Introduction

Rice is the world's most important staple food crop and is a primary food source for about half of the world's population. As farmland is decreasing and the global population increasing, there is an urgent need to secure grain production 1. Therefore, increasing grain yield has been the primary objective in current rice-breeding programs.

The shape of rice grain, a complex agronomical trait, plays an important role in determining the rice yield 2, 3. Currently, more than 22 quantitative trait loci (QTLs) related to rice grain width have been reported in different mapping populations 4, 5, 6, 7, 8, 9, 10, and several genes that regulate rice yield or grain shape have been identified, including Gn1a 11, Ghd7 12, GS3 13 and GW2 14. Gn1a encodes a cytokinin oxidase/dehydrogenase required for degrading the phytohormone cytokinin. Reduced expression of Gn1a causes cytokinin accumulation in inflorescence meristems and increases the number of reproductive organs, resulting in increased grain yield 11. Ghd7 encodes a CCT domain protein. Enhanced expression of Ghd7 under long-day conditions delays heading and increases plant height and panicle size, resulting in increased rice grain yield 12. GS3, a major QTL for grain length and weight and a minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein 13. GW2, a QTL for rice grain width and weight, encodes a previously unknown RING-type E3 ubiquitin ligase that negatively regulates cell division by targeting its substrate(s) for degradation by the ubiquitin-proteasome pathway 14. In Arabidopsis, DA1, which encodes a predicted ubiquitin receptor, sets the final seed and organ size by restricting the period of cell proliferation. The mutant protein encoded by the da1-1 allele has negative activity toward DA1 and a DA1-related protein, and overexpression of da1-1 cDNA dramatically increases the seed and organ size of wild-type plants 15. In addition, plant hormones and their receptors 16, 17, 18, 19, as well as the nutritional synthesis pathway 20, 21, have also been implicated in the development of grain in crops and other plant species. For example, the rice dwarf11 mutant has a defect in a novel cytochrome P450, which shares homology with enzymes involved in brassinosteroid biosynthesis, and bears seeds of reduced length 17. In maize, the gln1-4 mutant displays reduced kernel size owing to the loss-of-function in glutamine synthetase isoenzymes 21. Despite these efforts, the mechanisms that establish the final size of grain or seeds are still poorly understood.

In addition, human civilization has had a major impact on the domestication of crop plants. Regarding the origin of indica and japonica rice subspecies, one school of thought holds that indica and japonica rice originated independently from a wild ancestor 22, 23, 24; others favor the notion that the japonica were derived from the indica 25. Recent functional genomic studies have helped to excavate a number of domestication genes 26, 27, 28, 29. For example, sh4 and qSH1 are essential for effective field harvest because of their ability to reduce grain shattering 29, 30. In addition, grain or fruit size was important in the evolution of wild species because of the continuous selection for large ones during early domestication. In tomato, fw2.2 changes the fruit weight by almost 30% and seems to have been responsible for a key transition during domestication 27.

We previously reported a major QTL on chromosome 5, qGW5, which is associated with reduced grain width not only in the isogenic Asominori background but also in the recombinant background of Asominori and IR24 under multiple environmental conditions 31. In this study, we report the fine mapping, cloning and initial characterization of this gene. We also provide evidence that GW5 is associated with rice domestication. Our work sheds light on the mechanisms of grain development and domestication of rice.

Results

Characterization of the narrow rice line, CSSL28

Grain width is quantitatively inherited; therefore, it is difficult to analyze the mechanism of grain formation using conventional methods. To dissect the loci that control grain width into several single genes, 71 recombinant inbred lines (RILs) were derived from a cross between Asominori and IR24 by single-seed descent 32. Sixty-six chromosome segment substitution lines (CSSLs) with largely Asominori background, named CSSL1-66, were produced by nonselectively crossing and backcrossing 19 selected RILs with Asominori to produce the BC3F1 generation 33. One of these CSSLs, CSSL28 (narrow-grain rice), shows a slender-grain phenotype, because it harbors a chromosomal segment between RFLP (restriction fragment length polymorphism) markers C263 and R2289 (Figure 1). This segment derives from IR24 (narrow-grain rice) and is substituted into the Asominori (wide-grain rice) genomic background. CSSL28 shows a 16.4% reduction in grain width of paddy rice, and its 1 000-grain weight is reduced by 18.7% compared with that of Asominori, mainly because of the reduced grain width (Table 1).

