Rice pollen hybrid incompatibility caused by reciprocal gene loss of duplicated genes

Abstract

Genetic incompatibility is a barrier contributing to species isolation and is caused by genetic interactions. We made a whole genome survey of two-way interacting loci acting within the gametophyte or zygote using independence tests of marker segregations in an F₂ population from an intersubspecific cross between O. sativa subspecies indica and japonica. We detected only one reproducible interaction, and identified paralogous hybrid incompatibility genes, DOPPELGANGER1 (DPL1) and DOPPELGANGER2 (DPL2), by positional cloning. Independent disruptions of DPL1 and DPL2 occurred in indica and japonica, respectively. DPLs encode highly conserved, plant-specific small proteins (∼10 kDa) and are highly expressed in mature anther. Pollen carrying two defective DPL alleles became nonfunctional and did not germinate, suggesting an essential role for DPLs in pollen germination. Although rice has many duplicated genes resulting from ancient whole genome duplication, the origin of this gene duplication was in recent small-scale gene duplication, occurring after Oryza-Brachypodium differentiation. Comparative analyses suggested the geographic and phylogenetic distribution of these two defective alleles, showing that loss-of-function mutations of DPL1 genes emerged multiple times in indica and its wild ancestor, O. rufipogon, and that the DPL2 gene defect is specific to japonica cultivars.

Species diversification and species maintenance are key topics in evolutionary biology (1, 2). Identification of genes responsible for barriers to gene flow between two species provides insight into molecular mechanisms of reproductive isolation (3) and relationships between evolution of barrier genes and speciation. Despite the importance of such studies, only a handful of genes responsible for genetic incompatibilities in intrinsic postzygotic barriers have been identified from a few well-characterized model species, including mouse, fly, yeast, Arabidopsis, and rice (3, 4), where map-based cloning is possible.

Asian cultivated rice, Oryza sativa, is one such well-characterized species with a small genome size (∼400 Mb), and genome sequences available for its two differentiated subspecies, indica and japonica (5, 6). Both indica and japonica comprise a large variety of strains showing a wide range of genetic diversity and structured populations, which seem to be in the early stages of speciation (7–9). In the process of rice breeding, various genetic incompatibilities in intersubspecific hybrids between indica and japonica have been reported (ref. 7; see also Oryzabase at www.shigen.nig.ac.jp/rice/oryzabase/top/top.jsp). Despite many attempts to isolate the genes responsible for these hybrid incompatibilities, only a few genes have been isolated (10–13), and the relationships between genetic incompatibilities and rice differentiation is poorly understood.

To investigate the molecular mechanisms of reproductive isolation and the status of evolutionary differentiation within Oryza, we developed a method to map reproductive barriers by regression analysis of allele frequencies using an indica-japonica F₂ population, and mapped more than 30 reproductive barriers (14). These mapped barriers are due to interactions between genetic factors of the parents of the hybrid, and detection of interacting loci is the first step toward isolation of the causal genes. In this study, detection of the two-way interaction and positional cloning of the detected interacting loci pair were carried out using the same cross-combination used previously for mapping reproductive barriers (14). Evolutionary analysis of the causal genes was performed to investigate the establishment of this genetic incompatibility in rice.

Results and Discussion

Detection of the Interacting Loci Causing Genetic Incompatibility.

