Introduction

Transmission ratio distortion (TRD) refers to a naturally occurring phenomenon in which the two alleles at a heterozygous locus are not transmitted equally to the progeny, and this leads to a deviation in the genotype frequencies from the expected Mendelian ratios. TRD is induced by a variety of mechanisms, such as non-random chromosome segregation during meiosis (Birchler et al., 2003; Fishman and Saunders, 2008), preferential gamete dysfunction in hybrids (Lyttle, 1991; Moyle and Graham, 2006; Long et al., 2008; Chen et al., 2008; Tao et al., 2009a, 2009b; Phadnis and Orr, 2009) and preferential gamete success during fertilization (Price, 1997; Fishman et al., 2008). Because TRD can dramatically alter the frequency of alleles in a population by disrupting proper Mendelian segregation, it has been hypothesized that TRD is a driving force in the emergence of a reproductive barrier (Frank, 1991; Hurst and Pomiankowski, 1991). With regard to the process of TRD-mediated reproductive barrier formation, Frank (1991) and Hurst and Pomiankowski (1991) independently proposed that the genes responsible for gamete dysfunction in hybrids and consequently induced TRD are fixed rapidly in a population due to their ‘selfish nature,’ but that they may easily become suppressed within a population to alleviate their deleterious effects on fertility. As a result, two allopatric populations might evolve different TRD systems. If these populations later hybridize, normally suppressed TRD within one population will be re-expressed in hybrids of individuals from each population, leading to hybrid sterility, which acts as a reproductive barrier between the two allopatric populations (Frank, 1991; Hurst and Pomiankowski, 1991).

In plants, TRD has been detected many times in interspecific and intraspecific hybrids (Morishima et al., 1992; Koide et al. 2008b; and references therein). Among them, TRD occurring in either the male (mTRD) or female (fTRD) gametes has been frequently reported and some of the genes causing sex-specific TRD have been cloned (Chen et al., 2008; Long et al., 2008). On the other hand, there are few reports on sex-independent TRD (siTRD), which results from preferential transmission of both male and female gametes carrying one of the two alleles in the heterozygote (Rick, 1966; Koide et al., 2008c). Little is known about the genetic basis and evolutionary history of siTRD, although siTRD exerts the strongest effect on segregation distortion among these types of TRD.

We previously reported S6 locus-mediated siTRD in a hybrid of Asian cultivated rice (Oryza sativa) and wild rice (Oryza rufipogon; Sano, 1992; Koide et al., 2008a). Asian cultivated rice and wild rice belong to the same biological species, forming a primary gene pool (O. sativa-O. rufipogon complex) according to the classification system for gene pools (Harlan 1975). Thus, this provides an opportunity to examine the genetic basis of intraspecific TRD. We observed a reduction in seed setting among the F1 plants derived from a cross between T65wx (O. sativa ssp. japonica) and a near-isogenic line (NIL; designated as NIL-S6 in this study) carrying a segment of chromosome 6 derived from a strain of O. rufipogon (Ruf-S6 in this study; Sano, 1992). When the F1 hybrids were reciprocally crossed with T65wx, the resultant BC1F1 progeny plants exhibited a reduced seed-setting rate, whereas the F2 progeny plants derived from self-pollination of the F1 hybrid plants exhibited a normal seed-setting rate (Sano, 1992).

This phenomenon is due to an interaction between a gene designated S6 in the chromosomal segment derived from Ruf-S6, and its opposing allele (S6a) in T65wx. The S6 allele acted as a ‘gamete eliminator,’ and was transmitted more frequently than S6a through both the male and female gametes in heterozygotes (S6/S6a). Female gametes possessing the S6a allele were aborted in the heterozygotes, causing a reduced seed-setting rate (Sano, 1992; Koide et al., 2008a). In contrast, no defect was observed in the pollen grains of the heterozygotes, although male gametes possessing the S6a allele were rarely transmitted to the next generation (Sano, 1992; Koide et al., 2008a). We have also revealed that Asian rice strains frequently harbor an additional allele (S6n), which however, does not induce any preferential abortion in heterozygotes (S6/S6n and S6a/S6n) at the S6 locus (Koide et al., 2008a), as shown by test-cross experiments and subsequent genetic mapping using NILs that carry the genetic background of T65wx. The presence of the S6n allele, which modifies the effect of the S6 allele in heterozygotic state at the S6 locus, suggested that S6 locus-mediated siTRD was caused by the allelic differentiation at the S6 locus occurred during the evolution of Asian rice.