Figure 1
figure 1

qGW5 controls grain width and weight in CSSL28. (A) Genotype of the narrow grain line, CSSL28 (a segment substitution between the RFLP markers C263 and R2289 on chromosome 5). The black bar indicates the fragment from IR24 in the Asominori genomics background. (B) Phenotype of the paddy rice grain, IR24 (top), Asominori (middle) and CSSL28 (bottom). (C) Phenotype of the brown rice grain. Scale bar: 3 mm.

Full size image
Table 1 Phenotypic analysis of Asominori and CSSL28
Full size table

GW5 is associated with a 1 212-bp deletion in the Asominori cultivar

For genetic analysis and isolation of the dominant narrow-grain QTL, named GW5, the F2 population was constructed from a cross between Asominori and CSSL28, and the QTL was dissected into a single gene 31. By means of linkage analysis using the genotype data of both the GW5 gene and simple sequence repeat (SSR) markers, the GW5 gene was mapped to an interval between markers RM3328 and RMw513 in 805 homozygotes (Figure 2A). Thus, the GW5 gene was mapped to a genomic region 2.7 cM in length, located 2.3 cM from RM3328 and 0.37 cM from RMw513. In this region, three bacterial artificial chromosome (BAC) contigs were found – OJ1725_E07, OJ1097_A12 and B1007D10 (Figure 2B) – according to the Nipponbare genome (http://www.gramene.org).

Figure 2
figure 2

Fine mapping of GW5. This map was constructed on the basis of publicly available rice sequences. The CAPS and Indel markers developed in this work are indicated. (A) GW5 was mapped to a region between markers RM3328 and RMW513 in 805 recessive individuals. (B) Three BAC contigs, OJ1725_E07, OJ1097_A12 and B1007D10, cover the GW5 locus in the Nipponbare genomic sequence. (C) GW5 was narrowed down to a 21-kb genomic DNA region between CAPS markers Cw5 and Cw6, and it co-segregated with the Indel1 marker in 2 180 recessive individuals. (D) Compared with IR24, 1.2-kb genomic DNA is deleted in Asominori. (E) Three ORFs are predicted in the candidate region harboring GW5 (http://softberry.com). The predicted ORF2 is located in the deleted region in Asominori.

Full size image

For high-resolution mapping of GW5, five new cleaved-amplified polymorphic sequence (CAPS) markers were developed on one of the BACs (OJ1097_A12). The GW5 locus was pinpointed into an interval between the CAPS markers Cw5 and Cw6 in 2 180 wide-grain homozygous individuals from the BC4F2 population (Figure 2C and Table 2). Within this region, the 21-kb fragment of Asominori and the corresponding 22.2-kb fragments of CSSL28 and IR24 were sequenced. Further comparison of these sequences showed that the Asominori genome harbored a 1 212-bp deletion (Figure 2D). Meanwhile, two Indel (insertion or deletion) markers – Indel1, including 775 bp of the 1.2 kb fragment, and Indel2, containing the aforementioned 1.2-kb fragment and totaling 1 897 bp in length – were designed (Table 3) and used to confirm the presence of GW5. Genetic analysis revealed that these two Indel markers co-segregated with GW5 in 2 180 wide-grain progenies (Figure 2C).

