Index INTRODUCTION MATERIALS AND METHODS Plant materials Phenotypic evaluation DNA analysis Data analyses RESULTS Mapping of a hybrid male sterility gene, S24 Identification of an epistatic locus Selective pollen abortion by S24 and S35 DISCUSSION References

INTRODUCTION

Hybrid sterility is one of the postzygotic barriers, in which the hybrids survive but are unable to reproduce their offspring. Such reproductive barriers generally hinder transferring of genes for useful characters in the cross-breeding of crops. Despite the importance of these barriers, little is known about the kind of molecules involved in hybrid sterility and how they arise in the course of speciation.

Asian cultivated rice (Oryza sativa L.) is classified into two major groups, japonica and indica, which were first named by Kato (1930) and correspond to the subspecies category in taxonomy. Their F₁ hybrids show normal vegetative growth and no significant disturbances in chromosome pairing at meiosis, but they show various forms of postzygotic reproduction isolation (RI) such as hybrid sterility and hybrid breakdown in their progeny. Most inter-subspecific hybrid sterility is caused by disharmonious interactions between nuclear genes derived from their respective parents, excluding a case of cytoplasmic-nuclear interactions causing male sterility (Shinjyo, 1975). Earlier investigations characterized well the genetic basis of the nuclear gene interaction, and a number of gene loci were identified by mapping with visible morphological markers and isozymes. Based on genetic studies, two major genetic models for rice F₁ hybrid sterility were developed. One is an allelic interaction at a single locus (Kitamura, 1962; Ikehashi and Araki, 1986) and the other is epistatic interaction at two independent loci (Oka, 1974; Sato and Morishima, 1987). Many of the previous rice studies till date support the monogenic model. However, the two-locus Dobzhansky-Muller model that hybrid inviability is caused by deleterious epistatic interaction (Dobzhansky, 1936; Muller, 1942) is still widely accepted in animals and plants, and is also applicable to hybrid sterility. In addition, hybrid sterility shows typical polygenic inheritance hence it is conceivable that the interaction among the polygenes is related to the sterility mechanism. Recent studies have mapped epistatic genes underlying hybrid sterility and inviability in diverse animal species such as Drosophila (Noor et al., 2001; Tao et al., 2003) and in plants such as Mimulus (Sweigart et al., 2006). Moreover, some of the major sterility factors have been successfully cloned and characterized at the molecular level (Trachtulec et al., 1997; Ting et al., 1998; Barbash et al., 2003). Although a number of RI genes have been identified in plants, the molecules underlying the RI mechanism remain largely unelucidated. Recent work in rice has elucidated that hybrid female sterility gene, S5, encodes an aspartic protease, but the detailed mechanism causing hybrid female sterility is still unknown (Chen et al., 2008).

We now have the entire genomic sequences of a typical japonica variety Nipponbare (International Rice Genome Sequencing Project, 2005) and an indica variety 93-11 (Yu et al., 2002). Complete genome sequences of these subspecies revealed a large number of DNA markers on the basis of their sequence differences. The DNA markers accelerate genotyping in individuals of hybrid progeny and fine-mapping for gene isolation. As a result of this advance, a growing number of quantitative traits loci (QTL) for agronomic traits and hybrid sterility have been identified using the various cross combinations of rice (Liu et al., 2001). Some QTLs related to hybrid sterility showed interaction with other QTLs in these studies (Wu et al., 1995; Song et al., 2005). However, QTLs are fraught with ambiguity regarding map position and effect. Epistatic interactions among genes including the interaction with environmental factors are thought to make a substantial contribution to variation of pollen and seed sterility in hybrid progeny. To get a comprehensive insight into the RI system, these epistatic and minor factors should also be identified and characterized precisely.

