Lost genome segments associate with trait diversity during rice domestication

Abstract

Background: DNA mutations of diverse types provide the raw material required for phenotypic variation and evolution. In the case of crop species, previous research aimed to elucidate the changing patterns of repetitive sequences, single-nucleotide polymorphisms (SNPs), and small InDels during domestication to explain morphological evolution and adaptation to different environments. Additionally, structural variations (SVs) encompassing larger stretches of DNA are more likely to alter gene expression levels leading to phenotypic variation affecting plant phenotypes and stress resistance. Previous studies on SVs in rice were hampered by reliance on short-read sequencing limiting the quantity and quality of SV identification, while SV data are currently only available for cultivated rice, with wild rice largely uncharacterized. Here, we generated two genome assemblies for O. rufipogon using long-read sequencing and provide insights on the evolutionary pattern and effect of SVs on morphological traits during rice domestication.

Results: In this study, we identified 318,589 SVs in cultivated and wild rice populations through a comprehensive analysis of 13 high-quality rice genomes and found that wild rice genomes contain 49% of unique SVs and an average of 1.76% of genes were lost during rice domestication. These SVs were further genotyped for 649 rice accessions, their evolutionary pattern during rice domestication and potential association with the diversity of important agronomic traits were examined. Genome-wide association studies between these SVs and nine agronomic traits identified 413 candidate causal variants, which together affect 361 genes. An 824-bp deletion in japonica rice, which encodes a serine carboxypeptidase family protein, is shown to be associated with grain length.

Conclusions: We provide relatively accurate and complete SV datasets for cultivated and wild rice accessions, especially in TE-rich regions, by comparing long-read sequencing data for 13 representative varieties. The integrated rice SV map and the identified candidate genes and variants represent valuable resources for future genomic research and breeding in rice.

Keywords: Association; Domestication; Oryza; Phenotype; Structure variation.

Fig. 1

Genomic landscape of the JX1 and SL1 genomes. a Genome size of cultivated and wild rice based on genome survey results. b Distribution of k-mers for JX1 and SL1. c Phenotype of the plant and panicle of JX1 and SL1. d, e Genome-wide chromosome heatmap of the Hi-C data for JX1 and SL1. f SV number distribution and gene, transposable element, and GC content densities in 100-kb sliding windows

Fig. 2

SV discovery. a Histograms illustrating frequencies of different sizes of SVs. b SVs from each sample were merged using a nonredundant strategy starting with J-SN265 and iteratively adding unique calls from additional samples. The growth rate of the nonredundant SVs declines as the number of samples increases. SVs shared among all samples are shown as red portions of each bar. c Stacked bar graphs showing the proportion of repeat types for all annotated insertions and deletions. Count, the proportion of individual repeat annotations; bp, the proportion of cumulative repeat sequence length; Other, other repeat types. d Structural variation density from repeat/non-repeat genomic regions in continuous 200-kb windows. Significance was tested by Fisher’s exact test; ***, p < 0.001. e Frequency for each variant type (insertion and deletion). Compared to deletions, a greater proportion of insertions were shared among all rice accessions. f, g Comparisons of the TE number (f), GC content (f), single-nucleotide mutation rate (f), and gene expression level (g) between SV and non-SV regions. Data are presented as means ± 95% CI

Fig. 3

Feature of SVs associated with domestication. a Distribution of F_ST values between O. rufipogon, based on SVs within 20-kb windows, with japonica and indica. The corresponding plots for SNPs are provided in b. c Box plots of SV length under the selective sweeps detected by F_ST analyses or not

Fig. 4

Scatter plots showing SNP (a,b) and SV (c,d) occurrence frequencies in japonica and indica subspecies

Fig. 5

GWAS of grain length, grain width, the ratio of grain length to grain width, grain weight, primary branch, secondary branch, plant height, flag leaf length, flag leaf width, and the angle of flag leaf for 649 rice accessions based on SVs. a Plots of significant associated sites with the nine traits on chromosomes. b Manhattan plot of grain width using SV-GWAS on chromosome 5. c SV results in the presence and absence of Nipponbare (NIP) and 9311. d Haplotype frequencies in cultivated and wild rice. e Comparison of the grain width variation between the two haplotypes of the 1200-bp SV

Fig. 6

GWAS on grain length in natural and BCF₅ rice populations. a Manhattan plot of grain length using SV-GWAS on chromosome 5. b SV results in the presence-and-absence of Nipponbare (NIP) and wild rice (SL1). c Haplotype frequencies in cultivated and wild rice. d Comparison of the grain length variation between the two haplotypes of the 8598-bp SV. e Phenotypic characterization of two lines in the BCF₅ population. f Scanning electron microscopic analysis of the outer spikelet hull surfaces of two lines in the BCF₅ population. g Expression profiling of LOC_Os03g27510