Table 2 Recombination events between GW5 and molecular markers
Full size table
Table 3 A list of primers used in this work
Full size table

On the basis of the available sequence annotation (http://www.softberry.com; http://www.ncbi.nlm.nih.gov/BLAST), two open reading frames (ORFs) were predicted in the 21.0-kb target region of Asominori, and three ORFs were predicted in the corresponding region of CSSL28. However, no sequence differences were found between these lines for ORF1 and ORF3. ORF1 encodes a protein showing high similarity to ubiquitin-protease-like protein, whereas ORF3 encodes an unknown protein containing a calmodulin-binding motif. Interestingly, we found that the 22.2-kb region of CSSL28 contains a third ORF between ORF1 and ORF3, designated ORF2, which is located in the deleted region of Asominori (Figure 2E). Motif scan analysis showed that the predicted product of ORF2 harbors a nuclear localization signal (NLS) and an arginine-rich domain (http://hits.isb-sib.ch/cgi-bin/motif_scan). These results suggest that ORF2, which is present in CSSL28 but absent in Asominori, likely corresponds to GW5.

The deletion prevails in wide-grain rice cultivars and defines a domestication-related rice gene

To further confirm the identity of GW5, we randomly selected 46 rice lines, including Asominori, IR24 and 44 other cultivars, to determine the co-segregation relationship between the absence of the candidate GW5 gene and the wide-grain phenotype (Table 4). The 46 rice lines were divided into two groups: group I contained 23 narrow-grain varieties (grain width ranging from 2.40 to 2.85 mm) and group II consisted of 23 wide-grain ones (3.30–3.92 mm) (Table 4). Segregation analysis of the Indel1 marker showed that the 775-bp target fragment was observed only in group I cultivars, but not in group II cultivars (Figure 3A and 3B). Similarly, PCR analysis with the Indel2 marker detected a 1 897-bp fragment in group I cultivars, but only a 697-bp product in group II cultivars (Figure 3C and 3D). These results suggested that the 1.2-kb fragment containing the GW5 gene was absent in all 23 wide-grain cultivars examined here but was present in the narrow-grain ones. The strict correlation between the grain-width phenotype and deletion of the 1.2-kb genomic DNA strongly supports the notion that the candidate ORF2 located on the deleted 1.2-kb fragment represents GW5.

Table 4 The 46 rice cultivars analyzed in this work and their places of origin
Full size table
Figure 3
figure 3

Genotyping analyses in 46 rice cultivars. (A) PCR products of the Indel1 marker in 23 narrow-grain rice lines; (B) PCR products of the Indel1 marker in 23 wide-grain rice varieties; (C) PCR products of the Indel2 marker in 23 narrow-grain rice lines; and (D) PCR products of the Indel2 marker in 23 wide-grain rice varieties. The DNA sample number is listed in Table 3.

Full size image

To address how this deletion in GW5 prevails in rice domestication, we further analyzed the 46 cultivars. Group I contained 21 indica species (grain width ranged from 2.40 to 2.84 mm) and two slender japonica varieties (one 2.74 mm wide, the other 2.85 mm wide), whereas most japonica (3.30–3.92 mm) fell within group II, with the exception of two wide indica lines (3.46 and 3.48 mm). PCR analysis showed that GW5 was deleted in all group II lines, including both japonica and indica subspecies. We thus conclude that the deletion is highly correlated with the grain-width phenotype among the japonica rice in group II. Our results suggest that GW5 was kept in most japonica cultivars during rice domestication and that this gene may underlie the relatively wide- and short-grain phenotype 23.

GW5 is expressed in slender-grain rice

The sequence of 9 311 (a narrow-grain indica rice cultivar) in the cognate region is similar to that in the GW5 region in CSSL28. Analysis of the indica rice genome sequence showed that GW5 represents a single-copy gene without any expressed sequence tags (ESTs) or cDNA support. However, we noticed that expression of the candidate ORF was detected in the tiling microarray database of 9 311 (GenBank Acc: CL971152) 34. To further confirm the expression of GW5, we performed RT-PCR analysis using mRNAs derived from the young panicle of CSSL28 (Figure 4A). We obtained RT-PCR products with primer combinations of F6 and R2, as well as with F6 and R3 (Figure 4B and Table 3). These PCR products were confirmed by sequencing analyses. Therefore, GW5 is expressed in CSSL28.