Understanding epistasis for hybrid sterility will aid in unveiling molecular networks for gametogenesis in addition to the mechanism of postzygotic RI in plants. An essential step towards the detailed understanding of the molecules underlying hybrid sterility is to identify the causal genes on a chromosome map. This study aimed to dissect and characterize an epistatic interaction underlying hybrid male sterility found in the inter-subspecific indica/japonica cross of rice. Here, we report novel epistatic loci, termed S24 and S35, which were found using chromosome segment substitution lines (CSSLs) and near-isogenic lines (NILs). We demonstrated that epistasis is an important genetic factor to generate wide phenotypic variation of hybrid sterility in rice. These data suggest that complex epistasis has been developed as a part of the postzygotic RI mechanism in rice.

MATERIALS AND METHODS

Plant materials

Recombinant inbred lines (Tsunematsu et al., 1996), which were derived from a japonica variety, Asominori, and an indica variety, IR24, were backcrossed with Asominori to produce a series of CSSLs named “AIS” (Kubo et al., 2002). BC₃F₁ and BC₃F₂ plants were genotyped using RFLP markers evenly scattered over the rice genome (Kubo et al., 2002). The RFLP genotypes of BC₃F₁ and BC₃F₂ were utilized as reference data to analyze the genes in this study. The BC₃F₂ and their derivative progeny were used for analysis of pollen sterility in this study. Pollen-sterile segregants in the BC₃F₃ population were backcrossed with Asominori and the resulting BC₄F₁ populations were used for segregation analysis. To map and characterize interacting factors, near-isogenic line (NIL) for the targeted genes (S24 and S35) were developed by the cross between two different CSSLs and marker-assisted selection (MAS). The procedure for developing NIL carrying S24 and S35 genes is summarized in Fig. 1. The CSSLs used in this study were AIS 34, a CSSL carrying an IR24 segment of chromosome 5 and AIS 1, a CSSL carrying an IR24 segment of chromosome 1. After a backcross of AIS 1 with Asominori, the resultant F₁ plant was crossed with AIS 34. The obtained NIL-F₁ with double heterozygous genotype for S24 and S35 were self-pollinated to produce NIL-F₂ seeds. The NIL-F₂ individuals were phenotyped and genotyped by using SSR markers. Selected NIL-F₂ plant with S24-ir/S24-ir S35-as/S35-ir genotype was self-pollinated and the progeny (NIL-F₃) was used for mapping of S35 locus.

View Details

Fig. 1 Crossing design to characterize epistasis between S24 and S35 genes. “AIS 1” and “AIS 34” are CSSLs carrying IR24 homozygous segments of chromosomes 1 and 5, respectively, with Asominori genetic background. The Asominori chromosome is shown in white and the IR24 chromosome is shown in black. The heterozygous segment of AIS 34 observed in the BC3F2 generation (short arm of chromosome 8) was not traced in later generations and is shown in gray. AIS 34 in BC3F5 generation was used for crossing to produce NIL-F1. The NIL-F2 plants were genotyped with the SSR markers RM13 (linked to S24 locus) and RM1 (linked to S35 locus) and the selected S24-ir/S24-ir S35-as/S35-ir plants were self-pollinated to obtain NIL-F3 population for S35 segregation.

Phenotypic evaluation

Pre-flowering panicles from each individual in the population were collected and fixed in a formalin/acetic acid/alcoholic (FAA) solution to examine the pollen phenotype. Fixed samples were stored in a 70% ethanol solution. Pollen grains from 3 to 6 anthers one day before anthesis were stained with 1% iodine-potassium iodide (I₂-KI) and 1% aceto-carmine solutions. More than 200 pollen grains were scored for each individual. Stained pollen grains with a normal size were considered fertile. Stained small pollen grains and empty pollen grains were considered to be sterile. Seed-set was counted on three panicles for each plant. Seed fertility scores were determined based on filled spikelets divided by the total number of spikelets.

