and Asp and Gly residues in motif B2 are involved in nucleotide contacts (Asp 76 and Gly 80 in SshTopVIB). These residues are part of the GHKL domains involved in ATP binding and the nucleotide-binding pocket [8]. Our analysis showed that these residues were conserved in the rice TopVIB protein (Asn 93 , Asp 187 and Gly 191 in OsTopVIB, respectively). The middle region of OsTopVIB matched with the helix-two turns-helix domain characterized by 9-to 12-amino acid insertion peptides, which functions as a linker to position the N-and C-terminal domains in archaea [8]. A postulated motif B4 was identified in OsTopVIB; this motif B4 in archaeal TopVIB is related to the ATPase domain of GyrB, which is referred to as a transducer domain and has been proposed to mediate intersubunit communication by structurally transforming signals from the ATP binding site of the GHKL domain to the DNA binding and cleavage domains of the holoenzyme [2,8].

OsSpo11-4 interacts with OsTopVIB
Considering that archaeal TopVI functions as an A 2 B 2 heterotetramer and rice genome encodes Spo11/TopVIA and TopVIB homologs, we first examined interactions among OsSpo11s and OsTopVIB using yeast two-hybrid assay. The results demonstrated that both OsSpo11-4 and OsTopVIB can strongly self-interact, which suggests that each may form a homodimer. OsSpo11-1 had no detectable self-interaction. OsSpo11-5 had undetectable self-interaction, but the N-terminal extension sequence (OsSpo11-5N) of OsSpo11-5 can strongly interact with the full-length protein and relatively weakly with the C-terminal part (OsSpo11-5C) of the protein. In addition, relatively weak interactions were detected in OsSpo11-5N/ OsSpo11-1 and OsSpo11-5C/OsSpo11-4 ( Figure 3A and B). Importantly, this analysis revealed that OsTopVIB interacts with OsSpo11-4 strongly but not with OsSpo11-5 and OsSpo11-1.
We further validated the interaction between OsSpo11-4 and OsTopVIB by GST pull-down assay. The purified pET-OsSpo11-4 (purified under native conditions) and pET tag (from pET-32a vector, control) were precleared with GST preabsorbed in GST affinity resin, then yeast expressed, GST affinity resin-binding GST-OsTopVIB fusion protein was incubated with the precleared pET-OsSpo11-4 or pET tag. Western blot analysis revealed that pET-OsSpo11-4 (about 72 kD) could interact with GST-OsTop-VIB ( Figure 3C).
We also examined the interaction of OsSpo11-4 and OsTop-VIB using purified OsTopVIB proteins and cell protein extracts from rice flowers ( Figure 3D). The OsTopVIB-GST fusion or GST (control) was incubated with protein extracts from flowers at the meiosis stage. After the resulting resin was washed to remove unspecified proteins, specifically bound proteins were analyzed by SDS-PAGE and western blotting. An antibody against GST detected the GST band (about 26 kD) in GST pulldowns and detected the GST-OsTopVIB band (about 103 kD) in GST-OsTopVIB pulldowns, which indicates that GST and GST-OsTopVIB, respectively, were bound to the GST affinity resin. The antibody against OsSpo11-4 detected the OsSpo11-4 band (55 kDa) in the GST-OsTopVIB pulldowns but not in the GST control ( Figure 3D). These results indicated possible interaction of OsTopVIB and OsSpo11-4 in vivo. In short, the yeast two-hybrid and pull-down results suggest that OsSpo11-4 and OsTopVIB may interact to form a functional complex in vivo. Gaps are shown by dashes. Black boxes indicate conserved residues, and grey boxes indicate similar residues. The respective amino acid position of each sequence is given on the right. The active tyrosine residue and the DXD sequence identified in archaeal TopVIA are marked with asterisks in motif I and motif V, respectively. Sequences used here are OsSpo11-5C (accession No. AY154916), OsSpo11-1 (GU170363) and