A selfish genetic element confers non-Mendelian inheritance in rice

See allHide authors and affiliations

Science  08 Jun 2018:
Vol. 360, Issue 6393, pp. 1130-1132
DOI: 10.1126/science.aar4279

Sterility in rice via toxin and antidote

Crossing wild and domestic rice often results in hybrid sterility. Such genetic barriers can prevent the movement of potentially beneficial genes from wild rice into domestic varieties. To understand the barriers preventing gene flow, Yu et al. mapped a quantitative trait locus (QTL) that determines sterility between wild-type and domestic rice. This QTL encodes two open reading frames (ORFs) that are both expressed during gametogenesis. The ORFs encode a toxin, which affects the development of pollen, and an antidote, which is required for pollen viability. Thus, selfish genetic elements can underlie evolutionary strategies that facilitate reproductive isolation.

Science, this issue p. 1130


Selfish genetic elements are pervasive in eukaryote genomes, but their role remains controversial. We show that qHMS7, a major quantitative genetic locus for hybrid male sterility between wild rice (Oryza meridionalis) and Asian cultivated rice (O. sativa), contains two tightly linked genes [Open Reading Frame 2 (ORF2) and ORF3]. ORF2 encodes a toxic genetic element that aborts pollen in a sporophytic manner, whereas ORF3 encodes an antidote that protects pollen in a gametophytic manner. Pollens lacking ORF3 are selectively eliminated, leading to segregation distortion in the progeny. Analysis of the genetic sequence suggests that ORF3 arose first, followed by gradual functionalization of ORF2. Furthermore, this toxin-antidote system may have promoted the differentiation and/or maintained the genome stability of wild and cultivated rice.

Identifying so-called speciation genes that cause reproductive isolation is a central goal in evolutionary biology. Postzygotic reproductive isolation (PRI), embodied by hybrid sterility, inviability, or weakness, drives speciation and maintains species identity by restricting gene flow between populations (1, 2). The Bateson-Dobzhansky-Muller model postulates that hybrid reproductive isolation results from deleterious interactions among at least two loci from evolutionarily divergent populations (3). Accumulating evidence suggests that selfish genetic elements (SGEs), DNA sequences that gain a transmission advantage relative to the rest of the genome, could drive genome evolution by causing hybrid incompatibilities and segregation distortion in different organisms (410); however, the role of SGEs in genome evolution and their underlying molecular mechanisms have remained obscure.

We performed quantitative trait locus (QTL) analysis of a backcross F1 (BC1F1) population derived from the cross between two highly divergent rice species—wild rice, O. meridionalis accession 82031 (Mer), and O. sativa ssp. japonica, Dianjingyou1 (DJY1)—and detected four major QTLs controlling hybrid male sterility (fig. S1, A and B). Of these, qHMS7 is located near several previously identified QTLs (fig. S1C).

To clone the causal gene(s) for qHMS7, we developed a near-isogenic line for qHMS7 (NIL-qHMS7) (fig. S1D). Intriguingly, the NIL-qHMS7 plant can only be maintained in a heterozygous status at the qHMS7 locus in the DJY1 background. Examination of the self-pollinated progeny of NIL-qHMS7 (BC6F2) revealed a bimodal distribution for pollen fertility (fig. S2). Genotyping revealed that the semi-sterile plants were of the DJY1/Mer genotype (D/M) and the fully fertile plants were of the DJY1/DJY1 genotype (D/D) at the qHMS7 locus. No Mer/Mer type (M/M) plant was detected. The seed-set rate of all 50 randomly selected plants was above 85% (fig. S3), indicating that this locus did not affect female gametes and seed setting. These results suggest that qHMS7 acts as a single locus conferring male semi-sterility and that the Mer-type pollens were not transmissible to the progeny. This notion was further supported by a gametophytic transmission assay using three genetic populations of DJY1 and NIL-qHMS7 (table S1). Histological and cellular examination of various stages of pollens revealed that Mer-type pollens aborted before the tricellular stage in NIL-qHMS7 (Fig. 1, A to E, and figs. S4 to S7), possibly owing to defects in the second mitosis that produces tricellular mature pollen grains.

