Report

An SNP Caused Loss of Seed Shattering During Rice Domestication

See allHide authors and affiliations

Science  02 Jun 2006:
Vol. 312, Issue 5778, pp. 1392-1396
DOI: 10.1126/science.1126410

Abstract

Loss of seed shattering was a key event in the domestication of major cereals. We revealed that the qSH1 gene, a major quantitative trait locus of seed shattering in rice, encodes a BEL1-type homeobox gene and demonstrated that a single-nucleotide polymorphism (SNP) in the 5′ regulatory region of the qSH1 gene caused loss of seed shattering owing to the absence of abscission layer formation. Haplotype analysis and association analysis in various rice collections revealed that the SNP was highly associated with shattering among japonica subspecies of rice, implying that it was a target of artificial selection during rice domestication.

Cultivation of major cereals likely started about 10,000 years ago (14). During this domestication, ancient humans subjected several key events to selection. These included increase in the number of seeds, improvement of fertility, change in plant architecture, change in seed shape, adaptation of flowering time to local areas, loss of seed color, and loss of seed shattering.

Recent studies in rice have revealed that several independent domestication events might have occurred to establish cultivated rice (3, 57). The archaeological record reveals that japonica rice, a subspecies of Oryza sativa, was bred about 10,000 years ago in the upstream regions of the Yangtze River in southwest China (3, 8, 9).

The loss of seed-shattering habit is thought to be one of the most important events in rice domestication, because the “easy-to-shatter” trait in wild relatives results in severe reduction in yield. Over the course of human history, distinct grain-threshing systems have been developed in several different eras in local areas of the world, in accordance with the degree of seed shattering. In current rice-breeding programs, this seed-shattering habit is still a target, especially in the construction of new indica (another subspecies of O. sativa) cultivars. Thus, seed-shattering habit is one of the most important agronomic traits in rice cultivation and breeding.

It is difficult to obtain shattering-related mutants and reveal the underlying molecular mechanisms, because most rice cultivars have somehow lost the seed-shattering habit. We therefore used natural variations in seed shattering among cultivars. Generally, indica cultivars exhibit relatively strong seed shattering, whereas some japonica cultivars do not exhibit it at all (Fig. 1A). We first performed a QTL (quantitative trait locus) analysis between a shattering-type indica cultivar, Kasalath, and a nonshattering-type japonica cultivar, Nipponbare. The seed-shattering degree (breaking tensile strength) of each grain was measured (10), and the average value was scored for QTL analysis.

Fig. 1.

qSH1 is required for formation of the abscission layer at the base of the rice grain. (A) Seed-shattering habits of rice panicles. Photos taken after grabbing rice panicles. (Left) Nonshattering-type cultivar, Nipponbare. (Right) Shattering-type cultivar, Kasalath, in which the seed has shattered. (B) Chromosomal locations of QTLs for seed-shattering degree, based on an F2 population from a cross between Nipponbare and Kasalath. Positions of circles indicate positions of QTLs, and circle size indicates the relative contribution of each QTL. Red circles, Nipponbare alleles contributing to nonshattering habit; blue circles, Kasalath alleles contributing to nonshattering. qSH1 is marked on chromosome 1 with the nearest DNA marker (C434). (C) Non–seed-shattering habits of Nipponbare, Kasalath, and NIL(qSH1). Breaking tensile strength upon detachment of seeds from the pedicels by bending and pulling was measured (10). Increase in value indicates loss of shattering. NIL(qSH1), a nearly isogenic line carrying a Kasalath fragment at the qSH1 locus in the Nipponbare background, as shown in fig. S1A. (D) Photo of a rice grain. White box indicates position of abscission layer formation. (E to G) Nipponbare. (H to K) Kasalath. (L to N) NIL(qSH1). (E), (H), and (L) Longitudinal sections of positions corresponding to white box in (D). Arrows point to the partial abscission layer of Kasalath in (H), the complete abscission layer of NIL(qSH1) in (L), and the corresponding region of Nipponbare in (E). (F), (I), and (M) Scanning electron microscope (SEM) photos of pedicel junctions after detachment of seeds. (G), (J), (K), and (N) Close-up SEM photos corresponding to white boxes in (F), (I), and (M). (G) Broken and rough surface of Nipponbare when forcedly detached. (N) Peeled-off and smooth surface of NIL(qSH1) upon spontaneous detachment. In Kasalath, rough center surface (K) and smooth outer surface (J) are observed. Scale bars: 500 μm in (E), (H), and (L); 100 μm in (F), (I), and (M); 10 μm in (G), (J), (K), and (N).