OsSpo11-4 (GU177866) from Oryza sativa; AtSpo11-1 (AJ251989), AtSpo11-2 (AJ251990) and AtSpo11-3 (AL162973) from Arabidopsis thaliana; ScSpo11 (P23179) from Saccharomyces cerevisiae; SpRec12 (P40384) from Schizosaccharomyces pombe; NcSpo11 (CAB88597) from Neurospora crassa; DmSpo11 (AAC61735) from Drosophila melanogaster; CeSpo11 (CAA92974) from Caenorhabditis elegans; MmSpo11 (Q9WTK8) from Mus musculus; HsSpo11 (Q9Y5K1) from Homo sapiens; SsTopVIA (O05208) from Sulfolobus shibatae and MjTopVIA (Q57815) from Methanobacterium janaschii. doi:10.1371/journal.pone.0020327.g001

OsSpo11-4 is able to catalyze double-strand DNA cleavage in vitro
To address the enzymatic properties of OsSpo11s and OsTopVIB, we used a eukaryotic yeast expression system for generation of soluble proteins that are often critical for the structure and activity of eukaryotic proteins. We purified OsSpo11-1, OsSpo11-4, OsSpo11-5 and OsTopVIB by GST affinity chromatography ( Figure 4A, 4C and 4D). These native proteins were further collected by removing the GST tag ( Figure 4B). The purified native OsSpo11-1, OsSpo11-4, Os-Spo11-5 and OsTopVIB proteins were incubated with kDNA to assay the activity of decatenation, which is specifically catalyzed by type II DNA topoisomerase. kDNA is a catenated DNA extracted from the kinetoplast of insect trypanosome Crithidia fasciculate, which is composed of the aggregation of interlocked DNA circles with high molecular size. These high-molecular-size networks cannot migrate from the loading pore. When the networks were cleaved, minicircular DNAs (cleaved and resealed) or linear DNAs (lkDNA, unresealed) were released and quickly moved into the gel. As shown in Figure 5A, the decatenated kDNA marker (dkDNA, prepared by decatenation reaction catalyzed by topoisomerase II) showed relative positions of open circular nicked DNA (OC) and closed circular monomers (CC). Among the examined proteins, only OsSpo11-4 could catalyze the catenated kDNAs into free linear forms completely, which have a molecular size equal to that of the lkDNA marker. Moreover, the decatenation reaction rate was proportioned to the concentration of OsSpo11-4 ( Figure 5C), these results suggest that OsSpo11-4 can specifically cleave kDNAs to produce double-strand breaks. OsSpo11-4 had the enzymatic activity alone and appeared to be independent of OsTopVIB. This finding is inconsistent with archaeal TopVI, in which catalytic subunit TopVIA performs decatenation of tangled DNA in the presence of the B subunit [1,2]. Furthermore, when kDNA was decatenated by archaeal (S. shibatae) TopVI, open circular and relaxed, covalently closed circular DNA rings were produced, which indicates that archaeal TopVI can re-ligate the broken dsDNA ends to covalently closed DNA rings [1,2], whereas OsSpo11-4 appeared not to have this activity.
We also examined the cleavage activity of these proteins using the pUC18 plasmid as reaction substrate. Consistent with the above data, only OsSpo11-4 had the activity to cleave pUC18 plasmids to linear DNAs, and other proteins had no detectable activity ( Figure 5B). Together, these results clearly indicate that OsSpo11-4 itself can cleave dsDNA, and its enzymatic activity appears independent of OsTopVIB.
Although the DNA cleavage activity of OsSpo11-4 was independent of OsTopVIB, it was strictly Mg 2+ -dependent. OsSpo11-4 created DSBs on kDNA substrates only in the presence of Mg 2+ , which leads to accumulation of linearized kDNA ( Figure 5D). The reaction rate enhanced when the concentration of the Mg 2+ increased up to 10 mM; however, concentrations of Mg 2+ higher than 10 mM did not further increase the amount of kDNA, on the contrary, the cleavage activity of OsSpo11-4 declined when exposed to higher levels of Mg 2+ ( Figure 5E).

