Genome-wide comparative diversity uncovers multiple targets of selection for improvement in hexaploid wheat landraces and cultivars

Abstract

Domesticated crops experience strong human-mediated selection aimed at developing high-yielding varieties adapted to local conditions. To detect regions of the wheat genome subject to selection during improvement, we developed a high-throughput array to interrogate 9,000 gene-associated single-nucleotide polymorphisms (SNP) in a worldwide sample of 2,994 accessions of hexaploid wheat including landraces and modern cultivars. Using a SNP-based diversity map we characterized the impact of crop improvement on genomic and geographic patterns of genetic diversity. We found evidence of a small population bottleneck and extensive use of ancestral variation often traceable to founders of cultivars from diverse geographic regions. Analyzing genetic differentiation among populations and the extent of haplotype sharing, we identified allelic variants subjected to selection during improvement. Selective sweeps were found around genes involved in the regulation of flowering time and phenology. An introgression of a wild relative-derived gene conferring resistance to a fungal pathogen was detected by haplotype-based analysis. Comparing selective sweeps identified in different populations, we show that selection likely acts on distinct targets or multiple functionally equivalent alleles in different portions of the geographic range of wheat. The majority of the selected alleles were present at low frequency in local populations, suggesting either weak selection pressure or temporal variation in the targets of directional selection during breeding probably associated with changing agricultural practices or environmental conditions. The developed SNP chip and map of genetic variation provide a resource for advancing wheat breeding and supporting future population genomic and genome-wide association studies in wheat.

Keywords: SNP genotyping; breeding history; polyploid wheat; selection scans; wheat improvement.

Fig. 1.

Population structure of a worldwide collection of wheat accessions. The DAPC of populations of spring (triangles) and winter (circles) wheat from the Pacific Northwest of the United States (PNW), eastern United States, midwestern United States, central North America, China, Europe, and southeastern and western Australia. The SNP discovery panel and landraces are shown by red and yellow circles, respectively.

Fig. 2.

Wheat-genome selection scans. (A) Distribution of extreme F_ST and PHS values across the wheat genome. The map locations of Ppd-B1, Ppd-D1, Sr36, Rht-B1, Rht-D1, Vrn-A1, Vrn-B1, Vrn-D1, and FT QTL genes are shown on top. Chromosome boundaries are shown by vertical gray lines. F_ST was estimated between landraces and cultivars [F_ST(L/C)], between spring and winter wheat [F_ST(S/W)], and between populations within spring [F_ST(S)] and winter wheat [F_ST(W)]. (B) Distribution of single-locus χ² tests of segregation distortion across the wheat genome. The horizontal dashed line corresponds to the significance threshold P ≤ 1e⁻¹⁰. (C) Distribution of SNP density of across the wheat genome calculated for 10-cM sliding windows with 2-cM overlaps. (D) The graphical genotype of the Sr36 genomic region with SNP alleles shown in red and green.

Fig. 3.

Comparison of allele frequencies between cultivars and landraces. (A) Distribution of minor allele-frequency (MAF) counts in landraces and cultivars at combined, synonymous, and nonsynonymous SNP loci. (B) Distribution of joint allele-frequency density between cultivars and landraces.

Fig. 4.

Impact of improvement on LD and effective population size. (A and B) Boxplots showing the interquartile range of genetic distances for all pair-wise comparisons of either linked (A) or neighboring (B) SNP pairs grouped on the basis of the extent of LD (r²). Open and gray boxplots correspond to cultivars and landraces, respectively. Wilcoxon signed-rank test: NS, nonsignificant; *P ≤ 0.05; ***P ≤ 0.001. (C) Demographic model used to assess the size of the improvement bottleneck (α). N_A, N_L, and N_C are population sizes of ancestral, landrace, and cultivar populations. The population of cultivars underwent a population bottleneck T₁ generations ago. The population remains at α * N_A size for T_b = T₁ − T₂ generations. The density plot shows the posterior distribution of α with the peak at 0.94. For details see SI Methods.

Fig. 5.

Cumulative frequency of SNP alleles subjected to selection in wheat populations. Cumulative allele frequency distribution in the spring (Left) and winter (Right) wheat from different geographic regions was calculated for SNP alleles identified in selection scans. (Lower) Population-specific frequencies of the two SNP alleles showing evidence of selection.