For fine mapping of a reproductive barrier as a single genetic factor, and understanding its molecular mechanisms of hybrid incompatibility, it is essential to know interacting loci and where they act. In the early stages of speciation, as in japonica-indica, genetic incompatibility caused by simple two-way interactions would be expected to be more common, because two-way interactions represent the smallest number of differences, which could cause an incompatibility between them. To detect two-way interactions between segregating genotypes at different loci, χ² independence tests of all codominant marker combinations were performed for a high-density linkage map that was made using an F₂ population from a cross between japonica cultivar Nipponbare and indica cultivar Kasalath (15) and used for mapping reproductive barriers (14). Contour plots of χ² values are shown in black contours of the upper-left part of Fig. 1. High χ² values caused by linkage were observed at diagonals between corresponding chromosomes. There were 14 nonlinked chromosomal pair regions where χ² peak values were more than 20 and the highest peak value was 33.6 between chromosome 9 and 12. The probability of a χ² peak value of 20 with 4 degrees of freedom in a χ² distribution is less than 5 × 10⁻⁴ for a single test. However, this analysis involves a great degree of multiple testing. A Bonferroni correction for multiple testing cannot be applied in this case, because each test was dependent by linkage. To find true interactions, reproducibility of the independence tests of marker segregation was tested using 286 different F₂ plants from the same cross-combination (red contours of upper-left part of Fig. 1). Although allele frequency curves were similar between different populations (Fig. 1 A and B), reproducible peaks of χ² independence test values were rare in nonlinked regions, suggesting most of the peaks of the independence tests were false or unstable. A reproducible peak was detected between previously mapped reproductive barriers on chromosomes 1 and 6. The detected two-way interaction acts within segregating gametophyte or zygote; however, the previous analysis (14) could not distinguish between these possibilities. To determine where the interaction occurred, χ² independence tests of marker segregation were performed for reciprocal backcross populations between Nipponbare and (Nipponbare × Kasalath) F₁ that was made by using Nipponbare as maternal parent. Both contour plots are shown in the lower-right part of Fig. 1. Only when F₁ was used as pollen parent, a high χ² peak, 39.6, was observed between marker S11214 on chromosome 1 and marker S1520 on chromosome 6 (blue contour in Fig. 1), suggesting reduction of pollen transmission having Kasalath allele at S11214 and Nipponbare allele at S1520.

Fig. 1.

χ² values of independent segregation between markers and allele frequencies along chromosomes. Allele frequencies of marker genotypes were plotted along linkage maps made by two different F₂ populations from a cross between Nipponbare and Kasalath. Red, green, and dark blue represent Nipponbare homozygote, Kasalath homozygote, and heterozygote, respectively. (A) Previously constructed high-density linkage map using 186 F₂ plants (15). (B) Newly constructed linkage map using different 288 F₂ plants. χ² values of independence tests between marker segregations in four different populations were expressed in contours along previously constructed linkage map. (Left Upper) Black contours are χ² values from 10 to 35 in intervals of 2.5 for independent segregation between all combinations of 994 codominant markers of the previously developed F₂ population. Red contours are χ² values from 10 to 35 in intervals of 2.5 for 182 markers of the newly developed F₂ population. (Right Lower) Blue contours are from 7.5 to 45 in intervals of 2.5 for 159 markers of 236 backcross plants made using F₁ as a paternal parent. Magenta contours are from 7.5 to 45 in intervals of 2.5 for those of 234 backcross plants made by F₁ as a maternal parent. Diagonals are overlapping contours of the four different populations.

Map-Based Cloning.

To identify the responsible genes near the marker S11214 and S1520 on chromosomes 1 and 6, linkage analysis was carried out for each locus as a single factor by fixing the nontransmittable genotype at the other locus. Genotypes of mapping loci were determined by allele frequency tests of their self-pollinated progeny (Fig. S1). Fine mapping using more than 10,000 self-pollinated and backcrossed progeny indicated that candidate genes were located within regions of 59 kb on Nipponbare chromosome 1 and 11 kb on Nipponbare chromosome 6, including nine and two predicted genes, respectively (Fig. 2 A and B and Fig. S1). A pair of genes, one from each region, shared a high degree of homology with each other (Fig. 2 A and B). Because these homologous genes showed nucleotide sequence differences between Nipponbare and Kasalath, they were regarded as primary candidates. LOC_Os01g15448 on chromosome 1 was designated DOPPELGANGER1 (DPL1) and LOC_Os06g08510 on chromosome 6 as DOPPELGANGER2 (DPL2).

Fig. 2.

