Presence-absence variation in A. thaliana is primarily associated with genomic signatures consistent with relaxed selective constraints

Abstract

The sequencing of multiple genomes of the same plant species has revealed polymorphic gene and exon loss. Genes associated with disease resistance are overrepresented among those showing structural variations, suggesting an adaptive role for gene and exon presence-absence variation (PAV). To shed light on the possible functional relevance of polymorphic coding region loss and the mechanisms driving this process, we characterized genes that have lost entire exons or their whole coding regions in 17 fully sequenced Arabidopsis thaliana accessions. We found that although a significant enrichment in genes associated with certain functional categories is observed, PAV events are largely restricted to genes with signatures of reduced essentiality: PAV genes tend to be newer additions to the genome, tissue specific, and lowly expressed. In addition, PAV genes are located in regions of lower gene density and higher transposable element density. Partial coding region PAV events were associated with only a marginal reduction in gene expression level in the affected accession and occurred in genes with higher levels of alternative splicing in the Col-0 accession. Together, these results suggest that although adaptive scenarios cannot be ruled out, PAV events can be explained without invoking them.

Keywords: Arabidopsis; adaptive evolution; exon deletion; presence–absence variation; transposable elements; whole genome evolution.

Fig. 1.

Distribution of E-PAV genes (n = 330)—those with at least one, but not all, exons missing in at least one accession—by GOslim categories for molecular function (a), biological process (b) and cellular component (c), and by family (d). Both expected and observed number of E-PAV genes per category represented on each bar. Where there is a significant enrichment (P ≤ 0.05) between the amount of observed and expected E-PAV genes for a particular category, an asterisk is shown over the bars. Only categories with at least one E-PAV gene are shown.

Fig. 2.

Genetic features associated with intact (having no exons under P/A variation), E-PAV (having at least one, but not all, exons missing in at least one accession), and CDS-PAV (having the entire CDS missing in at least one accession) genes. From left to right, top to bottom: dN/dS, age, expression level, Tajima’s D, number of paralogs, and tau. See supplementary table S3, Supplementary Material online, for values of means and statistical analysis.

Fig. 3.

Genomic context for intact (having no exons under P/A variation), E-PAV (having at least one, but not all, exons missing in at least one accession), and CDS-PAV (having the entire CDS missing in at least one accession) genes. Averaged values for the genes in each set are given for, from top to bottom, the intergenic distance, the percentage of TE bases in the nongenic sequence of a 10-kb window centered on that gene’s midpoint, and the percentage of recombinogenic motifs in the genic sequence of a 1-kb window centered on that gene’s midpoint. See also supplementary tables 3 and 4 (Supplementary Material online) for the values of specific TE families and other window sizes.

Fig. 4.

Distribution of Z scores for standardized transcript abundance data in the affected accession. Data show that 210 genes that have one or more missing exons in only one of 17 A. thaliana accessions (relative to Col-0).