The Evolution of Sex-Independent Transmission Ratio Distortion Involving Multiple Allelic Interactions at a Single Locus in Rice

Abstract

Transmission ratio distortion (TRD) is frequently observed in inter- and intraspecific hybrids of plants, leading to a violation of Mendelian inheritance. Sex-independent TRD (siTRD) was detected in a hybrid between Asian cultivated rice and its wild ancestor. Here we examined how siTRD caused by an allelic interaction at a specific locus arose in Asian rice species. The siTRD is controlled by the S₆ locus via a mechanism in which the S₆ allele acts as a gamete eliminator, and both the male and female gametes possessing the opposite allele () are aborted only in heterozygotes (). Fine mapping revealed that the S₆ locus is located near the centromere of chromosome 6. Testcross experiments using near-isogenic lines (NILs) carrying either the S₆ or alleles revealed that Asian rice strains frequently harbor an additional allele () the presence of which, in heterozygotic states ( and ), does not result in siTRD. A prominent reduction in the nucleotide diversity of S₆ or carriers relative to that of carriers was detected in the chromosomal region. These results suggest that the two incompatible alleles (S₆ and ) arose independently from and established genetically discontinuous relationships between limited constituents of the Asian rice population.

TRANSMISSION ratio distortion (TRD) refers to a naturally occurring phenomenon in which the two alleles at a heterozygous locus do not transmit equally to the progeny (Crow 1988; Lyttle 1991; Temin et al. 1991). TRD is induced by a variety of mechanisms, including the nonrandom segregation of chromosomes during meiosis (Pardo-Manuel de Villena and Sapienza 2001; Birchler et al. 2003; Fishman and Willis 2005), preferential dysfunction of gametes in hybrids (Lyttle 1991; Temin et al. 1991; Silver 1993; Moyle and Graham 2006; Úbeda and Haig 2005), and preferential success of gametes in fertilization (Price 1997; Diaz and MacNair 1999). TRD is detected not only within a given population but also between populations and/or species. Genomewide surveys have frequently revealed significant TRD in intra- and interspecific hybrids (Moyle and Graham 2006). Because TRD dramatically alters the frequencies of alleles in a population by disrupting Mendelian segregation, it has been hypothesized that TRD is the driving force that contributes to the rise of reproductive barriers (Frank 1991; Hurst and Pomiankowski 1991; Orr and Irving 2004).

In plants, preferential dysfunction of gametes can take place in either the male (Cameron and Moav 1957; Loegering and Sears 1963; Sano 1983) or the female (Maguire 1963; Scoles and Kibirge-Sebnya 1983), or in both (Rick 1966; Endo and Tsunewaki 1975; Sano et al. 1979; Finch et al. 1984). Among these types, sex-independent TRD (siTRD) exerts the strongest effect on segregation distortion, since the abortion of both male and female gametes carrying one of the two alleles is induced by the presence of the opposite allele in the heterozygote (Rick 1966; Endo and Tsunewaki 1975; Sano et al. 1979; Finch et al. 1984).

In rice, extensive studies have been carried out on the genetic basis of isolating barriers in intra- and interspecific hybrids, since barriers often present a serious problem in hybridization efforts in breeding programs. A variety of genetic mechanisms have been proposed to explain hybrid sterility (reviewed by Koide et al. 2008a), including those involving combinations of chromosomes bearing cryptic structural differences (Li et al. 1997), nuclear and cytoplasmic genomes of different origin (Shinjo 1984), recessive alleles of duplicate genes (Oka 1974, 1988), two complementary dominant genes (Chu and Oka 1972; Li et al. 1997), and different alleles in a single gene (Ikehashi and Araki 1986; Morishima et al. 1992). Among these, the interaction between alleles is most frequently observed as the cause of preferential dysfunction of gametes leading to TRD (Ikehashi and Araki 1986; Morishima et al. 1992; Koide et al. 2008a). Recent genomewide surveys have revealed the presence of a number of chromosomal regions that are involved in TRD in inter- and intraspecific crosses in rice (Xu et al. 1997; Harushima et al. 2001). However, there has been no report to date that addresses how each TRD system has arisen and become fixed in the rice populations.

We have previously reported that a gamete eliminator (S₆) functions in a hybridization between Oryza sativa and O. rufipogon (Sano 1992). The hybrid plants between T65wx (O. sativa ssp. japonica) and near-isogenic lines (NILs) carrying a segment of chromosome 6 derived from W593 (O. rufipogon) in the genetic background of T65wx exhibited a reduced rate of seed setting. When the hybrids were reciprocally crossed with T65wx, all the resultant progeny exhibited a reduced seed-setting rate, while the progeny derived from self-pollination of the hybrid plants exhibited a normal seed-setting rate (Sano 1992). This phenomenon was due to an interaction between the gene designated S₆ in the chromosomal segment derived from W593 and its opposite allele () in T65wx. The S₆ allele acts as a “gamete eliminator,” and both male and female gametes possessing the allele are aborted only in the heterozygote () (Sano 1992). This locus thus affords an opportunity to examine the genetic basis and evolution of the siTRD system. In the present study, we focus on siTRD caused by the allelic interaction of the S₆ locus and report the first detailed characterization of the siTRD system. We describe the presence of an additional allele () that induces no preferential abortion in heterozygotes with either the S₆ or allele, which was revealed by testcrosses examining allelic distribution in wild and cultivated rice accessions. We also report the histological analysis of gametogenesis, genetic mapping of the S₆ gene, and the analysis of nucleotide diversity around the S₆ locus. On the basis of the findings, the involvement of the siTRD system in the formation of isolating barriers in the evolution of the Asian rice population is discussed.