Fig. 1 Identification of qHMS7.

(A to D) Morphology (upper panels) and stained pollen (lower panels) of DJY1, Mer, hybrid F1 (Mer × DJY1), and NIL-qHMS7. Fertile pollen is stained dark and sterile pollen is not stained. Scale bars: 50 cm in the upper panels and 50 μm in the lower panels. (E) Pollen fertility of DJY1, Mer, hybrid F1, and NIL-qHMS7, shown as mean ± SD (n = 5, 5, 7, and 10 plants examined, respectively). **P < 0.01 (versus DJY1, by Student’s t test). (F) ORFs predicted in the fine-mapped region. The insertion of ORF3 in DJY1 is indicated with a gray triangle. Arrows indicate the orientation of ORFs.

The gene(s) underlying qHMS7 locus was delimited to a 31.6-kb genomic interval by the use of a map-based cloning strategy (fig. S8 and table S2). Sequencing analysis identified three genes in the qHMS7 region in DJY1 (ORF1D, ORF2D, and ORF3), but only two genes in Mer (ORF1M and ORF2M) (Fig. 1F). ORF1D, ORF1M, and ORF3 are predicted to encode homologous, grass family–specific proteins with a mitochondrial targeting signal at the N terminus, whereas ORF2D and ORF2M are predicted to encode a ribosome-inactivating protein (RIP) domain (11) containing protein conserved in monocots, with a putative bipartite nucleus localization signal and a nucleus exporting signal. The ORF2M and ORF2D proteins differ at 11 polymorphic sites (figs. S9 to S13).

Quantitative reverse transcription–polymerase chain reaction analysis revealed that expression of both ORF2 and ORF3 was significantly higher in stages 11 to 13 of anther development (fig. S14, A and B). ORF2 expression was lower in NIL-M/M plants [near-isogenic line with M/M genotype at the qHMS7 locus in DJY1 background (see below)] at mature pollen stage, compared to that in DJY1 (fig. S14C). As expected, ORF3 expression in DJY1 was about twice as high as that in NIL-qHMS7 at mature pollen stage, and no expression of ORF3 was detected in NIL-M/M plants (fig. S14D). Transient expression of green fluorescent protein fusions of ORF2D, ORF2M, and ORF3 showed that ORF2D and ORF2M are localized to both nucleus and cytoplasm (fig. S15), whereas ORF3 is localized to the mitochondria (fig. S16).

Hybrid sterility is often controlled by toxin-antidote or killer-protector systems that subvert Mendel’s law of segregation (12). The selective abortion of Mer-type pollens in NIL-qHMS7 but not in Mer suggests that ORF3 may encode a protein that functions as an antidote to a pollen-killing toxin produced in NIL-qHMS7. To test this notion, we transformed an intact ORF3 genomic fragment from DJY1 into D/M-type calli. As expected, primary (T0) transgenic plants harboring a single-copy transgene of ORF3 (genotype D/M; ORF3/-) showed a partial restoration of pollen fertility (~75%), compared to the transgene-negative plants (~50% fertility). No effect on pollen fertility was observed in the T0 transformants of ORF1D or ORF2D (table S3). In addition, T1 plants (derived from single-copy T0 transformants of ORF3) of the genotype (D/M; −/−) were semi-sterile, whereas the T1 plants of the genotype (D/M; ORF3/-) showed ~75% fertility, and the T1 plants of the (D/M; ORF3/ORF3) genotype showed normal fertility (~95%). Notably, we recovered T1 plants of the (M/M; ORF3/-) and (M/M; ORF3/ORF3) genotypes, and both were fully fertile (Fig. 2 and table S4). These results indicate that ORF3 has a protective function for Mer-type pollen carrying it (thus acting in a gametophytic manner).

Fig. 2 ORF3 performs a pollen-protection function in a gametophytic manner.