Fine mapping of the reproductive barrier genes. The candidate regions were delimited within 59 kb on chromosome 1 (A) and within 11 kb on chromosome 6 (B) of Nipponbare. Numbers in blue between dashed lines indicate number of recombinant plants between the selection markers and that of the analyzed segregants. Cosegregated markers with the barrier loci and two highly similar genes on chromosome 1 and 6 were indicated in magenta and green, respectively. (C) Schematic gene structure and alignment of DPLs of Nipponbare (N) and Kasalath (K). Each transcript sequence was determined by 5′ and 3′ RACE experiments. Red and yellow boxes indicate the coding and untranslated regions, respectively. Gray lines between gene structures show identical bases. The inserted TE in DPL1-K⁻ is indicated by a triangle. The base change from A to G in the second intron of DPL2-N⁻ was marked with an arrowhead and was predicted as a loss of branch site for splicing (Table S2). Arrows indicate positions of gene-specific primers for DPL1 and DPL2 used for RT-PCR analysis. The empty boxes of DPL1-K⁻ were predicted gene structure based on the DPL1-N⁺, because no DPL1-K⁻ transcript was detected in any tissue.

Gene Structure and Expression Analysis Suggested Loss-of-Function Alleles of DPLs.

Sequence analysis of the Nipponbare and Kasalath genomes and their transcripts suggested that alleles on Nipponbare chromosome 1 (DPL1-N⁺) and Kasalath chromosome 6 (DPL2-K⁺) had the same coding sequence structure (Fig. 2C). DPL1-N⁺ and DPL2-K⁺ encode highly similar, small, unknown proteins of 94 and 95 amino acids, respectively (Fig. S2). In contrast, alleles on Kasalath chromosome 1 (DPL1-K⁻) and Nipponbare chromosome 6 (DPL2-N⁻) had structural differences from the above two alleles. There were no other similar genes in the japonica or indica genome. DPL1-K⁻ had a 518-bp insertion in the predicted coding sequence, and the transcript could not be detected in any tissues by RT-PCR (Fig. 3E). The inserted fragment had diagnostic terminal inverted repeats, and it was widely dispersed in the rice genome, suggesting it was a transposable element (TE). The transcript of DPL2-N⁻ was a readthrough product of the second intron, generating a premature stop codon (Fig. 2C and 3E). Transcripts of DPL1-N⁺, DPL2-K⁺ and DPL2-N⁻ were detected in all tissues examined (Fig. S3A). Markedly, the expressions in anther at tricellular stage were 10-fold higher than in the other tissues. High expressions of DPLs in pollen were also observed in in-situ hybridization experiments (Fig. S3B). Anti-DPL antibodies recognized small proteins of 10.3 and 10.4 kDa for DPL1-N⁺ and DPL2-K⁺ in extracts from Nipponbare and Kasalath mature anthers, respectively (Fig. 3F). DPL2-N⁻ protein (12.6 kDa) was not detected in extracts from Nipponbare anthers. The lack of DPL1-K⁻ transcript and the absence of DPL2-N⁻ protein suggested that both DPL1-K⁻ and DPL2-N⁻ were loss-of-function alleles.

Fig. 3.

Pollen phenotype and expression analysis of Nipponbare (N), Kasalath (K), and near-isogenic lines (NILs). Pollen grains were stained with I₂KI (A) and DAPI (B) solutions. (C) Pollen germination on a germination medium. [Scale bars, 80 μm (A and B) and 160 μm (C).] (D) Pollen germination rate of Nipponbare and Kasalath (white bars) and five NILs (black bar) on the pollen germination medium. Error bars indicate SDs, which were estimated from six and four individuals of Nipponbare and Kasalath, respectively. All NILs has DPL1-K⁻/DPL1-N⁺ in Nipponbare background (DPL2-N⁻/DPL2-N⁻). (E) RT-PCR of DPL expression in mature anther of Nipponbare and Kasalath. Actin was used as control. Gene-specific primers are shown in Fig. 2C. (F) Immunoblot analysis of DPL proteins. Predicted molecular weights of DPL1-N⁺, DPL2-K⁺, and an N-terminal hexahistidine tag recombinant protein are 10.3, 10.4, and 12.8 kDa, respectively. The lane indicated by R, N, and K are recombinant protein, extracts from mature anther of Nipponbare, and extracts from mature anther of Kasalath, respectively.