MATERIALS AND METHODS

Genetic stocks:

Three NILs, T65wx, T65S₆ (W593), and T65WxSe1 (Pat), were made and used for genetic mapping (Table 1). T65S₆ (W593) and T65WxSe1 (Pat) harbor the short arm of chromosome 6 from W593 (O. rufipogon from Malaysia) and Patpaku (ssp. indica of O. sativa from Taiwan), respectively, in the background of T65wx. The detailed genotypes of these three NILs are described in Matsubara et al. (2003). To examine the allelic distribution at the S₆ locus, two other NILs, and T65st₁S₆, were also established and used for the analysis (Table 1). These two NILs carried stripe1 (st1) as a visible marker, which is linked with the S₆ locus (P = 0.13, Sano 1992). carried the allele from T65wx. T65st₁S₆ was selected from an F₂ population of × T65S₆ (W593). To fine-map the S₆ gene, another NIL, A58S₆ was established using A58 (ssp. japonica of O. sativa) as the recurrent parent because of the short life cycle of A58 (Table 1). A58S₆ carries the S₆ allele introduced from T65S₆ (W593).

TABLE 1.

Genetic stocks of NILs and recurrent parents used in this study

	Genotype
Lines	st1	S6	Donor (species)	Backcross generation	Reference
T65S6 (W593)a	+	S6	W593 (O. rufipogon)	BC8	Sano (1992)
T65 WxSe1 (Pat)b	+	c	Patpaku (O. sativa ssp. indica)	BC8	Dung et al. (1998)
A58S6	+	S6	W593 (O. rufipogon)	BC4 and BC5	This study
	st1		F1176 (O. sativa ssp. japonica)	BC7	This study
T65st 1S6	st1	S6	F1176 (O. sativa ssp. japonica) and W593 (O. rufipogon)	BC7	This study
T65wx	+		Kinoshita-mochi (O. sativa ssp. japonica)	BC12	Oka (1974)
A58	+		Kokusyokuto-2 (O. sativa ssp. japonica)

^aFormerly named W593A (Sano 1992).

^bFormely named T65 (Wx-pat) (Dung et al. 1998).

^cThe allelic state was determined in this study.

Seeds were germinated in petri dishes at 30° in late April, and each of the seedlings was transplanted into plastic pots and grown in a greenhouse. The plants were placed in short-day fields (10.5 hr) 8 weeks after sowing due to photoperiod sensitivity, as needed.

Cytological observations:

Spikelets were sampled from panicles before heading. Samples were fixed in FAA (formalin:glacial acetic acid:70% ethanol = 1:1:18) and stored in 70% ethanol until use. Pollen fertility was estimated from the percentages of pollen grains stainable with potassium iodine solution (I₂-KI). Ovaries were dehydrated in a graded ethanol-butanol series, embedded in Paraplast Plus (Oxford Labware, St. Louis), and then cut into 10-μm thick sections. Sections were stained with safranin and fast green (Sylvester and Ruzin 1993) and observed under light microscopy (BH-2, Olympus, Tokyo).

Mapping of the S₆ gene:

To map the S₆ locus, a total of 1886 segregating plants of A58 × A58S₆ were genotyped. Since the S₆ allele kills female gametes possessing the allele in heterozygotes (), the seed-setting rate was analyzed to detect the heterozygotes of according to the method described below. In addition, three recombinant lines (P-1, P-2, and P-3) were obtained from the derivatives of T65wx × T65WxSe1 (Pat) to determine the allelic state at the S₆ locus of Patpaku.

For genotyping the alien segments with molecular markers, genomic DNA was isolated from a small piece of frozen leaf according to the method of Monna et al. (2002) with slight modifications. Six markers on chromosome 6 (Wx, OsC1, Hd1, RG264, RM3498, and G2028) were used according to the method of Matsubara et al. (2003). Two microsatellite markers (RM3183 and RM3498) were selected from the public database (http://www.gramene.org/). In addition, five cleaved amplified polymorphic sequence (CAPS) markers (S14439, P139, R111C, G05, and C133A) were designed on the basis of sequences in the public database (accession nos. AP003763, AP003574, AP005656, AP005967, and AP005450). The primers for PCR amplification were as follows: 5′-ccg aaa aga gtc ctc cga ag-3′ and 5′-cca cct aag aag cca gca cc -3′ for S14439; 5′-gaa atg cca ctg gcc tac at-3′ and 5′-ttc agg cga gca att tag gt-3′ for P139; 5′-tca ggg cta atc aat ggc gaa g-3′ and 5′-tta gtg gat gcc tgg acg atg a-3′ for R111C; 5′-cca ttc ctc cgt cca aac aca t-3′ and 5′-ccc aaa tca cac aca tcg tgc t-3′ for G05; and 5′-cct aaa cgc aag cca ctg tc-3′ and 5′-gca ttg cat gtt cag ttt tc-3′ for C133A. To detect the polymorphisms in the CAPS markers (S14439, P139, R111C, G05, and C133A), the amplified products were digested with HpyCH4, AluI, HinfI, Sau3A, and MspI, respectively. The recombination values were estimated on the basis of the maximum likelihood method.

Survey for allelic distribution at the S₆ locus:

To examine the distribution of the S₆ allele in Asian rice strains, 9 strains of O. sativa and 14 strains of O. rufipogon were surveyed (Table 2). The strains of O. sativa included 3 strains of ssp. japonica and 6 ssp. indica. Since intervarietal crossings frequently produce semi-sterility as well as TRD in rice (Oka 1988), the siTRD specific to the S₆ locus was evaluated on the basis of the level of seed setting of F₁ plants and distorted segregation for a marker (st₁) linked to S₆. The rate of seed setting was determined by counting the number of fertile and sterile spikelets of two panicles for each plant. Strains carrying the S₆ allele are expected to yield a low seed-setting rate in F₁ hybrids and a segregation distortion for st₁ in F₂ when crossed with , while strains carrying the allele are expected to exhibit these abnormalities when crossed with T65st₁S₆. Furthermore, when no difference is found between the crosses with both and T65st₁S₆, the strain is expected to carry an additional allele (), the presence of which in heterozygotic states ( and ) does not result in siTRD, as reported in the tomato by Rick (1966, 1971). To determine the allelic state of W1807 (O. rufipogon), segregation distortion was examined for a molecular marker, P139, closely linked to S₆ (Figure 2).

