Bivalent Formation 1, a plant-conserved gene, encodes an OmpH/coiled-coil motif-containing protein required for meiotic recombination in rice

Highlight OsBVF1 is a novel gene that is required for meiotic recombination in the monocot rice.


Introduction
Meiosis is a specialized form of cell division that halves the chromosome number of diploid cells in producing haploid cells; it is highly conserved for sexual reproduction in most eukaryotes (Gerton and Hawley, 2005;Ramesh et al., 2005). It comprises two rounds of cell division, meiosis I and meiosis II, and each round can be divided into four stages: prophase, metaphase, anaphase, and telophase. Prophase I is a relatively long phase taking up 85-95% of the total time of meiosis, and has been further divided into five stages: leptotene, zygotene, pachytene, diplotene, and diakinesis (Wang et al., 2014b). Homologous chromosome (homolog) interaction is the crucial event during meiotic prophase I, including pairing, synapsis, recombination, and segregation. Proper interaction not only ensures the subsequently accurate segregation between homologs, but also redistributes the genetic alleles among the progeny, which has a great impact in biological diversity.
In the last three decades, molecular genetic studies have identified many genes involved in different meiotic processes in a variety of model species, such as Saccharomyces cerevisiae, Drosophila melanogaster, Caenorhabditis elegans, and several higher plants (Zickler and Kleckner, 1998Osman et al., 2011;Ma et al., 2014;Mercier et al., 2015). In the dicot model plant Arabidopsis, so far, more than 80 meiosis-related genes have been identified (Osman et al., 2011;Wang et al., 2014b;Mercier et al., 2015). By contrast, only ~30 meiotic genes in the monocot model plant rice (Oryza sativa L.) have been cloned and functionally studied . For example, OsMEL1/2 and OsAM1 are required for the initial meiotic events and their mutations cause the failure of meiotic entrance or arrest at an early stage (Nonomura et al., 2007;Che et al., 2011;Nonomura et al., 2011). It has been well studied that meiotic recombination is initiated by the programmed formation of double strand breaks (DSBs) catalysed by SPO11, which is an evolutionarily conserved type II topoisomerase in eukaryotes (Keeney et al., 1997;Grelon et al., 2001;Stacey et al., 2006;Yu et al., 2010;An et al., 2011). In rice, two SPO11 homologs, OsSPO11-1 and OsSPO11-4, were identified as being required for DSB formation (Nonomura et al., 2004a;Yu et al., 2010;An et al., 2011). In addition, in yeast, there are at least eight genes involved in this process (Keeney, 2008). In Arabidopsis, AtPRD1/2/3, AtDFO, AtPCH2, and MTOPVIB were required for DSB formation (De Muyt et al., 2007;De Muyt et al., 2009;Zhang et al., 2012;Lambing et al., 2015;Vrielynck et al., 2016). By contrast, only OsPAIR1, OsCRC1, OsSDS, and OsMTOPVIB were characterized as being DSB formation related in rice (Nonomura et al., 2004a;Miao et al., 2013;Wu et al., 2015;Fu et al., 2016;Xue et al., 2016). It seems that divergence of regulation of meiotic progression exists between rice and Arabidopsis.
After DSB formation, further resection of a single end produces 3′ end overhang, which is protected by replication protein A (RPAs) proteins (Iftode et al., 1999). Three RPA proteins were discovered to have a role in meiotic recombination in rice (Chang et al., 2009;. Further single end invasion is facilitated by RecA homologs; several rice RecA members were identified, such as OsDMC1, OsRAD51, OsRAD51C, and OsXRCC3 (Ding et al., 2001;Deng and Wang, 2007;Rajanikant et al., 2008;Tang et al., 2014;Zhang et al., 2015), suggesting that this process is conserved. As a consequence, repair of DSBs yields crossovers (COs) or noncrossovers (NCOs). Most organisms have two types of COs, of which the interference-sensitive CO (class I) depends on ZMM proteins, while the interference-insensitive CO (class II) is MUS81 dependent (Hollingsworth and Brill, 2004). In rice, several ZMM proteins such as OsMSH4, OsMSH5, OsMER3, OsHEI10, and OsZIP4 are involved in the class I CO pathway (Wang et al., 2009;Shen et al., 2012;Wang et al., 2012a;Luo et al., 2013;Zhang et al., 2014), but the MUS81 homolog has not yet been characterized. In addition, several proteins required for meiotic chromosome segregation have been isolated in rice, such as OsSGO1 (Wang et al., 2011b), OsREC8 (Shao et al., 2011), and OsBRK1 (Wang et al., 2012b).
The synaptonemal complex (SC) forms between homologous chromosomes and is important for the maturation of some recombination intermediates by stabilizing the paired chromosomes (Page and Hawley, 2004;Kleckner, 2015, 2016). The SC is a tripartite structure consisting of two parallel lateral elements and a central element. The rice PAIR2 and PAIR3 are axial elements, while OsCRC1 and OsZEP1, the homolog of ZIP1 in Saccharomyces cerevisiae and ZYP1 in Arabidopsis, are central elements of the SC (Sym et al., 1993;Wang et al., 2010;Wang et al., 2011a;Higgins et al., 2005;Nonomura et al., 2007;Yuan et al., 2009;Miao et al., 2013). Interestingly, unlike other species, partial loss of function of the rice ZEP1 has a distinct role in increase of COs Wang et al., 2015), suggesting that different plant species may have the specific factors controlling meiosis.
In this study, we identified a sterile rice mutant with meiotic defects and isolated a gene (named Bivalent Formation 1, OsBVF1) by map-based cloning that encodes a conserved protein with a putative coiled-coil motif and an outer membrane protein H (OmpH) motif. In the bvf1 mutant, meiotic DSB formation failed to be detected, thereby resulting in the failure of synapsis. At diakinesis, unlike the wild type (WT) that formed 12 bivalents, bvf1 produced 24 univalents and had improper chromosome segregation in both anaphase I and II. Further analysis showed that installation of the central element, OsZEP1, of the SC was also defective. Taken together, our results reveal a new protein that is required for meiotic DSB formation and the subsequence synapsis and recombination in rice.