Five QTLs were detected on five chromosomes of rice in an F2 population of a cross between Kasalath and Nipponbare (Fig. 1B). Nipponbare alleles at three QTLs on chromosomes 1, 2, and 5, and Kasalath alleles at two other QTLs on chromosomes 11 and 12, all contributed to shattering reduction, suggesting that loss of seed shattering may occur independently in japonica and indica.

The QTL with the largest effect, termed QTL of seed shattering in chromosome 1 (qSH1), explained 68.6% of the total phenotypic variation in the population (Fig. 1B). We therefore made a near-isogenic line (NIL) (fig. S1A) that contained a short chromosomal segment from Kasalath at the qSH1 region in a Nipponbare genetic background. The NIL exhibited the formation of a complete abscission layer between pedicel and spikelet at the base of the rice seed (Fig. 1, D and L to N) and had a stronger seed-shattering phenotype than either Kasalath or Nipponbare (Fig. 1C). In contrast, no abscission layer was observed in Nipponbare at all (Fig. 1, E to G). This indicated that a mutation in the qSH1 gene alone resulted in complete loss of the abscission layer in the Nipponbare genetic background and that the Kasalath allele of qSH1 could rescue it. Kasalath could form a partial abscission layer only at the peripheries in the transverse plane (Fig. 1, H to K), perhaps because of the presence of the minor QTLs (Fig. 1B).

A large-scale linkage analysis of 10,388 plants segregating at the qSH1 region (fig. S1, B and C) was performed for the fine mapping of qSH1. We finally succeeded in mapping the functional natural variation in 612 bp between the flanking markers qSH1-F and qSH1-H and found only one single-nucleotide polymorphism (SNP) within this region (Fig. 2A). We confirmed this result using several recombinant homozygous plants in the progeny (Fig. 2A). Gene prediction for the qSH1 region in both Nipponbare (11) and Kasalath genome sequences showed no distinct open reading frame (ORF) in the SNP region. However, located 12 kb away from the SNP, we found one ORF (locus ID Os01g0848400 in the Rice Annotation Project DataBase) for a rice ortholog of the Arabidopsis REPLUMLESS (RPL) (12, 13) gene (Fig. 2B and fig. S2). The RPL gene encodes a BEL1-type homeobox (14, 15) and is involved in the formation of a dehiscence zone (or abscission layer) alongside the valve in the Arabidopsis fruit (silique). Because the fruit originates from the carpels in Arabidopsis, the botanical origin of the dehiscence zone in Arabidopsis fruit does not correspond to that of the abscission layer in rice seeds. However, it was still possible that this RPL ortholog was the qSH1 gene. To confirm this, we introduced ten 10- to 26-kb Kasalath genomic fragments scanning the predicted ORF and the SNP regions into the nonshattering Nipponbare cultivar (Fig. 2C and figs. S3 and S7). Only transgenic lines that contained the Kasalath fragment with both the ORF and the SNP exhibited complete seed shattering, although one fragment (termed sub51), which contained a full ORF region but not the SNP, partly complemented the phenotype (fig. S3). The other fragments were not able to complement it, even if they contained the entire ORF region or the SNP region. These results indicated that both the ORF and the SNP regions were required for full shattering function.

Fig. 2.

Fine mapping, identification of FNP, cloning of qSH1, and qSH1 expression. (A) Top left: C434 was the nearest marker to qSH1 upon the rough mapping. Markers C283 and R3265 were used to select recombinants near the qSH1 locus. (Bottom) Graphical genotypes of four selected recombinant homozygous lines and their nonshattering degrees. B and A are homozygous for Kasalath and Nipponbare, respectively. NIL(qSH1), B homo, A homo, and Nipponbare are control lines. (Top right) The single SNP in the 612-bp region. The typical RY repeat position is underlined. (B) Neighbor-joining phylogenetic tree of BEL1-type homeobox genes found in Arabidopsis and rice genomes. STM and OSH1 are outgroups. The region contains only the homeobox domains used for generating the tree (figs. S2C and S6). Rice and Arabidopsis genes are in red and blue type, respectively. (C) Complementation test for qSH1 gene. A 26-kb Kasalath fragment (TAC9) in TAC vector, pYLTAC7 (30), was transformed into Nipponbare. (Top) Nonshattering degrees of T0 plants were measured. (Bottom) Dots and crosses indicate DNA markers used to confirm the transformed and nontransformed parts, respectively, of the 26-kb fragment in each line. Lines 203 and 204 were partly transformed, because these lines lost the ORF region upon transformation. (D) to (I) In situ analysis of qSH1 expression. An 870-bp fragment hybridized specifically to the qSH1, not to a paralog in Fig. 2B, was used as a probe for this analysis. qSH1 expression was detected at shoot apical meristems in both Nipponbare (D) and NIL(qSH1) (G) upon floral transition (stage In1) (16). At the flower-formation stage, qSH1 expression was detected at the anther regions in both NIL(qSH1) and Nipponbare (E, H) and at the provisional abscission layer position only in NIL(qSH1) (H for stage In7, I for stage In8) and not in Nipponbare (E for stage In7, F for stage In8). Scale bars: 100 μm in (D), (F), (G), and (I); 200 μm in (E), and (H). Arrows point to the meristems in (D) and (G) and to the (provisional) abscission layers in (E), (F), (H), and (I).

