Hybrid male sterility in rice controlled by interaction between divergent alleles of two adjacent genes

Abstract

Sterility is common in hybrids between divergent populations, such as the indica and japonica subspecies of Asian cultivated rice (Oryza sativa). Although multiple loci for plant hybrid sterility have been identified, it remains unknown how alleles of the loci interact at the molecular level. Here we show that a locus for indica-japonica hybrid male sterility, Sa, comprises two adjacent genes, SaM and SaF, encoding a small ubiquitin-like modifier E3 ligase-like protein and an F-box protein, respectively. Most indica cultivars contain a haplotype SaM⁺SaF⁺, whereas all japonica cultivars have SaM⁻SaF⁻ that diverged by nucleotide variations in wild rice. Male semi-sterility in this heterozygous complex locus is caused by abortion of pollen carrying SaM⁻. This allele-specific gamete elimination results from a selective interaction of SaF⁺ with SaM⁻, a truncated protein, but not with SaM⁺ because of the presence of an inhibitory domain, although SaM⁺ is required for this male sterility. Lack of any one of the three alleles in recombinant plants does not produce male sterility. We propose a two-gene/three-component interaction model for this hybrid male sterility system. The findings have implications for overcoming male sterility in inter-subspecific hybrid rice breeding.

Hybrid sterility is the most common form of postzygotic isolation mechanisms between species or subspecies, provides an initial force driving genetic differentiation and speciation, and plays an important role in maintaining species identity (1, 2). Hybrid sterility in plants has been known for a long time (3–11). Cultivated rice comprises two species, Oryza sativa L. (Asian rice) and O. glaberrima Steud (African rice); O. sativa is classified further into two major types or subspecies, indica and japonica, which have been referred to as “Hsien” and “Keng” since the Han dynasty in China (12), based largely on the low affinity of fertility in their hybrids and other morphological features (3, 13, 14). Hybrids between the rice species or subspecies have significant hybrid vigor, or heterosis, that provides great potential for further productivity increases in rice. However, the partial or complete sterility of the hybrids forms a reproductive barrier hindering the utilization of heterosis in inter-(sub)specific hybrid rice breeding.

A number of loci conferring hybrid male or female sterility (or, in a few cases, both) have been identified in rice (8–10, 15–18); however, only a gene encoding an aspartic protease at the locus S5 (9) conditioning embryo-sac sterility in indica-japonica hybrids has been cloned recently (19). Plant hybrid sterility is thought to be caused by interactions between alleles (called “pollen killer,” “egg killer,” or “gamete eliminator”) at certain heterozygous loci derived from divergent populations (5–10). Several genetic models have been proposed for plant hybrid sterility, including the one-locus allelic interaction (one-locus sporogametophytic interaction) model (7–10), and the duplicate gametophytic lethal model (7). The one-locus allelic interaction model, which can explain the genetic behavior of most hybrid sterility loci, proposes that an allelic interaction of a hybrid sterility gene in a heterozygote causes selective abortion of gametes carrying a given allele, in most cases a japonica allele in the indica-japonica or O. glaberrima-japonica hybrids, thus resulting in poor transmission of this allele into the progeny. However, the molecular genetic mechanism for controlling the allele-specific gamete killing remains unknown.

In this study we cloned an indica-japonica hybrid male sterility locus, Sa, finding that it comprises two adjacently located genes, SaM and SaF, encoding a small ubiquitin-like modifier (SUMO) E3 ligase-like protein and an F-box protein, respectively. We show that allele-specific hybrid male sterility is controlled by direct and indirect interactions among three divergent alleles of the genes, suggesting that intercellular protein transport may occur during the early microspore development.

Results

Genetic Effect of Sa on Hybrid Male Sterility.

To study rice hybrid sterility, several near-isogenic lines (NILs) were developed that contained chromosomal segments from indica cultivars in the genetic background of a japonica variety, Taichung 65 (T65) (7). Using T65 and NIL E4, a major locus conferring hybrid male sterility, Sa, was identified (15, 16). T65 with a locus-genotype Sa^j/Sa^j and E4 with Saⁱ/Saⁱ were fully (>90%) male fertile (pollen fertile, MFF), but their F₁ hybrid (Saⁱ/Sa^j) exhibited male semi-sterility (MSS) in which ≈ 50% of pollen was sterile (Fig. 1 A–C). The male developmental defect caused by heterozygous Sa appeared in microspores of the early uni-nucleate stage (20), but female fertility was unaffected (15, 16). Of 666 analyzed F₂ plants, only three (0.45%) were of Sa^j/Sa^j; the others were either Saⁱ/Sa^j or Saⁱ/Saⁱ, with distribution being equal (Fig. 1 D–F). This result showed that the heterozygous Sa locus has a strong effect, impairing the development of almost all male gametophytes carrying Sa^j. In typical indica-japonica hybrids the accumulative effect of multiple hybrid sterility loci usually leads to serious decreases in both male fertility and seed-setting rate (16).