TABLE 2.

Seed setting and segregation patterns of the marker st1 observed in testcrosses for allelic identification in the S₆ locus

				Crossing with T65st1S6		Crossing with
Species	Strain	Subspecies or type	Origin	F1 seed setting (A)a	F2st1 (%)	F1 seed setting (B)a	F2st1 (%)	A/B	Putative allele
O. sativa	T65	ssp. japonica	Taiwan	34.1	58.4b,**	72.3	23.9	0.47
	Koshihikari	ssp. japonica	Japan	41.5	60.2b,**	88.0	22.6	0.47
	A58	ssp. japonica	Japan	46.2	57.3b,**	86.0	22.8	0.53
	PTB10	ssp. indica	India	16.7	20.3	14.0	17.8	1.19
	IR36	ssp. indica	Philippines	80.2	17.1**	76.7	17.2	1.05
	Patpaku	ssp. indica	Taiwan	72.7	23.3	67.0	21.3	1.09
	Acc27590	ssp. indica	Bangladesh	82.1	—	79.0	—	1.04
	Acc27591	ssp. indica	Bangladesh	31.7	9.2**	44.1	24.1	0.72
	444	ssp. indica	India	65.2	17.8	54.9	26.5	1.19
O. rufipogon	W107	Annual	India	42.1	29.7	55.1	20.5	0.76
	W2002	Annual	Myanmar	80.3	21.6	74.4	22.2	1.08
	W630	Annual	Myanmar	47.6	12.5**	42.7	—	1.11
	W2048	Perennial	China	20.4	45.1**	25.7	23.5	0.79
	W1718	Perennial	China	52.5	—	40.8	—	1.29
	W1943	Perennial	China	71.0	—	73.3	—	0.97
	W1944	Perennial	China	31.1	18.4	23.5	26.3	1.32
	W1945	Perennial	China	54.6	20.9	47.6	20.5	1.15
	W1952	Perennial	China	47.6	13.6*	42.3	—	1.13
	W1681	Perennial	India	82.1	—	74.9	—	1.10
	W593	Perennial	Malaysia	81.4	15.0*	39.1	2.7**	2.08	S6
	W172	Perennial	Thailand	67.2	11.2	25.5	3.6**	2.64	S6
	W1294	Perennial	Philippines	67.9	20.6	28.0	2.6**	2.43	S6
	W1807c	Perennial	Sri Lanka	77.1	26.7	26.6	0.1**	2.90	S6

* and ** indicate significant deviation from Mendelian inheritance at 5% and 1% levels, respectively.

^aThe rate of seed setting was determined by counting fertile and sterile spikelets of two panicles for each plant.

^bWhen more tightly linked marker, R111C, was used, the frequency of the W593-derived allele was close to 100% in the F₂ populations derived from A58 × A58S₆ and T65wx × T65S₆ (W593) (see text).

^cThe allelic state of W1807 was analyzed by crossings the line with T65WxS₆ (W593) and T65wx. The frequency of the alleles in the progeny was determined by the marker P139 linked with the S₆ locus.

Figure 2.—

Localization of the S₆ gene on the rice linkage map. (A) Location of the S₆ gene on choromosome 6. Genetic distance was estimated on the basis of 1886 segregating plants of A58 × A58S₆ (BC4 and BC5) except for the genetic distance between Hd1 and RM3183, which was estimated using 216 BC4 F₂ plants from the same population. (B) Graphical genotypes of chromosome 6 of recombinant plants and the rates of seed setting of the respective plants through self-fertilization. The physical map of chromosome 6 is based on the Rice Genome Research Program (http://rgp.dna.affrc.go.jp). The map position of each marker along chromosome 6 (AP008212) is shown in parentheses (in megabases). A solid circle represents the centromere. Chromosomal regions derived from W593 (carrying the S₆ allele) and A58 (carrying the allele) are indicated by thick and thin lines, respectively. Open circles represent the A58-derived alleles. The estimated range of the position of the S₆ gene is shown by a bar below the graphical genotypes. The rate of seed setting was determined by counting the number of fertile and sterile spikelets of two panicles for each plant. The data represent the average values obtained from at least two individual plants for each genotype. The rate of seed setting is shown as a percentage. Solid and open bars indicate seed fertility above and below 50%, respectively. (C) Regional recombination frequency on chromosome 6. The frequency of recombination (in centimorgans/megabases) was estimated using the mapping populations. The average recombination value of chromosome 6 (3.87 cM/Mb; Wu et al. 2003) is shown by a dashed line.

Diversity survey:

