A complete chromosome substitution mapping panel reveals genome-wide epistasis in Arabidopsis

Abstract

Chromosome substitution lines (CSLs) are tentatively supreme resources to investigate non-allelic genetic interactions. However, the difficulty of generating such lines in most species largely yielded imperfect CSL panels, prohibiting a systematic dissection of epistasis. Here, we present the development and use of a unique and complete panel of CSLs in Arabidopsis thaliana, allowing the full factorial analysis of epistatic interactions. A first comparison of reciprocal single chromosome substitutions revealed a dependency of QTL detection on different genetic backgrounds. The subsequent analysis of the complete panel of CSLs enabled the mapping of the genetic interactors and identified multiple two- and three-way interactions for different traits. Some of the detected epistatic effects were as large as any observed main effect, illustrating the impact of epistasis on quantitative trait variation. We, therefore, have demonstrated the high power of detection and mapping of genome-wide epistasis, confirming the assumed supremacy of comprehensive CSL sets.

Fig. 1. A complete set of Arabidopsisthaliana chromosome substitution lines.

The complete panel of 32 CSLs can be divided into two reciprocal recurrent backgrounds (vertical dashed line) and subgroups of parental genotypes, single CSLs and CSLs in which two chromosomes are exchanged (horizontal dashed lines). Arabidopsis genomes of each of the CSLs are represented by five homozygous chromosomes derived from either the Col-0 (orange) or Ler (purple) accession.

Fig. 2. Mapping and validation of single-chromosome substitution effects.

Flowering time (A) and main stem length (B) of sCSLs and their recurrent parents. Each dot represents the spatial corrected trait value of an individual of the genotype indicated below the x-axis. Horizontal bars indicate BLUPs with 95% confidence intervals shown as vertical bars (Supplementary Table S15). Asterisks denote significant effects. QTL plots for flowering time (C) and main stem length (D) as mapped in a RIL population. –log10(P) values for each chromosome are displayed in different colours, while the horizontal red dashed line represents the significance threshold. Support intervals for the QTLs are indicated by coloured bars according to effect sign (orange: +^Col-0, and purple: +^Ler). The x-axis indicates chromosome numbers below a rug profile of the marker positions in cM distance. Heatmap plots of the effect strength of reciprocal chromosome five introgression NILs on flowering time (E) and chromosome two introgression NILs on main stem length (F). In both panels the upper row represents NIL mapping in a Col background, whereas the lower row represents NIL mapping in a Ler background. Vertical lines indicate marker positions in cM. Colour intensity from yellow to red specifies the strength of significant effects. Dashed lines with brackets below the heatmap indicate support intervals. FLC and ERECTA tick marks indicate the position of obvious candidate genes explaining variation in flowering time and main stem length, respectively.

Fig. 3. Detection of interchromosomal interaction effects in a complete CSL panel.

Notched box-and-whisker plots of the complete panel of CSLs for flowering time (A) and main stem length (B). Each dot represents the spatial corrected trait value of an individual plotted in relation to all other individuals (grey boxes) or categorised according to its genotype for the chromosome indicated on the X-axis (orange boxes: Col; purple boxes: Ler). C–G Regression predicted effect plots of epistatic interactions identified with backward selection models. C–E Two-way interactions explaining variation in flowering time. F Two-way interaction explaining variation in main stem length. G Three-way interaction explaining variation in main stem length. Error bars represent the 95% confidence intervals of the predicted effect.