DNA analysis

DNA samples for RFLP analysis were extracted from frozen leaf samples using the CTAB method (Murray and Thompson, 1980). The isolated DNA (2.0 μg) was digested with restriction enzymes (Bgl II, Dra I, EcoR V, Hind III and Kpn I), separated by 0.8% agarose-gel electrophoresis and blotted onto Hybond N⁺ membranes (GE healthcare) by capillary transfer mediated in 0.4 M NaOH solution. Blotted membranes were rinsed in 2 × SSC, dried, and baked at 120°C for 20 min. DNA clones previously mapped by Tsunematsu et al. (1996) and Harushima et al. (1998) were used. DNA labeling, hybridization and signal detection were conducted using the ECL detection system (GE healthcare). DNA samples for PCR analyses were crudely extracted using 0.25 M NaOH buffer followed by neutralization with 0.1 M Tris-HCl. After 10-fold dilution of the crude extract, the diluted solutions (5.0 μl) were used as DNA templates for the PCR reaction. SSR analysis was carried out as per the protocol of Panaud et al. (1996). The original sources and motifs for the SSR markers used in this study can be found in Panaud et al. (1996) and McCouch et al. (2002).

Data analyses

Recombination values were estimated using the maximum likelihood equation (Allard, 1956). Obtained values were converted into genetic map distances (cM) using the Kosambi function (Kosambi, 1944).

RESULTS

Mapping of a hybrid male sterility gene, S24

Both parents, Asominori and IR24, were fertile for pollen and seed (> 90% fertility) when grown in paddy field conditions at Fukuoka, Japan. Reciprocal F₁ hybrids exhibited approximately 40% pollen fertility and 60% seed fertility (Table 1). Similar fertility rate in the reciprocal F₁ hybrids suggested that nuclear genes and not cytoplasmic factors were responsible for the hybrid male sterility. Segregation of pollen sterility was investigated using the BC₃F₂ populations, which were developed for the production of a series of CSSLs carrying the IR24 donor segment with Asominori genetic background (Kubo et al., 2002). We found five BC₃F₂ populations that showed segregation of pollen sterility. All of these populations carried an IR24 segment of the short arm of chromosome 5, suggesting the location of the hybrid male sterility gene to be at the chromosomal region. To confirm and characterize the gene causing hybrid male sterility, a BC₄F₁ population was developed by backcross of the BC₃F₃ segregant that showed pollen sterility. The BC₄F₁ showed segregation of pollen sterility in which the sterile phenotype showed continuous variation ranging from 10.3 to 49.6% (Fig. 2, A and B). Staining with an aceto-carmine solution revealed three types of pollen grains; normal tricellular, bicellular, and unicellular pollen in a matured anther from the sterile plants (Fig. 2A, a–c). This observation suggests that the sterile pollen grains halt their development at the unicellular or bicellular stage after meiosis. The BC₄F₁ population (N = 71) segregated into 37 pollen sterile and 34 fertile plants, with a clear bimodal distribution for pollen sterility (Fig. 2B). The observed segregation ratio fitted to the 1:1 ratio expected for monogenic inheritance (χ² = 0.13). All BC₄F₁ plants were genotyped using RFLP markers located on the retained heterozygous region in the BC₃F₁ plant. The parental BC₃F₁ plant of the BC₄F₁ population possessed the heterozygous donor segment on chromosomes 1, 5, 6, and 10 (Fig. 2C). The BC₄F₁ population recovered Asominori genome at RFLP loci G342 on chromosome 6 and R1629 on chromosome 10. However, G359 on chromosome 1 and R830 on chromosome 5 segregated for heterozygous and Asominori homozygous genotypes. The R830 marker revealed that 36 out of 37 sterile plants were heterozygous genotypes, whereas all fertile plants were of the Asominori homozygous genotype (Fig. 2B). This result indicated that the pollen sterility gene was tightly linked to R830 on chromosome 5 and caused pollen sterility in the heterozygous state. We named this gene S24. Linkage analysis showed that S24 was located between R830 and R3166 with a distance of 1.4 and 2.8 cM, respectively (Fig. 2D). Thus, the segregation of pollen sterility was mainly caused by a single locus, S24, in the backcross population.