Nine strains of O. sativa and 14 strains of O. rufipogon were used for the analysis of the DNA polymorphisms (Table 2). The nucleotide sequences of five regions (P139, F06, D11, R111C, and C133A) around the S₆ gene were determined by direct sequencing of the polymerase chain reaction (PCR) products. The following primers were used for the PCR: 5′-ccg aaa aga gtc ctc cga ag-3′ and 5′-cca cct aag aag cca gca cc-3′ for F06; and 5′-gaa atg cca ctg gcc tac at-3′ and 5′-ttc agg cga gca att tag gt-3′ for D11. These primers were designed on the basis of sequences in the public database (accession nos. AP003545 and AP003512). One loosely linked locus, Hd3a, which is located on the short arm of chromosome 6, and two unlinked loci, sd1 (semi-dwarf1), and qSH4, which are located on chromosome 1 and chromosome 4, respectively, were used as references. The nucleotide sequences of sd1 and qSH4 were determined according to the methods of Nagano et al. (2005) and Onishi et al. (2007), respectively. For qSH4, the combined sequences of the two regions (43A12-161k and 04A17-9k) were used. For Hd3a, the primers used were 5′-agc tag ata gct gcc tct atc aca gta t-3′ and 5′-cta gct tca tga gag acc tta gcc-3′, 5′-ccc tgc acc aca cac agt tc-3′ and 5′-tgt ctg aac ctg caa tgt at-3′ or 5′-agc tag ata gct gcc tct atc aca gta t-3′ and 5′-tat ata tgt tgt gtg tcg aga atc att tc-3′. These primers were designed on the basis of sequences in the public database (accession no. AP007223). Each accession was sequenced in both directions using a Big Dye Terminator Cycle sequencing kit (Applied Biosystems, Foster City, CA) on an ABI310 automatic sequencer (Applied Biosystems). The DNA sequences determined in this study are available from DDBJ/GenBank/EMBL (accession nos. AB433361–AB433506, AB433508–AB433511, and AB433513–AB433529). The sequence alignment was done using the CLUSTAL W computer program (Thompson et al. 1994) with minor modifications by visual inspection. Molecular population genetic analysis was conducted using DnaSP version 3.14 (Rozas and Rozas 1999). We compared the level of nucleotide diversity per silent site based on π (Nei 1987) and θ_w (Watterson 1975) and calculated Tajima's D statistic for testing neutrality (Tajima 1989). Among the five regions around the S₆ gene, D11 and C133A had portions of protein coding regions predicted from the public database (Rice Genome Research Project, http://rgp.dna.affrc.go.jp). D11 contained a part of the first exon and all of the second and third exons of a putative 2′-hydroxyisoflavone reductase. C133A contained the fifth exon of the putative mlo2 protein. To rule out selection acting on the functional regions, the silent sites were used to compare nucleotide diversity.

RESULTS

Effects of allelic interaction at the S₆ locus on gametogenesis:

A previous study indicated that the S₆ gene, derived from the wild rice (O. rufipogon) strain W593, induced the preferential abortion of both male and female gametes possessing its allelic alternative () from cultivated rice (O. sativa) strain T65 only in the heterozygote, and as a result, no allele was transmitted to the progeny (Sano 1992). To examine the female gametogenesis in the heterozygote () which was causing the siTRD, histological investigations were carried out (Figure 1). If dysfunction of the female gametes carrying the allele is induced in the presence of the S₆ allele in the heterozygote, half of the embryo sacs should degenerate. Of 42 ovules, 23 in fact did exhibit an abnormality in the embryo sac structure, while the remaining 19 ovules had a mature seven-celled structure like that found in the parental strain (T65wx) (Figure 1). The frequency of arrested embryo sacs was 0.55 (23/42), indicating that the embryo sacs carrying the allele were aborted, which is consistent with the notion that transmission of the allele is extremely reduced in the heterozygote (). On the other hand, no abnormalities were detected in mature pollen grains in the heterozygote (), as also in T65wx. To confirm the TRD through male gametes, the segregation of F₂ plants derived from T65wx × T65S₆ (W593) was analyzed by using a molecular marker, R111C, tightly linked with the S₆ locus (see below). Almost all F₂ plants (84/98) were homozygotes for the W593-derived allele and no homozygotes for the T65wx-derived allele were detected, indicating that transmission of the allele is reduced not only through female but also male gametes, as has been inferred from backcrosses (Sano 1992).

Figure 1.—

Embryo sac abnormality in the heterozygotes of . (A) Normal (i) and abnormal (ii) embryo sacs in T65wx/T65S₆ (W593). Arrow heads indicate nuclei. EN, egg nucleus; SY, synergid cell; PN, polar nuclei; AN, antipodal cell nuclei. Bar, 20 μm. (B) Frequencies of normal and abnormal embryo sacs in T65wx/T65S₆ (W593). ns indicates nonsignificant deviation from 1:1 ratio.

Fine mapping:

To roughly map the S₆ gene, 216 segregating plants from A58 × A58S₆ were genotyped using four molecular markers (Hd1, RM3183, C133A, and RM3498) (Figure 2A). A58 and A58S₆ carry the and S₆ alleles, respectively (Table 1). The rate of seed setting of each F₂ plant was scored to examine the phenotypic segregation of the S₆ locus, since the dysfunction of female gametes in the heterozygote () is reflected in the productivity of the F₂ plants. The F₂ plants were clearly segregated into two distinct groups: one exhibiting <50% (32.8 ± 4.4%) seed-setting rate and the other exhibiting >70% (80.2 ± 0.4%), and no F₂ plant exhibited a seed-setting rate between these values. Plants of the former and the latter groups were classified as heterozygotes and homozogotes, respectively. The allelic states of the above markers indicated that all plants classified as homozygotes were those of the S₆ allele. As a consequence of this analysis, the estimated recombination values of RM3183-S₆, S₆-C133A, and C133A-RM3498 were calculated to be 0.0040 ± 0.0010, 0.0070 ± 0.0014, and 0.0201 ± 0.0023, respectively (Figure 2A). The frequencies of the W593-derived alleles were significantly higher than those expected from Mendelian inheritance, especially at RM3183 (208/216, 96.3%) and C133A (206/216, 95.4%), which confirmed the notion that the S₆ allele is preferentially transmitted through both male and female gametes.

For fine mapping, 1670 additional segregating plants were surveyed with four additional markers (S14439, P139, R111C, and G05). However, no recombination was detected between S14439 and G05, indicating that the frequency of recombination is extremely low near the centromere. The ratio of genetic distance to physical distance (cM/Mb) greatly decreased with decreasing distance from the centromere, confirming the restriction of recombination near the centromere (Figure 2C). Due to a deficit of recombinants in this region, the S₆ locus was delimited to a region of >8000 kb between RM3183 and C133A, even with a large mapping population (Figure 2B).