It is conceivable that changes in genetic factors that positively or negatively control S6 locus-mediated siTRD occurred during the evolution of Asian rice, and such changes might have affected the presence or absence of reproductive barrier between constituents of the Asian rice population. With such possibilities in mind, in this study, we compared the effect of S6 locus-mediated TRD between two F2 populations that were produced using a NIL and its original wild strain as respective parental strains for crossing, and examined whether there are genes which modify the effect of S6 locus-mediated siTRD that exist in the genetic background of Asian rice strain. We also examined the extent of male- and female-specific TRD by reciprocal backcross experiments. Based on the results, together with those of subsequent genetic and cytological analyses, we report the involvement of unlinked modifiers and sex-specific mechanisms in this phenomenon.

Materials and methods

Genetic stocks

Three lines, T65wx, Ruf-S6 and NIL-S6 were used. T65wx carries wx (waxy) gene as a genetic marker in the genetic background of Taichung 65 (O. sativa ssp. japonica). Ruf-S6 is a perennial type strain of O. rufipogon, W593. NIL-S6 carries the short arm and a portion of the long arm of chromosome 6 from Ruf-S6 in the genetic background of T65wx (Sano, 1992; Matsubara et al., 2003; Koide et al., 2008a formally named as T65S6 [W593]). T65wx harbors the S6a allele at the S6 locus (near the centromeric region of chromosome 6), whereas Ruf-S6 and NIL-S6 harbor the S6 allele at the S6 locus (Koide et al., 2008a). Although T65wx harbors wx gene from Kinoshita-mochi (Oka, 1974; derived from BC12), wx gene does not affect S6 locus-mediated TRD.

Genetic crosses and genotyping to detect S6 locus-mediated TRD

To examine the effect of S6 locus-mediated TRD on linked loci on chromosome 6, a total of 98 F2 segregating plants derived from T65wx × NIL-S6 were genotyped using 15 DNA markers from chromosome 6 (Wx, E12, R1962, RM204, RM314, OsC1, RM276, RM539, Hd1, R538, R111C, R32, RM3498, G2028, and RM1340). Additionally, to examine the effect of S6 locus-mediated TRD in the hybrids between O. sativa and the original wild strain of O. rufipogon, a total of 103 F2 segregating plants derived from T65wx × Ruf-S6 were genotyped using eight DNA markers from chromosome 6 (E12, RM204, RM276, Hd1, R111C, RM3, RM3498, and RM1340).

To further characterize the S6 locus-mediated TRD in the cross of T65wx × Ruf-S6, transmission of the S6 allele through males (that is, mTRD) and females (that is, fTRD) was assessed by reciprocal backcross experiments. To estimate the degree of mTRD, F1 plants (T65wx × Ruf-S6) were used as the pollen parents and pollinated to female T65wx and Ruf-S6 plants. On the other hand, to estimate the degree of fTRD, F1 plants (T65wx × Ruf-S6) were used as the female parents and pollinated with male T65wx and Ruf-S6 plants. The segregation ratio at the S6 locus was estimated from that of the tightly linked DNA marker R111C.

For genotyping, genomic DNA was isolated from a small piece of frozen leaf according to the method of Monna et al. (2002), with slight modifications. Three Indel markers (Wx, OsC1 and Hd1), three restriction fragment length polymorphism markers (R538, R32 and G2028) and a cleaved amplified polymorphic sequence (CAPS) marker, E12, from chromosome 6 were used for genotyping according to the method of Matsubara et al. (2003). A CAPS marker, R111C, was used according to the method of Koide et al. (2008a). Seven microsatellite markers (RM204, RM314, RM276, RM539, RM3498, RM3 and RM1340) were selected from a public database (http://www.gramene.org). Additionally, one CAPS marker, R1962, was designed based on a sequence from the public database (acc. no. AP006554). The sequences of the primers used for a CAPS marker, R1962, were 5′-gct tgg att atg aca ttt ag-3′ and 5′-tga agc aag gaa caa aca-3′. To detect the polymorphism, the amplified products were digested with TaqI. The recombination values were estimated, based on the maximum likelihood method (Allard, 1956).

