Rice Domestication by Reducing Shattering

See allHide authors and affiliations

Science  31 Mar 2006:
Vol. 311, Issue 5769, pp. 1936-1939
DOI: 10.1126/science.1123604


Crop domestication frequently began with the selection of plants that did not naturally shed ripe fruits or seeds. The reduction in grain shattering that led to cereal domestication involved genetic loci of large effect. The molecular basis of this key domestication transition, however, remains unknown. Here we show that human selection of an amino acid substitution in the predicted DNA binding domain encoded by a gene of previously unknown function was primarily responsible for the reduction of grain shattering in rice domestication. The substitution undermined the gene function necessary for the normal development of an abscission layer that controls the separation of a grain from the pedicel.

Cereals, the world's primary food, were domesticated from wild grass species. Because wild grasses naturally shed mature grains, a necessary early step toward cereal domestication was to select plants that could hold on to ripe grains to allow effective field harvest (1, 2) (fig. S1). The selection process might have been mainly unconscious because grains that did not fall as easily had a better chance of being harvested and planted in the following years. Consequently, nonshattering alleles had an increased frequency and eventually replaced the shattering alleles during domestication. The finding that one locus accounted for most phenotypic variance of grain shattering between a cereal crop and its wild progenitor suggested that the domestication process could have been initiated quickly by selection at the locus (35). The molecular genetic basis of the selection, however, has not been characterized.

Rice (Oryza sativa) was domesticated from one or both of two closely related species—O. nivara and O. rufipogon—distributed from southeastern Asia to India (6, 7). Our recent genetic analysis of an F2 population derived between O. sativa ssp. indica and the wild annual species O. nivara identified three quantitative traitloci(QTL)—sh3, sh4, and sh8—responsible for the reduction of grain shattering in cultivated rice (5). Of these QTL, sh4 explained 69% of phenotypic variance, and the other two explained 6.0% and 3.1% of phenotypic variance. The sh4 allele of the wild species caused shattering and was dominant.

Two previous QTL studies using crosses between O. sativa ssp. indica and the wild perennial species O. rufipogon detected four and five shattering QTL (8, 9). Both studies identified a QTL at the same location of sh4 with either the largest or nearly largest phenotypic effect among the detected QTL. Moreover, genetic analyses between O. sativa ssp. japonica and O. rufipogon and two other closely related wild species O. glumaepetula and O. meridionalis all found that a single dominant allele from each of the three wild species was responsible for grain shattering (10, 11). This locus, named Sh3, was mapped to the same chromosomal location as sh4.

Our QTL analysis located sh4 between simple sequence repeat (SSR) markers RC4-123 and RM280 (5), which had a physical distance of about 1360 kb in the O. sativa genome (12) (Fig. 1A). Because of the large and dominant effect of the O. nivara allele, we were able to phenotypically distinguish F2 individuals that were homozygous recessive (ss) from those that had at least one O. nivara allele of sh4 (ns and nn), regardless of the genotypes at the remaining two QTL of small effect. After evaluating a total of 489 F2 plants genotyped at the three shattering QTL, we consistently found that plants with the ns and nn genotypes at sh4 shed all mature grains when hand tapped, whereas plants with the ss genotype at sh4 did not shed grains or only partially shed mature grains under vigorous hand shaking.

Fig. 1.

Molecular cloning of sh4. (A) Chromosomal location of sh4 determined by QTL mapping. Dotted lines indicate 1-lod (logarithm of the odds ratio for linkage) supporting interval. (B) Fine mapping of sh4. Vertical lines indicate SSR and SNP markers. Numbers above lines: markers numbered consecutively according to the order of evaluation; numbers below lines: the number of recombinants left in the chromosomal interval still containing sh4 after the evaluation of the marker. White horizontal arrows indicate the orientation and size of open reading frames between markers 5 and 6. The mutation responsible for nonshattering was mapped to between markers 9 and 11, in a predicted gene with two exons (black bars) and an intron (gray bar). The start and stop codons of the gene are labeled. Lines below illustrate two constructs made for gene transformation; red and blue represent sequences of O. nivara and O. sativa, respectively. (C) Seven mutations found between the mapping parents are labeled a through g. Variation at these sites is compared between rice cultivars and wild A-genome species in the phylogenetic context; s and n represent sequences of O. sativa and O. nivara parents, respectively. The number of accessions of a species with the same combination of sequences is indicated. (D) sh4 protein sequence of O. sativa. Mutations between the mapping parents are indicated: red for substitution (mutation d; K in O. nivara) and blue for insertion/deletion (mutation b and c; deletion in O. nivara). The predicted Myb3 DNA binding domain is underlined. The predicted nuclear localization signal is colored green. The sh4-GFP construct for subcellular localization encodes a recombined protein beginning from the sh4 N terminus to the amino acid E (colored orange) followed by GFP. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