Figure 4
figure 4

GW5 expression analyses. (A) The sequence of the predicted ORF2. The primers (F6, R2 and R3) used for RT-PCR analysis and the predicted NLS are denoted by arrows and a line, respectively. In addition, amino-acid residues 20-71 formed an arginine-rich region. (B) RT-PCR analysis of ORF2 using mRNA derived from the young panicle of CSSL28. Lane 1: molecular marker. Lane 2: a 140-bp RT-PCR product using primers F6 and R3 (marker 1). Lane 3: a 200-bp RT-PCR product using primers F6 and R2 (marker 02). Lane 4: control (no template was added in the PCR reaction).

Full size image

GW5 physically interacts with polyubiquitin

Sequence analysis indicated that GW5 encodes a novel protein without significant homology to any proteins of known biochemical function. The protein is predicted to contain an NLS and an arginine-rich domain. To test the functionality of the predicted NLS, we fused the coding region of GW5 to the N-terminus of GFP. Transient expression in onion epidermal cells showed that the GW5-GFP fusion protein is exclusively localized to the nucleus (Figure 5).

Figure 5
figure 5

Sub-cellular localization of GW5. (A) and (C) Bright-field images of onion epidermal cells. (B) GFP by itself (pCAMBIA 1302 vector) localizes to both cytoplasm and nucleus. (D) GW5-GFP fusion protein is localized only to the nucleus. Scale bars: 50 μm.

Full size image

In an effort to identify the functional partners of GW5, we carried out a yeast two-hybrid screen using full-length GW5 as the bait. A prey library was constructed using mRNAs derived from the young panicle of IR24 before heading, at which time GW5 expression was detected. We repeatedly found that polyubiquitin protein interacts with GW5. This result was found 14 times in about 200 candidate positive clones on synthetic growth medium without leucine, tryptophan, histidine and adenine. X-gal filter lift assays also detected a clear interaction between GW5 and polyubiquitin (Figure 6B). This result suggests that the GW5 protein may play an important role in regulating the grain shape through involvement with the ubiquitin-proteasome pathway.

Figure 6
figure 6

GW5 interacts with polyubiquitin in yeast two-hybrid assay. (A) AH109 yeast cells expressing pGBKT7-GW5/pGADT7 (top), pGBKT7/pGADT7-polyubiquitin (middle) and pGBKT7-GW5/pGADT7-polyubiquitin (bottom) were selected on synthetic growth medium without Leu and Trp. (B) X-gal filter lift assay showing that GW5 interacts with polyubiquitin.

Full size image

Discussion

In this study, we report the fine mapping and cloning of GW5, a major QTL controlling grain width and weight in rice. We identified a 1 212-bp deletion between the CAPS markers Cw5 and Cw6 on chromosome 5 in Asominori rice (japonica, wide grain) in comparison with IR24 rice (indica, slender grain). In this deleted region, we identified an ORF that is expressed in CSSL28. Furthermore, through genotyping analyses of 46 randomly selected rice cultivars, we detected a high correlation between grain shape and GW5; that is, loss of GW5 occurred in the wide-grain lines, including most japonica and a few indica rice, whereas GW5 was present in the slender-grain rice, including most indica and a few japonica rice cultivars (Figure 3 and Table 4). Together, these findings strongly support the notion that deletion of GW5 is responsible for the change in grain width and weight and that the ORF we identified in this deleted genomic region represents the GW5 gene.

In addition, it is believed that wild ancestors with small seeds are usually favored by natural selection because small seed size is frequently associated with a higher number of seeds per plant, early maturity and wider geographic distribution 35. Domestication of wild rice probably started about 9 000 years ago, and O. sativa evolved and differentiated into two subspecies, indica and japonica 23. As numerous single-nucleotide polymorphisms (SNPs) and Indels were identified between 9 311 (indica) and Nipponbare (japonica) 36, it was suggested that chromosomal rearrangements might have occurred during the domestication of these two subspecies, which caused significant divergence between the compositions of the two corresponding genomes. In our study, the deletion of GW5 was detected in most japonica and a few indica cultivars with the wide-grain phenotype. Our data are in agreement with the hypothesis that indica and japonica rice originated independently from a mutual wild ancestor.