Experimental materials
The bvf1 mutant was identified from the japanica cv Nipponbare (Nip) mutant library induced by 60 Co γ-ray radiation in our laboratory. The mapping populations were constructed by crossing the heterozygote (BVF1/bvf1) with indica cv Huanghuazhan (HHZ), and backcrossed with HHZ. All the materials were planted in fields in Guangzhou from spring to autumn (two growth seasons). For the recombinant screening, germinated seeds were planted in 96-well plates, and 3-week-old seedlings were used for high-throughput DNA preparation as described previously . Detected recombinant plants were planted in field or buckets.

Observation of pollen viability
Spikelets with mature pollen at the heading stage were collected and fixed in 70% ethanol. Then pollen grains were dissected out of anthers in 1% I 2 -KI solution. The strained pollen grains were firstly observed under a microscope (Olympus CX31), and then pictures were taken under an Axio Observer Z1 fluorescence microscope (Zeiss, Oberkochen, Germany).

Observation of meiotic chromosome morphology
Young panicles (4-8 cm in length) of both WT and bvf1 mutant were collected and fixed in Carnoy's solution (ethanol:glacial acetic acid (v:v) 3:1) at room temperature in less than 24 h (Cheng, 2013).
The fixed panicles were washed with 70% ethanol three to five times until the glacial acetic acid faded and then stored in it at 4 o C. Pollen mother cells (PMCs) undergoing meiosis was squashed in water or phosphate-buffered saline (PBS). The slides with PMCs were then moved to a hot block at 45 o C, mixing the cells with a few drops of 65% glacial acetic acid and heating for 1 min. Before the drop dried, previously frozen Carnoy's solution was added to the center of the drop to separate the cells (Wang et al., 2014a). After the liquid dried, 4,6-diamidino-2-phenylindole (DAPI) in anti-fade solution (Vector Laboratories, Burlingame, CA, USA) was added to the slide and covered up for observation. Chromosome images were captured under the Axio Observer Z1 fluorescence microscope.