With the reliable phenotyping method available, we grew ∼12,000 F2 seedlings and screened for recombinants between RC4-123 and RM280 (13). Plants with the genotype of ss at one marker and ns at the other were selected, and a total of 134 individuals were grown for phenotypic evaluation. By progressively examining SSR and SNP (single-nucleotide polymorphism) markers between RC4-123 and RM280, we finally mapped the mutation responsible for the derivation of nonshattering in cultivated rice to a 1.7-kb region of a gene with a previously unknown function (Fig. 1B and table S1). The gene is predicted to be a transcription factor, and its coding region is physically located between 34,014,305 and 34,012,126 base pairs (bp) on assembly LOC_Os04 g57530 of rice chromosome 4 (The TIGR Rice Genome Annotation Database).

The comparison of the 1.7-kb sequences between the mapping parents revealed seven mutations (Fig. 1C). These include one mutation in the intron: (a) a 1-bp substitution; three mutations in the first exon: (b) a 15-bp or five–amino acid insertion/deletion, (c) a 3-bp or one–amino acid insertion/deletion, and (d) a 1-bp or an amino acid substitution; and three mutations 5′ upstream of the start codon: (e) a 1-bp substitution at site -55, (f) a 3-bp insertion/deletion between sites –343 and –344, and (g) an 8-bp insertion/deletion between sites –558 and –559.

To assess the polymorphism and evolutionary direction of these mutations, we sequenced this 1.7-kb region from an additional 14 rice cultivars representing the diversity of O. sativa (14), 21 accessions of O. nivara covering the distributional range of the wild species (15), 6 accessions of O. rufipogon, and 1 accession of each of the four remaining wild A-genome species (Fig. 1C and table S2). The cultivars were polymorphic for mutation f, i.e., some of the cultivars had the same sequence as O. nivara. At the remaining six mutation sites, all cultivars shared the same sequences, which were different from those of the O. nivara parent.

Surprisingly, three accessions of O. nivara had the same sequences as O. sativa at these six sites. It was then found that plants grown from these accessions had the nonshattering phenotype. Greenhouse observations indicated that these accessions had additional characteristics of cultivated rice that were not found in O. nivara, such as upright tillers, short awns, and/or photoperiod sensitivity. This suggests that the three accessions are weedy rice that has received and fixed the sh4 allele from cultivars.

The remaining accessions of the wild species with confirmed shattering differed invariably from the cultivars by one mutation, d, which was a nucleotide substitution of G for T or an amino acid substitution of asparagine for lysine in O. sativa. At the remaining five mutation sites, sequence polymorphism was found within the wild species (Fig. 1C). That is, some of the wild accessions shared the same sequence with cultivated rice at these sites but had the shattering phenotype. The results thus indicate that the amino acid substitution at site d was selected for the development of nonshattering cultivars during rice domestication.

The Blast search of the GenBank for protein sequences identified three predicted genes that are most similar to sh4. These include a rice gene (XP_469180) with 32% amino acid sequence identity with sh4, and two Arabidopsis genes (NP_174416 and NP_181107) with 32% and 29% identity with sh4. None of these genes has been functionally characterized. The two Arabidopsis genes were predicted to be transcription factors (16), and one of them had a cDNA sequence (AAT99796) in the database. The next most similar group of genes was also from rice and Arabidopsis but had only 20 to 22% amino acid sequence identity with sh4.

