Waxy phenotype evolution in the allotetraploid cereal broomcorn millet: mutations at the GBSSI locus in their functional and phylogenetic context

Abstract

Waxy mutants, in which endosperm starch contains ~100% amylopectin rather than the wild-type composition of ~70% amylopectin and ~30% amylose, occur in many domesticated cereals. The cultivation of waxy varieties is concentrated in east Asia, where there is a culinary preference for glutinous-textured foods that may have developed from ancient food processing traditions. The waxy phenotype results from mutations in the GBSSI gene, which catalyzes amylose synthesis. Broomcorn or proso millet (Panicum miliaceum L.) is one of the world's oldest cultivated cereals, which spread across Eurasia early in prehistory. Recent phylogeographic analysis has shown strong genetic structuring that likely reflects ancient expansion patterns. Broomcorn millet is highly unusual in being an allotetraploid cereal with fully waxy varieties. Previous work characterized two homeologous GBSSI loci, with multiple alleles at each, but could not determine whether both loci contributed to GBSSI function. We first tested the relative contribution of the two GBSSI loci to amylose synthesis and second tested the association between GBSSI alleles and phylogeographic structure inferred from simple sequence repeats (SSRs). We evaluated the phenotype of all known GBSSI genotypes in broomcorn millet by assaying starch composition and protein function. The results showed that the GBSSI-S locus is the major locus controlling endosperm amylose content, and the GBSSI-L locus has strongly reduced synthesis capacity. We genotyped 178 individuals from landraces from across Eurasia for the 2 GBSSI and 16 SSR loci and analyzed phylogeographic structuring and the geographic and phylogenetic distribution of GBSSI alleles. We found that GBSSI alleles have distinct spatial distributions and strong associations with particular genetic clusters defined by SSRs. The combination of alleles that results in a partially waxy phenotype does not exist in landrace populations. Our data suggest that broomcorn millet is a system in the process of becoming diploidized for the GBSSI locus responsible for grain amylose. Mutant alleles show some exchange between genetic groups, which was favored by selection for the waxy phenotype in particular regions. Partially waxy phenotypes were probably selected against-this unexpected finding shows that better understanding is needed of the human biology of this phenomenon that distinguishes cereal use in eastern and western cultures.

Fig. 1.

Panicum miliaceum starch granules. Starch was stained with Lugol’s solution and observed using a light microscope. The scale is indicated. (A) Material scraped from mature seeds. The genotypes and accessions shown are i: S₀/L_C, MIL-4 #1 (nonwaxy, dark blue–black staining); ii: S_-15/L_C, line P017-10-2 (partially waxy, some granules staining red and some granules staining paler blue–purple, indicating the presence of amylose); iii: S_-15/L_f, MIL-82 #1 (waxy, granules stain red, some darkly. The characteristic blue of amylose staining is absent). (B) Pollen squashed on a microscope slide to release some of the starch granules within. The genotypes and accessions shown are i: S₀/L_C, MIL-4 #1 (blue–black staining, amylose present); ii: S_-15/L_f, MIL-70 #1 (red staining, amylose free).

Fig. 2.

Biochemical properties of endosperm starch in six GBSSI genotypes of P. miliaceum. (A) Amylose content. (B) Starch swelling power. (C) GBSSI protein content. (D) Starch synthase activity.

Fig. 3.

Sections of grains with low-amylose content. Dry-cut sections of 1.5 μm of mature endosperm of genotype S_-15/L_C stained with Lugol’s solution. Examples of the outer endosperm (OE; left panel), including the subaleurone cells (sa); mid endosperm further into toward the center of the grain (ME, middle panel); and the central endosperm (CE; right panel) are shown. The scale is indicated.

Fig. 4.

Microsatellite genotype clusters defined by InStruct. Proportional allocations for each plant sample to each gene pool for the InStruct K = 2 (A) and K = 7 (B) models. Alleles at the GBSSI-S and GBSSI-L loci are shown.

Fig. 5.

Microsatellite genotype clusters and GBSSI allele distribution. For each sample, the pie chart shows the proportional allocation to each gene pool under the K = 7 model. The alleles at the GBSSI-S and GBSSI-L loci are shown superimposed.

Fig. 6.

Dendrograms showing microsatellite genotype clusters and GBSSI alleles. Neighbor-joining tree showing relationships among samples based on microsatellite genotypes, using Nei’s genetic distances (Nei et al. 1983). Branches are colored according to the highest proportional allocation to the gene pools identified under the InStruct analysis in the K = 7 model (even where this is <50%). The GBSSI genotype is shown for each sample at the GBSSI-S and GBSSI-L loci. Where multiple individuals share a microsatellite and GBSSI genotype, the number of individuals is indicated in brackets. The genotype of individuals heterozygous for the GBSSI-L locus is shown as both alleles separated by /.