Pollen Having DPL1-K⁻ and DPL2-N⁻ Failed to Germinate, but Could Be Rescued by Transformation with Functional DPLs.

In a functional assay of pollen, near isogenic lines (NILs) having heterozygous DPL1 (DPL1-K⁻/DPL1-N⁺) in Nipponbare background (DPL2-N⁻/DPL2-N⁻) were used to eliminate phenotypic effects of other reproductive barriers between Nipponbare and Kasalath. By examining the transmission rate of pollen to the progeny of self-pollinated plants, it could be inferred that pollen possessing DPL1-K⁻ and DPL2-N⁻ together was not transmitted to the next generation. Therefore, half of the pollen grains from the NILs possessing both DPL1-K⁻/DPL1-N⁺ and DPL2-N⁻/DPL2-N⁻ should have some defect in function. The mature pollen of the NILs did not show any defect in pollen shape, number of nuclei, or starch accumulation (Fig. 3 A and B). Examination of pollen germination activity in vitro revealed that about half of the NILs pollen grains could not germinate, whereas almost all of the pollen grains of Nipponbare and Kasalath germinated, and pollen tubes were elongated (Fig. 3 C and D). This suggests that the pollen carrying both DPL1-K⁻ and DPL2-N⁻ failed to germinate.

Gene expression, structure, and pollen function analyses suggested that DPL1-K⁻ and DPL2-N⁻ are loss-of-function alleles. Therefore, pollen having these two alleles should be rescued by transformation with a functional allele, DPL1-N⁺ or DPL2-K⁺. When DPL1-N⁺ was introduced into NILs possessing DPL1-K⁻/DPL1-K⁻ and DPL2-K⁺/DPL2-N⁻ (Fig. S4A), self-pollinated progeny carrying homozygous DPL1-K⁻ and homozygous DPL2-N⁻ could be generated by rescued pollen possessing DPL1-K⁻ and DPL2-N⁻ alleles (Table S1). All of these DPL1-K⁻/DPL1-K⁻;DPL2-N⁻/DPL2-N⁻ plants contained the DPL1-N⁺ transgene with no exception, whereas no progeny having the same allele combination were produced by the NILs transformed with vector only. Similarly, when NILs possessing both DPL1-K⁻/DPL1-N⁺ and DPL2-N⁻/DPL2-N⁻ were transformed with DPL2-K⁺ (Fig. S4B), those progeny that had homozygous DPL1-K⁻ and homozygous DPL2-N⁻ always contained transgenic DPL2-K⁺ (Table S1). These findings clearly showed that a functional DPL1-N⁺ or DPL2-K⁺ allele is essential for pollen transmission. In conclusion, DPL1-K⁻ and DPL2-N⁻ are revealed loss-of-function alleles that act as a pair of hybrid incompatibility genes by their combination in the pollen. This is a typical case of the Bateson–Dobzhansky–Muller (BDM) incompatibility for barrier formation by genetic incompatibility between species (1, 2), and an example of the “set of duplicate gametic lethal” model (7).

Origin of the Genetic Incompatibility by DPL Genes in Rice.