OsSPO11-4 is expressed preferentially in flowers
Semi-quantitative RT-PCR was used to evaluate mRNA levels of OsSPO11-4 in different rice tissues, including 2-week-old leaves, young roots, young shoots and flowers at different meiotic stages. OsSPO11-4 was expressed at the highest level in flowers, in which pollen mother cells were at the meiotic phase, and at relatively lower levels in roots, buds and leaves, with no detectable expression in mature pollen grains ( Figure 6A). The levels of and bait (BD-fusion) constructs were cotransformed into yeast strain AH109. A, X-gal staining results and a comparison with the positive control (con) supplied by the Yeast Two Hybrid Kit. These colonies were first screened by growth on QDO (SD/-Ade/-His/-Leu/-Trp) medium lacking adenine, histidine, leucine and tryptophan, and then analyzed by X-gal staining for 7 h. B, Interaction evaluated according to X-gal staining results: strong (++), weak (+) or no (2) interaction. s5, OsSpo11-5; s5N, N-extension of 412 amino acids of OsSpo11-5; s5C, C-terminal 302AAs TopVIA region of OsSpo11-5; s1, OsSpo11-1; s4, OsSpo11-4; VIB, OsTopVIB. C and D, Pull-down assay. C, Purified pET-OsSpo11-4 or pET tag (from pET32a vector) expressed in E. coli incubated with GST-OsTopVIB-bound resin. The pulldowns were examined by western blot analysis with an antibody against pET-Spo11-4. Lane 1, purified pET tag alone. Lane 2, precleared pET tag using GST-binding resin. Lane 3, resin-bound GST-OsTopVIB incubated with precleared pET tag. Lane 4, purified pET-OsSpo11-4 fusion protein alone. Lane 5, precleared pET-OsSpo11-4 fusion protein using GST-binding resin. Lane 6, resin-bound GST-OsTopVIB incubated with precleared pET-OsSpo11-4 fusion protein. D, Proteins extracted from rice flowers at the male meiosis stage incubated with resin-bound GST (con) or GST-OsTopVIB (VIB). The pulldowns were subjected to SDS-PAGE and then Western blot analysis with an antibody against GST (GST Ab) or OsSpo11-4 (S4 Ab). doi:10.1371/journal.pone.0020327.g003 OsSPO11-4 mRNA appeared to be developmentally dependent in flowers, showing the highest levels at the pre-meiotic and meiotic stages, and the gradually decreased levels with advancing development ( Figure 6A).
We further used RNA in situ hybridization to examine the temporal and spatial expression patterns of OsSPO11-4 in flowers using DIG-labeled antisense and sense (control) probes. In flowers, the OsSpo11-4 antisense probe detected strong signals in premeiotic and meiotic pollen mother cells, tapetal cells and meiotic ovaries. No signals were detected in spores at the uninucleate microspore and subsequent stages or in other flower organs such as lemma and palea. Sense probes did not detect signals ( Figure 6B). The expression pattern suggests that OsSpo11-4 might function in pre-meiotic and meiotic pollen mother cells.

OsSpo11-4 is required for efficient meiosis
To address the in vivo function of OsSpo11-4, we obtained RNAi lines of OsSpo11-4 using gene-specific cDNA fragments. The presence of the transgene in hygromycin-resistant rice planets was examined by PCR with primer pairs localized to the spacer and inserted cDNA sequences on the OsSPO11i vector. The PCRpositive transgenic lines showed decreased seed setting rate to different degrees as compared with wild-type plants ( Figure 7A). Furthermore, 6 lines representing different sterile phenotypes (L11, L19, L28, L39, L39, L45), identified to have a single-copy insertion of the transgene by Southern blot hybridization, were used to generate T1 plants. T1 generations showed stable and heritable sterile phenotypes (17% seed setting rate in one line, ranging from 50% to 58% for 4 lines and 72% in 1 line, with 91% in the wild type) ( Figure 7B). The endogenous transcripts of OsSPO11-4 were considerably downregulated in OsSpo11-4i lines as compared with the wild-type control ( Figure 7D). To evaluate whether the sterile phenotype involved pollen abortion, we examined the viability of pollen grains from OsSpo11-4i lines at maturity using Alexander staining, whereby viable pollen grains are stained red and non-viable ones stained green [37] (Figure 7C). Most of the wild-type pollen grains (96%, n = 1525) were viable, whereas a high proportion of pollen grains from OsSpo11-4i lines were non-viable (31,76% variable in different lines), which suggests that pollen abortion in OsSpo11-4i lines related to the sterile phenotype.