Fig. 1.

Hybrid male sterility controlled by the Sa locus. (A–F) Pollen phenotypes and genotypes of Sa in T65 and E4, and their F₁ and F₂ plants. The genotypic frequencies (%) were determined using an Sa-cosegregation marker G02–76.3 (see Fig. 2A). (G–L) Pollen phenotype and genotype of the genes of a recombinant and transgenic plants. M⁺, M⁻, F⁺, and F⁻ denote SaM⁺, SaM⁻, SaF⁺, and SaF⁻, respectively. tM⁻/−, tM⁺/−, and tF⁺/− indicate hemizygotes of the transgenes. (Space bars, 50 μm.)

Identification of Two Adjacent Genes at the Sa Locus.

Sa was mapped primarily on chromosome 1 (21). We further located this locus to a 370-kb region with 200 F₂ plants of the T65 x E4 cross (Fig. 2A). We then used the flanking markers A10-2 and G02-148 to screen ∼ 10,500 F₂ plants and obtained 322 recombinants. The high-resolution mapping was accomplished with 11 polymorphic markers developed in this region (supporting information (SI) Table S1). Three crosses between certain recombinants and T65 (Fig. 2A) were conducted to obtain new recombinants for the genetic dissection.

Fig. 2.

Identification of two genes in the Sa locus. (A) Top: Primary mapping of Sa. The numbers of the markers following G02- indicate their positions (kb) in a P1 artificial chromosome clone P0013G02. Middle: Fine mapping with key recombinants obtained from 10,500 F₂ plants and crosses of H70 x T65 (H70T65), A408 x T65 (A408T65), and C513 x H70 (C513H70). The SaF and SaM alleles of E4 and T65 were determined by the functional SNPs G02-69.8 and G02-74.6, respectively, and G02-76.3 was used for segregation analysis of the recombinant's progenies. MFF, male full fertility; MSS, male semi-sterility; asterisk indicates the chromosome eliminated in male sterility. Bottom: Structures of SaF and SaM alleles. The fifth intron of the SaM-T65 allele is shown by a dotted line. ATG, start codon; TAG/TGA, stop codons. (B) Sequence and structural variations of the proteins. (C) Alignment of conserved domains among SaM, yeast Mms21, and BAB02569 of Arabidopsis thaliana. The underlined region in SaM⁺ contains the self-inhibitory domain.

Based on the two recombinants G392 and H122, Sa was delimited to an ∼ 10-kb region containing two predicted genes, which we named SaF and SaM (Fig. 2A). However, analysis of other recombinants showed that the induction of male sterility by this locus could not be explained by the single-gene hypothesis. Hereafter, the alleles of SaF-E4, SaF-T65, SaM-E4, and SaM-T65 are designated SaF⁺, SaF⁻, SaM⁺, and SaM⁻, respectively. Four recombinants (A408, H70, C513, and LS38) with heterozygous SaF and homozygous SaM⁺ or SaM⁻ were fully male fertile; three other recombinants (G735, H70T65, and A408T65) with homozygous SaF⁻ and heterozygous SaM also showed full fertility, not semi-sterility. On the other hand, G392 and H122 containing heterozygous SaF and SaM and an individual H71 with homozygous SaF⁺ and heterozygous SaM were consistently semi-sterile. A genotype (C513H70) with heterozygous SaF and SaM in a recombined arrangement between them (repulsion phase) exhibited male semi-sterility, with the SaF⁺SaM⁻ allele set to be eliminated in the male. This genetic dissection demonstrated that the male semi-sterility occurred only when both the SaM alleles (SaM⁺/SaM⁻) and at least one SaF⁺ allele (SaF⁺/SaF⁻ or SaF⁺/SaF⁺) were present, and the pollen grains carrying SaM⁻ were aborted. SaF⁻ had no role in male sterility, as shown in G735, H70T65, and A408T65.