To estimate the origin of this genetic incompatibility by independent disruption of DPL genes in rice, evolution of DPL genes was investigated by sequence and expression analysis of 43 well-diverged wild accessions and cultivars (Table S2) belonging to eight rice species (16). All accessions/cultivars had both DPL1 and DPL2 orthologs, and both DPLs seemed to be functional in 16 accessions/cultivars, suggesting that the gene duplication of DPL occurred in a common ancestor of rice, and both DPLs were functional before rice differentiation (Fig. S5 and Table S2). Using nucleotide variation in the coding regions of both DPL1 and DPL2 (Fig. S5), phylogenetic relationships between O. sativa and O. rufipogon, which is a wild progenitor of O. sativa (7, 17), were calculated (Fig. 4A). In O. sativa and O. rufipogon, there were three types of DPL1 coding sequences: functional Nipponbare type, nonfunctional Kasalath type (TE inserted), and functional Kasalath type; and two types of DPL2 coding genome sequences (not transcribed sequences): Nipponbare type and Kasalath type. O. rufipogon accessions and O. sativa cultivars could be classified into four groups: group I, both DPLs were Nipponbare type sequences (10 japonica cultivars and two O. rufipogon); group II, Nipponbare type DPL1 and Kasalath type DPL2 (five indica cultivars and two O. rufipogon accessions); group III, nonfunctional Kasalath type DPL1 and Kasalath type DPL2 (five indica cultivars and four O. rufipogon accessions); and group IV, functional Kasalath type DPL1 and Kasalath type DPL2 (four O. rufipogon accessions). Although O. sativa and O. rufipogon are classified based on whether they are domesticated, all four groups contained O. rufipogon accessions, and these two species were not distinguished by DPL gene sequences. Classification by DPL coding sequences was in good agreement with phylogenetic analyses of O. sativa and O. rufipogon by p-SINE insertion patterns (18, 19). Indica-japonica differentiation was estimated to have occurred 0.4–0.2 mya and before domestication (9, 20, 21). DPL2-N⁻-type gene disruption by intron readthrough product was observed in japonica cultivars, but not in O. rufipogon accessions, in group I (Table S2). DPL2-N⁻-type disruption occurred in a common ancestor of tropical and temperate japonica cultivars (Fig. 4A). However, the insertion of the TE in DPL1 (DPL1-K⁻) was detected in five indica cultivars and four accessions from O. rufipogon belonging to group III (Table S2). We have investigated the reproductive barrier caused by DPL1 and DPL2 as being between japonica and indica. However, indica cultivars and O. rufipogon accessions of group II have functional DPLs (Table S2 and Fig. S5). Recently, distinct population structure within indica cultivars was detected and aus and indica groups were designated (8, 22). All three aus cultivars (Kasalath, Dular, and N22) were in group III (Table S2). DPL1-K⁻-type disruption occurred in a common ancestor of aus and O. rufipogon in group III (Fig. 4A).

Fig. 4.

Evolutionary history of gene duplication and disruption of DPLs. (A) Phylogenetic relationships of eight species from Oryza sativa complex and Oryza officinalis complex. A clade consisting of O. sativa and O. rufipogon (shown as black node) was divided into four groups according to the nucleotide variations in the coding regions of DPLs as shown in Fig. S5. The groups having disrupted DPLs are shown as arrows and colored text background. Groups containing both disrupted and intact DPLs are outlined in color. (B) The geographic distribution of examined O. rufipogon accessions and O. sativa cultivars. Filled and open circles indicate O. rufipogon accessions and O. sativa cultivars, respectively. The accessions/cultivars belonging to groups I–IV are colored as blue, yellow, magenta, and lime, respectively, as in Fig. S5. The location of origin of seven accessions/cultivars from China was only shown as within the country borders, as precise information was not available. (C) Historical representation of the independent duplications of the DPL genes after Oryza-Brachypodium and before sorghum-maize differentiations. Predicted upstream and downstream neighboring genes to the DPL ortholog in each species are shown. The syntenic genes showing similarity with each other are shown as the same color. Arrowheads indicate predicted gene duplication points, respectively.

The geographic distributions of examined O. rufipogon accessions and O. sativa cultivars were investigated to estimate where the disruptions of DPLs occurred. The two O. rufipogon accessions belonging to group I were only distributed in China (Fig. 4B). However, the four O. rufipogon accessions of group III having DPL1-K⁻ type were distributed in Thailand, Myanmar, and eastern India. These features suggested that these loss-of-function alleles appear to have arisen in populations similar to those currently found in the geographic regions stated above.

