High-density haplotyping with microarray-based expression and single feature polymorphism markers in Arabidopsis

Abstract

Expression microarrays hybridized with RNA can simultaneously provide both phenotypic (gene expression) and genotypic (marker) data. We developed two types of genetic markers from Affymetrix GeneChip expression data to generate detailed haplotypes for 148 recombinant inbred lines (RILs) derived from Arabidopsis thaliana accessions Bayreuth and Shahdara. Gene expression markers (GEMs) are based on differences in transcript levels that exhibit bimodal distributions in segregating progeny, while single feature polymorphism (SFP) markers rely on differences in hybridization to individual oligonucleotide probes. Unlike SFPs, GEMs can be derived from any type of DNA-based expression microarray. Our method identifies SFPs independent of a gene's expression level. Alleles for each GEM and SFP marker were ascertained with GeneChip data from parental accessions as well as RILs; a novel algorithm for allele determination using RIL distributions capitalized on the high level of genetic replication per locus. GEMs and SFP markers provided robust markers in 187 and 968 genes, respectively, which allowed estimation of gene order consistent with that predicted from the Col-0 genomic sequence. Using microarrays on a population to simultaneously measure gene expression variation and obtain genotypic data for a linkage map will facilitate expression QTL analyses without the need for separate genotyping. We have demonstrated that gene expression measurements from microarrays can be leveraged to identify polymorphisms across the genome and can be efficiently developed into genetic markers that are verifiable in a large segregating RIL population. Both marker types also offer opportunities for massively parallel mapping in unsequenced and less studied species.

Figure 1.

Comparison of SFPdev distributions between RIL and parental GeneChips. The distribution of SFPdev values (x-axis) in the RIL population (gray bars) are compared to distributions in the parental Bay-0 and Sha GeneChips from the factorial experiments (black and white bars, respectively) for SFP marker At1g74090-8 with 24.3% missing data (when scored with the parental min-max algorithm).

Figure 2.

Heat map showing a matrix of pair-wise recombination values for SFPs and microsatellite markers assayed on 148 RILs. Pair-wise comparisons between markers were used to assign recombination scores, which are plotted as a heat map matrix of all markers (top to bottom) against all markers (left to right). Lowest recombination scores, suggesting marker linkage, are indicated by red boxes, while blue boxes indicate high recombination scores, suggesting no linkage. The 599 SFP and 38 microsatellite markers are ordered by their physical location in the Col-0 genomic sequence; heavy lines demarcate the five linkage groups. (Supplemental Fig. 5a shows a high resolution version of this figure, including marker details.)

Figure 3.

Haplotypes of 148 RILs plus parental genotypes. Each column represents a RIL (first 148 columns) or parental genotype (last four columns). Rows correspond to SFP and microsatellite markers, arranged (top to bottom) in physical order based on the Col-0 genomic sequence. The five Arabidopsis linkage groups are laid end-to-end and are separated by horizontal gray lines. Red boxes indicate Sha genotypes, blue boxes indicate Bay-0 genotypes, and gray boxes indicate markers scored as missing data. The final column depicts the physical distances between the markers. Below the haplotypes, the proportions of Sha (red bars) and Bay-0 (blue bars) alleles in each RIL are depicted. (A high resolution version of this figure, with marker names included, is available in Supplemental Fig. 6a).