SaM Encodes a SUMO E3 Ligase-like Protein, and SaF Encodes an F-Box Protein.

The SaM⁺ allele encodes a protein of 257 aa with similarity to the C-terminal region including the SP-RING-like domain of Mms21 (Fig. 2 B and C), a SUMO E3 ligase in yeast (22). An SNP, the marker G02–74.6, is present in the SaM⁻ allele. This G-to-T mutation changes the 3^′ splicing site of the corresponding fifth intron of SaM⁺, which results in a truncated 217-aa protein, by removing the entire sixth exon (80 bp) and creating a new stop codon in the seventh exon (Fig. 2 A–E). A database search showed that SaM is unique in rice; its nearest homolog in Arabidopsis thaliana encodes an unknown protein with a 95-aa region similar to the Mms21 domain in SaM⁺ (Fig. 2C).

SaF encodes a 476-aa protein with an F-box and a plant-specific F-box protein domain (FBD). F-box proteins mediate protein–protein interactions, but the function of the FBD is unclear. The SaF protein is homologous to a ribosomal RNA apurinic site-specific lyase (RALyase) in wheat (23) with 33% identity in the overall sequences and 64% and 44% identities in the F-box and FBD, respectively (Fig. 2C). About 20 members of the F-box/FBD subfamily are predicted in rice, and they have up to 75% identity to RALyase and up to 40% identity to SaF. In the coding region of SaF, only one SNP was found between SaF⁺ and SaF⁻, which results in a Phe-to-Ser substitution in position 287 (Fig. 2 A and B).

SaM was expressed in all tissues tested as examined by RT-PCR, albeit at different levels (Fig. 3A). This analysis showed that only the cDNA of SaM⁺ was detectable in the purified pollen of the F₁ plants, providing molecular evidence for the selective abortion of SaM⁻-carrying pollen. SaF also was expressed constitutively (Fig. 3A). Immunoblotting confirmed that SaM⁻ encoded a truncated protein (Fig. 3B).

Fig. 3.

Expression of SaM and SaF. (A) RT-PCR assay of the genes in T65, E4, and its F₁. Full-length cDNAs of the SaM alleles and a cDNA segment (480 bp) of the SaF alleles are shown. The anthers included pollen, and pure pollen grains were separated from the anther wall tissue. (B) Immunoblot detection of SaM⁻ of T65 (lanes 1–3) and SaM⁺ of E4 (lanes 4–6) prepared from leaf (lanes 1 and 4), anther (lanes 2 and 5), and young panicle (lanes 3 and 6), as probed with an antibody to the N-terminal region.

SaF Interacts with SaM⁻ but not with SaM⁺.

A bacterial two-hybrid (B2H) interaction assay demonstrated that both SaF⁺ and SaF⁻ interacted physically with SaM⁻ but not with SaM⁺ (Fig. 4A). Therefore, the amino acid substitution in SaF⁻ did not affect its physical interaction with SaM⁻ but impaired the biological function of SaF⁻ for male sterility. The interactions were confirmed by a bimolecular fluorescence complementation (BiFC) assay (24), which also showed that the interacting proteins targeted to the cytoplasm and nuclei (Fig. 4C). Because both SaM⁺ and SaM⁻ are required for male sterility (see below), we tested whether their protein products interact directly. The results showed that no physical interactions occurred between them or between the homozygous protein molecules (Fig. 4A). To find out whether the interaction between SaM⁺ and SaF is suppressed by an extra domain in SaM⁺, we prepared four deletion-constructs of SaM⁺ for the B2H assay. Only one construct with deletion to an amino acid position 202, near the divergent point of SaM⁻, was able to interact with SaF (Fig. 4B). Therefore, we determined that a self-inhibitory domain within the 203–218 region of SaM⁺ (see Fig. 2C) blocks the interaction, probably by affecting the protein's structure.

Fig. 4.

Interaction between SaF and SaM proteins. (A) Bacterial two-hybrid assay. SaM⁺(M⁺) and SaM⁻ (M⁻) were expressed in a bait vector, pBT, and SaF⁺(F⁺) and SaF⁻ (F⁻) were expressed in a prey vector, pTRG. CK+, positive control. (B) Mapping of a self-inhibitory domain in SaM⁺. Numbers of the deletion constructs indicate the lengths of the truncated proteins. (C) Interaction between SaM⁻ and SaF⁺(or SaF⁻) and the targeting to cytoplasm and nuclei in onion epidermal cells, as assayed by BiFC. Top: YFP signals on the basis of protein interaction. Bottom: Bright field of the same cells. (Scale bars, 100 μm.)