Expression vector construction
Total RNA from spikelets of WT rice were extracted. Total RNA (1 μg) was reverse transcribed by using M-MLV Reverse Transcriptase (Promega, Madison, WI, USA) with Oligo-T (18) as primer, the products of which were taken as the template used afterwards. The ORF sequence of Os05g0251400 was amplified by primers pOX-BVF1-F/R (see Supplementary Table S1 at JXB online) and ligated into a binary vector so as the ORF was under the control of the ubiquitin promoter. The green fluorescent protein (GFP) fusion vectors were constructed with the Ω-PCR procedure (Chen et al., 2013) with primers GFP-BVF1/BVF1-GFP. The fluorescence images were captured using an LSM 7 DUO Confocal Microscope (Zeiss).

Rice transformation and genotyping
By Agrobacterium (stain EHA105)-mediated transformation, the vector constructs were transferred into callus induced from seeds of heterozygous mutant plants. Positive transformants were screened by PCR amplification with HPT primers and vector-specific primer pOX-T (see Supplementary Table S1 at JXB online), respectively. The endogenous genotypes of the transformants were identified by a semi-nested PCR with specific primers BVF1(F)/(R)/(R2) (Supplementary Table S1).

Identification and characterization of a sterile rice mutant
We created a mutant library of a japonica cultivar Nipponbare by 60 Co γ-ray radiation. By screening the mutant library, we obtained a sterile mutant, named bivalent formation 1 (bvf1) according to our later observation that the mutated causal gene affects bivalent formation in meiosis. The mutant had as normal vegetative growth as the WT plants, but with no seed setting at the reproductive growth stage (Fig. 1A, D). Further characterization showed that the mutant exhibited smaller anthers and completely sterile pollen grains (Fig. 1B,  C). When the mutant plants were pollinated with WT pollen grains, no seed was produced, suggesting that the female gametes were also sterile. The segregation of fertile (104) to sterile (34) individuals in the progeny of self-fertilized mutant heterozygotes fitted the 3:1 ratio (Supplementary Table S2), indicating that a single recessive gene is responsible for the male and female sterile phenotypes.

Meiosis is defective in pollen mother cells of bvf1
It is known that defective mutation of many meiotic genes causes male and female sterility in both human and plants (Székvölgyi and Nicolas, 2010;Luo et al., 2014). To explore the possibility for the sterility in bvf1, we observed the meiotic chromosome behavior of pollen mother cells (meiocytes) using chromosome spreads stained with DAPI at different meiotic stages in both WT and bvf1. As shown in Fig. 2, in WT, at leptotene, the chromosomes began to condense and displayed a thread-like feature under microscopy ( Fig. 2A). At zygotene, the homologous chromosomes aligned together and began to pair with each other (Fig. 2B). At pachytene, the homologs were stabilized by the synaptonemal complex (SC) and displayed thick thread-like chromosomes (Fig. 2C). At diakinesis, following the disassembly of the SC, the 12 pairs of homologs (also called bivalents) physically associated by chiasma and sister chromatid cohesion were observed (Fig. 2D). At metaphase I, all bivalents were aligned at the equatorial plate pulled by spindles (Fig. 2E), thereby resulting in the subsequent segregation to the opposite poles (Fig. 2F). Finally, the two dyads simultaneously underwent meiotic II cell division and formed the tetrad microspores (Fig. 2G, H).
Compared with the WT chromosome morphology, no obvious difference was observed in bvf1 from leptotene to zygotene (Fig. 2I, J). At pachytene, unlike WT with fully synapsed homologs, the bvf1 chromosomes condensed and aligned together, but did not show thick chromosomes (n=82 meiocytes) (Fig. 2K), suggesting a defect in synapsis. From diplotene to diakinesis, in contrast to the WT that formed 12 bivalents, bvf1 had 24 univalents (n=110 meiocytes) (Fig. 2L), suggesting a failure of crossover formation. Due to the recombination defect, the 24 univalents were not well aligned at the equatorial plate at metaphase I (n=167 meiocytes) (Fig. 2M), and showed an uneven segregation to the two poles at anaphase I. Moreover, 70.9% bvf1 meiocytes at anaphase I (n=79) had lagging chromosomes (Fig. 2N). At meiosis II, due to the unequal segregation of chromosomes, the bvf1 meiocytes produced abnormal tetrads with uneven chromosome numbers and micronuclei (Fig. 2O, P). The failure of bivalent formation and the aberrant chromosome segregation provides an explanation for the complete sterility in bvf1. Together, these results indicate that OsBVF1 is required for normal bivalent formation in rice meiosis.