Cytological observations and pollen tissue PCR

Spikelets were sampled from the panicles before heading. The samples were fixed in formalin: glacial acetic acid: 70% ethanol (1:1:18) and stored in 70% ethanol. The ovaries were dehydrated in a graded ethanol–butanol series, embedded in Paraplast Plus (Oxford Labware, St. Louis, MO, USA), and then cut into 10-μm thick sections. The sections were stained with safranin and Fast Green (Sylvester and Ruzin, 1993) and observed by light microscopy (BH-2, Olympus, Tokyo, Japan).

To examine whether the S6 locus-mediated mTRD occurred before or after pollen grain production, pollen grains from heterozygous plants were genotyped according to the method of Petersen et al. (1996) with modifications. A total of 2–3 μg of pollen grains were collected from F1 plants derived from T65wx × NIL-S6 at the flowering stage and transferred to tubes containing 32.7 μl of H2O, 5 μl of 10 × Takara Ex Taq buffer (Takara Bio, Otsu, Japan), 5 μl of 50% dimethyl sulfoxide, 2.5 mM of each deoxyribonucleotide triphosphate, 1 μl of a 20 pM solution of each primer and 0.3 μl of Takara Ex Taq DNA polymerase (5 U μl−1). The CAPS marker R111C was used for genotyping. PCR was performed for 30 cycles (1 min at 96 °C, 1 min at 56 °C and 1 min at 72 °C), followed by 10 min at 72 °C. For polymorphism detection, the amplified products were separated electrophoretically on a 2.5% agarose gel in 1 × TAE buffer and the DNA fragments were detected by staining with ethidium bromide.

Results

Effects of the genetic background on S6 locus-mediated TRD

To examine the effect of genetic background on the strength of S6 locus-mediated siTRD, we analyzed the difference in TRD at the S6 locus between two F2 populations derived from crosses of T65wx × NIL-S6 and T65wx × Ruf-S6. To compare the effect of S6 locus-mediated TRD, we used the DNA marker R111C, which is tightly linked with the S6 locus (Koide et al., 2008a).

Although TRD was detected in both crosses, the effect was different. In the F2 population derived from T65wx × NIL-S6, almost all of the plants (84/98) were homozygous for the O. rufipogon-derived allele (S6). No homozygote for the O. sativa-derived allele (S6a) was detected (Table 1), indicating that transmission of the S6a allele was reduced in both the female and male gametes (that is, siTRD), consistent with previous data (Sano, 1992; Koide et al., 2008a). However, in the F2 population derived from T65wx × Ruf-S6, the numbers of homozygotes for the O. rufipogon-derived allele (S6), heterozygotes and homozygotes for the O. sativa-derived allele (S6a) were 48, 49, and 6 respectively (Table 1). The segregation ratio of the F2 plants was close to 1:1:0 in this cross.

Table 1 Frequencies of each allele of a DNA marker (R111C) in the F2 plants from the crosses of T65wx × NIL-S6 and T65wx × Ruf-S6, and the BC1F1 plants
Full size table

Such a difference in the segregation ratio between the two cross combinations can be explained by either of the following models: (1) the degree of S6 locus-mediated TRD was changed by unlinked genes when the original wild strain of O. rufipogon (Ruf-S6) was used; (2) a novel TRD which tends to transmit the O. sativa-derived allele (S6a) and counteracts the over-transmission of the S6 allele occurred at a locus linked to S6 when the original wild strain of O. rufipogon (Ruf-S6) was used. To examine these two possibilities, the segregation ratio at markers on chromosome 6 was analyzed using two F2 populations derived from crosses of T65wx × NIL-S6 and T65wx × Ruf-S6 (Figure 1). In both cases, strong TRD was detected only near the centromeric region where S6 is located. Moreover, with an increase in the genetic distance from the centromeric region, the degree of TRD decreased. If other loci on chromosome 6 were to affect the segregation pattern, the pattern of reduction in TRD should be affected near the causative loci. Thus, these results suggest that no novel TRD occurred on chromosome 6, but the degree of the S6 locus-mediated TRD was changed by unlinked genes when the original wild strain of O. rufipogon (Ruf-S6) was used as one of the parents. In addition, in both populations, TRD was detected even at distal DNA marker loci 50 cM distant from R111C, indicating that the S6 locus-mediated TRD affected most of this chromosomal region, irrespective of the genetic background.

