Genetic regulation of transcriptional variation in naturalArabidopsis thalianaaccessions

Abstract

An increased knowledge of the genetic regulation of expression in Arabidopsis thaliana is likely to provide important insights about the basis of the plant’s extensive phenotypic variation. Here, we reanalysed two publicly available datasets with genome-wide data on genetic and transcript variation in large collections of natural A. thaliana accessions. Transcripts from more than half of all genes were detected in the leaf of all accessions, and from nearly all annotated genes in at least one accession. Thousands of genes had high transcript levels in some accessions but no transcripts at all in others and this pattern was correlated with the genome-wide genotype. In total, 2,669 eQTL were mapped in the largest population, and 717 of them were replicated in the other population. 646 cis-eQTLs regulated genes that lacked detectable transcripts in some accessions, and for 159 of these we identified one, or several, common structural variants in the populations that were shown to be likely contributors to the lack of detectable RNA-transcripts for these genes. This study thus provides new insights on the overall genetic regulation of global gene-expression diversity in the leaf of natural A. thaliana accessions. Further, it also shows that strong cis-acting polymorphisms, many of which are likely to be structural variations, make important contributions to the transcriptional variation in the worldwide A. thaliana population.

Transcriptome-based prediction of complex traits in maize

“…To determine the degree to which the importance of a transcript correlated with the importance of trans- regulatory variants, significant eQTLs (multiple testing corrected p < 0.05) were identified for each transcript using the linear regression (modelLINEAR) approach from MatrixeQTL implemented in R. Benjamini-Hochberg false discovery rate correction was used to adjust p for multiple testing and eQTLs were considered significant if adjusted p < 0.05. The distance for considering eQTL as cis was 1 mega base 50 , however, because <0.1% of eQTL identified were cis , all eQTL were analyzed together. The importance of an eQTL or the neighboring genetic marker located within a 2kb window of the eQTL with the greatest average importance score was compared to the importance of the transcript with the eQTL in question (T:eQTL pair).…”

Dual-trait pleiotropic analysis in highly stratified natural populations using genome-wide association summary statistics

“…We extracted the expression levels of 19 genes within a ± 20kb window around the top associated SNP using RNA-seq gene expression measurements from 140 accessions [14]. Among these, the distributions of 14 gene expression phenotypes significantly deviate from normality (Kolmogorov-Smirnov test statistic > 0.8), and these genes were filtered out due to potential unreliable measurements [18]. The remaining 5 genes were passed onto eQTL mapping at the discovered locus (Materials & Methods).…”

“…We downloaded this data together with their corresponding whole-genome SNP genotypes available as a part of the 1001 Genomes project [9, 10] to replicate our SMR findings. Following the quality control procedure in [18], we removed two accessions from the data (Alst 1 and Ws 2) due to missing genotype data and two accessions (Ann 1 and Got 7) due to their low transcript call rate (16,861 and 18,693 genes with transcripts as compared to the range of 22,574 to 26,967 for the other the accessions). The final dataset used for eQTL mapping included 1,347,036 SNPs with MAF above 0.05 and call-rate above 0.95 for 140 accessions, and corresponding RNA-seq derived FPKM-values for 33,554 genes.…”

“…Expression values for 19 genes around 20kb up/downstream of the top associated SNP were extracted from [14]. 14 genes did not pass Kolmogorov-Smirnov test (ks test statistics ¡ 0.8) were filtered out due to potential unreliable measurement mentioned in [18]. The remaining five genes were subsequently passed onto eQTL mapping using qtscore procedure in GenABEL [23].…”

Gene co-expression network connectivity is an important determinant of selective constraint

“…A recent reanalysis [71] of two existing datasets assaying gene expression among natural accessions of A. thaliana [72,73] observed that thousands of genes displayed clear present/absent expression among accessions. In contrast, when filtering our data using a similar approach, we did not find any genes displaying this pattern of expression variation (S3 Fig), an observation that we also confirmed in an independent P. tremula dataset [74], albeit containing substantially fewer genotypes.…”