Map-based cloning of OsBVF1
To isolate the gene conferring the mutant phenotype, we crossed the heterozygous BVF1/bvf1 plants (male and female fertile) with an indica rice variety, HHZ. The F 1 plants were further backcrossed with HHZ. By linkage analysis using 10 sterile F 2 plants and a set of polymorphic markers covering the whole genome, a region on the short arm of chromosome 5 was found to link with bvf1. Then we used a total of 775 F 2 and BC 1 F 2 plants and a number of molecular markers on this region (Supplementary Table S3) to primarily map the locus on a region of ca 2821 kb (Fig. 3A). Through further screening of new recombinants in the segregated F 3 and F 4 populations with the markers 507966 and 510787, OsBVF1 (Osbvf1) was further delimited to an 84-kb region between two markers, 509167 and 509251 (Fig. 3A), a region that includes seven annotated genes. Then we amplified these genes by PCR for subsequent sequencing (Fig. 3A). A single base deletion in the third exon of the gene Os05g0251400 was detected in bvf1 (Fig. 3B), which caused a frame-shift and a premature stop codon (Fig. 3B). Because no other mutations in the other genes within the 84-kb region were found, we considered Os05g0251400 as the candidate gene for OsBVF1.
To verify the function of Os05g0251400, we constructed a binary vector (pOX-BVF1) with the whole 1115-bp open reading frame sequence (AK103883) of Os05g0251400 driven by the maize ubiquitin promoter. This construct was used to transform calli induced from immature seeds of the heterozygous BVF1/bvf1 plants. By genotyping of the endogenous Os05g0251400 in the transgenic-positive transgenic (T 0 ) plants with the mutation site-specific primer set 1400-T (Supplementary Table S1), four out of 17 T 0 plants were found to have homozygous Osbvf1, and they all showed recovered fertility and normal seed-setting ( Fig. 3C and Supplementary  Fig. S1). In the T 1 generation of these four plants, the segregants with and without the transgenes co-segregated with the fertile and sterile phenotypes (Supplementary Table  S4 and Supplementary Fig. S1). Therefore, we conclude that the single-base deletion in Os05g0251400 is responsible for this sterile mutation of the target gene.
Sequence analysis (www.ncbi.nlm.nih.gov/, last accessed 13 March 2017) showed that OsBVF1 encodes a hypothetical protein of 286 amino acids (aa) (protein Accession No.: NP_001055029) with a putative conserved OmpH (outer membrane protein H) domain from the 62nd to 152nd aa, and this protein is unique in the rice genome (Fig. 4A). The mutation in the bvf1 allele produces a truncated protein of 99 aa. By running the 'COILS' program (http://www.ch.embnet.org/ software/COILS_form.html, last accessed 13 March 2017) using OsBVF1 as query, it is predicted that OsBVF1 also can form two coiled-coil motifs in the central region (54-81 aa and 90-124 aa) ( Supplementary Fig. S2), which partially overlaps with the OmpH domain ( Fig. 4A and Supplementary Fig. S3). Thus, OsBVF1 encodes a new protein with a unique OmpH domain coupling with the coiled-coil motif in rice.