Both SaM Alleles Are Required, and SaM⁻ Is a Gametophytic Determinant for the Selective Pollen Killing.

Genetic analysis showed that microspores containing SaM⁺ developed normally in F₁ or recombinants with the genotype SaM⁺SaF⁺/SaM⁻SaF⁻ or SaM⁺SaF⁺/SaM⁻SaF⁺, whereas those carrying SaM⁻ aborted. To test the function of the alleles and identify the molecular mechanism for this allele-specific gamete selection, we transferred SaM⁻ (denoted tSaM⁻) and SaM⁺ (tSaM⁺) into E4 and T65, respectively. Three E4 transformants (T₀ generation) with a single tSaM⁻ insertion showed typical male semi-sterility (see Fig. 1G). Furthermore, the transgene segregated in the presence: absence ratio 1:1 (Table 1) and co-segregated with male sterility in the T₁ generation (Table S2). This test also confirmed that SaF⁻ is not necessary for male sterility.

Table 1.

Functional analysis of SaM⁺, SaM⁻, and SaF⁺ in rice by transformation

Plant (line)	Genotype (no. plants)	Phenotype	Segregation in Progenies
F1	M+F+/M−F−	MSS	M−/M−:M+/M−:M+/M+ = 3:325:338 [χ2(1:2:1) = 337.4***]
#1-1	M+F+/M+F+//tM−/−	MSS	tM−:− = 61:48 [χ2(1:1) = 1.6, χ2(3:1) = 21.1***]
#1-2	M+F+/M+F+//tM−/−	MSS	tM−:− = 72:58 [χ2(1:1) = 1.5, χ2(3:1) = 26.7***]
#1-3	M+F+/M+F+//tM−/−	MSS	tM−:− = 111:138 [χ2(1:1) = 2.9, χ2(3:1) = 122.9***]
#2	M−F−/M−F−//tM+/− (19)	MFF
#2-1	M−F−/M−F− (4)	MFF
#2-2	M−F+/M−F− (7)	MFF
#2-3	M−F−/M−F−//tM+/− (3)	MFF
#2-4	M−F+/M−F−//tM+/− (8)	MCS
#3	M+F−/M−F−	MFF	M−/M−:M+/M−:M+/M+ = 90:195:97 [χ2(1:2:1) = 0.44]
#3-1	M+F−/M−F−//tF+/−	MSS	M−/M−:M+/M−:M+/M+ = 2:96:100 [χ2(1:2:1) = 93.2***]
#3-2	M+F−/M−F−//tF+/−	MSS	M−/M−:M+/M−:M+/M+ = 3:62:60 [χ2(1:2:1) = 52.0***]
#3-3	M+F−/M−F−//tF+/−	MSS	M−/M−:M+/M−:M+/M+ = 20:83:65 [χ2(1:2:1) = 24.1***]
#3-4	M+F−/M−F−//tF+/−	MSS	M−/M−:M+/M−:M+/M+ = 7:45:27 [χ2(1:2:1) = 11.6**]

Note: #1-1 to #1-3, tSaM⁻ T₀ plants of E4; #2, tSaM⁺ T₀ plants of T65; #2-1 to #2-4, F₁(T₁) plants from a cross between #2 and a recombinant C513 (M⁻F⁺/M⁻F⁻); #3, the recombinant G735; #3-1 to #3-4, selected tSaF⁺ T₀ plants of G735 line with heterozygous SaM. MCS, male complete sterility. tM⁻(tM⁺, tF⁺) and − indicate the presence and absence of the transgene, respectively.

**, P < 0.005;

***, P < 0.0001.

All 19 T65-T₀ plants with tSaM⁺ were fully male fertile (see Table 1 and Fig. 1H). We then crossed them with a recombinant (C513) (see Fig. 2A) carrying SaM⁻SaF⁺/SaM⁻SaF⁻. As expected, F₁(T₁) plants without tSaM⁺ and/or SaF⁺ were all fully male fertile (see Table 1). However, segregants with SaM⁻SaF⁺/SaM⁻SaF⁻//tSaM⁺ were highly male sterile without or with very little viable pollen, rather than exhibiting semi-sterility (see Fig. 1I), because all microspores carried SaM⁻. These results confirmed that SaM⁻ (tSaM⁻) determined which microspores were aborted, whereas SaM⁺(tSaM⁺), like SaF⁺ (see below), functioned in a sporophytic manner to induce the sterility of the SaM⁻-carrying microspores.