To determine whether the sterile pollen grains resulted from meiotic defects in RNAi plants, we examined the meiotic chromosome behavior using DAPI-stained chromosome spreads of male meiocytes from the RNAi line L19, along with the wildtype control (Figure 8). In wild-type plants ( Figure 8A-I), chromosomes in leptotene cells appear as thin threads. Homologous chromosomes begin to associate side by side at zygotene and fully synapse and condense into thick threads at pachytene. Synapsed homologous chromosomes begin to separate at diplotene, and 12 bivalents become highly condensed at diplotene and diakinesis. Thereafter, 12 highly condensed bivalents align on equator plates at metaphase I and are subject to reductional division at anaphase I. The segregated univalents in each pole were partially decondensed at telophase I, finally generating dyads. During meiosis II, the 2 daughter cells divide simultaneously with parallel orientations of spindles, finally separating to generate 4 haploid tetrads.
Male meiocytes from OsSpo11-4i lines did not display obvious aberrance from premeiotic interphase to middle zygotene ( Figure 8A1 and 8B1) as compared with wild-type male meiocytes ( Figure 8A and 8B). The detectable defects appeared in pachytene male meiocytes; 20.4% of the examined male meiocytes at this phase (n = 298) had chromosome segments ( Figure 8C1). For male meiosis entering into late diplotene and diakinesis, 12 bivalents appeared in wild-type male meiocytes ( Figure 8D and 8E); however, 40.7% (n = 452) of OsSpo11-4i male meiocytes had more than 12 distinguishable chromosomes ( Figure 8D1 and 8E1), which indicates the presence of univalents or chromosome fragments. At metaphase I, these aberrant chromosomes dispersed throughout the nucleus rather than aligning on the metaphase plate ( Figure 8F1); lagging or/and unequally segregated chromosomes were observed at anaphase I ( Figure 8G1). A mixture of normal and aberrant dyads existed at telophase I ( Figure 8H1). These defects led to aberrant chromosome behavior and unequal separation of chromosomes at meiosis II, generating triads and polyads with variable chromosome contents ( Figure 8I1 and 8I2).

Discussion
The archaeal TopVI, a heterotetramer composed of two A subunits (TopVIA) and two B subunits (TopVIB), has the ability to pass DNA double strands through each other. When the enzymatic reaction occurs, TopVI cleaves one DNA duplex to open a DNA gate and captures another DNA duplex to pass though the gate, then rejoins the two ends of the cleaved DNA duplex to close the gate; this topoisomerase activity enables TopVI to solve topological problems of DNA during replication, transcription, recombination and chromosome segregation [38]. Yeast and currently sequenced animal genomes contain one homologue of TopVIA, namely Spo11, and no homologue of TopVIB [27]. The Spo11 protein maintains the activity of creating double-strand breaks but does not rejoin the DNA breaks after cleavage [39]; Spo11 plays a conserved role in initiation of homologous recombination during meiosis, which requires its cleavage activity [4,9]. Different from the yeast and animal genomes, the completely sequenced genomes of higher plants have at least three Spo11/ TopVIA homologues and one TopVIB homologue [27]. Previous studies have revealed three Spo11/TopVIA homologues in Arabidopsis [21,22] and their corresponding proteins in rice [23]. In this study, we identified two novel Spo11/TopVIA homologues (OsSpo11-4 and OsSpo11-5), which exist only in the rice genome. Therefore there are at least five Spo11/TopVIA homologues in rice. These plant Spo11/TopVIA homologues share low similarity with each other (20,30%), suggesting they might be different in