View Details

Table 1 Pollen and seed fertility (mean ± SD%) of Asominori, IR24 and F1 hybrid

View Details

Fig. 2 Identification of S24 locus causing hybrid male sterility. (A) Photomicrograph of pollen grains of pollen-sterile segregants stained with aceto-carmine solution (a-c). Tricellular (a), bicellular (b), and unicellular pollens (c) were observed in a single anther from the sterile plants. SC: sperm cell, VC: vegetative cell. Arrow heads indicate pollen nuclei. (B) Frequency distribution of pollen fertility in BC4F1 population, classified by RFLP genotype of R830 on chromosome 5. (C) Graphical genotype of BC3F1 plant, a progenitor of the mapping population (BC4F1). Horizontal lines on the graphical genotype indicate positions of 116 RFLP loci that were used for whole-genome genotyping in BC3F1 (Kubo et al., 2002). RFLP markers (G359, R830, G342, and R1629) used for genotyping of BC4F1 plants were shown. (D) Linkage map showing the location of S24 for hybrid male sterility. Left, RFLP framework map of chromosome 5 constructed by Harushima et al. (1998); right, S24 map constructed from the BC4F1 population (N = 71) in this study.

Identification of an epistatic locus

The fertile class showed a mean value of 97.7% pollen fertility with a small variation ranging from 90 to 100%, whereas the pollen sterile class showed a wide sterility variation ranging from 10.3 to 49.6%. Besides, the pollen sterile class was classifiable into two subgroups according to the pattern of starch accumulation in sterile pollen grains (Fig. 3A, a–c). Sterile subgroup I (Fig. 3A, b) showed a high proportion of small pollen grains in which starch accumulation was occurred compared to empty pollen grains, whereas sterile subgroup II (Fig. 3A, c) was rich in empty pollen grains. Such sterility variation was presumed to be due to segregation of another genetic factor located on the genetic background or to environmental effect. To address the question, we examined the relationship between the variation in the sterile class and G359, the RFLP marker located at the other heterozygous region of chromosome 1. In the Asominori homozygous genotype for S24, there was no remarkable phenotypic difference between the two genotypes at G359 in pollen fertility. However, in the heterozygous genotype for S24, we found significant differences of pollen sterility between heterozygotes (15.4%) and Asominori homozygotes (34.6%) for G359 (Fig. 3B). Two-way ANOVA showed interaction between S24 and G359 (P < 0.0001) indicating that the G359-neighboring sterility gene interacted with the S24 locus. The new gene interacting with S24 was designated as S35. The S35 gene decreased pollen fertility by about 20% and the phenotypic effect of S35 was visually recognizable because of the increase of empty pollen grains under microscopic observation (compare Fig. 3A, b to Fig. 3A, c). Further we examined whether these pollen sterility genes affect seed fertility. Excluding the double heterozygous plants for S24 and G359, the BC₄F₁ plants showed nearly complete fertility (Fig. 3B). The double heterozygote partial seed sterility ranged from 58.4 to 85.3% (Mean value 74.4%). These results indicate that high pollen sterility was caused by the interaction between S24 and S35 leading to partial seed sterility. The pollen semi-sterility due to the single effect of the S24 gene had low effect on seed fertility.

View Details

Fig. 3 Effect of the genotype at S24 and G359 loci on pollen and seed fertility in the BC4F1 population. (A) Photomicrograph of pollen grains of pollen-sterile segregants stained by I2-KI solution (a–c). a, Pollen grains from the fertile plant; b and c, Sterile pollen grains from the different plant individuals. (B) Pollen and seed fertility for each genotype combination at S24 and G359 loci. Mean and standard deviation (%) for pollen (left) and seed fertility (right) are shown by bar chart. Chromosome segments of IR24 are shown in black and those from Asominori in white.

To precisely map and characterize the S35 gene, a NIL that introgressed S35 and S24 segments was developed and evaluated for pollen sterility. Consistent with the above results, the double heterozygous (S24-as/S24-ir S35-as/S35-ir) and S24 heterozygous plants (S24-as/S24-ir S35-as/S35-as and S24-as/S24-ir S35-ir/S35-ir) showed pollen sterility in the NIL-F₂ generation (Fig. 1 and Fig. 4). Interestingly, the NIL-F₂ having S24-ir/S24-ir S35-as/S35-ir genotype showed pollen semi-sterility, even though the S24-ir/S24-ir plants showed fertility with either S35 homozygous background (Fig. 4). Excluding some possible recombinant plants, the other genotypic classes were fertile.