Figure 1
figure 1

Map position and TRD of markers on chromosome 6 in the F2 populations. (a) Physical map of the DNA markers on chromosome 6, based on Rice Genome Research Program data (http://rgp.dna.affrc.go.jp). The solid circle represents the centromere. (b) Frequency of each allele of the DNA markers along the genetic linkage map of chromosome 6 in F2 populations derived from T65wx × NIL-S6 (n=98) and T65wx × Ruf-S6 (n=103). The position of each marker was determined, based on the genetic distance (in cM) from R111C. The frequencies of the O. rufipogon homozygous genotype (solid squares), heterozygous genotype (open circles) and O. sativa homozygous genotype (open squares) are plotted at the marker positions.

Full size image

The degree of S6 locus-mediated mTRD depends on the female parent

The segregation ratio of homozygotes for the O. rufipogon-derived allele (S6), heterozygotes and homozygotes for the O. sativa-derived allele (S6a) at R111C was close to 1:1:0 in the F2 plants derived from T65wx × Ruf-S6, as mentioned above (Table 1). This result suggests that the transmission of the S6a allele was reduced through female or male gametes (fTRD or mTRD), or that transmission of the S6a allele was partially reduced through both female and male gametes. To examine which type of TRD occurred in the progeny of the cross between O. sativa (T65wx) and O. rufipogon (Ruf-S6), we carried out backcrossing experiments. Using F1 plants as the female parents, the degree of fTRD was estimated from the segregation ratio of BC1F1 plants. In contrast, the degree of mTRD was estimated using F1 plants as the male parents.

All of the BC1F1 plants were heterozygous or homozygous for the O. rufipogon-derived allele (S6) at R111C, when F1 plants were used as the female parents and crossed with T65wx or Ruf-S6, respectively (Table 1). Thus, the proportion of the transmission of S6 through female gametes was 100%, indicating complete fTRD. Similarly, when T65wx plants were used as the female parents and crossed with F1 plants, almost all of the BC1F1 plants (25/26) were heterozygous (Table 1), indicating mTRD. In contrast, when Ruf-S6 plants were used as the female parents and crossed with F1 plants, the transmission ratio of S6 through male gametes was 70% (19/26; Table 1), indicating incomplete mTRD. There was a significant difference in the transmission ratios of S6 through male gametes between the two BC1F1 populations (P=0.049 by Fisher's exact test), indicating that the degree of S6 locus-mediated mTRD varied depending on the background genotype of the female (pistil) parent. These results suggest that the degree of S6 locus-mediated mTRD was partly suppressed by unlinked modifier(s) in the progeny of the cross between O. sativa (T65wx) and O. rufipogon (Ruf-S6), whereas that of fTRD was not suppressed. Moreover, these results also suggest that heterozygotes (S6/S6a) produced both S6 and S6a pollen grains of normal fertilization potential.

Abortion occurs after meiosis in female gametogenesis, but not in male gametogenesis

Our backcross experiments suggested that S6 locus-mediated preferential abortion occurred in female gametes, whereas it did not occur in pollen grains in the heterozygotes (S6/S6a). To test this possibility, cytological observations were performed and the specific developmental stage at which the abnormality occurred was determined (Figure 2). Abnormal ovules were detected in the heterozygotes: bi-nucleate embryo sacs with a single enlarged nucleus (Figure 2a), tri-nucleate (Figure 2b) and penta-nucleate embryo sacs (Figure 2c) were observed in the abnormal ovules. This indicates that a defect in the S6a female gametophyte in the heterozygotes occurred during the mitotic stage; thus, the S6 locus-mediated fTRD occurred after meiosis.

Figure 2
figure 2

Embryo sacs at different developmental stages in the S6/S6a heterozygotes and S6a/S6a homozygotes. (a–c) Abnormal embryo sacs in the S6/S6a heterozygotes. (a) Abnormal bi-nucleate embryo sac with enlarged nuclei (arrowhead). (b) Abnormal tri-nucleate embryo sac. (c) Abnormal penta-nucleate embryo sac. (d–g) Normal embryo sac development in the S6a/S6a homozygotes. (d) A functional megaspore. (e) A bi-nucleate embryo sac. (f) A tetra-nucleate embryo sac. (g) An embryo sac after the third division. EN, egg nucleus; SY, synergid; PN, polar nuclei; AN, antipodal cell nuclei. Bar=20 μm.