Examination of the sh4 protein, using programs Prosite and PredictNLS, identified a Myb3 DNA binding domain and a nuclear localization signal (Fig. 1D), suggesting that sh4 is a transcription factor. To test this hypothesis, we fused the gene for a green fluorescent protein (GFP) with sh4 to make sh4-GFP, which was driven by a Ubi promoter in the plasmid construct. The construct was introduced into a japonica cultivar, Taipei 309, a rice strain tested as suitable for gene transformation (17). The nuclear localization of GFP-tagged sh4 was determined (Fig. 2). This result supports the bioinformatic prediction that sh4 is a transcription factor.

Fig. 2.

Subcellular localization of sh4. Roots of rice cultivar Taipei 309 transformed with Ubi::sh4-GFP were stained with 4′-6-diamidino-2-phenylindole (DAPI) and observed under various conditions. (A) A differential interference contrast image of epidermal cells. (B) The same cells showing the DAPI-stained nuclei. (C) The same cells showing the nuclear localization of sh4-GFP fluorescence. (D) The merged image. Bar, 10 μm.

Reverse transcription–polymerase chain reaction (RT-PCR) detected the expression of sh4 at the flower and pedicel junction, where mature grains separate from the mother plant (Fig. 3A). Gene expression was not detected in the remaining parts of flowers or pedicels or in the leaves. We amplified, using RT-PCR, the entire coding region of sh4 cDNA from both mapping parents. The comparison of the cDNA sequences showed that the intron was spliced from the same position as predicted by rice genome annotation.

Fig. 3.

Expression of sh4 and flower and grain detachment. (A) RT-PCR results, using total RNA isolated from the flower and pedicel junction (FP), from the remaining portions of pedicel (P) and unopened flower (F), and from leaves (L). Above the flower, the separation location of a mature grain from the pedicel of O. nivara is shown. (B) Real-time RT-PCR estimate of relative expression of sh4, using RNA isolated from leaves (L) and the flower/grain and pedicel junction ∼3 days before flower opening (–3), the day of flower opening (1), and every 3 days thereafter during seed development. (C) Force required to pull flowers or grains away from pedicels on the day of flower opening (1) and every 3 days thereafter during seed development. (D) Force required to pull away grains of transformants containing the empty plasmid (control), expressing construct 1, and expressing construct 2. Error bars, ±SD.

We conducted real-time PCR to compare the relative levels of sh4 expression at various stages of flower and seed development in both mapping parents (Fig. 3B). Although there was a trend of increased gene expression as seeds matured, a substantial increase began 12 days after pollination. The expression in O. sativa continued to increase on day 18, while the measurement for O. nivara was no longer possible due to shattering.

We measured the strength of flower and grain attachment to the pedicel at the corresponding developmental stages. Flowers and grains were pulled away at the interface where a mature grain separates from the pedicel, and the force required was measured. For the first 9 days after pollination, the force was not significantly different between the developmental stages in either species (Fig. 3C). The force began to decrease in both species from day 12. The decline continued at a much faster rate in O. nivara than in O. sativa. On day 18, shattering prevailed in O. nivara, which left few grains to measure. In O. sativa, the force measured on day 18 was about half of that required at the earlier stages; it then decreased at a rather slow pace but did not reach the level permitting grain shattering in O. nivara.

We conducted rice transformation to confirm the gene function and to test the role of the amino acid substitution. We made two constructs that had the O. sativa promoter and recombined coding regions between the mapping parents. The two constructs differed only at the mutation site d. Construct 1 contained O. nivara sequence from the 3′ nontranslated region to the inclusion of mutation site d and O. sativa sequence from mutation site e to the 5′ regulatory region. Construct 2 contained O. nivara sequence from 3′ to the inclusion of mutation site c and O. sativa sequence from mutation site d to the 5′ regulatory region (Fig. 1B). The plasmids were introduced into Taipei 309.

The expression of the introduced constructs in the transgenic plants was verified by RT-PCR identification of the 18-bp deletion of the O. nivara sequence at mutation sites b and c. Transformants expressing construct 1 showed significantly reduced strength of grain attachment to pedicel (Student's t test, P = 0.003), whereas no significant difference was found between transformants expressing construct 2 and the control (t test, P = 0.5) (Fig. 3D). The results thus support the finding made from the genetic mapping and sequence comparison that the amino acid substitution at site d was primarily responsible for the reduction of grain shattering in rice domestication.

