A Killer-Protector System Regulates Both Hybrid Sterility and Segregation Distortion in Rice

See allHide authors and affiliations

Science  14 Sep 2012:
Vol. 337, Issue 6100, pp. 1336-1340
DOI: 10.1126/science.1223702


Hybrid sterility is a major form of postzygotic reproductive isolation that restricts gene flow between populations. Cultivated rice (Oryza sativa L.) consists of two subspecies, indica and japonica; inter-subspecific hybrids are usually sterile. We show that a killer-protector system at the S5 locus encoded by three tightly linked genes [Open Reading Frame 3 (ORF3) to ORF5] regulates fertility in indica-japonica hybrids. During female sporogenesis, the action of ORF5+ (killer) and ORF4+ (partner) causes endoplasmic reticulum (ER) stress. ORF3+ (protector) prevents ER stress and produces normal gametes, but ORF3– cannot prevent ER stress, resulting in premature programmed cell death and leads to embryo-sac abortion. Preferential transmission of ORF3+ gametes results in segregation distortion in the progeny. These results add to our understanding of differences between indica and japonica rice and may aid in rice genetic improvement.

Reproductive isolation is both an indicator of speciation and a mechanism for maintaining species identity. The Dobzhansky-Muller model (1) suggests that hybrid incompatibility results from deleterious interactions between independently evolved loci from diverged populations. Studies in animal models such as Drosophila and mice have identified several of such interactive genes that cause hybrid incompatibility and segregation distortion (2, 3). In plants, hybrid sterility is a major form of postzygotic reproductive isolation, and several genes have been identified that conform to the Dobzhansky-Muller model for reproductive isolation (47). Hybrid sterility between indica and japonica subspecies of cultivated rice (Oryza sativa L.) is one example of postzygotic reproductive isolation in plants (810). Genetic analyses of indica-japonica hybrids have identified a large number of loci conditioning hybrid sterility (10). Several genes for indica-japonica hybrid sterility (1113) and interspecific hybrid sterility between O. sativa and O. glumaepatula (14) were recently cloned, aiding in our understanding of the biological processes of hybrid sterility in rice species.

S5 is a major locus for hybrid sterility in rice that affects embryo-sac fertility, as identified in a number of studies across a range of germplasms (11, 1519). The S5 locus has three alleles, an indica allele S5-i, a japonica allele S5-j, and a neutral allele S5-n (15). Hybrids of genotype S5-i/S5-j are mostly sterile, whereas hybrids of genotypes consisting of S5-n with either S5-i or S5-j are mostly fertile (1517). The S5 region has been mapped (18) and covers up to five open reading frames (ORF1 to ORF5). Transformation studies of ORF3 to ORF5 (11) from an indica variety into a japonica variety showed reduced fertility, due to embryo-sac abortion, for transformants harboring indica ORF5, whereas the fertility of transformants of ORF3 and ORF4 was not affected. The indica and japonica alleles of ORF5, which encodes an aspartic protease, differ by two nucleotides, whereas the wide compatibility allele has a large deletion in the N terminus of the predicted protein, causing subcellular mislocalization of the protein (11).

Because segregation distortion has been observed in progenies of indica-japonica hybrids (12, 19, 20), we assayed S5 genotypes of 195 seedlings from a BC6F1 plant BL(BL/NJ), a near isogenic line (NIL) heterozygous for the S5-i allele from an indica variety Nanjing 11, backcrossing successively with a japonica-variety Balilla. The resulting progeny showed genotypes deviating from the expected 1:2:1 ratio (Table 1). A maximum likelihood estimate for the frequency of S5-j transmitted via the female gametes was 0.1, compared with the expected 0.5. Similar segregation distortion was also observed in progenies from other indica-japonica crosses (Table 1).

Table 1

Segregation distortion at the S5 locus detected in F2 seedlings from various crosses.

View this table:

Because it is difficult to explain the hybrid sterility and segregation distortion by ORF5 alone, we determined genomic sequences of ORF1 to ORF4 for Nanjing 11 (indica), Balilla (japonica), Dular, and 02428 (the latter two are wide compatibility varieties that can produce highly fertile hybrids in crosses with either indica or japonica). Sequence polymorphisms with predictable functional changes among the genotypes in the predicted proteins were observed in ORF3 and ORF4 but not ORF1 or ORF2 (fig. S1). By investigating the transcripts, we observed that the translation start codons of ORF4 and ORF5 were located only 0.8 kb away, but the genes were transcribed in opposite directions. The ORF4 sequence of Balilla and 02428 was predicted to encode a protein with a transmembrane domain and had no homology with any known proteins (fig. S2). An 11–base pair (bp) deletion predicted to cause premature termination of the predicted protein and a loss of the putative transmembrane domain (fig. S2) was detected in ORF4 of Nanjing 11 and Dular relative to Balilla and 02428 (fig. S1). ORF3 was mapped 11.7 kb away from ORF4 and showed homology to a heat shock protein Hsp70 gene. The ORF3 sequences of Balilla and Dular have a 13-bp deletion relative to the other two genotypes (fig. S1), which results in a frameshift in the C terminus of the protein (fig. S3). On the basis of the sequence differences in these ORFs (fig. S1), we designated the ORF3 allele from Nanjing 11 and 02428 as ORF3+ and the other allele as ORF3–; the ORF4 allele from Balilla and 02428 as ORF4+ and the other one as ORF4–; and the ORF5 allele from Nanjing 11 as ORF5+, the one from Balilla as ORF5–, and those from Dular and 02428 as ORF5n.

We tested the effect of ORF3 on hybrid sterility by crossing a transgenic BalillaORF3+ plant [in which ORF3+ from Nanjing 11 was transformed into Balilla and showing normal fertility (Table 2)] with BL(NJ/NJ), a NIL in which the S5 fragment contains ORF3 to ORF5 from Nanjing 11 (ORF3+, ORF4–, and ORF5+) introgressed into a Balilla background (ORF3–, ORF4+, and ORF5–). A Balilla and BL(NJ/NJ) cross typically produces hybrid with reduced fertility (Table 1). However, BL/NJORF3+ plants from this cross showed 71.5% spikelet fertility, compared with 50.3% of the BL/NJ plants (Table 2). This rescue was confirmed in the progeny of heterozygous plants BL/NJORF3+, in which the fertility of BL/NJORF3+ plants (75.1%) was much higher than that of BL/NJ plants (46.8%). Therefore, we inferred that ORF3+ rescued fertility of the indica-japonica hybrid, presumably by protecting the gametes from the killing effect of ORF5+. Comparison between normal and ORF3+-rescued plants [BL/BL or NJ/NJ versus BL/NJORF3 (Table 2)] showed that the fertility-protecting effect of ORF3+ is only partial. We suspect that the independent transmission of the transformed hemizygous ORF3+ relative to the host S5 locus explains these observations because we would expect only approximately half of the gametes to inherit the ORF3+ transgene, which would be protected from killing by ORF5+.

Table 2

The effects of ORF3 and ORF5 on spikelet fertility.

View this table:

To support our hypothesis, we crossed BalillaORF3+ carrying a transformed ORF3+ and BalillaORF5+ carrying a transformed ORF5+ (Table 2). The progeny plants lacking any transgene were fully fertile, as expected, as were the ones carrying ORF3+ alone. Transgenic ORF5+ plants were sterile, whereas the addition of transgene ORF3+ rescued the fertility of the plants.

Transforming ORF4+ into BL(NJ/NJ) resulted in no fertility reduction in BL(NJ/NJ)ORF4+ transformants (Table 3), as expected, because of the presence of the protector ORF3+ in the introgressed fragment. Also, the fertility of hybrids involving the Dular fragment (ORF3–, ORF4–, and ORF5n) from the BL(BL/DL) × BL(BL/NJ) cross was normal, regardless of whether the allelic fragment was indica or japonica (Table 3). However, in the F1 progeny from a BL(NJ/NJ)ORF4+ × BL(DL/DL) cross, individuals with the transgene BL(DL/NJ)ORF4+ exhibited reduced spikelet fertility (39.3%), compared with the transgene-negative plants BL(DL/NJ) or the parental BL(NJ/NJ)ORF4+. Furthermore, statistically significant segregation distortion at the S5 locus was observed in the F2 progeny produced from BL(DL/NJ)ORF4+ plants (Table 3), in which the NJ fragment (estimated frequency 0.654) was favored at the cost of the DL fragment (0.346). Consequently, the frequency of DL/DL homozygote was deficient compared with the expected 1:2:1 ratio. Thus, the addition of ORF4+ in this hybrid resulted in the death of gametes with the Dular fragment.

Table 3

The effects of ORF4 and ORF5 on spikelet fertility and segregation distortion.