In addition to the Kasalath type, three different kinds of disruptions of the DPL1 genes were predicted (Fig. 4A and Fig. S5). All examined accessions of O. glaberrima and O. barthii had a long deletion, including the whole second exon of the DPL1 gene. W1236 from O. rufipogon in group IV had a nucleotide substitution at the 21st polymorphic site, generating a premature stop codon, and W1169 from O. glumaepatula had a single nucleotide deletion at the 10th polymorphic site, resulting in a frame shift (Fig. S5). Such disruptions of DPL genes in other rice species suggest DPL genes have the potential to be reproductive barriers among these species.

Timing of the Duplications of DPL Genes.

DPL orthologs of other species were sought in public databases. DPL genes are widely found in angiosperms. Forty-two species have at least one DPL and their amino acid sequences are highly conserved (Fig. S2). To estimate the timing of the DPL duplication, syntenic conservation among grass species was studied by using a public database (Materials and Methods). Grass species experienced ancient whole-genome duplication, but rice chromosomes 1 and 6 do not share a homologous ancient grass chromosome (6, 23–27), and the duplicated segment between the DPL1 and DPL2 loci extended only from a site near the end of first intron to the polyadenylation site of these genes (Fig. 2C). Duplicated DPL genes of rice were not generated by whole-genome duplication in an ancestral grass. Brachypodium distachyon is the closest model plant to Oryza (23, 28) and has a single DPL. The regions around the DPL ortholog on chromosome 1 of B. distachyon had sequence similarity with the region flanking DPL2 on rice chromosome 6 (Fig. 4C). This synteny was also seen in other two grass species, in chromosome 10 of Sorghum bicolor and chromosome 9 of Zea mays. These syntenic conservations and the lack of synteny in the regions surrounding DPL1 suggested that DPL2 is the most ancient among the grasses. The similarity of the deduced amino acid sequences of DPL genes within O. sativa (98% identity) was higher than that between O. sativa and B. distachyon (88% identity). This finding suggested that DPL1 on rice chromosome 1 was generated by recent gene duplication after Oryza-Brachypodium differentiation (green arrowhead in Fig. 4C). As well as O. sativa, S. bicolor and Z. mays also have two DPLs (Fig. S2). The similarity of the deduced amino acid sequences of the DPLs on sorghum chromosome 2 and maize chromosome 2 (100% identity) was higher than that within the sorghum DPLs (89% identity). The regions around DPLs on chromosomes 2 and 10 of S. bicolor and chromosomes 2 and 9 of Z. mays do not share a homologous ancient grass chromosome, similar to the case in rice (26, 27). This suggested that the duplication might have occurred before sorghum-maize differentiation separately from the Oryza-Brachypodium lineage (blue arrowhead in Fig. 4C).

Genetic incompatibility caused by reciprocal loss or subfunctionalization of duplicated genes between isolated populations is described as BDM-type hybrid incompatibility. Because most species of flowering plants and vertebrates experienced whole-genome duplications, and species diversity increased after their polyploidization (29), this type of incompatibility has been suggested as a plausible cause of speciation (30–32). Nevertheless, only three examples of hybrid incompatibility caused by reciprocal gene loss have been identified, Bikard et al. (33) in Arabidopsis, Yamagata et al. (12), and this study in rice. Rice and Arabidopsis experienced ancient whole-genome duplications, once at ∼65 mya, and twice at ∼70 and ∼40 mya, respectively (29). A large number of paralogous genes have been lost since duplication, and a large number of surviving duplicated genes seem to have undergone subfunctionalization, according to transcriptome analysis (34, 35). Therefore, it would seem likely that duplicated genes causing hybrid incompatibility would have their origin in ancient whole-genome duplications; nevertheless, a common feature of identified genetic incompatibilities caused by duplicated genes [i.e., histidinol-phosphate aminotransferase genes on chromosomes 1 and 5 of Arabidopsis (33), mitochondrial ribosomal protein L27 genes on chromosomes 4 and 8 (12), and DPLs on chromosomes 1 and 6 of rice] is that they do not share ancestral chromosomes (6, 23, 25, 28, 35, 36). We performed a whole-genome survey for detection of two-way interaction, in which hybrid incompatibility in either gametophyte or zygote by reciprocal loss or subfunctionalization of essential genes can be detected with the same sensitivity. However, only a single pair was detected in this japonica-indica cross (Fig. 1). There was no other hybrid incompatibility caused by reciprocal gene loss or subfunctionalization, even though about 30 reproductive barriers were mapped (14) and rice has recent (5∼7 mya) duplicated segments between chromosome 11 and 12 (9, 37). Bikard et al. (33) also performed a similar whole-genome survey of two-way interaction in Arabidopsis intraspecies cross, and detected two interacting pairs (figure S1 in ref. 33). Taking these findings into account, genetic incompatibility caused by reciprocal disruptions of duplicated genes seems to be rare in Arabidopsis and rice-differentiating populations.