Earlier studies have determined that some domestication genes have key regulatory roles in plant morphology and architecture. For example, TB1, one of these key regulatory genes, controls the morphological transition from wild species to cultivated maize 26, and the major QTLs fw2.2 27, ovate 28 and qSH1 29 control tomato fruit size, the transition of tomato fruit from round to pear-shaped and seed shattering, respectively. The identification of GW5 and the earlier reported GW2 as major QTLs that control rice-grain width lends strong support to the notion that “major morphological changes during evolution and domestication can be attributable to a few major regulatory genes of large effects rather than to many genes, each of which contributes a small effect to the changes” 37.

The GW5 ORF encodes a predicted protein of 144 amino acids without significant sequence homology to any proteins of known molecular function. It contains a predicted NLS and an arginine-rich domain. Using a GFP fusion strategy, we showed that GW5 protein is exclusively localized to the nucleus. Furthermore, GW5 physically interacts with polyubiquitin, as shown by the yeast two-hybrid assay. These observations suggest that GW5 might be involved in the ubiquitin-proteasome pathway to regulate grain width and weight. It is particularly interesting to note that GW2, another QTL for rice-grain width and weight, encodes a RING-type protein with E3 ubiquitin ligase activity 14. Loss of GW2 causes increased cell numbers, a wider spikelet hull and an accelerated grain milk-filling rate, resulting in enhanced grain width, weight and yield. GW2 may negatively regulate cell division by targeting its substrate(s) for regulated proteolysis by proteasomes. It will be of great interest to determine the functional relationship between GW2 and GW5 in controlling seed development. GW5 may work together with GW2 in targeted degradation of certain substrates that promote cell division and grain growth. GW5 may also work together with the deubiquitinating enzymes and might be involved in reversing ubiquitination by substrate-specific cleavage of ubiquitin moieties 38, 39, 40. Recent studies have also pointed to a critical role of the ubiquitin pathway in grain development in other plant species 14, 15. For example, loss-of-function of a ubiquitin receptor significantly increased the seed and organ size in Arabidopsis 15. Further studies to determine the molecular function of GW5 will yield more insights into the mechanism of grain development, which could provide a potential tool for improving the crop grain yield.

Materials and Methods

Plant materials

CSSL28, harboring the GW5 gene, was used to build the BC4F2 population by backcrossing with Asominori followed by subsequent self-pollination. When the plants matured, seeds were collected from the primary panicles and dried for 72 h at 50 °C. Paddy rice and brown rice were used for evaluating grain size. A total of 2 180 wide-grain homozygous individuals were collected for fine mapping. These plants were collected in three consecutive years (2004-2006) at three locations (Nanjing, Hainan and Beijing). The widths of 20 rice grains randomly selected from each line were estimated using an electronic digital caliper (Guanglu Measuring Instrument Co. Ltd, China) with a precision of 0.1 mm. The plant materials and markers used in this study are listed in Table 2.

For co-segregation analysis of GW5 and rice grain shape, we selected Asominori, IR24 and 44 local varieties, including 21 normal japonica lines, two indica lines with the wide-grain phenotype, 21 normal indica lines and two slender japonica lines. These rice cultivars were supplied by the National Rice Germplasm Bank (Table 3).

DNA preparation, PCR protocol and DNA marker analysis

DNA was extracted from fresh leaves of BC4F2 individuals and 46 rice lines using the methods described earlier 41. PCR was performed using a procedure described earlier 42, with minor modifications. PCR products were separated on 8% non-denaturing polyacrylamide gels and detected using the silver staining method 43.

Primer sequences, map position and amplified length of new SSR, CAPS and Indel markers used for fine mapping of GW5 are listed in Table 3. New CAPS markers were developed by comparing the original length with CAPS length, using SNP2CAPS software 44. PCR products of CAPS and Indel markers were separated on 1-2% agarose gels.