Figure 5. Double-strand DNA cleavage catalyzed by purified OsSpo11 and OsTopVIB proteins. Each purified OsSpo11 and OsTopVIB protein or a combination shown above the image was added into a reaction mixture containing substrate, and purified GST was used as a control. After reaction, the mixture was subjected to agarose separation (for details, see ''Materials and methods''). dkDNA, decatenated kDNA Marker; lkDNA, linear kDNA marker; kDNA, kinetoplast DNA; s5, OsSpo11-5; s1, OsSpo11-1; s4, OsSpo11-4; VIB, TopVIB; GST, the GST protein control. A, kDNA as a substrate. dkDNA shows relative positions of open circular nicked DNA-OC, and relaxed, closed circular monomers-CC; dkDNA and lkDNA markers were provided by the Topoisomerase Assay Kit, kDNA was used as a catenated DNA reference after incubation in a reaction mixture without protein. B, pUC18 plasmid as a substrate. pUC18 refers to a reaction containing buffer and pUC18 plasmids only (c, circular pUC18 marker). EcoRI refers to pUC18 plasmids digested by EcoRI, which cuts pUC18 only once (l, linear pUC18 marker). C, DNA cleavage reaction rate is proportioned to the concentration of OsSpo11-4. s4 refers to only reaction buffer and OsSpo11-4 were added, while 1,6 refer to 0.5 mM, 0.4 mM, 0.3 mM, 0.2 mM, 0.1 mM and 0 mM OsSpo11-4 were added to the standard reaction, respectively. Cleavage activity was determined using kDNA decatenation assays. D, the effect of Mg 2+ on OsSpo11-4 activity. 0,5 refer to 0 mM, 2.5 mM, 5 mM, 7.5 mM, 10 mM and 12.5 mM Mg 2 were added to the standard reaction (kDNA decatenation), respectively. E, reaction rate quantification of 0,5 in panel D. Reaction rate was determined as a percentage of linear kDNA generated compared to the total kDNA added. doi:10.1371/journal.pone.0020327.g005 functions. Genetic and functional analysis of the three Spo11/ TopVIA homologues in Arabidopsis have demonstrated that AtSpo11-1 and AtSpo11-2 have a function similar to the yeast Spo11 protein [25,26,28] whereas AtSpo11-3 may interact with AtTopVIB to form a putative TopVI complex and functions in somatic endoreduplication [29,30,31]. The recent study in rice shows that OsSpo11-1 is necessary for meiotic pairing and crossover formation, indicating that Spo11-1 plays a conserved role in Arabidopsis and rice [35]. The function of Spo11-2 and Spo11-3 might also be highly conserved among plants; however, since OsSpo11-4 and OsSpo11-5 are present just in rice, they may have roles distinct from Spo11-1, Spo11-2 and Spo11-3.
Our enzyme activity analysis showed that OsSpo11-4 itself can catalyze DNA cleavage in vitro without the help of OsTopVIB and OsSpo11-4 had no detectable activity in resealing the broken ends in vitro, with or without OsTopVIB. This in vitro enzymatic activity of OsSpo11-4 is distinct from that of archaeal TopVIA, which cleaves DNA only in the presence of TopVIB and reseals the broken DNA ends after cleavage [1]; but it is similar to that of the Spo11 protein from yeast and animals, which produce doublestrand breaks in the absence of TopVIB and does not rejoin DNA breaks [39]. It is interesting that although OsSpo11-4 is more analogous to Spo11 than to TopVIA in terms of in vitro enzymatic features, it interacts with OsTopVIB, which is similar to TopVIA but different from Spo11. Our yeast two-hybrid assay revealed that both OsSpo11-4 and OsTopVIB can self-interact, and OsTopVIB interacts only with OsSpo11-4 among the 3 examined Spo11 proteins (the other two are OsSpo11-1 and OsSpo11-5). Further pull-down assay confirmed the interaction between OsTopVIB and OsSpo11-4. These results indicate that as a TopVIA homologue, OsSpo11-4 might combine with OsTopVIB to form a TopVI heterotetramer similar to that of archaea.