(A) A schematic of expected genotypes of the gametes from the ORF3 T0 transgenic plants (genotype D/M, ORF3/-) and the resulting T1 progeny from selfing. T1 plants of the (M/M; ORF3/-) and (M/M; ORF3/ORF3) genotypes are highlighted in green ovals. (B) Genotype and pollen fertility of the T1 plants from a selfed T0 transgenic plant (genotype D/M; ORF3/-; line Q040). Scale bars, 50 μm. The observed and expected plant numbers of the corresponding genotype among 162 T1 progeny are shown by N and (NE), respectively. Pollen fertility of the T1 plants of various genotypes is shown as mean ± SD. The genotype and pollen fertility of two additional T1 families are shown in table S4.

The selective abortion of Mer-type pollens in NIL-qHMS7 but not in Mer also suggests that the pollen killer is most likely encoded by the DJY1 allele(s) at the qHMS7 locus. Consistent with this notion, the T2 progeny derived from selfed (M/M; ORF3/-) type T1 plants segregated with an expected ratio of 1:2:1 for the genotypes (M/M; −/−), (M/M; ORF3/-) and (M/M; ORF3/ORF3), and they were all normally fertile (fig. S17A and table S5). To identify the gene(s) encoding the predicted toxin, we individually transformed “ORF1D,” “ORF2D,” and “ORF1D+ORF2D” into calli derived from the seeds of the NIL-M/M plants. Transgenic plants harboring “ORF2D” or “ORF1D+ORF2D” were completely male sterile, whereas the “ORF1D” transgenic plants or transgene-negative plants were normally fertile (fig. S17, B and C, and table S6). These observations suggest that ORF2D has a pollen-killing function and acts in a sporophytic manner (i.e., the pollen-killing function is determined by the parental genotype). This notion was further supported by the observed normal pollen fertility in transgenic NIL-qHMS7 plants in which ORF2D was knocked out via CRISPR-Cas9 (fig. S18).

To trace the evolutionary origins of ORF2 and ORF3, we compared genomic sequences among a diverse group of AA-genome rice species (table S7). A single haplotype of ORF3 was identified in 50.66% of O. rufipogon and 94.76% of Asian cultivated rice accessions, but not in other AA-genome rice species, including O. meridionalis, O. longistaminata, O. barthii, and O. glaberrima (Fig. 3A and table S7). In comparison, ORF2 was found in all accessions that we examined. Besides ORF2M and ORF2D, 25 additional haplotypes were identified (we collectively termed these haplotypes ORF2N) (fig. S19). As ORF3 is homologous to ORF1 and ORF1 was identified in all sequenced accessions, we speculated that ORF3 might have derived from an ancient gene duplication event in a subpopulation of O. rufipogon earlier than the functional ORF2D. As O. meridionalis is a basal lineage of AA-genome rice and ORF2M has the most amino acid substitutions compared to ORF2D, ORF2M might represent the ancestral sequence from the common ancestor(s) of AA-genome wild rice. The observation that all haplotypes of ORF2N either lack the protective ORF3 or contain polymorphisms suggests that they should be nonfunctional in pollen killing, similar to ORF2M. This notion was supported by the detection of putative QTLs for qHMS7 between DJY1 and different AA-genome species that carry ORF2M or ORF2N but lack ORF3 (table S8). We further deduced that it is likely that ORF2M gradually evolved into ORF2N and later evolved into ORF2D in some subpopulations of O. rufipogon through a multistep process (Fig. 3A and fig. S19). It should be noted that although the above data support such a gradual evolution model, the possibility of nonfunctional forms derived from functional ORF2 and ORF3 preexisting in the ancestors of wild rice (a degenerative model) cannot be completely excluded yet.

Fig. 3 The evolutionary origin and a genetic action model of ORF2 and ORF3.

(A) The ancestral wild rice has an ancestral genotype of ORF2, but lacks ORF3. On the basis of haplotype combinations of ORF2 and ORF3, AA-genome wild rice and cultivated rice can be grouped into four types (I to IV). Their frequencies are shown under each type. ORF2M, ORF2N, and ORF2D are shown in gray, purple, and wine red, respectively, and ORF3 is shown in green. (B) A diagram shows that in the hybrids of type IV O. rufipogon and Asian rice (genotype: 2D/2D; 3D/3D) with other ancestral AA-genome wild rice or cultivated rice (genotypes 2M/2M; -/- or 2N/2N; -/-), pollen of the genotype 2M; - or 2N; - is aborted owing to the lack of protective ORF3.