In situ hybridization analysis revealed that in the NIL the ORF was expressed at the inflorescence meristem in the stage of rachis meristem establishment [inflorescence stage 1 (In1)] (16) (Fig. 2G). It was also expressed at both the anther region and the provisional abscission layer at the base of the spikelet in the stage of floral organ differentiation (In7) (Fig. 2H) and in the stage of rapid elongation of the rachis and branches (In8) (Fig. 2I). The abscission layer was not yet observable in In7. On the other hand, in Nipponbare, the ORF was expressed in the same way as in the NIL (Fig. 2, D to F), except that it was not expressed at the provisional abscission layer in either In7 or In8 (Fig. 2, E and F).

These results, together with the complementation results, led us to conclude that this RPL ortholog was the qSH1 gene and that the identified SNP affected only the spatial mRNA expression pattern of qSH1 at the abscission layer. A quantitative RT-PCR for RNA samples of developing panicles supported this conclusion (fig. S8). Consistently, a cis-element search revealed that the Kasalath qSH1 allele contained a typical RY-repeat (17) that was a binding site of the ABI3 (VP1)–type (18, 19) transcription factor at the SNP site (Fig. 2A and fig. S4). Therefore, the change in the transcriptional control of key genes such as RPL and qSH1 could explain the difference in abscission layer formation between rice and Arabidopsis. Several genes downstream of RPL have been identified, such as SHP1/2 genes (20) belonging to the AG-clade MADS box genes in Arabidopsis. Phylogenetic analysis has revealed that SHP genes evolved after the eu-dicots separated from the common ancestors of eu-dicot and monocot plants (21), and all the Arabidopsis AG-clade MADS box genes are expressed in the carpel regions (22). In addition, it has been recently shown that two AG orthologs have evolutionally conserved functions with AG in rice (23). Thus, it is very likely that no functionally related ortholog of SHP1/2 exists in the rice genome. Therefore, it is also possible that qSH1 expression may lead to formation of the abscission layer at the base of the seed by a mechanism distinct from that of dehiscence zone formation in Arabidopsis fruit.

Hence, we believe that, like the RPL in Arabidopsis (24), qSH1 may have pleiotropic functions in the spikelet development and plant architecture of rice, as well as in abscission layer formation. Therefore, null or severe mutations in the ORF may cause serious defects in rice development to establish it as a cultivar. This type of SNP, which caused loss of qSH1 mRNA expression only at the abscission layer, could be survived during rice domestication. Similarly, it has been proposed that several maize domestication genes contain critical polymorphisms that have resulted from prehistoric artificial selection in the 5′ regulatory regions, although the functional nucleotide polymorphisms have not yet been identified (25, 26).

To address how the SNP in qSH1 prevailed during rice domestication, we next analyzed rice core collections (27) (Fig. 3, A and B). The results revealed that the SNP was highly associated with the degree of seed shattering among temperate japonica rice cultivars (a subgroup of japonica) (28) and implied that this SNP had been a target of artificial selection for nonshattering habit during rice domestication (Fig. 3E). All tested indica cultivars exhibited strong seed shattering; this result was consistent with the fact that they all contained the functional SNP (fig. S5C). Other QTLs need to be considered to explain the differences among tropical japonica cultivars (the other subgroup of japonica) (fig. S5D).

Fig. 3.