View this table:

To examine the role of ORF4+ in relation to ORF5+, we crossed BL(BL/DL) with BalillaORF5+ (Table 3). Among F1, the two genotypes with the transformed ORF5+ (BL/BLORF5+ and BL/DLORF5+) were mostly sterile, whereas their transgene-negative counterparts were fertile. This was also observed in the F2 segregants of this cross. Moreover, DL/DLORF5+, which is homozygous for the Dular genotype with an added ORF5+, showed normal fertility, and fertility of BL(BL/NJ) was not affected by the transformed ORF5+. Given that both Balilla and Dular had ORF3– and nonkiller ORF5, the fertility difference between BL/DLORF5+ and DL/DLORF5+ can be ascribed to differences in ORF4 between Balilla and Dular (figs. S1 and S2). Because both ORF5+ and ORF4+ are indispensable for gamete killing, ORF4 is apparently a partner in gamete killing with ORF5.

The results of genetic analysis for the S5-induced hybrid sterility presented above are schematically summarized in Fig. 1A. Female gametes in the indica-japonica hybrid are killed during sporogenesis by ORF5+ in partner with ORF4+ but protected by ORF3+.

Fig. 1

Schematic representation of the killer-protector system in an indica-japonica hybrid regulated by the S5 locus. (A) A genetic model depicting the process of megaspore formation and effects of the three genes, where 3+, 3–, 4+, 4–, 5+, and 5– represent ORF3+, ORF3–, ORF4+, ORF4–, ORF5+, and ORF5–, respectively, and colored blocks and circles represent the proteins. In the megaspore mother cell and daughter cells immediately after meiotic division, killing would not occur because of the presence of ORF3+. Killing would occur in the daughter cell carrying ORF3– and ORF4+ at a later stage of megaspore development. (B) Hypothetical molecular processes involving ER-stress and PCD. bZIP50-S, spliced bZIP50; ER, endoplasmic reticulum; PM, plasma membrane.

An expression database (21) indicated that both ORF4 and ORF5 show low expression—almost at the background level throughout the life cycle (fig. S4). ORF3 transcripts were more abundant, especially in developing panicles. Transient expression assays with rice protoplasts revealed that both ORF3+ and ORF3– proteins localized to the endoplasmic reticulum (ER) (fig. S5, A to H). ORF4+ localized to the plasma membrane and Golgi (fig. S5, Q to T), whereas ORF4– localized to the ER (fig. S5, I to P). ORF5 protein was found in the extracellular domain (11).

Microarray analysis of ovaries at the functional megaspore stage showed that ORF3 expression was statistically significantly higher in BalillaORF5+ transgene-positive than in -negative plants (table S1). We detected an ER stress-responsive UPRE-like cis-element (TGACGAGG) (22) in the promoter of ORF3 at –256 bp (fig. S1). Expression of a number of ER stress-responsive genes was also statistically significantly higher in BalillaORF5+ plants (table S1 and fig. S6A). This pattern of induction is highly similar to that observed in ER stress studies in rice (2325). As a response to stress (25), the ER stress sensor IRE1 transduces signals through the unconventional splicing of OsbZIP50 mRNA, which causes a frameshift producing a nuclear localization signal in the protein designated OsbZIP50-S, which regulates the expression of many ER stress-responsive genes, including ORF3. We confirmed that the spliced OsbZIP50-S mRNA was present in BalillaORF5+ plants (fig. S7). Taken together, these results suggested that introduction of ORF5+ into Balilla induced ER-stress in ovaries.

Bax inhibitor-1 (BI1) is a conserved ER-resident cell death suppressor in eukaryotes and plays an important role in modulating the ER stress-mediated programmed cell death (PCD) pathway both in Arabidopsis and rice (26, 27). Our analysis (fig. S6B) showed that OsBI1 was up-regulated in BalillaORF5+ plants. OsKOD1 [an orthologous gene of kiss of death (KOD), which acts as a PCD-inducer in Arabidopsis] (28) and Hsr203j (the commonly used cell death marker) (24, 29) were also up-regulated by ORF5+. Expression of OsCP1, which acts as an executor of the PCD process in rice tapetum development (30, 31), was elevated by ~12-fold in ovaries of BalillaORF5+ plants. In addition, many differentially expressed genes between BalillaORF5+ transgene-positive and -negative plants revealed in the microarrays (table S2) were PCD-related, such as the cytochrome P450, LTPL, and GDSL gene families (31). These results suggested that ORF5+-induced ER stress might provoke abnormal PCD in embryo-sac development. We performed a terminal deoxynucleotidyl transferase-mediated deoxy-uridine 5′-triphosphate (dUTP) nick-end labeling (TUNEL) assay for PCD by detecting DNA fragmentation during female sporogenesis (fig. S8). No cellular abnormality or TUNEL signal was observed before meiosis (fig. S8, A and B). Cellular abnormality was observed in megasporocyte undergoing meiosis and afterward in BalillaORF5+ (fig. S8, C to J). TUNEL signal occurred earlier and stronger in BalillaORF5+ than in Balilla (fig. S8, C to J). Thus, premature PCD occurred during female sporogenesis in BalillaORF5+, resulting in embryo-sac abortion. In contrast, introduction of ORF3+ into BalillaORF5+ plant by crossing BalillaORF5+ with homozygous BalillaORF3+ restored normal expression levels of the ER stress-responsive and PCD-related genes in the hybrid (fig. S9). This implies that ORF3+ is a suppressor of ORF5+-induced ER stress and subsequent PCD.