SaF⁺ Is a Component of the Sa Male Sterility System.

To test the function of SaF⁺ for the male sterility, we transferred tSaF⁺ into the progeny (three genotypes) of G735 (SaM⁺SaF⁻/SaM⁻SaF⁻), a full-fertile recombinant (see Fig. 1J). Eighteen T₀ plants were obtained, but only the four plants that had heterozygous SaM (and tSaF⁺) showed male semi-sterility (see Table 1 and Fig. 1K). Furthermore, the segregation for SaM in their T₁ progeny was distorted from the Mendelian ratio (see Table 1), similar to the distortion observed in the F₂ of T65 x E4. Together with the genetic analysis and protein interaction assays, this evidence indicates that SaF⁺ is a component of this hybrid male sterility system. Several genetic mechanisms, such as hybrid sterility, pollen tube competition (certation), specific pollen–pistil interactions, and hybrid breakdown or weakness, have been proposed as genetic mechanisms for the segregation distortion reported in plant hybrids (25, 26). Our results indicate that the segregation distortion of the Sa region is a consequence of hybrid male sterility caused by the allele-specific gamete selection.

Divergence of SaM and SaF Arose in Wild Rice Species.

To trace the origins of the variation in SaF and SaM, we investigated the SNPs G02-69.8 and G02-74.6 (see Fig. 2A) in 13 wild species and in cultivated rice. The nucleotides “T” and “C” at G02-69.8 were variably present in populations of nine wild species (including the common wild rice, O. rufipogon Griff) and in indica cultivars (Table S3). For G02-74.6, only the nucleotide “G” was detected in all tested wild species and indica, with the exception of O. rufipogon, which was “G”-only in accessions distributing in South and Southeast Asia but contained both “G” and “T” (54 of 110 accessions carried “T”) in the accessions from southern China. In summary, three haplotypes of the linked orthologs, SaM⁺SaF⁺, SaM⁺SaF⁻, and SaM⁻SaF⁻, were present in the O. rufipogon populations (Fig. 5A). Most indica cultivars (95 of 106 accessions) contained SaM⁺SaF⁺, but 11 accessions carried SaM⁺SaF⁻. All the tested 108 japonica cultivars possessed SaM⁻SaF⁻. These results indicate that the variation in SaF occurred before the split of most, if not all, of the Oryza species, whereas the mutation in SaM most likely arose in an O. rufipogon population with SaM⁺SaF⁻ in southern China and generated the haplotype SaM⁻SaF⁻ (Fig. 5A).

Fig. 5.

Hybrid male sterility/fertility expression in rice hybrids. (A) Evolution of the SaM and SaF orthologs in wild and cultivated rice species. Asterisk indicates the occurrence of the SNPs at G02-69.8 and G02-74.6. Other wild species and other possible sequence variations are not shown. SA/SEA and SC indicate accessions from south Asia/Southeast Asia and south China, respectively. (B) A two-gene/three-component interaction model. It is assumed that the selective transport of the SaF⁺ and SaM⁺ proteins, shown by the same marks for the alleles without the horizontal or vertical lines, occurs at the tetrad stage (showing two cells only). Arrows of broken lines shown in H71 and E4-T0 indicate the transport of protein is not necessary for the male sterility.

Discussion

The genetic behavior of most identified hybrid sterility loci in plants, including Sa, fits the one-locus allelic interaction model at the general genetic level (7–10). However, the molecular mechanism, that is, how an allele (in many cases from indica or O. glaberrima) of a heterozygous locus acts at the molecular level as a killer to eliminate the gametophytes carrying another allele (e.g., from japonica), is a longstanding question. This study shows that Sa is a complex locus comprising two adjacent genes, SaM and SaF, and that hybrid male sterility is controlled by a special interaction between these two genes.

A Molecular Genetic Model for Rice Hybrid Male Sterility.

