Evolution of Cis-Regulatory Elements and Regulatory Networks in Duplicated Genes of Arabidopsis

Abstract

Plant genomes contain large numbers of duplicated genes that contribute to the evolution of new functions. Following duplication, genes can exhibit divergence in their coding sequence and their expression patterns. Changes in the cis-regulatory element landscape can result in changes in gene expression patterns. High-throughput methods developed recently can identify potential cis-regulatory elements on a genome-wide scale. Here, we use a recent comprehensive data set of DNase I sequencing-identified cis-regulatory binding sites (footprints) at single-base-pair resolution to compare binding sites and network connectivity in duplicated gene pairs in Arabidopsis (Arabidopsis thaliana). We found that duplicated gene pairs vary greatly in their cis-regulatory element architecture, resulting in changes in regulatory network connectivity. Whole-genome duplicates (WGDs) have approximately twice as many footprints in their promoters left by potential regulatory proteins than do tandem duplicates (TDs). The WGDs have a greater average number of footprint differences between paralogs than TDs. The footprints, in turn, result in more regulatory network connections between WGDs and other genes, forming denser, more complex regulatory networks than shown by TDs. When comparing regulatory connections between duplicates, WGDs had more pairs in which the two genes are either partially or fully diverged in their network connections, but fewer genes with no network connections than the TDs. There is evidence of younger TDs and WGDs having fewer unique connections compared with older duplicates. This study provides insights into cis-regulatory element evolution and network divergence in duplicated genes.

Figure 1.

Regulatory protein footprint numbers differ in classes of duplicates. A, Box plots show the number of footprints within 500 bp of the transcription start site in different classes of duplicates as well as Brassicaceae-specific genes (BSG) and singletons. Purple Xs show the mean (Kruskal-Wallis test and Mann-Whitney U test P values are significant for all comparisons, P < 0.01). B, Box plots show the difference in the number of footprints between two genes in a duplicate pair within each duplicate class. Purple Xs show the mean (Kruskal-Wallis test and Mann-Whitney U test P values are significant for all comparisons, P < 0.01).

Figure 2.

WGD and TD pairs have differing wiring outcomes following duplication. A, Network rewiring between pairs may not change after duplication (conserved), pairs may have some common and some unique connections (partly diverged), or they may have no connections in common (diverged). Examples from each scenario from the two classes of duplicates are presented; there are no conserved examples from the WGDs. Within the specific examples, purple and blue circles represent WGD and TD pairs, respectively. Red and green circles are genes and interactions unique to each gene within the pair, and gray circles are shared genes and interactions. Lines with arrows indicate regulatory interactions, with the arrow pointing toward the target gene. PPD1, PEAPOD1; PPD2, PEAPOD2. B, Examples of a partially diverged WGD and TD pair. Purple circles are WGD pair, and blue are TD pair. Red and green circles are genes and interactions unique to each gene within the pair, and gray circles are shared genes and interactions. Lines with arrows indicate regulatory interactions, with the arrow pointing toward the target gene. C, Connectivity composition in the WGD, tandem, and other duplicates sets. The number of gene pairs is in brackets. D, The number of unique edges between members of a duplicate pair.

Figure 3.

Relationships between divergence in connectivity and duplication age. Box plots of the average number of unique edges between duplicate pairs within binned K_s ranges. A, TDs. B, Other duplicates. C, WGDs (asterisks indicate Mann-Whitney U test P values < 0.05).