Unlike maize and wheat, where artificial selection is associated with an almost uniform increase in seed or grain size, domesticated rice exhibits dramatic phenotypic diversity for grain size and shape. Here we clone and characterize GS3, an evolutionarily important gene controlling grain size in rice. We show that GS3 is highly expressed in young panicles in both short- and long-grained varieties but is not expressed in leaves or panicles after flowering, and we use genetic transformation to demonstrate that the dominant allele for short grain complements the long-grain phenotype. An association study revealed that a C to A mutation in the second exon of GS3 (A allele) was associated with enhanced grain length in Oryza sativa but was absent from other Oryza species. Linkage disequilibrium (LD) was elevated and there was a 95.7% reduction in nucleotide diversity (theta(pi)) across the gene in accessions carrying the A allele, suggesting positive selection for long grain. Haplotype analysis traced the origin of the long-grain allele to a Japonica-like ancestor and demonstrated introgression into the Indica gene pool. This study indicates a critical role for GS3 in defining the seed morphologies of modern subpopulations of O. sativa and enhances the potential for genetic manipulation of grain size in rice.

GRAIN SIZE 3 (GS3) is a cloned gene that is related to seed length. Here we report the discovery of new deletion alleles at the GS3 locus, each of which confer short seed. We selected ten short seeded cultivars from a collection of 282 diverse cultivars. Sequence analysis across the GS3 gene in these ten cultivars identified three novel alleles and a known allele that contain several independent deletion(s) in the fifth exon of GS. These independent deletion variants each resulted in a frameshift mutation that caused a premature stop codon, and they were functionally similar to one another. Each coded for a truncated gene product that behaved as an incomplete dominant allele and conferred a short seeded phenotype. Haplotype analysis of these sequence variants indicated that two of the variants were of japonica origin, and two were from indica. Transformation experiments demonstrated that one of the deletion alleles of GS3 decrease the cell number in the upper epidermis of the glume, resulting in a significant reduction in seed length. The multiple and independent origins of these short seeded alleles indicate that farmers and early breeders imposed artificial selection favoring short seeds.

Stigma exsertion is an important trait that contributes to the improvement of seed production in hybrid rice. We demonstrate that GS3, one of the genes regulating seed length, also regulates stigma length and participates in stigma exsertion in rice. GS3 mRNA is expressed in the basal part of the young stigma, and a nonsense mutation in the second exon of GS3 causes an increase in cell number, resulting in elongation of the stigma. Manipulation of GS3 should contribute to the improvement of hybrid seed-production efficiency.

Postzygotic reproductive isolation maintains species integrity and uniformity and contributes to speciation by restricting the free gene flow between divergent species. In this study we identify causal genes of two Mendelian factors S22A and S22B on rice chromosome 2 inducing F1 pollen sterility in hybrids between Oryza sativa japonica-type cultivar Taichung 65 (T65) and a wild relative of rice species Oryza glumaepatula. The causal gene of S22B in T65 encodes a protein containing DUF1668 and gametophytically expressed in the anthers, designated S22B_j. The O. glumaepatula allele S22B-g, allelic to S22B_j, possesses three non-synonymous substitutions and a 2-bp deletion, leading to a frameshifted translation at the S22B C-terminal region. Transcription level of S22B-j and/or S22B_g did not solely determine the fertility of pollen grains by genotypes at S22B. Western blotting of S22B found that one major band with approximately 46 kDa appeared only at the mature stage and was reduced on semi-sterile heterozygotes at S22B, implying that the 46 kDa band may associated in hybrid sterility. In addition, causal genes of S22A in T65 were found to be S22A_j1 and S22A_j3 encoding DUF1668-containing protein. The allele of a wild rice species Oryza meridionalis Ng at S22B, designated S22B_m, is a loss-of-function allele probably due to large deletion of the gene lacking DUF1668 domain and evolved from the different lineage of O. glumaepatula. Phylogenetic analysis of DUF1668 suggested that many gene duplications occurred before the divergence of current crops in Poaceae, and loss-of-function mutations of DUF1668-containing genes represent the candidate causal genetic events contributing to hybrid incompatibilities. The duplicated DUF1668-domain gene may provide genetic potential to induce hybrid incompatibility by consequent mutations after divergence.