Full size image

On the other hand, no developmental defect was observed in the mono-, bi- and tri-nucleate stages of pollen development in the heterozygotes (S6/S6a). To examine the genotype of mature pollen grains produced in the heterozygotes (S6/S6a), pollen tissue PCR was carried out. DNA fragments that corresponded to both genotypes were amplified by PCR from pollen grains, as were amplified from leaf DNA (Figure 3), indicating that the heterozygotes (S6/S6a) produced both S6 and S6a pollen grains. Taken together, these results indicate that the preferential abortion of gametes occurred after meiosis in the S6 locus-mediated fTRD, whereas no detectable abnormality occurred in the S6 locus-mediated mTRD.

Figure 3
figure 3

Genotype of pollen grains from a heterozygote as determined using the marker R111C. S6a/S6a, S6/S6, and S6/S6a indicate homozygotes for the O. sativa-derived allele, homozygotes for the O. rufipogon-derived allele and heterozygotes, respectively.

Full size image

Discussion

Chromosomal regions affected by the TRD caused by allelic interactions at the S6 locus

The S6 locus has been mapped to a region including the centromere of chromosome 6 (Koide et al., 2008a). In the present study, we found that the degree of TRD caused by the S6 locus decreased along with the genetic distance from the centromeric region in the F2 population derived from the cross between T65wx and NIL-S6 (Figure 1). If other hybrid sterility loci on chromosome 6 were to affect the segregation pattern in this cross combination, the pattern of the reduction in TRD should be affected near the causative loci. A clear reduction pattern in TRD towards the distal end of chromosome 6 was observed, indicating that the segregation distortion caused by the S6 locus was independent of that caused by other hybrid sterility loci, as had been previously suggested (Koide et al., 2008a). Moreover, a similar pattern of reduction in TRD was observed in the F2 population derived from the cross between T65wx and Ruf-S6 (Figure 1). These results suggest that the S6 locus is the causal factor of TRD on DNA marker loci on chromosome 6 in both of the F2 populations derived from T65wx × NIL-S6 and T65wx × Ruf-S6.

In Mimulus, Fishman and Willis (2005) examined the pattern of the reduction in TRD by developing NILs with a meiotic drive locus, D, from M. guttatus. The D allele exhibited a nearly 100% transmission advantage via female meiosis in hybrids with M. nasutus (Fishman and Willis, 2005). The effect of the TRD caused by the D locus was observed even at a locus 55 cM away. Similarly, the effect of the strong TRD induced by an alien 5B chromosome was observed at a locus 50 cM from the most distorted locus in wheat (Kumar et al., 2007). The chromosomal ranges affected by the S6 locus were comparable to those affected by the most distorted locus in Mimulus and wheat, suggesting that strong TRD often affects a locus 50 cM distant.

fTRD, governed by the centromeric region, occurred after meiosis

In this study, the most severe TRD was observed at R111C near the centromere. This result is comparable to that from genetic mapping using a segregating population consisting of a large number of individual plants (Koide et al., 2008a). Several examples of TRD near centromeric or neocentromeric regions have been reported in Mimulus and maize (Dawe and Cande, 1996; Yu et al., 1997; Fishman and Willis, 2005; Fishman and Saunders, 2008). In Mimulus, because the D locus near the centromere caused significant fTRD without an increase in ovule or seed mortality, it was suggested that fTRD is a consequence of the preferential transmission of chromosomes with a centromere containing the D allele during asymmetric female meiotic division processes (Fishman and Willis, 2005; Malik, 2005). The Ab10/knob system in maize involves the genetic activation of neocentromeric knob regions that competitively bind microtubules and orient the carrier chromatids toward the outer spindle poles at meiosis II (Dawe and Cande, 1996; Yu et al., 1997). In both cases, the fTRD, which is governed by the centromeric or neocentromeric region, occurs during meiosis, with no deleterious effect on female gametes.