We propose a two-gene/three-component interaction model for this hybrid male sterility system (Fig. 5B). The absence of any one of the three alleles, SaM⁺, SaM⁻, or SaF⁺, fails to produce male sterility (see Fig. 2A and Table 1). In an F₁ plant, the linked allele sets (SaM⁺SaF⁺, SaM⁻SaF⁻) are separated from each other in the haploid microspores. Therefore, the SaF⁺ and SaM⁺ proteins may need to be transported from their own microspores to those carrying SaM⁻ for interaction. The SaF⁺-SaM⁻ complex may interact further with SaM⁺ indirectly to trigger a specific sterility process (Fig. 5B). Because the male developmental defect appears at the early uni-nucleate microspore stage (20), the protein transport may occur at the tetrad stage, through cytoplasmic channels existing between tetrad cells (27). However, SaM⁻ should be unable to move to the microspores carrying SaM⁺ to cause sterility, probably because of the loss of a necessary domain in the truncated region. The selective protein transport and the specific SaF⁺-SaM⁻ interaction restrict the sterility process in the SaM⁻-containing microspores, thereby resulting in allele-specific pollen killing. This model also can explain the induction of male sterility in other recombinants and in the transgenic plants (Fig. 5B). In some of these plants, SaF⁺ (tSaF⁺) and/or SaM⁺ (tSaM⁺) co-exist with SaM⁻ (tSaM⁻) in microspores, and the transgenes can function by ectopic expression to cause male sterility. Therefore, the molecular effect of the “allelic interaction” of gene sets does not necessarily require genetic allelism (i.e., location at the same position of the chromosomes). In conclusion, SaM⁻ acts as a gametophytic factor in the male sterility system, whereas SaM⁺ and SaF⁺ play their roles in any microspores in which they are located. On the other hand, the blocking of the SaF⁺–SaM⁺ interaction by the self-inhibitory domain in SaM⁺ may be an important mechanism to prevent triggering the sterility process in SaM⁺-carrying microspores, thus facilitating its transmission to hybrid progenies and avoiding male sterility in indica cultivars.

Intercellular protein trafficking through plasmodesmata is an important direct-signaling process for the development of organs such as shoot apical meristem and endodermis; often, specific domains are required for the targeted protein transports (28, 29), although this process has not been reported in developmental microspores. Therefore, our studies provide an ideal system for studying cell-to-cell interaction by protein trafficking in the early microspore development. Plants seem to have complex SUMOylation systems, and SUMO modification of proteins regulates a number of biological processes in plants (30). Although the activity of SaM as a SUMO E3 ligase needs to be tested, it is possible that a SUMOylation signaling pathway regulates the sterility process in microspore development of indica-japonica hybrids.

Sa Hybrid Male Sterility and the indica-japonica Differentiation.

Dobzhansky-Muller-type hybrid sterility in animals has been shown to result from the divergence of underlying genes that have primary functions (31, 32). The longstanding polymorphisms of the SaF and SaM orthologs among and within the Oryza species suggest that the varied alleles might have primary functions. We show that the incompatible haplotype SaM⁻SaF⁻ was generated by a two-step mutation process in different wild rice species (Fig. 5A). The intermediate haplotype SaM⁺SaF⁻ might have acted as a buffer to avoid the elimination of SaM⁻SaF⁻ from the wild rice populations when SaM⁻ arose. This process is similar to but distinct from the stepwise mutation model for the generation and maintenance of incompatible alleles of hybrid sterility genes in animals (33). The absence of SaM⁻ in indica rice suggests that this sterility system eliminated this allele during evolution; the fixation of SaM⁻SaF⁻ in japonica rice could be attributed to random genetic drift accompanied by its geographical isolation from indica or to an adaptive advantage of this haplotype for japonica. Therefore, this male sterility system might have contributed to the reproductive barrier and the genetic differentiation between the subspecies.

Early reports suggested that indica originally was domesticated from O. rufipogon and that japonica was further generated from indica (12, 34). Recent studies propose that indica and japonica originated independently from different populations of O. rufipogon (35, 36). Our results seem to corroborate the hypothesis of independent domestication (Fig. 5A). Furthermore, the data suggest that indica also evolved from different pre-differentiated populations of O. rufipogon and provide molecular evidence that indica and japonica originated from O. rufipogon in south or Southeast Asia and south China, respectively (36).

Compatible Haplotype Useful for Hybrid Rice Breeding.