Materials and Methods

Detection of Two-Way Interactions Within Gametophyte or Zygote.

Construction of a high-density linkage map using 186 F₂ plants from an O. sativa ssp. japonica cv. Nipponbare, O. sativa ssp. indica cv. Kasalath cross was described previously (14, 38). A total of 994 codominant markers that were mapped to different locations using more than 175 plants were selected and used to test independence of marker segregation by χ² analyses. To confirm independence tests of marker segregation, another linkage map of 182 markers was constructed using 288 F₂ plants from the same cross, including newly developed codominant PCR markers (Table S3) and restriction fragment length polymorphism (RFLP) markers that showed a peak χ² value >20 in the high-density map. Reciprocal backcross populations between Nipponbare and Nipponbare × Kasalath F₁, made using Kasalath as pollen parent, were developed. The sizes of backcross populations made using F₁ as maternal and paternal parent were 234 and 236, respectively. To perform χ² tests for independence of segregation, 159 markers covering the whole genome were used for both populations.

Map-Based Cloning of DPLs.

For high-resolution mapping of DPL1 gene on chromosome 1 or DPL2 gene on chromosome 6, recombinant plants that had recombination near the RFLP markers S11214 or S1520 and were homozygous at the interacting locus were selected among ∼4,000 and 6,000 descendants from the above populations, respectively. PCR markers distinguishing Nipponbare and Kasalath were developed for selection and fine mapping based on Nipponbare and 93-11 genome sequences (5, 6) as shown in Table S3. PCR for selecting recombinants was performed with plant crude DNA extracts (39, 40). Genotypes at the barrier loci of the selected recombinants were determined by their progeny tests (Fig. S1). Putative causal genes within each candidate region were identified by reference to the Rice Genome Annotation (http://rice.plantbiology.msu.edu/index.shtml).

Full-Length cDNA Cloning.

Total RNA was extracted from mature anthers with the Concert Plant RNA Reagent (Invitrogen). Full-length cDNA sequences of DPLs were determined using the SMART RACE cDNA Amplification Kit (Clontech) with mRNA isolated using Dynabeads Oligo (dT)₂₅ (Invitrogen). The determined sequences are available from DNA Data Bank of Japan (DDBJ accession nos. AB535626–AB535628).

Complementation Test.

Kasalath and Nipponbare BAC clones, including DPL2 and DPL1, were obtained from the National Institute of Agrobiological Sciences (Tsukuba, Japan) and the BAC/EST Resource Center at the Clemson University Genomics Institute (Clemson, SC), respectively. A genomic DNA fragment of DPL1-N⁺ containing a region extending from 4.4 kb upstream of the transcription start site to 1.6 kb downstream of the polyadenylation site and a DPL2-K⁺-containing region extending from 2.5 kb upstream to 1.7 kb downstream were isolated from the BAC clones. Each fragment was subcloned into the binary vector (pBGH1) (41). The constructs were transferred into suitable NILs (Fig. S4) by Agrobacterium-mediated transformation (42). The genotype of each transgenic plant and the self-pollinated progeny were checked by PCR using gene-specific primers (Table S3).