Our combined results suggest that the toxin-antidote system encoded by ORF2 and ORF3 constitutes an SGE in rice that confers a transmission advantage to the ORF3-carrying male gametes (fig. S20). In the hybrids of different types of wild rice and cultivated rice, pollens of the genotype (2M; -) or (2N; -) are aborted owing to the lack of protective ORF3 (table S8, Fig. 3B, and fig. S20). Our results suggest that qHMS7 may play a role in promoting the differentiation and/or maintaining the genome stability of wild and cultivated rice by restricting gene flow between them. Further, the observed nearly complete male sterility of the DJY1 × Mer hybrid suggests that the combined action of qHMS7 and other hybrid incompatibility loci could effectively prevent gene flow between rice populations (Fig. 1C and fig. S1B). These findings are consistent with and in support of the proposition that SGEs may affect the formation of new species (13, 14).

RIPs are toxic RNA N-glycosidases that affect translation processes and have been implicated in apoptotic pathways in mammalian cells and antiviral, antifungal, and insecticidal activities in plants (11). Recent studies also demonstrated that RIPs can induce apoptosis through mitochondrial cascade independent of translation inhibition (15). The finding that ORF2 encoding a RIP domain–containing protein localized to the nucleus and cytoplasm and that ORF3 encoding an antidote localized to the mitochondria reaffirms a critical role of mitochondria and possibly energy supply for male sterility regulation in plants (16). Although the exact spatiotemporal interaction between the “toxin” and “antidote” remains to be elucidated, the identification of qHMS7 as an SGE regulating PRI in rice adds new evidence supporting a general role of such elements in eukaryote genome evolution and speciation. Notably, several previously reported genes (such as S5, Sa, and Sc) regulating PRI in rice also embody characteristics of SGEs (7, 8, 10, 17). As wild rice is an important germplasm resource for hybrid rice breeding, our findings may also offer approaches to overcome male sterility in wild rice–cultivated rice hybrids, thus facilitating utilization of the strong hybrid vigor. In addition, a toxin-antidote system might be modified to fend off infestations of weedy rice, which causes economic losses and poses a threat to ecosystems and biodiversity. Thus, continued efforts to identify SGEs and elucidate their mechanisms have important implications in agriculture.

Supplementary Materials

Materials and Methods

Figs. S1 to S20

Tables S1 to S9

References (1832)

References and Notes

Acknowledgments: We thank Q. Yang and L. Han for providing rice germplasm and K. Olsen for discussion. Funding: Supported by National Key Research and Development Program of China (2016YFD0100301), National Natural Science Foundation of China (U1502265), and the Agricultural Science and Technology Innovation Program of the Chinese Academy of Agricultural Sciences. Author contributions: X.Y. and Z.Z. performed all experiments and analyzed data. X. Zheng, J.Z., W.K., P.W., W.B., H. Zheng, H. Zhang, J. Li, J. Liu, Q.W., L.Z., K.L., Y.Y., X.G., J. Wang, Q.L, F.W., Y.R., S.Z., X. Zhang, Z.C., C.L., S.L., X.L., Y.T., L.J., and S.G. performed some of the experiments. J. Wan, H.W., C.W., and D.T. supervised the project and wrote the paper. Competing interests: J. Wan, X.Y., Z.Z., X. Zhang, L.J., S.Z., X.L., S.L., and Y.T. are inventors on patent application (no. 201810164610.3) filed in the State Intellectual Property Office of China by Nanjing Agricultural University that covers ORF3 and ORF2 sequences of the qHMS7 locus and their applications. Data and materials availability: All data are available in the manuscript or the supplementary materials. All of the DNA sequences obtained in this study have been deposited in GenBank (accession numbers MG865759 to MG865763).

Stay Connected to Science

Navigate This Article