A genome resource for green millet Setaria viridis enables discovery of agronomically valuable loci

Abstract

Wild and weedy relatives of domesticated crops harbor genetic variants that can advance agricultural biotechnology. Here we provide a genome resource for the wild plant green millet (Setaria viridis), a model species for studies of C₄ grasses, and use the resource to probe domestication genes in the close crop relative foxtail millet (Setaria italica). We produced a platinum-quality genome assembly of S. viridis and de novo assemblies for 598 wild accessions and exploited these assemblies to identify loci underlying three traits: response to climate, a 'loss of shattering' trait that permits mechanical harvest and leaf angle, a predictor of yield in many grass crops. With CRISPR-Cas9 genome editing, we validated Less Shattering1 (SvLes1) as a gene whose product controls seed shattering. In S. italica, this gene was rendered nonfunctional by a retrotransposon insertion in the domesticated loss-of-shattering allele SiLes1-TE (transposable element). This resource will enhance the utility of S. viridis for dissection of complex traits and biotechnological improvement of panicoid crops.

Fig. 1

a, S. viridis in its common, highly disturbed habitat next to a road. b, Diversity panel resequencing statistics: average library coverage, contig N50 (Kb), assembly size and number of genes considered present per library. The red vertical line in the lower right panel represents the number of genes necessary for a library to be included for PAV analysis (n = 39,000).

Fig. 2. PAV and SNP diversity of subpopulations.

a, Geographic distribution and population assignment of North American accessions based on SNP data. b, Geographic distribution and population assignment of Eurasian accessions based on SNP data. c, PCA of SNP data, showing placement of North American samples among Eurasian ones. d, PCA of PAV variants, excluding admixed individuals. e, PCA of SVs, excluding admixed individuals. *, accession is admixed; black rim on circle, accession is in its native range.

Fig. 3. GWAS and cloning of Les1.

a, Manhattan plot of GWAS results, showing a single QTL (peak −log₁₀ P > 30) for seed shattering on chromosome 5. The red line indicates experiment-wise P = 0.01 level after Bonferroni correction. The GWAS used a univariate mixed linear model from GEMMA, with centered kinship matrix. Wald test P value was used for assessing significant peaks, but other P value estimates give similar results. b, Zoom-in on the chromosome 5 peak. Larger dots represent missense SNPs identified by snpEff. Different colors of missense SNPs indicate PROVEAN score range (blue for ≥2.5, green for ≤2.5 and ≥4.1, red for ≤4.1; −2.5 and −4.1 represent 80% and 90% specificity). Lower scores indicate higher likelihood of deleterious effects of the mutation. c, Table of Les1 alleles in S. viridis (Sv) and S. italica (Si), with structural characteristics, background line and shattering phenotype. d, Sanger sequence validating the position of the adenine insertion (frameshift) in SvLes1-CRISPR1.

Fig. 4. Phenotypic characterization of SvLes1-CRISPR1 mutants and naturally occurring alleles.

a, Force required to break the AZ (tensile strength in grams of force) in SvLes1-1, SvLes1-2 and SvLes1-CRISPR1 lines, measured every 2 d starting at 8 d after heading. For each genotype, n = 3 plants (biological replicates) with two inflorescences per plant (technical replicates) and 20 spikelets per inflorescence. Data are presented as median with 25th and 75th percentiles (boxes); whiskers reach values up to 1.5× the interquartile range above and below the hinges; filled circles are outliers. Significance values computed by analysis of variance for each date. *P = 0.0017; ***P < 2.2 × 10⁻¹⁶. b, Image of S. viridis high-shattering (SvLes1-1 in ME034; left two panicles) and SvLes1-CRISPR1 mutant (right) panicles 7 weeks after heading. Scale bar, 1 cm. c, Seed dispersal distances in a wind tunnel, measured from four independent high-shattering plants (SvLes1-1 in ME034) and eight SvLes1-CRISPR1 plants at week 6. Data are presented as means ± 1 s.d. (boxes); lines are ± 2 s.d.; filled circles are outliers.

Fig. 5. Phenotype and mapping of small leaf angle and loss of ligule.

a, A10.1 (left) and TB159 (right) showing small leaf phenotype in TB159 relative to A10.1. Scale, 2 cm. b, Junction of leaf and blade for A10.1 (left) and TB159 (right) showing well-developed ligule in A10.1 and lack of ligule in TB159. Scale, 5 mm. c, BSA mapping result, red arrow indicating QTL. Wald test P value was used to assess significance, but other P value estimates give similar results. d, Synteny analysis around SvLg2 and maize lg2 locus. Screenshot from analysis with CoGE (https://genomevolution.org/r/1ei1s).