Association of qSH1 haplotypes with degree of shattering. (A) The four genomic regions with DNA polymorphisms at the qSH1 locus are shown as thick black rectangles. Japonica has two subgroups, tropical japonica and temperate japonica (28). Only polymorphisms found in the population of temperate japonica cultivars are presented in (B), (C), and (D), with the exception of SNP3. SNP3 is present to show the lack of polymorphism at this site in all japonica cultivars tested, although the SNP3 found in Kasalath caused one amino acid change in qSH1, which was the sole amino acid change found between the ORFs of Nipponbare and Kasalath (fig. S5). SNP3 was not a target for human selection during rice cultivation. (B) Cultivars in temperate japonica core collections selected by genome-wide RFLP (27) and/or SSR (simple repeat sequence) analysis. U, upland-type cultivars. (C) Temperate japonica cultivars of Chinese origin. N and Y, North and Yunnan, respectively. These cultivars were assigned to temperate japonica by genome-wide RFLP analysis. No indications of RFLP and/or SSR in the core column mean the cultivars were not selected as core collections. (D) Nipponbare and Kasalath controls. At right in (B), (C), and (D), nonshattering degrees were also examined. (E) Statistical analysis of the association of seed shattering with genotype. ANOVA analysis was done with data shown in (B). AT repeat 2 and qSH1/SNP, but not SNP7, showed significant associations with seed shattering in temperate japonica cultivars. Standard errors are also shown in the graph.

Rice cultivation likely started about 10,000 years ago; paddy-style rice cultivation is believed to have started in the Yangtze River region of China about 7000 years ago and to have been imported into Japan about 3000 years ago (3, 8). We therefore analyzed rice cultivars of Chinese origin (Fig. 3C) and found that the nonshattering SNP at qSH1 might have been used in ancient China 3,000 to 10,000 years ago, most likely about 7,000 years ago upon the establishment of paddy-style rice cultivation.

Crop domestication might have proceeded during relatively short periods (less than 10,000 years) through the occurrence of nucleotide polymorphisms, such as by spontaneous mutation, recombination, and fixation in populations. Because rice is a self-pollinated plant, such newly occurring nucleotide polymorphisms would have easily become fixed in individuals. If such individuals propagated and contributed to the establishment of cultivated rice, we should be able to follow step by step the haplotype changes that occurred during rice domestication. Therefore, we examined the haplotypes around the qSH1 gene in the rice collections (Fig. 4A). The identified SNP was likely to be assigned as a mutation that occurred in early domesticates of japonica subspecies (Fig. 4B) but not as a preexisting natural variation. In the hypothetical process of evolution of qSH1, the SNP distribution clearly revealed a strong selection by ancient humans for the SNP during rice domestication (Fig. 4B). In addition, the estimated haplotype of the common ancestor at the qSH1 locus was found in a wild rice accession, W1943 (Fig. 4B), which is closely related to the japonica subspecies (6). We could therefore follow how domestication proceeded at the level of DNA sequence change, from ancestors to cultivated rice. Many agronomic traits are related to domestication events and could have been the targets of artificial selection during domestication. Therefore, this type of evolutionary analysis may give us some insights into the domestication process and could reveal practical, useful allele information for future breeding in cereals (29). For instance, introgression of the Nipponbare qSH1 allele into indica cultivars would reduce the seed-shattering degree and could improve yield.

Fig. 4.

Haplotypes at qSH1 and hypothetical evolutionary process in japonica. (A) Eight SNPs (including the qSH1/SNP) and five SSRs from four genomic fragments were examined. On the basis of the results in 80 lines (including 43 cultivars from core collections), 18 haplotypes found at qSH1 were presented (fig. S5). Results including the indica cultivars are shown only in fig. S5; we did not examine accurate haplotypes of some indica cultivars because of a lack of PCR amplification fragments. (B) Hypothetical process of evolution at qSH1 during japonica rice domestication. Nine mutations and two recombination events are enough to explain the natural variations at qSH1 in japonica. Among 75 cultivars tested (52 temperate japonica and 23 tropical japonica), 27 contained the nonshattering T allele at qSH1/SNP but represented only two haplotypes (asterisks) among 13 haplotypes found in the 75 cultivars, suggesting strong selection by humans during domestication. Numbers under haplotypes indicate corresponding numbers of rice accessions. Numbers of accessions in RFLP and SSR core collections, in that order, are indicated in parentheses. Red circles highlight the mutation position in the haplotypes.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1126410/DC1

Materials and Methods

Figs. S1 to S8

References

References and Notes

View Abstract

Navigate This Article