On the basis of these results, we hypothesize (Fig. 1B) that the activity of extracellular ORF5+ produces a molecule that is sensed by plasma membrane–localized ORF4+ and eventually triggers ER stress. The ER stress subsequently actuates the IRE1-mediated splicing of OsbZIP50 mRNA, producing OsbZIP50-S, a transcription factor that turns on expression of ER stress-responsive genes, including ORF3. The ER stress is resolved in the presence of ORF3+, thus producing normal female gametes. Whereas in the absence of ORF3+, unresolved ER stress induces PCD-related genes, causing anomalous PCD, which leads to embryo-sac abortion, despite the presence of OsBI1. Thus, proteins encoded by ORF3, ORF4, and ORF5 are elements involved in different stages of the ER stress-induced PCD pathway regulating hybrid fertility.

We obtained sequences of ORF3, ORF4, and ORF5 for 82 accessions of O. sativa, O. rufipogon, and O. nivara from 16 countries over a diverse geographical area (table S3). Nineteen haplotypes were identified containing single-nucleotide polymorphisms (SNPs) and insertions/deletions (Indels) in the coding sequence of ORF3 (fig. S10). Four of the haplotypes could be placed in the ORF3– allele group, and 15 classified into the ORF3+ allele group. Five of the 18 haplotypes detected for ORF4 were classified into the ORF4– group and 13 into the ORF4+ group. Together with the three allelic groups identified in ORF5 (ORF5+, ORF5–, and ORF5n) (32), there are a total of 12 possible combinations formed of the three genes. We observed 9 of the 12 combinations (table S4). ORF3+, ORF4+, and ORF5+ was the most common, especially in the two wild rice species O. rufipogon and O. nivara. This represents a balance between killing and protecting the gametes, according to our genetic model. The sequence of the outgroup O. glumaepatula suggested that ORF3+, ORF4+, and ORF5+ is the ancestral type. The suicidal combination of ORF3–, ORF4+, and ORF5+, which would not be able to survive in nature, was not observed in the sample, although it would be the easiest to be generated at the population level through mutation and/or recombination. Two other combinations (ORF3–, ORF4–, and ORF5+ and ORF3–, ORF4–, and ORF5–) were not detected because of either their rarity or the source of sample. This result was congruent to the proposed genetic model of the killing-protecting system. The typical indica-like (ORF3+, ORF4–, and ORF5+) and japonica-like (ORF3–, ORF4+, and ORF5–) types were found in wild rice accessions, suggesting that the ancestors of indica and japonica rice probably originated before domestication. Because indica and japonica rice (also indica-like and japonica-like wild rice) have distinct ranges of distribution worldwide, geographical isolation might have played an important role in maintaining distinctions of the rice groups as well as the killer-protector system. This killer-protector system may have a profound implication in the evolution and diversification of rice. Reproductive isolation enforced by the killer (ORF4+, ORF5+) would have promoted genetic differentiation between indica and japonica rice, which appears to be a major source of genetic diversity in the rice gene pool, whereas the protector (ORF3+) and nonkiller combinations of ORF4 and ORF5 would allow for hybridization and gene flow, thus providing a coherent force at the species level.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 to S7

References (3341)

References and Notes

  1. Acknowledgments: We thank D. S. Brar of the International Rice Research Institute for providing the rice seeds, and S. Luan of University of California, Berkeley, USA for discussion. This research was supported by grants from the National Natural Science Foundation (31130032 and 30921091), the 863 Project (2012AA100103), and the 111 Project (B07041) of China. All of the DNA sequences obtained in this study have been deposited in the GenBank from accession codes JX138498 to JX138505. A patent for the ORF5 sequence has been approved by the State Intellectual Property Office of China (ZL200710053552.9).
View Abstract

Stay Connected to Science

Navigate This Article