Root restriction accelerates genomic target identification in quinoa under controlled conditions

Abstract

Quinoa (Chenopodium quinoa) is a nutritious and resilient crop that displays a high genetic and phenotypic variation. As the popularity of this crop increases, there is a growing need to integrate classic and modern breeding tools to favor its improvement. We tested root restriction as a method to reduce plant size and enable high-throughput phenotypic screening of large sets of quinoa plants under controlled conditions. We verified how increasing root restriction does not affect the prediction of field behavior with respect to other standard greenhouse cultivation procedures. We then combined the phenotypic information obtained with our root restriction system with whole-genome re-sequencing data to characterize a quinoa diversity panel of 100 accessions and showed that phenotypic data obtained from root-restricted plants provide real insights into quinoa genetics. Finally, we carried out a genome-wide association study (GWAS) and identified a previously described locus for betalain biosynthesis, as well as other candidate loci linked to betalain biosynthesis and seed size. Overall, we showed that a phenotyping system based on root restriction can aid the identification of genomic targets in quinoa, which can complement and inform field trials for certain traits. This work supports further breeding and faster improvement of quinoa.

FIGURE 1

Impact of root restriction on plant phenotypes. (a) Danish line Titicaca grown in a standard 2‐L container and a root‐restriction container. (b) Principal component analysis for combined overview of phenotype scores across experimental settings. Red circles: restricted‐1, pink circles: restricted‐2, yellow squares: standard, blue crosses: field. (c) Broad sense heritability computed with Cullis (orange) Piepho (blue) methods. Abbreviations: Weeks to flowering (WTF), dry biomass post‐harvest (PB), total seed weight per plant (SW), seed diameter (SD). (d‐g) One‐to‐one comparison matrixes for the phenotypic scores between different settings. For each trait, scores histograms, scatter distributions, Lowess polynomial regression and correlation eclipses are plotted. Significant Pearson (* p < 0.05, ** p < 0.01, *** p < 0.001) coefficients are highlighted in colors (red >0.7, dark orange >0.4, light orange >0.2). Abbreviations: greenhouse restricted‐1 (GRR1), restricted‐2 (GRR2), restricted‐1/2 (GRR12) and standard (GS), open field (OF).

FIGURE 2

Characterization of a root‐restricted quinoa diversity panel. (a) Genotype likelihood‐based population structure, visualizing principal components PC1, PC2 (left) and PC2, PC3 (right). Colors and shapes refer to 6 sub‐population clusters identified by k‐means. (b) Suggested geographical origin of each germplasm subpopulation, based on k‐means clustering and existing IPK annotation. Symbols and colors match k‐means clusters. Arrows show previously described domestication trajectories. (c) Admixture analysis using k = 2 (highland, coastal) and k = 6 (k‐means clusters), performed across all 124 accessions used to construct the population structure. Numbers indicate putative k‐means sub‐populations, colors indicate NGSadmix groups. Abbreviations: Inter‐andean valley (IAV) (d) PCA of combined phenotypic scores. Colors and shapes match the k‐means clusters. (e‐j) For each trait: boxplots representing the distribution of phenotypes (y‐axis) across individuals within each k‐means cluster (x‐axis). Kruskal‐Wallis significance scores are reported (NS: non‐significant, * p < 0.05, ** p < 0.01, **** p < 0.0001). Abbreviations: Weeks‐to‐flowering (WTF), Plant dry biomass after harvest (PB) expressed in g, Stem betalain content (BC) scored between 1–5 for individual replicates and normalized on a scale from 0 to 1, Seed weight per plant expressed in grams (SW), Hundred‐kernel weight (HKW) expressed in g, Seed diameter (SD) expressed in mm.

FIGURE 3

GWAS analysis by linear mixed model analysis using the likelihood ratio test in GEMMA. (a) Betalain content in stem (BC) was scored by visually assessing stem pigmentation. (b) Quantile‐quantile plot for stem pigmentation association analysis, comparing the distribution of observed p‐values against an expected normal distribution. (c) Manhattan plot for stem pigmentation association analysis depicting association p‐values as ‐log10(P) for all individual SNPs across the 18 chromosomes. Red and blue lines represent the Bonferroni and suggestive significance thresholds, respectively, adjusted for the actual number of independent tests. Black line represents the significance threshold for False Discovery Rate (FDR)‐adjusted p‐values. Local Manhattan plots of loci of interest are also shown. SNPs found within gene coding sequences are highlighted as follows: nonsense (magenta), missense (yellow), splicing (purple). Among the SNPs lying upstream/downstream (black) or in introns (green), only the ones above FDR significance threshold are shown. Linkage disequilibrium (LD) heatmaps for each region are included above the graphs, with red representing higher LD. (d) Quinoa accessions showing contrasting seed diameter (SD). (e) Quantile‐quantile plot for seed diameter association analysis. (f) Manhattan plot for seed diameter association analysis.