One question raised is whether the function of OsSpo11-4 is more similar to that of yeast and animal Spo11 or to that of archaeal TopVIA. It is possible that the enzymatic activity of OsSpo11-4 in vivo is the same as that in vitro and OsSpo11-4 plays a role analogous to Spo11, which is creating double-strand breaks to initiate meiotic recombination; as an OsSpo11-4 interactive protein, OsTopVIB might be unnecessary for DNA cleavage but function as an accessory factor similar to Mei4, Ski8/Rec103, Xrs2, Rec102, Rec104, Rec114, Mer2/Rec107, Mre11 and Rad50 in budding yeast, which are essential for other processes during meiotic recombination initiation [40,41,42,43,44]. Similar to this hypothesis, the Arabidopsis Spo11-2 protein, which can interact with AtTopVIB in yeast two-hybrid assay, is required for meiotic recombination [22,26]. Alternatively, OsSpo11-4 might function as the A subunit of a putative topoisomerase similar to archaeal TopVI, but the full enzymatic activity in vivo requires the help of other proteins besides OsTopVIB. Similar to the second hypothesis, in Arabidopsis, RHL1, BIN4 and MID have been identified as components of the TopVI complex constituted by AtSpo11-3 and AtTopVIB on the basis of their interaction with AtSpo11-3 [32,33,34].
Our cytological analysis demonstrated that downregulated OsSpo11-4 mediated by RNAi led to aberrant meiosis in rice and a proportion of the RNAi meiocytes showed chromosome fragmentation from the pachytene stage. During meiotic recombination, double-strand breaks are generated by Spo11 and then processed and repaired by Rad51, Dmc1 and other accessory proteins [45,46], thus chromosome fragmentation may result from failure to repair DSBs rather than from failure to generate DSBs. Chromosome fragmentation has been observed in mutants defective in genes required for DSB repair, for example, RAD51C in Arabidopsis [47]. Therefore, meiotic chromosome fragmentation caused by knockdown of OsSpo11-4 indicates that OsSpo11-4 is possibly involved in processing or repairing DSBs rather than DSB formation during meiotic recombination, which indirectly supports the second hypothesis. Considering that OsSpo11-4 can catalyze DNA cleavage and interact with OsTopVIB, we propose that OsSpo11-4 and OsTopVIB might form a topoisomerase complex similar to archaeal TopVI and play a role in decatenation or/and solving DNA topological problems arising during meiotic DSB processing or repair. Further analyses of osspo11-1 osspo11-4 double mutants and identification of components of the putative TopVI complex (consisting OsSpo11-4 and OsTopVIB) will be important for understanding the in vivo biochemical and biological functions of the OsSpo11-4 protein.
It remains an open question why the two additional Spo11/TopVIA homologues, OsSpo11-4 and OsSpo11-5, are required for rice.

Plant materials
Seedlings of rice Zhonghua 10 (O. sativa L. ssp. japonica) were planted as described previously [48]. Flowers at different developmental stages were collected. Young leaves were harvested from 3-week-old plants. Young roots and buds were taken from seedlings germinated on sterile-water-soaked papers.

Phylogenetic analysis
Phylogenetic analyses were performed to infer the evolutionary relationships of the Spo11, TopVIA, and TopVIB homologues. Homologues of Spo11, TopVIA, and TopVIB in archaea and eukaryotes were identified from public databases using Blast searches. These protein sequences were used as queries for BlastP searches of the NCBI nonredundant database, and TBlastN searches of the NCBI databases of expressed sequence tags (ESTs) and high-throughput genome sequences. Homology of these proteins was conformed by multiple sequence alignments using ClustalX (version 1.8) and phylogenetic analyses using Phyml software (version 3.0, only for Spo11/TopVIA).
Yeast SP-Q01 cells harboring recombinant plasmids from EMM (Q-bio gene) plates were inoculated in EMM liquid media at 30uC. After reaching a mid-log phase (OD600 = 0.5), cells were harvested by centrifugation at 12006g and washed once with an extraction buffer (0.2 M Tris-HCl, pH 8.0, 10 mM EDTA, 150 mM ammonium sulfate, 50% glycerol, 1 mM PMSF, 2 mM DTT). The pelleted cells were resuspended in 500 ml PBS buffer (0.14 M NaCl, 2.7 mM KCl, 10.1 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3), and then broken by vortexing at 4uC for 5 min in the presence of acid-washed glass beads (425-600 mm in diameter, Sigma). Supernatant was collected by centrifugation at 95006g. The GST fusion proteins were purified by GST affinity chromatography with Glutathione Sepharose 4B (Amersham) according to the manufacturer's instructions. Finally, the purified native proteins were obtained by removing GST tags with use of biotinylated thrombin (Novagen) (2.5 units/25 mg GST fusion proteins) following the manufacturer's method.