In the S6 locus-mediated fTRD system, approximately half of the ovules exhibited an abnormality in embryo sac structure during female gametogenesis, and the seed-setting rate was reduced in heterozygotes (S6/S6a; Koide et al., 2008a), indicating that fTRD occurred post-meiosis, which is different from that mediated by the D locus in Mimulus or the Ab10/knob system in maize. By cytological observation, bi-nucleate embryo sacs with a single-enlarged nucleus, tri-nucleate embryo sacs and penta-nucleate embryo sacs were found in the abnormal embryo sacs produced by the heterozygotes (S6/S6a; Figure 2), indicating that an abnormality in nuclear division or migration occurred during the second or third round of mitosis after meiosis.

Mutations affecting female gametogenesis after the mono-nucleate stage have been reported in Arabidopsis and maize (Sheridan and Huang, 1997; Drews et al., 1998). In Arabidopsis hdd (hadad) mutants, female gametophytes are arrested at the bi-, tetra- or octa-nucleate stage (Drews et al., 1998). In lo2 (lethal ovule2) mutants in maize, nuclear division is affected and embryo sacs are arrested at the mono-, bi- or tetra-nucleate stage, and in some cases, the nuclei enlarge dramatically, suggesting a failure of entry into the prophase (Sheridan and Huang, 1997). In the embryo sacs of the lo2 mutants, abnormal behavior of the tubulin cytoskeleton was also observed. The failure to display a normal pattern of cytoskeleton behavior in the mutant embryo sacs was suggested to be an indirect result of abnormal interactions between defective nuclei lacking normal nuclear surface features and microtubule components of the microtubular cytoskeleton that are required for normal spindle orientation and nuclear migration (Huang and Sheridan, 1994; Sheridan and Huang, 1997).

The phenotype observed in the S6 locus-mediated fTRD system is similar to the hdd mutants in Arabidopsis and lo2 mutants in maize. In all cases, embryo sacs are arrested during mitotic division. Moreover, in the cases of S6 and lo2, enlarged nuclei in the abnormal embryo sacs were observed. Based on the fact that the abnormalities in the embryo sacs of the S6/S6a heterozygotes were similar to those in the hdd and lo2 mutants, and given that S6 was mapped to a region including the centromere where the attachment of microtubules to the kinetochore occurs during mitosis, it appears likely that S6 is located close to the centromere and that its location and/or function disrupts the normal relationship between microtubules and the centromeric region. Detailed analyses of the behavior of the chromosomes or cytoskeleton during mitosis will help advance our understanding of the molecular mechanisms underlying the S6 locus-mediated preferential abortion of female gametes.

Genetic mechanisms controlling the degree of mTRD

In this study, differences in the degree of TRD at the S6 locus were observed between two F2 populations derived from crosses between T65wx and a NIL (NIL-S6), and between T65wx and the original wild strain (Ruf-S6). siTRD was observed in the F2 population derived from T65wx × NIL-S6, whereas the degree of TRD was reduced in the F2 population derived from T65wx × Ruf-S6. The segregation ratio of homozygotes for the O. rufipogon-derived allele (S6), heterozygotes and homozygotes for the O. sativa-derived allele (S6a) was close to 1:1:0 in this latter population (Table 1). Because NIL-S6 and Ruf-S6 are of different genetic backgrounds, the effect of S6 locus-mediated siTRD may be due to differences in genes in the respective genetic backgrounds. Moreover, backcrossing experiments revealed that the degree of mTRD was reduced only when Ruf-S6 was used as the female (pistil) parent, whereas transmission of the S6 allele through the female parent (fTRD) was 100% when T65wx or Ruf-S6 was used as the male (pollen) parent (Table 1). Transmission of the S6a allele from male T65wx × Ruf-S6 plants was observed following crosses with female Ruf-S6 pistils (Table 1), and pollen grains carrying the S6a allele were detected by tissue PCR in the heterozygotes (Figure 3). Thus, the heterozygotes produced not only S6, but also S6a pollen grains with normal fertilization potential, consistent with previous cytological observations of normal mature pollen grains in S6/S6a heterozygotes (Koide et al., 2008a). This suggests that the mTRD observed in the cross between the T65wx × Ruf-S6 male and T65wx female was not due to the dysfunction of pollen grains carrying the S6a allele, and occurred after pollen grain production.