View Details

Fig. 4 Frequency distribution of pollen fertility of the nine genotypes classified by RM13 and RM1 in a NIL-F2 population (N = 117). An asterisk denotes that these plants were supposed to be recombinants between S24 and RM13 because they are heterozygous at another PCR marker located on the opposite side of the S24 locus (data not shown).

To perform linkage analysis of S35, NIL-F₃ population was developed by the self-pollinating the plant that was IR24 homozygous for RM13 and heterozygous for RM1 (Fig. 1). The NIL-F₃ (N = 147) segregated into 74 semi-sterile (30–70% fertility) and 73 fertile (75–100% fertility) (Fig. 5, A and B). The semi-sterile and the fertile groups corresponded to heterozygous and IR24 homozygous alleles at SSR marker locus RM6324, respectively. The S35 locus was located between RM6324 and RM8105 on chromosome 1 by linkage analysis (Fig. 5C). The observed segregation ratio of 1 (S35-as/S35-as) : 74 (S35-as/S35-ir) : 72 (S35-ir/S35-ir) at RM6324 locus significantly deviated from the expected 1:2:1 segregation ratio (P < 0.001), which was biased toward IR24 alleles. No segregants with Asominori homozygous genotype at both RM6324 and RM8105 loci were observed, indicating that the male gametes having S35-as allele would be aborted.

View Details

Fig. 5 Mapping of S35. (A) Photomicrograph of pollen grains of pollen-sterile segregants stained by I2-KI solution. Pollen grains from the fertile plant (a) and from the sterile plant (b). (B) Frequency distribution of pollen fertility in the S35 segregating population (NIL-F3, N = 147). Bar color denotes the genotype frequency of SSR marker RM6324 located on chromosome 1. Black, gray, and white bars indicate heterozygous, homozygous for IR24 allele, and homozygous for Asominori allele, respectively. (C) Linkage map showing the position of S35 constructed from the NIL-F3 population. S35 was located on the short arm (upper) of chromosome 1.

Selective pollen abortion by S24 and S35

We next investigated genotypes of aborted male gametes by reciprocal crosses between the sterile individuals and Asominori. When the pollen semi-sterile S24-as/S24-ir S35-as/S35-as plants were pollinated with Asominori pollen, 58 semi-sterile and 58 fertile progeny that fitted to the expected 1:1 ratio were observed. However, when the semi-sterile plants were used as the pollen parent, all progeny were semi-sterile, indicating that only the male gametes carrying S24-ir alleles were fertile (Table 2). Additionally, we also examined frequency of male gametes in all the other pollen-sterile NIL-F₂ with different two-locus genotype classes, i.e., (I) S24-as/S24-ir S35-as/S35-ir, (II) S24-as/S24-ir S35-as/S35-as, (III) S24-as/S24-ir S35-ir/S35-ir, and (IV) S24-ir/S24-ir S35-as/S35-ir. The pollen-sterile plants in each genotype class were used to pollinate Asominori, and the obtained progeny were genotyped with RM13 and RM1 to investigate frequency of the male gamete genotype. In the highly sterile NIL-F₂ with double heterozygous genotype for S24 and S35 loci (genotype class I), the male gamete with the S24-ir S35-ir genotype showed the highest transmission ratio (79.6%), and the other gametes with the Asominori allele at either locus showed a lower transmission ratio (9.6, 7.2, and 3.6% in S24-ir S35-as, S24-as S35-ir, and S24-as S35-as, respectively) (Table 3). The male gametes with Asominori alleles for both S24 and S35 loci transmitted at the lowest ratio (3.6%). The two other NIL-F₂ with S24 heterozygous alleles (genotype class II and III) obviously showed reduced transmission of the Asominori allele at S24 independent of S35 genotype (transmitted at lower than 10% frequency). In addition, the NIL-F₂ with S24-ir/S24-ir S35-as/S35-ir genotype (genotype class IV) also showed reduced transmission of the Asominori allele at S35. Thus, elimination of the male gametes having the S24-as or S35-as allele is a common feature of all sterile phenotypes. It should be noted that remarkable elimination of the male gametes with S24-ir S35-as allele (9.6% in genotype class I) was observed in the double heterozygous NIL-F₂ (S24-as/S24-ir S35-as/S35-ir), even though it was fertile in S24-as/S24-ir S35-as/S35-as zygote (93.6% in genotype class II). This result was consistent with the result that NIL-F₂ having S24-ir/S24-ir S35-as/S35-ir genotype showed pollen semi-sterility and indicated that S35 caused male gamete abortion by interaction with the S24-ir allele.