The haplotype SaM⁺SaF⁻ found in some indica cultivars (Table S3) is expected to be compatible with japonica cultivars as well as with the indica ones with SaM⁺SaF⁺ because of the absence of SaF⁺ or SaM⁻ in their hybrids (Fig. 5A) and thus can be defined as a neutral or compatible allele-complex, Saⁿ. Therefore, this complex and the incompatible ones Saⁱ (SaM⁺SaF⁺) and Sa^j (SaM⁻SaF⁻) form a tri-haplotypic system controlling the hybrid male sterility and fertility in O. sativa. Rice cultivars with Saⁿ and other compatible alleles of hybrid sterility loci, such as S5ⁿ (9, 19), are valuable germplasm for overcoming the reproductive barrier in the breeding of inter-subspecific hybrid rice. Therefore, our findings have important implications for genetic improvements in rice, providing molecular markers for screening Sa-compatible germplasm and possibly creating transgenic Sa-compatible lines by suppressing the expression of the Sa gene(s).

Materials and Methods

Rice Materials.

Rice materials for the gene cloning and functional analysis included the T65 and E4 (6) and their hybrid progeny prepared in this study. The wild and cultivated rice species used for haplotype survey (Table S3) were provided by the International Rice Research Institute and other institutions in China.

Map-Based Cloning.

The pollen phenotypes (MFF, MSS) of rice were observed in a 1% I₂-KI (potassium iodide) solution under a microscope (Leica DMLB). F₂ plants of T65 x E4 were used for the mapping. Hybrid plants producing fully fertile pollen (> 95% pollen fertile) and semi-fertile pollen (≈ 50% of pollen grains sterile) represent the homozygote and heterozygote for Sa, respectively. Key recombinants (see Fig. 2A) were confirmed further by investigating the phenotypic and genotypic segregations in their progenies. Genotyping with SNP markers was done as described (37).

Genomic sequences of the SaM alleles and SaF⁺ (including their promoters and downstream sequences) were PCR-amplified with the primers SaM-p1/SaM-p2 for SaM and SaF-p1/SaF-p2 for SaF⁺ (all primers are listed in Table S4) and were cloned into a binary vector pCAMBIA1300 (Cambia). The constructs were transferred into suitable rice lines (see Table 1) by Agrobacterium-mediated transformation (38).

B2H and BiFC Assays.

The coding regions of SaM⁺ and SaM⁻ cDNAs were amplified with SaM-p5/SaM-p6 and SaM-p5/SaM-p7, respectively, and were cloned into a bait vector pBT of the BacterioMatch two-hybrid system (Stratagene). The coding regions of SaF⁺ and SaF ⁻ cDNAs were amplified with SaF-p3/SaF-p4 and cloned into a prey vector pTRG. To check for possible interactions between the heterozygous or homozygous SaM⁺ and SaM⁻, the genes amplified with SaM-p8/SaM-p9; SaM-p10/SaM-p11 were cloned into pTRG. The deletion constructs of SaM⁺ were prepared by inverse PCR of the gene's construct in pBT with an anchor primer (SaM-pa12) in combination with four inverse-directed primers (SaM-pr13–SaM-pr16), respectively, followed by digestion with a restriction endonuclease XhoI and self-ligation. The normal expression of the SaM fusion proteins in Escherichia coli was confirmed by immunoblot (Fig. S1).

For the BiFC assay, the SaM⁻-coding cDNA amplified with SaM-p17/SaM-p18 was cloned into the vector pUC-SPYCE (24); the SaF⁺- and SaF⁻-coding cDNAs amplified with SaF-p5/SaF-p6 were cloned into the vector pUC-SPYNE (24). The combined constructs were co-transferred into onion epidermal cells by a helium-driven accelerator (BioRad). After culture for 10 to 15 h, the BiFC-based YFP expression was viewed with a confocal laser scanning microscope system TCS SP2 (Leica).

RT-PCR.

Mature pollen grains were separated from anther wall tissue as described (39). Total RNAs were extracted from rice tissues and pollen with TRIZOL (Invitrogen). Reverse transcription reactions for SaM and SaF were performed with SuperScript III kit (Invitrogen), and expression of the target genes was assayed by RT-PCR with SaM-p19/SaM-p20 for SaM and SaF-p7/SaF-p8 for SaF.

Immunoblot.

A peptide antigen corresponding to residues 1–25 of SaM was synthesized (Invitrogen) and used for immunizing rabbits. Detection of SaM⁺ and SaM⁻ with the antiserum was carried out as described (37).