Distribution of the S₆ and S₆^a alleles in Asian rice accessions:

To examine the allelic distribution at the S₆ locus, two NILs carrying the S₆ or allele, namely, the T65st₁S₆ or lines, respectively, were established (Table 1). The allelic distribution at the S₆ locus in 23 strains of Asian cultivated rice and wild rice were investigated by testcrosses with these NILs (Table 2). Because st1 is linked with the S₆ locus, and its phenotype in homozygotes is manifested at the young seedling stage, it was used as a visible marker of the segregation distortions caused by the S₆ locus.

The F₁ hybrids between three strains of O. sativa ssp. japonica and T65st₁S₆ all exhibited a low seed-setting rate, but the F₁ hybrids between these strains and did not, but rather, exhibited a normal rate (Table 2). In addition, a marked excess of st₁ homozygotes in the F₂ populations was observed when the three strains were crossed with T65st₁S₆. These results indicate that the three strains carry the allele at the S₆ locus. In contrast, the F₁ hybrids between the four wild strains (W593, W172, W1294, and W1807) and either or T65wx exhibited a low seed-setting rate but the F₁ hybrids between these strains and T65st₁S₆ or T65S₆ (W593) did not, indicating that the four wild strains carry the S₆ allele at the S₆ locus. A marked distortion of the linked st1 marker in the F₂ populations supported the conclusion that the four wild strains harbor the S₆ allele.

The other six O. sativa ssp. indica and nine O. rufipogon strains (W107, W2002, W630, W1718, W1943, W1944, W1945, W1952, and W1681) produced a pattern different from that described above (Table 2). There was no substantial difference in the seed-setting rate in the F₁ hybrids derived from the two testcrosses, although the seed-setting rate varied depending on the crossing parents. A marked excess of st1 homozygotes was detected in neither of the F₂ populations derived from the two testcrosses. Therefore, it was suspected that these strains carry an additional allele (; “n” refers to “neutral”; Rick 1966, 1971), which induces no preferential abortion in the heterozygotes ( and ). Subsequent analysis indicated that the estimated chromosomal location of the gene overlapped with that of the S₆ gene, suggesting that the gene is allelic to the S₆ gene (see Figure 3).

Figure 3.—

Localization of the gene on the rice linkage map. (A) Location of the gene on chromosome 6. The physical map of chromosome 6 is based on the Rice Genome Research Program (http://rgp.dna.affrc.go.jp). The map position of each marker along chromosome 6 (AP008212) is shown in parentheses (in megabases). Graphical genotypes of NILs used for mapping the gene are shown below the linkage map. Chromosomal regions derived from Patpaku (carrying the allele) and T65wx (carrying the allele) are indicated by thick shaded lines and thin solid lines, respectively. A solid circle represents the centromere. Open circles represent T65wx-derived alleles. The estimated ranges of the positions of the and S₆ genes are shown by solid bars below the graphical genotypes. (B) The rate of seed setting in F₁ plants derived from crosses between NILs for the gene and S₆ or carriers. For the analysis of seed setting, see legend to Figure 2. The rate of seed setting is shown as a percentage (average of two individual plants). Solid and open bars indicate rates of seed setting above and below 50%, respectively.

The F₁ hybrids between the remaining strain W2048 and the two tester lines exhibited a low seed-setting rate. When the seed-setting rate of a hybrid is low, TRD would be expected to be a more reliable indicator for determining the allelic state at the S₆ locus. Unlike the other wild rice strains, W2048 displayed the segregation distortion: a marked excess of st1 homozygotes was detected in the F₂ population derived from a cross between W2048 and T65st₁S₆, but not that from a cross between W2048 and . These results indicate that W2048 carries the allele. Taken together, these results reveal that plants carrying the allele are present at a high frequency and are widely distributed in Asia. The S₆ allele was detected in four strains from South to Southeast Asia, and the allele was detected in three cultivated (O. sativa ssp. japonica) and one wild (O. rufipogon) rice strain from East Asia.

The additional allele (S₆ⁿ) at the S₆ locus:

As mentioned above, of the 23 Asian wild rice and cultivated rice strains examined, 15 strains exhibited no TRD, as evaluated by the st₁ phenotype in the crosses with either T65st₁ S₆ or (Table 2), suggesting the presence of an additional allele () at the S₆ locus, which induces no preferential abortion in the heterozygotes ( and ), as proposed in the tomato (Rick 1966, 1971). To test this possibility, a NIL, T65WxSe1 (Pat), was characterized. Because the testcross indicated that Patpaku carried the gene (Table 2), the chromosomal fragment transgressed from Patpaku should contain the gene, if is allelic to the S₆ gene (Figure 3). Hybrid plants between T65WxSe1 (Pat) and T65wx ( carrier), and those between T65WxSe1 (Pat) and T65S₆ (W593), both exhibited a high seed-setting rate (88.0 and 85.4%, respectively). This shows that neither the S₆ nor the alleles induces abortion of gametes carrying the Patpaku-derived allele in the heterozygotes. The fragment introduced from Patpaku was further segmented by selfing the heterozygote plants. Three F₃ lines with distinct homozygous introgressions (P-1, P-2, and P-3; Figure 3) in the S₆ region were selected on the basis of the six markers (Wx, OsC1, Hd1, RG264, RM3498, and G2028) and were used to map the gene. When the three lines were crossed with both the and S₆ carriers, only the hybrid plants between P-3 and T65S₆ (W593) exhibited a low seed-setting rate (44.0%) (Figure 3). This result indicates that if a plant harbors the chromosomal region between Hd1 and G2028 from Patpaku, the plant induces no preferential abortion in response to the S₆ and alleles. Since the region between Hd1 and G2028 contained the region in which the S₆ gene was located, we concluded that Patpaku possesses the allele at the S₆ locus.