GST pull-down assay
To generate a polyclonal antibody against OsSpo11-4, a 360-bp fragment of OsSpo11-4 cDNA corresponding to amino acids 1 to 120, which has no similarity to the other 2 OsSpo11 proteins, was amplified with primers P5 and P19 (59-ATACTCGAGGA-GAAACCTTGACTTCCT-39, XhoI), cloned into the EcoRI/SalI sites of pET-32a(+) (Novagen), and confirmed by sequencing. Escherichia coli BL21 (DE3) cells harboring the recombinant plasmid were cultured in LB medium at 37uC to reach an exponential phase, and thereafter the culture was induced immediately with 0.2 mM IPTG. The recombinant protein purified by Ni 2+ affinity chromatography (QIAGEN) was used to generate a polyclonal mice antibody against pET-OsSpo11-4.
The GST pull-down assay was performed as described [53,54]. Briefly, for pull-down assay with purified OsSpo11-4 protein, the full-length OsSpo11-4 ORF sequence amplified with primers P5 and P6 given above was cloned into the EcoRI/XhoI sites of pET-32a(+) and confirmed by sequencing. pET-OsSpo11-4 fusion protein and pET tag were purified from E. coli BL21 (DE3) cells harboring the recombinant plasmids with Ni 2+ affinity chromatography as described above. pET-OsSpo11-4 and pET tag were changed into 0.5% NP-40 lysis buffer [20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5% NP-40, 16cocktail (Roche)], and then were precleared by incubating with Glutathione Sepharose resin preabsorbed by excessive GST proteins on a rocker at 4uC for 2 h. The supernatants were collected and incubated with an equivalent amount of GST-OsTopVIB preabsorbed to Glutathione Sepharose resin on a rocker at 4uC for 6 h. The resulting resins were washed 3 times with the lysis buffer, and thereafter boiled in SDS-PAGE sample buffer to elute bound proteins (termed as pulldowns). The purified pET tag and pET-OsSpo11-4 fusion, the precleared pET tag and pET-OsSpo11-4 fusion, and the GST pulldowns were separated on SDS-PAGE and hybridized with an antibody against pET-OsSpo11-4 fusion protein. SDS-PAGE and western blot were performed as described previously [51].
For pull-down assay using protein extracts, purified yeastexpressed GST-TopVIB or GST (control) as described previously were bound to Glutathione Sepharose 4B resin. Proteins extracted from young rice flowers at the meiosis stage as described [55] were incubated with GST-OsTopVIB or GST preabsorbed to Glutathione Sepharose 4B resin on a rocker at 4uC for 6 h. The resulting Glutathione Sepharose resins were washed 3 times with 0.5% NP-40 lysis buffer and then boiled in SDS-PAGE sample buffer to elute bound proteins. GST and GST-OsTopVIB pulldowns were separated on SDS-PAGE and hybridized with monoclonal antibody against GST and antibody against OsSpo11-4, respectively.

DNA topoisomerase activity assay
The decatenase activity was examined by use of the Topoisomerase II Assay Kit (Topogen) in a standard reaction mixture (20 ml) containing 2 ml buffer A (50 mM Tris-HCl, pH 8.0, 120 mM KCl, 7.5 mM MgCl 2 , 0.5 mM DTT, and 30 mg/ml BSA), 2 ml buffer B (10 mM ATP), 100 nM of proteins and 0.2 mg of kinetoplast DNA (kDNA). The mixture was incubated at 32uC for 30 min, and then subjected to 1% agarose gel in 16TAE buffer with ethidium bromide included in the gel. For each assay, 3 independent biological replicates were performed.
The dsDNA cleavage activity was analyzed as described above, except pUC18 plasmid (TIANGEN) was used instead of kDNA as a reaction substrate.
The effects of Mg 2+ concentration on TopVI activity were determined using standard conditions with changing variable concentrations of Mg 2+ at a time. DNA bands were visualized by UV, photographed, and analyzed by image Pro-plus software (version 5.1). DNA cleavage reaction rate was determined as a percentage of linear kDNA generated compared to the total kDNA added.