View Details

Table 2 Segregation of pollen fertility in the progeny of reciprocal cross between pollen sterile plant (PS) and Asominori

View Details

Table 3 Allele frequency of male gametes in the sterile NIL-F2 plants

DISCUSSION

We conducted genome-wide identification of the postzygotic RI genes by using the CSSL series and the derivative NILs which contain small chromosome segments associated with the subject traits. The reciprocal sets of CSSLs derived from the indica/japonica cross have led to identification of the complex epistasis underlying hybrid breakdown (Kubo and Yoshimura, 2002, 2005). In the present study, we found another novel pair of genes responsible for hybrid male sterility. The epistasis was comprised of the two independent loci, S24 and S35, located on chromosomes 5 and 1, respectively. Segregating populations obtained from backcrosses or mutual crosses between two different CSSLs showed clear bimodal segregation for pollen sterility, leading to the discovery of the individual phenotypic effects of the S24 and S35 loci. Our genetic analysis showed that the S35 gene was dependent on the partner gene S24 to cause male sterility, whereas S24 caused male sterility independent of S35. The genetic interaction of these two loci did not correspond to a simple Dobzhansky-Muller model that explanation of deleterious interaction between two genes from the parents. The plant carrying S24-as/S24-ir genotype with IR24 genetic background did not show pollen semi-sterility. Indeed, we obtained the CSSL carrying Asominori homozygous alleles in IR24 genetic background (Kubo et al., 2002). These findings strongly suggest the possibility that a third sterility factor may be present on the Asominori genome, and that the hybrid male sterility could be caused by epistasis among multiple loci including the unidentified third gene. In Drosophila, many studies proposed hybrid male sterility due to polygenes (Tao et al., 2003; Moehring et al., 2006). We have previously demonstrated that epistasis at multiple loci causes female sterility in the hybrid progeny of rice (Kubo and Yoshimura, 2005). Together with the previous results of rice QTL analyses (Wu et al., 1995; Wang et al., 1998), these studies imply that complex gene interaction at multiple loci is the universal nature of gametogenesis, functioning as the postzygotic reproductive barriers that drive new-species evolution. Our observations also propose that diverse alleles at epistatic loci could bring about variable epistatic effects for sterility contributing to phenotypic variation for sterility in the hybrid progeny.

The frequency of the male gamete genotype of S24 and S35 loci in pollen semi-sterile or highly sterile plants also implied complexity of this genetic mechanism. In the double heterozygotes for S24 and S35 loci, only the male gamete with the S24-ir S35-ir genotype was likely to be fertile and normally transmitted to progeny through the male gametes. The other male gametes carrying S24-as S35-as, S24-ir S35-as, and S24-as S35-ir genotype seemed to be aborted at 90–95% frequencies in comparison to the 100% frequency of normal S24-ir S35-ir genotype gamete. Significantly, S24-ir S35-as gametes showed reduced transmission in the double heterozygote, even though most of them were normally developed in the S24-as/S24-ir S35-as/S35-as (Fig. 2, Table 2 and Table 3). This fact strongly suggested that the zygote genotype of S35 is important for expression of the male gamete abortion. The male gametes carrying japonica alleles were consistently eliminated at the S24 or S35 locus. Furthermore, among the four gamete genotype classes given by S24 and S35, the most significant elimination was observed in the male gamete with S24-as S35-as genotype in the double heterozygote plants. Male gamete eliminations with the japonica allele were commonly observed at these two loci. To our knowledge, over 30 hybrid sterility genes have been previously reported in the inter-subspecific hybrids of rice. Some of these reports showed predominant transmission of indica alleles at the reproductive barrier loci due to gamete abortion or selective fertilization. Together with our results in this study, the indica allele at the sterility locus is suggested to have a tendency to be transmitted preferentially with respect to the japonica allele. This tendency is highly interesting from the aspect of speciation and functional evolution of the RI genes.