Programmed organ detachment, such as the falling of old leaves, withered floral parts, and ripe fruits, is fundamental to plant function and adaptation and is regulated by an abscission zone at the juncture of the organ and the main body of the plant. The molecular genetic control of the abscission zone development, however, is poorly understood. The study of dicotyledonous plants such as bean, tomato, and Arabidopsis showed that an abscission zone encompassed several layers of small, densely cytoplasmic cells. In response to environmental and hormonal signals, the activation of abscission is coupled with cell expansion and secretion of hydrolytic enzymes that break the middle lamella between cell layers in the abscission zone (18).

In monocotyledons, including grasses, little is known about the development and function of abscission zones. Genes regulating the developmental processes have not been identified. Here we found that the abscission zone between a rice grain and the pedicel consists of mostly one layer of small, thin-walled cells. O. nivara has a complete layer of abscission cells between the grain and the pedicel, which is seen in a longitudinal section as continuous lines of abscission cells between the vascular bundle and the epidermis (Fig. 4A). O. sativa, however, has an incomplete abscission layer. In the longitudinal section, the line of abscission cells is discontinuous and completely absent near the vascular bundle, where they are replaced by thicker-walled cells similar to adjacent pedicel cells (Fig. 4, B and C). For both species, these anatomical features were seen in young flowers (flowers ∼15 days before opening were examined) and remained similar in mature grains.

Fig. 4.

Fluorescence images of longitudinal section of flower and pedicel junction. (A) O. nivara mapping parent, with complete abscission layer (al). (B) O. sativa ssp. indica mapping parent, with incomplete abscission layer. (C) O. sativa ssp. japonica Taipei 309, with incomplete abscission layer. (D) Transformant of Taipei 309 expressing construct 1, with improved abscission layer. f, flower side; p, pedicel side; v, vascular bundle. Bar, 50 μm.

Because japonica cultivars are generally harder to thresh than indica cultivars (19), Taipei 309 had a stronger grain attachment to the pedicel than the indica mapping parent (Fig. 3, C and D). Accordingly, the abscission layer of the japonica cultivar showed a higher degree of discontinuity and further retreat from the vascular bundle. The transgenic plants with the strength of grain attachment reduced to less than 100 g had substantially improved abscission layers that were more continuous and extended closer to the vascular bundles (Fig. 4D).

The results indicate that sh4 plays an important role in the establishment of the abscission layer from the early stage of flower development. The increased expression of sh4 in the late stage of seed maturation suggests that the gene may also play a role in the activation of the abscission process. One or both of the roles were undermined by the amino acid substitution of asparagine for lysine in cultivated rice.

In the process of rice domestication, human selection was likely to have favored mutations that reduced grain shattering but did not eliminate the formation or function of the abscission layer. In this way, grain loss due to shattering was largely prevented during harvest while a certain level of grain abscission was maintained so that the yield increase was not offset by creating difficulties in threshing. The inverse correlation between the expression level of sh4 and the strength of grain attachment in O. sativa at the late stage of grain maturation seems to suggest that the amino acid substitution did not knock out the gene function in cultivated rice.

The slower pace of increase in the level of sh4 expressed in O. sativa than in O. nivara during grain maturation might have been a result of selection in the regulatory region of the gene for a finer adjustment of the shattering/threshing balance during rice cultivation. A comparison of the regulatory sequences of sh4, the levels of gene expression, and the phenotypic difference among diverse rice cultivars should provide further insights into the genetic basis of agricultural selection continued through the history of rice cultivation.

Genetic analyses of crop domestication, especially the cloning of domestication-related genes, have shed a light on plant development and evolution. Mutations in regulatory genes were found responsible for drastic morphological modifications during maize and tomato domestication (2023). Here we show that the substitution of a neutral for a positively charged amino acid in a predicted DNA binding domain led to a physiological transition key to rice domestication. This is consistent with the finding that positively charged amino acids are critical residues on the surface of DNA binding domains (24). The cloning of sh4 opens opportunities for understanding the developmental pathway of programmed cell separation and seed dispersal in monocotyledonous plants and potentially for optimizing the methods of grain harvest.

Supporting Online Material

Materials and Methods

Fig. S1

Tables S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article