OsBVF1 is highly expressed in anther and its protein targets to nucleus
To examine the expression profile of OsBVF1, we performed a qRT-PCR experiment and found that OsBVF1 was expressed in various organs, with relatively higher level in anthers developing meiosis stages ( Fig. 4B and Supplementary Fig. S4). We also found that the mRNA level was obviously lower in bvf1 than in WT (Fig. 4B), probably due to degradation of the abnormal mutant mRNAs by the nonsense-mediated mRNA decay mechanism (Maquat, 2004). To investigate the subcellular localization of OsBVF1, we prepared a transient expression construct for an OsBVF1-GFP fusion protein. By co-transfer of the OsBVF1-GFP construct with a nucleus-localization marker construct expressing GHD-mCherry into rice protoplasts, we observed that the OsBVF1-GFP signal was mainly localized in nuclei, which overlapped with the GHD-mCherry signal (Fig. 4C), suggesting that OsBVF1 is a nucleus-localized protein.

OsBVF1 is indispensable for meiotic DSB formation
Meiotic recombination is initiated from the programmed DSB formation (Keeney et al., 1997). The formation of DSBs triggers the phosphorylation of the histone variant H2AX (γ-H2AX), which specifically marks DSBs and facilitates post-replication DNA repair (Dickey et al., 2009). To detect whether DSBs are formed in bvf1, we used immunofluorescence to examine the distribution of phosphorylated γ-H2AX with an anti-γH2AX antibody generated using the sequence from rice (Miao et al., 2013). To mark the chromosomes, we used OsREC8, a homolog of Arabidopsis meiotic specific cohesin SYN1 (Cai et al., 2003), which has a linear distribution pattern on chromosomes during early prophase I (Shao et al., 2011). As shown in Fig. 5, WT zygotene meiocytes showed dot-like signals of γH2AX (Fig. 5A), while no signals were detected in bvf1 (Fig. 5B), indicating that BVF1 is indispensable for rice meiotic DSB formation.
Following the DSB formation, the DSB ends are further processed by the MRX complex (Mre11/Rad50/Xrs2) and COM1/SAE2 (Mimitou and Symington, 2009). The rice OsMRE11 and OsCOM1 homologs have also been reported to participate in meiotic DSB repair (Ji et al., 2012(Ji et al., , 2013. We further examined the localization of OsMRE11 and Osbvf1 in developmental spikelets. The spikelets of 2-3 mm in length were at the PMC to meiosis stages. Actin 1 mRNA was used as the internal control. (C) The constructs expressing OsBVF1-GFP and a nuclear-localized fusion protein, GHD-mCherry, were co-transferred into rice protoplasts. Bars, 10 μm.
OsCOM1 in both WT and bvf1 mutant. Unlike WT with dotlike signals at pachytene chromosomes, we did not detect any signals of both proteins in bvf1 at a similar stage (Fig. 5C-F), supporting the idea that OsBVF1 functions upstream of DSB end procession. This hypothesis was further supported by the undetectable signal of the other downstream proteins OsDMC1  and OsMER3 (Wang et al., 2009) in bvf1 mutant meiocytes (Fig. 5G-J). Taken together, these results provide strong evidence to support the role of OsBVF1 in DSB formation during meiotic recombination.