Semi-quantitative RT-PCR and in situ hybridization
The transcript levels of OsSpo11-4 mRNA in different wild-type rice tissues and in RNAi transgenic plants were examined by semiquantitative RT-PCR with use of SuperScript II RNase H2 Reverse Transcriptase (Invitrogen) and primers P5 and P6 given above. The transcript of tubA gene (accession No. X91806) served as an internal control [50]. PCR was performed for 26 cycles.
To determine the expression profiles of OsSpo11-4 in flowers by in situ hybridization, flowers were fixed and further processed as described previously [50]. A 360-bp cDNA fragment spanning nucleotides 350 to 710 of OsSpo11-4 ORF, which showed no similarity to the other OsSPO11 genes, was amplified by use of primers P20 (59-CCCTGAACTTAACTTGCC-39) and P21 (59-AGATAATCCACCTTGACC-39), cloned into pGEM-T vector (Promega) and confirmed by sequencing. DIG-labeled sense and antisense RNAs were synthesized by in vitro transcription (Roche) of the linearized recombinant plasmid. Hybridization was performed as described [50].

Southern blot analysis of transgene in RNAi lines
An amount of 20 mg of genomic DNA from RNAi lines was digested completely with EcoRV, which has no cut site in the inserted hygromycin phosphotransferase gene sequence. The digested DNAs were separated by electrophoresis on a 0.8% agarose gel and transferred onto a Hybond N + nylon membrane (Amersham). Hybridization was performed at 65uC overnight with the a-32 P dCTP-labeled 515-bp fragment of the hygromycin phosphotransferase gene. The probe was labeled by use of a primer-a-gene labeling system (Invitrogen) following the manufacturer's instructions.

Real-time PCR expression analysis
The real-time PCR analysis was performed using primers specific to OsSPO11-4, P26 (GCTTATGATCGTCAGGT-TTCTTCAA), and P27 (GGGCCGGGACCTCTGATATA). The expression level of OsSPO11-43 in different RNA samples was calculated according to the internal standard gene, tubA, to normalize for variance in the quality of RNA and the amount of input cDNA. The relative expression of OsSPO11-4 in RNAi lines versus wild type was computed by the DDC T method (Applied Biosystems, USA).

Cytological analysis
For viability assays of mature pollen grains, flowers were randomly collected from RNAi and wild-type plants at the heading stage. Anthers of the sampled flowers were dissected, and pollen grains were stained in Alexander solution [37] and viewed on light microscopy. Meiotic chromosomes were spread and then stained with 49, 6-diamidino-2-phenylindole (DAPI) following the methods described previously [56,57]. Figure S1 Schematic genomic structure of OsSPO11-1, 4, 5 and OsTOPVIB genes. Exons are represented by black boxes. Start and stop codons are shown as arrows above the schematic sequence. (TIF) Figure S2 Multiple alignment of amino acid sequences of OsTopVIB and its homologues using ClustalX software (version 1.8). Gaps are shown by dashes. Black boxes indicate conserved residues, and grey boxes indicate similar residues. The respective amino acid position of each sequence is given on the right. These sequences are OsTopVIB (AY371050) from O. sativa; MtTopVIB (NP_276142) from Methanobacterium thermoautotrophicum; AfTopVIB (NP_069486) from Archaeoglobus fulgidus; SsTopVIB (O05207) from Sulfolobus shibatae and AtTop-VIB (AJ297843) from A. thaliana. (TIF) Figure S3 Western blot of yeast two hybrid proteins of OsTopVI. A, Western blot of expressed fusion proteins of the OsTopVI-pGADT7 constructs in yeast strain AH109 using anti-HA antibodies, con refers to protein expressed by the empty pGADT7 vector. B, Western blot of expressed fusion proteins of the OsTopVI-pGBKT7 constructs in AH109 using anti-myc antibodies, con refers to protein expressed by the empty pGBKT7 vector. s5, s5N, s5C, s1, s4 and VIB represent respective HA/myc fusion proteins. M: protein molecular weight. (TIF) Author Contributions