Comparative transcriptomics reveals patterns of selection in domesticated and wild tomato

Abstract

Although applied over extremely short timescales, artificial selection has dramatically altered the form, physiology, and life history of cultivated plants. We have used RNAseq to define both gene sequence and expression divergence between cultivated tomato and five related wild species. Based on sequence differences, we detect footprints of positive selection in over 50 genes. We also document thousands of shifts in gene-expression level, many of which resulted from changes in selection pressure. These rapidly evolving genes are commonly associated with environmental response and stress tolerance. The importance of environmental inputs during evolution of gene expression is further highlighted by large-scale alteration of the light response coexpression network between wild and cultivated accessions. Human manipulation of the genome has heavily impacted the tomato transcriptome through directed admixture and by indirectly favoring nonsynonymous over synonymous substitutions. Taken together, our results shed light on the pervasive effects artificial and natural selection have had on the transcriptomes of tomato and its wild relatives.

Keywords: abiotic stress; biotic stress; domestication.

Fig. 1.

Diversity in cultivated and wild tomatoes. (A) Bayesian relaxed-clock consensus chronogram, and examples of fruit and leaf divergence among tomato and wild relatives; nodes on the tree correspond to median branch lengths and blue bars represent 95% Bayesian confidence interval. (B) Distribution of mean distance to adjacent gene, larger distances are associated with centromeric sequences. (C) Single rate dS (gray) and single rate dN/dS (orange). (D) Frequency of expressed genes (red) and genes differentially expressed between tomato relatives (black). All plots reflect sliding windows (mean of 100 gene windows).

Fig. 2.

Evidence for increased nonsynonymous substitution rate in S. lycopersicum and S. galapagense. (A) Fraction of species-specific derived mutations in the coding regions that are nonsynonymous. (B) Distributions of dN/dS estimates from 1,000 bootstraps of the transcriptome-wide alignment for the whole tree (w.t.) and the branches labeled with red, blue, and yellow in Fig. 1. S. lyc, S. lycopersicum; S. gal, S. galapagense; S. pim, S. pimpinellifolium; S. chm, S. chmielewskii; S. hab, S. habrochaites; S. pen, S. pennellii.

Fig. 3.

Interspecific variation in expression. (A) Neighbor-joining tree built from the number of pairwise differentially expressed genes compared with (B) the unrooted genetic tree from Fig. 1A. The scale bar in A is for the number of differentially expressed genes and the scale bar in B is the expected number of substitutions per site. (C) Heatmap depicting scaled expression values of genes separated into two groups by significant contrasts (SI Appendix, SI Materials and Methods). The numbers correspond to the branch of the tree on which the changes are assumed to have occurred. (D) Product/substrate redox ratio of NAD(P)-linked reactions, calculated as described by refs. and . Black and gray indicate redox value of isocitrate dehydrogenase and malate dehydrogenase reactions, respectively. 2OG: 2-oxoglutarate. (E and F) Number of differentially expressed genes showing evidence of accelerated expression divergence (two-rate Brownian motion fit better than one-rate Brownian motion and OU) at ΔAIC > 4 or ΔAIC > 10, respectively. (G and H) Proportion of differentially expressed genes unique to each lineage (as determined by pairwise contrasts) showing evidence of accelerated expression divergence at ΔAIC > 4 or ΔAIC > 10, respectively. RvG indicates genes that show contrasting rates of evolution in the red and green fruited lineages.

Fig. 4.

Differential expression in S. lycopersicum and S. pennellii. (A) PCA factorial maps showing the largest components of variance, which separate samples by tissue (PC1 and PC2) and species (PC3). Triangles represent S. pennellii samples and circles S. lycopersicum. (B) Heat map comparing scaled expression values for S. lycopersicum (Top) and S. pennellii (Bottom) across tissues. Darker blue indicates higher expression. (C) PCA on log fold-changes by tissue. PC1 explains variance relating to global shifts in gene expression. (D) The remaining PCs describe tissue-specific shifts. Q3 and Q1 indicate the upper and lower quantile of the data, respectively (inf, inflorescence meristem; lf, leaf; rt, root; sd, aerial seedling; st, stem; veg, vegetative meristem).

Fig. 5.

Evolution of coexpression networks in S. lycopersicum and S. pennellii. (A) Heatmap depicting expression of genes assigned to the three modules in both species. Scaled log₂ expression values are shown with yellow and blue indicating high and low expression respectively. Green, yellow and purple bars indicate membership in the three identified transcription modules. (B) Global depiction of conserved coexpression network components. Three clear clusters that correspond to the three major modules are evident. (C) Comparison of connectivity (sum of the absolute correlation of expression with all other genes) for genes in the two species. Black indicates a low density of points and red indicates a high density. Connectivity is positively correlated, but the highest values are increased in S. pennellii. White dashed line indicates a slope of 1. (D) Intramodule connectivity for each module in each species. Yellow boxes are S. pennellii values and blue S. lycopersicum. (E) Fraction of differential edges specific to S. pennellii (yellow) and S. lycopersicum (blue) for each module. (F) Heatmap showing the change in connectivity for all gene pairs in the green module. Red indicates correlations that are found only in S. pennellii, blue indicates correlations found only in S. lycopersicum, and yellow indicates correlations of approximately the same strength in both networks.

Fig. 6.

cis and trans expression divergence. (A) Transgressive (light green) and S. pennellii-like (dark green) trans regulation relationships with genes on IL4-3. (B) Genotype by gene in IL4-3 (M82, blue and S. pennellii, green). (C) Log fold-change for differentially expressed genes between IL4-3 and M82. The black line indicates 0. (D) Correlation of log fold-change in expression between M82 and either the IL or PEN in sliding windows. A strong increase is seen on chromosome 4 indicating higher correspondence in gene expression between S. pennellii and the introgression line. (E) Heatmap showing median disimilarity (polymorphisms per 1,000 bp) between M82 and Heinz (green), M82 and S. pimpinellifolium (red), and Heinz and S. pimpinellifolium (blue). Increasing dissimilarity is indicated by lighter color; increasing similarity is indicated by darker color. Introgressions into one of the two cultivated lines show increased polymorphism rate between Heinz and M82 and decreased polymorphism between S. pimpinellifolium and the acceptor cultivated line (see chromosome 5). Shared introgressed segments show decreased polymorphism between S. pimpinellifolium and both cultivated lines. Many of these introgressions are larger in M82 (see chromosome 11). Large introgressions were frequently associated with centromeres, likely resulting from the increased linkage drag in these regions during breeding. Sliding window size for B, D, and E is 100 genes. (F) Number of expression changes in cis or trans to the IL4-3 introgression. Light green indicated transgressive changes and dark green indicates S. pennellii like expression. (G and H) Abundance of pectic monosaccharides galacturonic acid (GalA), glucoronic acid (GlcA), arabinose (Ara), and galactose (Gal) present in the walls (AIR, alcohol insoluble residue) of roots from M82 (black) and S. pennellii (white). Red asterisk: P < 0.001.