A plausible mechanism for the mTRD, which occurred after pollen grain production, is the difference in pollen performance, such as the ability of germination or the rate of pollen tube elongation, between the two types of pollen grains (that is, those carrying the S6 and S6a alleles). Further experiments on the ability of pollen germination or the rate of pollen tube elongation might reveal a difference between pollen grains carrying the S6 and S6a alleles. Pollen tube competition has been observed in diverse plant taxa (for example, Nelson, 1993; Ramsey et al., 2003; Rahme et al., 2009). In maize and rice, numerous loci for gametophyte factor (ga) have been reported. The ga allele is known to confer a pronounced advantage on fertilization as the result of competition among pollen grains, leading to mTRD in later generations. In the extreme case of pollen competition caused by the maize ga1 locus, the growth of ga1 pollen tubes is retarded or arrested, depending on the genotype of the female parent (Nelson, 1993). In the Silene genus, the effect of competition between the pollen grains from S. latifolia and S. dioica is also related to the genotype of the female parent (Rahme et al., 2009).

The degree of S6 locus-mediated mTRD was reduced only when plants with a Ruf-S6 genetic background were used as the female (pistil) parent in the backcross experiments (Table 1), suggesting that the difference in pollen performance is controlled by an interaction between the pollen (S6 or S6a) and pistil genotypes, and that the effects of the difference in pollen performance were weakened or partly suppressed by modifiers in the genetic background of the female Ruf-S6. To identify the modifier(s) involved in the suppression of mTRD, the development of recombinant inbred lines, each with different chromosomal segments in the genetic background, will be needed. A question arises as to how such a pattern of the difference in pollen performance and its modifier evolved in Asian rice population. It is tempting to speculate that O. rufipogon, which has a relatively higher outcrossing rate than O. sativa, might have traits suitable for outcrossing, such as high pollen competition ability and a capacity of stigmas to receive alien pollen. On the other hand, O. sativa, which is a predominantly selfing plant, might have lost such traits during the evolutionary process. Further analysis of the causative genes will help shed light on the evolution of mTRD and its modifier(s) in Asian rice.

We note that the result of our backcrossing experiments is not fully consistent with the segregation pattern observed in the F2 population derived from T65wx × Ruf-S6. In our experiments, approximately 27% of the S6a allele was transmitted to the progeny through male gametes when Ruf-S6 was used as the female (pistil) parent, whereas no S6a allele was transmitted to the progeny when T65wx or Ruf-S6 was used as the male (pollen) parent (Table 1). On the other hand, the segregation ratio of homozygotes for the O. rufipogon-derived allele (S6), heterozygotes and homozygotes for the O. sativa-derived allele (S6a) in the F2 population, was close to 1:1:0 (Table 1), suggesting that approximately 50% of S6a allele was transmitted to the F2 plants through male gametes. Moreover, a few homozygotes for S6a were observed in the F2 population, suggesting that the S6a allele was transmitted through both male and female parents, even though the transmission frequency was very low (Table 1). Although it is still unclear why the transmission ratio of the S6a allele in backcrossing was different from that in selfing, there are several possibilities that may explain the result. One simple explanation is that the number of samples in the backcross experiments might have not been large enough to detect transmission of S6a allele through the female parent. Alternatively, abnormalities, which induce failure in seed development and segregation ratio distortion in the subsequent generation, might have occurred after backcrossing. Another possibility is that a complex mechanism involving unknown factors in the genetic background, such as an epistatic interaction or a heterospecific gene interaction between male (pollen) and female (pistil) parents, might have reduced the degree of TRD in the F2 plants derived from T65wx × Ruf-S6.

Although the underlying mechanisms are unknown, these results show that the transmission of the S6 allele through female gametes (fTRD) was nearly complete, whereas the transmission of the S6 allele through male gametes (mTRD) changed depending on the genotype of the female (pistil) plants, suggesting the involvement of unlinked modifiers in this phenomenon. Furthermore, the results suggest that two different genetic mechanisms controlling mTRD and fTRD are involved in S6 locus-mediated siTRD, though it is unknown whether these two phenomena are governed by two tightly linked genetic components or the pleiotropic effects of a single gene. In combination with the observation that the degree of S6 locus-mediated TRD differed between different combinations of cultivated and wild rice strains (Table 1; Koide et al., 2008a), the finding of modifier(s) and sex-specific mechanisms in this study raises the possibility that multiple genetic factors affect the degree of siTRD mediated by the S6 locus, apart from the S6n allele. TRD of various degrees could have been established by different combinations of genes in Asian rice.