Gene Expression Analysis.

Total RNA was extracted from rice tissues with the Concert Plant RNA Reagent (Invitrogen), and cDNA was synthesized by SuperScript III reverse transcriptase (Invitrogen). The expressions of DPLs were checked by PCR using gene-specific primers (Table S3). For quantitative expression analysis of DPLs, 20 ng of cDNA was used with SYBR Premix Ex Taq (Takara) and gene-specific primers. Data were collected using the Takara Thermal Cycler Dice Real Time System in accordance with the instruction manual.

RNA in Situ Hybridization.

Anthers at bicellular stage were fixed in fixative solution (formalin:glacial acetic acid:50% ethanol = 5:5:90) for 24 h at 4 °C. Probes were prepared from a cloned plasmid (pCR II-TOPO plasmid; Invitrogen) containing DPL1-N⁺ cDNA fragment, which was amplified by PCR using the gene-specific primers RT1-1f and GSP1-4. The procedures for in situ experiments were performed as described by Kouchi and Hata (43).

Immunoblot Analysis.

Two peptides antigens of the residues 1–17 (MGSEDTKDMLKNVDWKT) and 24–40 (TDPSQPVVKKRLPKKIR) of DPL1-N⁺ common to all other DPLs of Nipponbare and Kasalath were synthesized and used for immunizing rabbits. Total protein preparation and Western blot detection were conducted following the method in ref. 44. To prepare recombinant DPL1-N⁺ protein, the 95-aa coding region was fused to a hexahistidine tag at the N-terminal, cloned into the expression vector pDEST17 (Invitrogen), and expressed in the BL21-AI Escherichia coli strain (Invitrogen). Recombinant protein and crude extracts of anther were electrophoresed on a 16% SDS/PAGE following the Tris-Tricine-SDS method (45).

Examination of Pollen Viability and Pollen Germinability.

Anther before flowering was fixed in an EAA solution (100% ethanol:acetic acid = 3:1) overnight at room temperature. The pollen maturity was investigated with I₂KI staining (0.2% iodine and 2% potassium iodine) and a solution of DAPI under a fluorescence microscope and bright field microscope. Pollen grains from dehisced anthers were immediately placed on a germination medium [1 mg H₃BO₃, 3 mg Ca(NO₃)₂, 1.7 g sucrose, and 0.07 g agarose in 10 mL sterilized water] and incubated at room temperature in a humid chamber. After 2–3 h, pollen grains were stained in a 0.005% aniline blue solution [0.005% aniline blue in 0.15 M K₂HPO₄ (pH 8.6)] and counted under a fluorescence microscope. More than 300 pollen grains were observed in each examination.

DPL Orthologs in Angiosperms.

Deduced amino acid sequences of DPL orthologs were predicted from deposited mRNA and genomic sequences in public databases (Fig. S2) and aligned by ClustalW in DDBJ (http://www.ddbj.nig.ac.jp). Deduced amino acid sequences of DPL1 and DPL2 from indica cultivar IR24, which has predicted functional DPL1 and DPL2, were used for the sequence comparison. The neighboring predicted upstream and downstream genes to the DPL ortholog of each species was sought in the Phytozome database (http://www.phytozome.net/index.php).

Evolutionary Analysis in Rice.

RT-PCR was performed on seedlings or mature anthers as described for gene expression analysis (Table S2). The determined nucleotide sequences were aligned with ClustalW. The phylogenetic tree, including O. sativa cultivars and O. rufipogon accessions, was constructed using the coding sequence of DPL1 and DPL2 by the NJ method (46). The relationship of other rice species is shown with reference to the phylogenetic tree constructed on the basis of p-SINE insertion patterns (47). All sequences are available from DDBJ (accession nos. AB534814–AB534897, AB560672, and AB560673).