Survey of nucleotide diversity in chromosomal regions around the S₆ locus:

Owing to the presence of the additional allele, association between genetic diversity and allelic state at the S₆ locus was analyzed on the basis of the relative nucleotide diversity between plants carrying the incompatible alleles (S₆ or ) and those carrying the additional allele (). We analyzed the nucleotide sequences at the five loci (P139, F06, D11, R111C, and C133A) encompassing the centromeric region (Figure 4), and compared Watterson's θ (θ_w) and π between these loci (Table 3). Silent sites were used for comparing nucleotide diversity to rule out selection acting on the functional regions. Tajima's D statistic revealed no significant deviations from neutrality at these loci (Table 3). The carriers showed values ranging from 7.78 × 10⁻³ (D11) to 10.50 × 10⁻³ (C133A) for θ_w, and from 6.75 × 10⁻³ (P139) to 8.66 × 10⁻³ (F06) for π, which is comparable to the values for the nucleotide diversity previously reported in nuclear genes of rice (Yoshida et al. 2004; Yoshida and Miyashita 2005; Rakshit et al. 2007; Zhu et al. 2007). At these five loci, the estimates of θ_w and π were lower in the S₆ and carriers than those in the carriers (Table 3 and Figure 4B). In particular, the estimates were zero at the R111C locus in the S₆ and carriers. Since the F06 and D11 loci were physically closer to the centromere than the R111C locus, but a higher level of diversity was detected at these loci compared to the diversity at the R111C locus, the low nucleotide diversities of the S₆ and carriers in the R111C locus is evidently not due solely to the effect of low recombination near the centromere. The relative diversities ( and ) increased depending on the physical distance from the R111C locus (Figure 4C). The levels of nucleotide diversity at the P139 and C133A loci were similar to those at the unlinked loci sd1 and qSH4 and the loosely linked locus Hd3a (Table 3), suggesting that the reduction in nucleotide diversity is specific to the chromosomal region around the R111C locus, and this is not accounted for by either demographic effects or selection acting on the sites. These results indicate an association between nucleotide diversity in the chromosomal region and the allelic state at the S₆ locus, and are consistent with the notion that the S₆ and alleles arose independently from the allele (discussed below).

Figure 4.—

Survey of the nucleotide diversity around the centromere on chromosome 6. (A) A physical map of the candidate region for the S₆ gene. The position of the centoromere is indicated by a solid circle. (B) Levels of nucleotide diversity (θ_w) for the S₆, , and carriers at five sites in the chromosomal region. (C) The relative diversities of the S₆ and carriers to the carriers at five sites in the chromosomal region.

TABLE 3.

Levels of nucleotide diversity

		S6 (n = 4)				(n = 4)				(n = 15)
No. of polymorphic positions	Nucleotide diversityb		Tajima's D	No. of polymorphic positions	Nucleotide diversityb		Tajima's D	No. of polymorphic positions	Nucleotide diversityb		Tajima's D
	Loci	Lengtha				(θw × 103)			(π × 103)				(θw × 103)	(π × 103)			(θw × 103)	(π × 103)
Chromosome 6
P139	748	3	2.18 ± 1.58	2.23 ± 0.59	0.16766	9	6.55 ± 3.97	6.01 ± 3.19	−0.82943	19	7.81 ± 3.27	6.75 ± 0.95	−0.55548
F06	530	1	0.98 ± 0.98	0.90 ± 0.48	−0.61237	6	5.95 ± 3.80	5.45 ± 2.89	−0.80861	15	8.37 ± 3.42	8.66 ± 0.99	0.13696
D11	688	1	1.72 ± 1.72	1.58 ± 0.84	−0.75455	4	6.89 ± 4.70	6.31 ± 3.35	−0.79684	8	7.78 ± 3.79	7.33 ± 1.52	−0.25041
R111C	590	0	0.00 ± 0.00	0.00 ± 0.00	—	0	0.00 ± 0.00	0.00 ± 0.00	—	17	8.86 ± 3.77	6.68 ± 0.64	−0.99328
C133A	892	7	4.59 ± 2.87	4.21 ± 1.88	−0.81734	6	2.65 ± 1.81	2.43 ± 1.29	−0.80861	32	10.50 ± 4.21	8.14 ± 1.87	−1.03951
Hd3a	2448	18	4.59 ± 2.63	4.45 ± 1.94	−0.31310	15	3.82 ± 2.22	3.50 ± 1.72	−0.84729	43	6.30 ± 2.44	5.88 ± 0.51	−0.28539
Chromosome 1
sd1	1883	9	3.92 ± 2.37	3.59 ± 1.90	−0.82943	7	3.04 ± 1.90	2.79 ± 1.48	−0.82407	36	11.05 ± 4.45	8.55 ± 1.48	−1.16932
Chromosome 4
qSH4	1734	18	5.75 ± 3.29	5.46 ± 1.76	−0.50485	18	5.72 ± 3.28	5.24 ± 2.78	−0.85194	45	8.10 ± 3.13	8.23 ± 0.99	−0.11812

Indels were ignored for computation.

^aNo. of nucleotide positions analyzed.

^bSilent substitutions including synonymous substitutions and changes in the noncoding positions were used for the analysis.

