A Multiparent Advanced Generation Inter-Cross to fine-map quantitative traits in Arabidopsis thaliana

Abstract

Identifying natural allelic variation that underlies quantitative trait variation remains a fundamental problem in genetics. Most studies have employed either simple synthetic populations with restricted allelic variation or performed association mapping on a sample of naturally occurring haplotypes. Both of these approaches have some limitations, therefore alternative resources for the genetic dissection of complex traits continue to be sought. Here we describe one such alternative, the Multiparent Advanced Generation Inter-Cross (MAGIC). This approach is expected to improve the precision with which QTL can be mapped, improving the outlook for QTL cloning. Here, we present the first panel of MAGIC lines developed: a set of 527 recombinant inbred lines (RILs) descended from a heterogeneous stock of 19 intermated accessions of the plant Arabidopsis thaliana. These lines and the 19 founders were genotyped with 1,260 single nucleotide polymorphisms and phenotyped for development-related traits. Analytical methods were developed to fine-map quantitative trait loci (QTL) in the MAGIC lines by reconstructing the genome of each line as a mosaic of the founders. We show by simulation that QTL explaining 10% of the phenotypic variance will be detected in most situations with an average mapping error of about 300 kb, and that if the number of lines were doubled the mapping error would be under 200 kb. We also show how the power to detect a QTL and the mapping accuracy vary, depending on QTL location. We demonstrate the utility of this new mapping population by mapping several known QTL with high precision and by finding novel QTL for germination data and bolting time. Our results provide strong support for similar ongoing efforts to produce MAGIC lines in other organisms.

Figure 1. The distribution of the number of distinct founder genomes contributing to a ML.

Each ML is descended through a funnel mating design from up to 16 distinct founder genomes. This histogram shows the fraction of lines descended from a given number of distinct founders.

Figure 2. Distribution of the minor allele frequency for the 1,260 genotyped SNPs.

(A) in the 19 founders; (B) in 527 MLs.

Figure 3. Distribution of allele and 10-SNP Haplotype sharing among the 19 founders and the MLs.

Sharing between and within F4 families are plotted separately.

Figure 4. Genome-wide properties of the MLs.

In each panel the x-axis represents the complete 120 Mb genome of A. thaliana, with vertical red lines marking the chromosome boundaries and the pale blue vertical bars indicating the centromeres. (A) Number of 10-SNP haplotypes observed among founders (black) and MLs (red) across the genome. (B) The maximum posterior founder probability, m_Li at locus L, averaged across all MLs i. (C) The maximum posterior founder probability, m_Li for the ML i = “ML-100”. The vertical grey lines indicate probable recombination breakpoints where the identity of the most probable founder changes. (D) The posterior founder probabilities for ML-100. The vertical axis represents the 19 possible founders, s, in alphabetical order. The probability for founder s at locus L is represented by a grey bar at coordinate (L,s), the shade of grey varying from white (P = 0) to black (P = 1). (E) The locus-specific power to detect a QTL explaining 10% of the phenotypic variance, from 40,000 simulations. In each simulation a 10% QTL was placed randomly along the genome. Successful detection is defined as the event that the genome-wide maximum in the genome scan for the QTL is within 3 Mb of the true QTL location. (F) The locus-specific median mapping error for the successfully detected QTL in (E).

Figure 5. Distribution of the decay in mean LD (R²) as a function of distance between SNPs in the MLs.

Figure 6. Genome-wide patterns of LD (R²) in the MLs.

The chromosome boundaries are marked by black lines. The intensity of the LD between SNPs at loci x,y is indicated by the colour in the corresponding x,y coordinate, using the scale indicated in the legend.

Figure 7. Distributions of the mapping error in QTL location for QTL in which logP_MAX is 5, 10, 15, or 20.

Each curve is estimated from simulations as described in Materials and Methods. The width of the corresponding confidence interval is twice the mapping error. The horizontal dashed lines cut the distributions at the 50% (lower) and 90% (upper) points.

Figure 8. Examples of QTL scans.

The orange bars indicate the 90% confidence intervals for the identified QTL. (A,B) show QTL scans for bolting time on Chromosome 4; (A) Fixed effects (black) and mixed effects (red) logP, and (B) Hierarchical Bayes percentage of QTL variance (maroon); (C) fixed effects logP for the binary phenotype erecta on chromosome 2.