OsBVF1 is dispensable for axial element installation, but required for the central element installation of SC
After the progression of meiotic recombination, the SC, a proteinaceous structure including lateral and central elements formed between homologs, is important for the stabilization of recombination intermediates and facilitates subsequent homolog recombination (Zickler and Kleckner, 1999). The rice axial element (AE) protein OsPAIR2 is the homolog of yeast HOP1 and Arabidopsis AtASY1 (Nonomura et al., 2007). We examined the localization patterns of OsPAIR2 in WT and bvf1 meiocytes and found a normal linear pattern overlapping with zygotene chromosomes between WT and mutant (Fig. 6A, B), implying that the assembly of AEs is probably unaffected in the mutant.
To investigate whether the installation of SC occurs in bvf1, we examined the localization of rice OsZEP1 , a homolog of Arabidopsis AtZYP1, the central element of the SC (Higgins et al., 2005) in WT and mutant. The immunostaining signals for OsZEP1 at pachytene showed linear signals along with the entire chromosomes in WT meiocytes (Fig. 6C). By contrast, no such immunostaining signals were observed in the bvf1 meiocytes (Fig. 6D). Thus, we conclude that OsBVF1 is required for the installation of the SC in rice, probably by an indirect effect due to lack of DSB formation in the mutant.
According to the amino acid similarity, we built a phylogenetic tree among 23 representative plant species (Fig. 7). The proteins were divided into four groups among eudicots, monocots, pteridophytes and streptophyta/algae. The data suggest that OsBVF1 and the homologs are plant-conserved and they should be derived from a common ancestor. It is notable that, except for Populus trichocarpa, all the examined species have only a single copy of BVF1 or its homologs (Fig. 7), suggesting that the homologous genes did not expand during the evolution of plants.

Identification of a new meiotic gene in rice
Most of the reported meiotic genes in plants are comparatively conserved from yeast to higher eukaryotes (Osman et al., 2011;Luo et al., 2014;Mercier et al., 2015;Zickler and Kleckner, 2016). According to homology alignment in terms of sequence identity or similarity, previous studies have identified several meiotic genes in rice, such as OsDMC1 (Ding et al., 2001), OsSPO11-4  and OsRAD21-4 (Zhang et al., 2006). Compared with yeast or fruit fly, plants with larger genome sizes are supposed to have more complicated meiotic regulation. As previously reported, some meiotic genes, such as OsAM1 (Che et al., 2011)/AtSWI1 (Mercier et al., 2001), OsMOF1 (He et al., 2016) and OsPAIR1 (Nonomura et al., 2004a)/AtPRD3 (De Muyt et al., 2009, were found to be plant specific, and the rice meiotic genes OsMEL1 (Nonomura et al., 2007) and OsMEL2 (Nonomura et al., 2011) have no homologs in other plant species. Therefore, meiotic control may vary somewhat among different species, even in plants. Thus, identification of more meiotic genes is necessary to expand our knowledge of meiosis. In this study, through mutant screening and map-based cloning, we identified OsBVF1 from rice. Both sequence alignment and functional characterization support the fact that OsBVF1 is a novel meiotic gene that encodes a coiled-coil motif-and OmpH domain-containing protein.