DISCUSSION

The siTRD caused by the S₆ locus and its independence from the effects of other loci associated with sexual affinity on the same chromosome:

TRD violates Mendel's rules by the preferential transmission of a particular chromosome or allele at the expense of its partner. Preferential gamete dysfunction has been frequently reported as a mechanism of TRD in a range of organisms (Lyttle 1991; Temin et al. 1991; Silver 1993; Úbeda and Haig 2005; Moyle and Graham 2006). Cytological observations revealed that about half of the female gametes degenerated in the heterozygote (Figure 1). However, no defect was detected in the pollen grains by this method, although genetic analysis indicated that the transmission of both male and female gametes carrying the allele to the next generation was very low (Sano 1992; this study). This probably indicates that functional difference(s) between the male gametes with different alleles, which accounts for the TRD, occurs after the production of pollen grains. No reduction in the seed-setting rate was observed when the heterozygote was used as a male parent for crosses, which indicates that no aberration occurred after fertilization, although almost no alleles were transmitted to the progeny (Sano 1992). Taken together, these observations suggest that the primary event(s) resulting in the TRD phenomenon through the male gametes in the heterozygote occurs during a prezygotic process after the production of pollen grains. A plausible explanation for these observations is that TRD through the male gametes is caused by pollen tube competition (certation). Alternatively, it is possible that the dysfunction of the male gametes occurred at an early stage of gametogenesis and the observed pollen grains consist only of carriers of the S₆ allele. In any event, the genetic effects of the S₆ locus on the siTRD were altered in neither the female nor male gametes, even after repeated backcrosses, which indicates the stability of the phenomenon.

Genetic mapping revealed that the S₆ locus is located near the centromere of chromosome 6. On the same chromosome, genes that affect the fertility of F₁ hybrids have been identified. Among these are the cim (the cross-incompatibility reaction in the male) gene and the two Cif (the cross-incompatibility reaction in the female) genes responsible for the abortion of hybrid seeds, especially in endosperm, observed in a cross between W593 and T65wx (Matsubara et al. 2003; Koide et al. 2008c), which are the same parental strains of the materials used in the present study. The cim and one of the Cif genes, Cif₂, have been mapped in a chromosomal region that overlaps with the S₆ locus (Matsubara et al. 2003; Koide et al. 2008c). Another Cif gene, Cif₁ is located on the short arm of chromosome 6. The plants carrying cim derived from T65wx exhibit cross-incompatibility when pollinated to the female plants carrying both Cif₁ and Cif₂ derived from W593. The genotypes of T65wx and W593 are cif₁ cif₂ cim/cif₁ cif₂ cim and Cif₁ Cif₂ Cim/Cif₁ Cif₂ Cim, respectively. These three genes (cim, Cif₁, and Cif₂) act sporophytically, so that the heterozygotes (Cif₁ Cif₂ Cim/cif₁ cif₂ cim) produce only gametes that are cross-compatible. As a consequence, the cross-incompatibility system controlled by the cim and Cif genes acts on the formation of hybrid seeds after fertilization (Matsubara et al. 2003; Koide et al. 2008c), whereas that by the S₆ locus acts on gametes produced in the F₁ plants. Therefore, the developmental phase of the S₆ locus-mediated abortion and consequent segregation pattern are both unequivocally distinguished from those mediated by the Cif/cim genes, although the possibility cannot be ruled out that abortions at different developmental phases are caused by pleiotropic effects of a single gene, which would resemble a phenomenon caused by the Tcb1 locus in maize (Kermicle 2006).

The hybrid sterility loci, S₁ (Sano 1990), S₅ (Yanagihara et al. 1995), S₈ (Singh et al. 2006), S₁₀ (Sano et al. 1994), and S₂₆ (Kubo and Yoshimura 2001), have also been mapped to chromosome 6, a mapping procedure carried out using different pairs of crossing parents. These genes were mapped in chromosomal regions apart from the S₆ locus, and segregation analyses of marker genes tightly linked with these loci also indicated that the segregation distortion caused by these loci was independent of that by the S₆ locus (our unpublished data). These results indicate that the siTRD phenomenon caused by the S₆ locus is not affected by the allelic state at the other hybrid sterility loci on the same chromosome. The independence of the S₆ locus-mediated siTRD ensures that the pattern of nucleotide diversity around the S₆ locus (Table 3) depends exclusively on the allelic state at the S₆ locus.

The S₆ⁿ allele is widely distributed in Asian rice and functions in a locus-specific manner:

We examined the allelic distribution at the S₆ locus in 23 strains of Asian rice (Table 2). In addition to the S₆ locus, we have also analyzed the allelic distribution at the S₁ locus that causes siTRD in the hybrids between the Asian and African rice species (Koide et al. 2008b). Although many loci causing TRD have been identified in the rice genome (see Koide et al. 2008a), no other attempts that analyze allelic distribution of a single locus in various rice accessions have been reported. A notable difference between the siTRD caused by the S₆ and S₁ loci is the presence or absence of an additional allele the presence of which, in heterozygotic states, does not result in siTRD; both the testcross experiments and subsequent genetic mapping indicated the presence of the allele at the S₆ locus (Table2; Figure 3), whereas no such additional allele has been detected regarding the S₁ locus (Koide et al. 2008b). The testcross experiments also indicated that the allele was distributed widely, and was predominant in Asian rice (Table 2). A phylogenetic analysis using the nucleotide sequence data from the five loci (P139, F06, D11, R111C, and C133A) around the S₆ locus showed that both the S₆ and carriers clustered, and these clusters were separated from each other on the phylogenetic tree (data not shown). These results suggest that the allele could be the ancestral state of the locus and the S₆ and alleles evolved from the ancestral allele independently.

It has been frequently reported in varietal crosses that hybrid sterility in rice is caused by a single locus (Kitamura 1962; Oka 1964; Ikehashi and Araki 1986; Yanagihara et al. 1995; Liu et al. 1997). In these loci, a wide array of compatible alleles conferring compatibility to a wide range of strains has been proposed, and some have actually been mapped (Ji et al. 2005; Qiu et al. 2005; Wang et al. 2006). However, regarding the wide compatibility proposed in rice, whether these compatible alleles act specifically on the alleles at their own locus or on sterility genes at different loci, as reported in wheat (Tsujimoto and Tsunewaki 1985; Friebe et al. 2003) and Drosophila (Lyttle 1991; Temin et al. 1991), is not known. In the testcross experiments, the hybrids between the carriers and the tester lines did not necessarily exhibit a higher seed-setting rate than the hybrids between S₆ or carriers and the tester lines (Table 2). These results suggest that, unlike alleles conferring wide compatibility, the allele does not reduce the level of hybrid sterility that might be caused by other sterility loci, but rather interacts specifically with the S₆ and alleles.