Primary mapping of the GW5 gene

A total of 805 BC4F2 plants were genotyped using eight SSR markers to construct a small-scale linkage map. Among these molecular markers, four were new SSRs developed in our laboratory. The 805 BC4F2 plants were classified according to their measured phenotypic performance as one of two types: slender-grain rice, such as IR24 or CSSL28, or wide-grain rice, such as Asominori. On the basis of the segregation of grain width within the 805 BC4F3 families, the BC4F2 population was further partitioned into three groups: slender plants (genotype GW5GW5), slender-wide plants with segregation (genotype GW5gw5) and wide-grain individuals without segregation (genotype gw5gw5). To combine the molecular marker data with the GW5 gene, small-scale genetic mapping of the GW5 gene was performed using the program Mapmaker/Exp 3.0.

High-resolution mapping

For high-resolution mapping of the GW5 gene, the bulked-extreme and recessive-class approach was used to calculate the recombination frequencies (c) between the GW5 gene, the five newly developed CAPS and the two Indel markers in the 2 180 homozygous wide-grain BC4F2 plants. Thus, c=(N1+N2/2)/N, where N is the total number of wide-grain plants surveyed, N1 is the number of wide-grain individuals with the banding pattern of the slender-grain parent and N2 is the number of wide-grain plants with heterozygous banding patterns 45.

mRNA isolation and RT-PCR

Total RNA was extracted from the young spike rice IR24 (narrow-grain line) about 5 days before heading, using an RNA prep pure Plant Kit (Tiangen Biotech Co., Ltd). The mRNA was then further purified using PolyATtract® mRNA isolation systems (Promega Corporation), according to the manufacturer's instructions. First-strand cDNA was synthesized using a PrimeScriptTM RT-PCR Kit (TaKaRa Biotechnology Co., Ltd). RT-PCR was carried out using the first-strand cDNA as the template with the primers F6 (5′-AGA CGG AGG AGG AGG AAC GGG CGG CCA GTG-3′) and either R2 (5′-TCG ATC CTA TTT TTC GAG CTG TTT GGG TAG-3′) or R3 (5′-GAA TAT TCT TCC CAG ATC CAG GAC GAG G-3′). The PCR products were cloned into the pMD18-T vector and sequenced at the Invitrogen Sequencing Facility.

Protein sub-cellular localization

To investigate the cellular localization of GW5, the full-length cDNA of GW5 was cloned into the pCAMBIA1302 vector at the NcoI and SpeI sites. The GFP-GW5 fusion gene was driven by the 35S promoter. The control vector pCAMBIA1302 and the GFP-GW5 fusion gene construct were bombarded into onion epidermal cells using a helium ballistics device (Bio-Rad PDS-1000). The samples were examined with a Zeiss LSM510 confocal laser microscope.

Yeast two-hybrid screen

The full-length GW5 coding region was introduced into the pGBKT7 vector at the EcoRI and SalI sites. A BD Matchmaker library construction and screening kit (Clontech Laboratories, Inc.) was used for yeast two-hybrid assays. All protocols were performed according to the manufacturer's user manual. The yeast strain Y187 transformed with pGBKT7-GW5 was mated with the yeast strain AH109 transformed with a rice young spike AD fusion library and the resulting progeny were selected on SD/-Leu/-Trp/-His and SD/-Leu/-Trp/-His/-Ade/X-gal plates.

All selected pGADT7 clones were sequenced at the Invitrogen Sequencing Facility to ensure that the prey proteins were in-frame fusions with the GAL4 AD domain and that the sequences were accurate.

Accession codes

Gene bank: 9311, Exon-trapped, CL971152.

Gene bank: IR24, GW5, DQ991205.

Note

During the preparation of our paper, we noticed a report of a newly identified QTL, qSW5 (a QTL for seed width on chromosome 5), the deletion of which resulted in a significant increase in seed size owing to an increase in cell number in the outer glume of the rice flower. qSW5 was mapped to the same genomic location as GW5. An 11.2-kb genomic fragment from the Kasalath cultivar, covering the GW5 region, restored the slender-grain phenotype in Nipponbare 46. These results further confirm that qSW5 corresponds to GW5.