The role of OsBVF1 in meiotic recombination
In this study, we provided several lines of evidence to support a role of OsBVF1 in rice meiosis. First, the rice OsBVF1 was required for fertility, and mutation of OsBVF1 caused male and female sterility; second, chromosome morphology analysis showed that bvf1 was defective in formation of wellsynapsed chromosomes and only produced univalents, suggesting a failure of synapsis and crossover formation; third, the meiotic recombination defect in bvf1 is likely caused by failure of DSB formation, which is supported by the observation of the disappearance of the marker for localization of DSB and other proteins required for meiotic recombination; fourth, the undetectable OsZEP1 signal in bvf1 suggests a role in SC formation, but the failure of the SC is likely a consequence of the initial defect in DSB formation (as reviewed in Gray and Cohen, 2016).
Sequence analysis showed that OsBVF1 has an OmpH domain and two coiled-coil motifs. It is reported that OmpHcontaining proteins may play roles as protein folding catalysts or as chaperones in extracytoplasmic compartments (Missiakas et al., 1996). In addition, the coiled-coil motifs play an important role in mediating subunit oligomerization in many proteins (Lupas, 1996;Mason and Arndt, 2004). Among the identified meiotic proteins, the central element of the SC shares a coiled-coil motif in the central region with one globular domain at each end, as with OsZEP1 in rice , ZYP1 in maize (Golubovskaya et al., 2011) and Arabidopsis (Higgins et al., 2005), and Zip1 in budding yeast (Sym et al., 1993;Page and Hawley, 2004). The coiled-coil motifs of the proteins form ladder-like or hinge-like parallel structures in the central region of the SC. Besides, some other meiotic proteins also contain the coiled-coil motifs, including OsPAIR1, OsPAIR3, OsAM1, OsSGO1, and OsHEI10 in rice (Nonomura et al., 2004a;Che et al., 2011;Wang et al., 2011aWang et al., , 2012aWang et al., 2011b), DSY2 in maize (Lee et al., 2015) and RED1 in budding yeast (Smith and Roeder, 1997), providing evidence that the coiled-coil motif is one of the important domains among proteins required for meiosis.
Since our yeast two-hybrid assay did not detect any interaction between OsBVF1 and OsPAIR2/3 (Nonomura et al., 2007;Yuan et al., 2009;Wang et al., 2011a), OsZEP1  and OsCRC1 (Miao et al., 2013), it is likely that OsBVF1 may not directly participate in SC assembly in rice.

The function of OsBVF1 and its homologs might be highly conserved in plants
Research into evolutionary biology has indicated that between 55 and 75 million years ago, plants had their genomes duplicated so as to increase the chance of survival (Lohaus and Van de Peer, 2016). Comparison of plant species showed that at least several million years ago, many monocot lineages, including wild rice, had already experienced two distinct paleopolyploidies (Jiao et al., 2014). Lots of genomic hints can verify this notion, for example, the highly conserved meiotic genes such as OsDMC1 (Ding et al., 2001;Metkar et al., 2004), OsRAD51 (Rajanikant et al., 2008) and OsPAIR2 (Nonomura et al., 2004b) all have two copies in the rice genome, with one of the copies silenced like OsPAIR2 or both functionally reserved. Interestingly, in the case of OsBVF1 and its homologs (orthologs), from algae to monocots and eudicots, only a single copy of the gene was reserved in all the genomes (except for P. trichocarpa) during plant genome evolution, suggesting that these orthologous genes may Fig. 7. Phylogenetic analysis of plant proteins homologous to OsBVF1. The tree was constructed from the alignment of full-length proteins in rice and some representative plant species by MEGA6 using the neighbor-joining (NJ) model. The protein accession numbers are given in the brackets after species names. *The protein sequences are not available in GenBank, and are given in Supplementary Fig. S4. be important for plant sexual reproduction. Therefore, we infer that OsBVF1 and its homologs in other plants may be highly conserved, with a primary role in reproduction. Both the present results and previous findings indicate that some plant-specific genes, including OsBVF1, have evolved in the regulation of plant-specific meiosis.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Genetic and phenotypic analyses of the OsBVF1transgenic plants. Fig. S2. Coiled-coil motif prediction of OsBVF1 and OsPAIR3 based on the web-tool COILS. Fig. S3. Comparison of OmpH and coiled-coil motif sequences of OsBVF1 and OsPAIR3. Fig. S4. Expression pattern of OsBVF1 according to the Rice Expression Profile Database. Fig. S5. Sequence alignment of OsBVF1 and its homologous proteins. Table S1. Primers used in the study. Table S2. Segregation of fertile and sterile plants in bvf1 M 3 lines. Table S3. Segregation of fertile and sterile plants in bvf1 mapping populations. Table S4. Sequences of OsBVF1 homologs of some plant species that are not available at GenBank.