Previous QTL studies of hybrid sterility showed that a QTL for seed sterility was detected on the short arm of chromosome 5 in the different indica/japonica cross combinations (Wu et al., 1995; Li et al., 1997; Wang et al., 1998). Recently, Zhao et al. (2007) reported S31, a new locus for female sterility at the same position of chromosome 5. The phenotypic effects of these sterility factors were quite different from S24 as they affect seed fertility (Wu et al., 1995; Li et al., 1997; Wang et al., 1998) and S31 specifically affects female gametes but not male gametes (Zhao et al., 2007). On the other hand, neither female gametes nor seed fertility appeared to be affected by a single effect of S24 gene in our own analysis. Because of such dissimilar phenotypes, these reported genes and S24 are unlikely to be the same locus. Another study by Wang et al. (2006) also reported fine-mapping of a pollen sterility gene on the same position of chromosome 5. This is more likely to be the same locus in terms of the similarity of the position and the phenotypic effect. Nevertheless, interestingly, S31 has been found by using our CSSL materials analyzed in this study (AIS line No. 34) (Zhao et al., 2007). We can interpret this phenomenon in the following way, 1) two tightly linked genes for male and female sterility existed in the chromosome region and were expressed in each experimental condition; 2) a single gene involved in both male and female gametogenesis expressed in either female or male gamete in the different environments. In either case, these facts imply that a major factor(s) or a gene cluster that plays a key role for rice gametogenesis is located at the region of chromosome 5. Further cloning analysis and comparing sequences of these sterility alleles would elucidate this complex problem.

Understanding the mechanisms and number of loci responsible for hybrid sterility remains a major challenge in three perspectives: including cross-breeding, molecular biology of gametogenesis and phylogeny of rice. The inter-subspecific hybrids between indica and japonica show a higher yield potential compared with their intra-subspecific hybrids due to a greater genetic divergence. However, postzygotic RI in the inter-subspecific F₁ hybrid hinders utilization of such heterosis in rice breeding. Various hybrid sterility genes, including gamete eliminator (Sano et al., 1994), pollen killer (Kitamura, 1962), and female gamete abortion (Ikehashi and Araki, 1986), have been already identified and mapped on the rice chromosome map. On the other hand, more than twenty major QTLs relating to useful agricultural traits have been mapped and some of them were cloned in rice (Yano et al., 2000; Xue et al., 2008). A comprehensive chromosome map showing the detailed positions of genes for all these characters will provide valuable information for effective breeding programs of rice varieties.

Little has been known of molecules and molecular networks essential for gametogenesis in plants. Recent pollen transcriptome studies have reported that 40% of about 1000 pollen-expressed mRNA were detected specifically in Arabidopsis pollen (Honys and Twell, 2004). In rice, 71 genes have been reported to express specifically in male gametophyte (Suwabe et al., 2008). If loss-of-function mutations occur at two related pollen specific genes in each parental species, these genes act as complementary genes causing hybrid male sterility. Thus, analysis of epistatic loci causing hybrid male sterility provides clues to clarify of the molecular pathway related to male gametogenesis. Moreover, it would provide important information to trace the evolution of rice species. The two major cultivated subspecies, indica and japonica, have been proposed to originate from two different ecotypes of the wild rice progenitor, O. rufipogon, although this hypothesis continues to be debated (Kovach et al., 2007). Investigation of the allele distributions in cultivated rice and its wild relatives will provide a hint to uncover the dynamics of such rice differentiation processes from the original wild rice to the current plethora of diverse cultivated varieties.

This study was supported by the Bio-oriented Technology Research Advancement Institution (BRAIN), Japan.