Chromosomal region-specific reduction in genetic diversity associated with allelic state:

The presence of the allele prompted us to survey the chromosomal region around the S₆ locus in terms of nucleotide diversity and examine the association between the nucleotide diversity and allelic state at the S₆ locus. Association-based mapping could identify regions that are associated with a particular phenotype (Thornsberry et al. 2001; Palaisa et al. 2003). If a new sequence variant underlying a given phenotype has emerged, then there would have been less time for recombination, and thus a significant association would be expected between phenotype and genotype in the vicinity of the causative mutation. As a result, a reduction in nucleotide diversity should occur around the causative mutation within a population of individuals exhibiting the same phenotype. If the S₆ and alleles have arisen independently from the allele, a reduction in nucleotide diversities of the S₆ and carriers relative to the diversity of the carriers should be observed near the locus, as the relative nucleotide diversities would tend to increase with physical distance from it due to the shuffling into different genetic backgrounds through recombination. In fact, the nucleotide diversities of the S₆ and carriers were zero at the R111C locus and the relative diversities (θ_w-S₆/θ_w- and ) increased depending on the physical distance from the R111C locus (Figure 4B). Thus, a clear association between nucleotide diversity and allelic state at the S₆ locus was detected by comparing nucleotide diversity of the S₆ or carriers with that of the carriers, which most probably reflects independent lineages of the S₆ and carriers. These results also suggest that the causative alleles are most likely located in a region close to the R111C locus among the loci examined.

The resolution of association analysis depends on the structure of the linkage disequilibrium (LD) across the entire genome. Garris et al. (2003) reported that significant LD was detected at sites which were at most 100 kb apart around the xa5 locus in cultivated rice strains. However, it has not established how far the LD extends in the region near the centromere where recombination rates are low. The results of the present study consequently suggest that association mapping is potentially useful even for chromosomal regions where recombination is restricted. Further analyses using more accessions and loci would reveal whether the local nucleotide diversity is directly associated with the S₆ locus-mediated phenotype.

Allelic differentiation at the S₆ locus and evolution of the siTRD system that serves as an isolating barrier:

The Dobzhansky–Muller model proposes that hybrid incompatibilities are generally caused by an interaction between genes that have functionally diverged in each population (Coyne and Orr 2004). This model assumes two or more interacting loci. TRD systems often involve alleles at a minimum of two closely linked loci, a distorter and its cis-acting target (Lyttle 1991). Examples of the TRD system of this type include the mouse t-haplotype (Silver 1993), segregation distorter (SD) of Drosophila (Temin et al. 1991), and Gametocidal 2 (Gc2) in wheat (Friebe et al. 2003).

Alternatively, the development of hybrid incompatibility might be explained by a simple one-locus stepwise mutation model (Nei et al. 1983). This model assumes changes in allelic state in two different populations: A₀ mutates to A₁ in one population and to A₋₁ in another population. If the heterozygotes for alleles that are two or more steps apart from each other are infertile, hybrid incompatibility appears in A₁A₋₁ despite the fact that A₁A₀ and A₋₁A₀ are completely fertile. This model was adopted for explaining allelic differentiation at the Gamete eliminater (Ge) locus in tomato (Rick 1971). Rick (1966, 1971) reported that there are three alleles (Ge^p, Ge^c, and Geⁿ) at the locus. Ge^p induces abortion of the male and female gametes carrying Ge^c in the heterozygote (Ge^p/Ge^c). On the basis of their geographical distribution, it was suggested that Ge^p and Ge^c have arisen from the neutral allele Geⁿ. Likewise, because of the presence of the allele at the S₆ locus in rice, the evolutionary process at the S₆ locus can be explained by the one-locus stepwise mutation model. The results obtained so far are also consistent with the notion that the allele is the ancestral state of the locus, as described above.

Recent studies have suggested that TRD, which is normally masked within a species, plays a role in establishing the sterility barriers between species (Dermitzakis et al. 2000; Tao et al. 2001; Orr and Irving 2004). There are at least two possible models for the formation of isolating barriers induced by normally masked TRD. Frank (1991) along with Hurst and Pomiankowski (1991) proposed that genes causing TRD are rapidly fixed in a population due to their “selfish nature,” but may often become suppressed within each population to alleviate their deleterious effects on fertility. These genes can become reexpressed in hybrids between individuals from allopatric populations and cause the sterility of hybrids.

Another possible explanation is that a transmission ratio distorter never did appear in the history of the two allopatric populations and TRD instead appeared as a consequence of an inappropriate interaction between genes that had diverged in these populations. These genes have no evident selfish behavior in so far as they are present in each population. In this model, genes of neither lineage pass through the adaptive valley represented by the driving genotype that results in TRD (Orr and Presgraves 2000). Assuming that the S₆ and alleles have arisen independently from the allele, the incompatible alleles (S₆ and ) could have become fixed in different populations without carrying any selective disadvantage, which would favor the latter explanation. The predominance of the allele is consistent with the fact that the Asian wild rice population consists primarily of plants with interfertile relationships (Morishima et al. 1963; Chu et al. 1969). The siTRD system mediated by the allelic interaction between the S₆ and alleles established genetically discontinuous relationships between limited constituents (e.g., between O. sativa ssp. japonica and O. rufipogon in Southeast Asia) of the existing rice population.