Phenotype loss is associated with widespread divergence of the gene regulatory landscape in evolution

Abstract

Detecting the genomic changes underlying phenotypic changes between species is a main goal of evolutionary biology and genomics. Evolutionary theory predicts that changes in cis-regulatory elements are important for morphological changes. We combined genome sequencing, functional genomics and genome-wide comparative analyses to investigate regulatory elements in lineages that lost morphological traits. We first show that limb loss in snakes is associated with widespread divergence of limb regulatory elements. We next show that eye degeneration in subterranean mammals is associated with widespread divergence of eye regulatory elements. In both cases, sequence divergence results in an extensive loss of transcription factor binding sites. Importantly, diverged regulatory elements are associated with genes required for normal limb patterning or normal eye development and function, suggesting that regulatory divergence contributed to the loss of these phenotypes. Together, our results show that genome-wide decay of the phenotype-specific cis-regulatory landscape is a hallmark of lost morphological traits.

Fig. 1

Comparative framework to detect sequence divergence of regulatory elements in snakes. a Comparison of the tegu genome to genomes of other sequenced reptiles. N(x)% graph showing the contig and scaffold size (y-axis), where x% of the genome consists of contigs and scaffolds of at least that size. b Phylogenetic tree of the limbless and limbed species included in our multiple genome alignment. Snakes are in orange, remaining Episquamata species in green, remaining Sauropsida species in brown, and outgroup species in gray. All animal illustrations were taken from phylopic.org. c Overview of the computational steps to identify CNEs that are specifically diverged in snakes. d Filled gray and green density curves depict global and local Z-score distributions of 141,610 CNEs, respectively (22,812 CNEs with missing genomic sequence are excluded). Dashed curves depict the Z-score null distributions obtained by simulating the evolution of CNEs under selection in all amniote species

Fig. 2

Snake-diverged CNEs are associated with limb genes and limb regulatory elements. a Snake-diverged CNEs are significantly associated with genes annotated with limb-related functions. Left: Bars depict Benjamini & Hochberg adjusted p-values derived by a one-sided Fisher’s exact test. Middle: Observed (orange vertical bar) and expected number (gray violin plots, based on 10,000 random subsets sampled from all CNEs) of snake-diverged CNEs associated with genes in each set (the thick box inside the violin plot indicates the first quartile, the median and the third quartile). The Z-score measuring the number of standard deviations that the observed number is above the random expectation is indicated. Right panel: Many snake-diverged CNEs (dots) are far away from the transcription start site of genes in these sets. CNEs associated with the same gene have the same color. b Snake-diverged CNEs significantly overlap regulatory elements active in embryonic limb tissue of tegu lizard, green anole lizard, and mouse. Orange bars correspond to limb regulatory datasets. ATAC-seq datasets generated in this study are highlighted in green. Remaining visual representation as in a. All animal illustrations were taken from phylopic.org. c Genome browser screenshot shows that snake-diverged CNEs (orange) overlap tegu limb-specific ATAC-seq peaks. ATAC-seq signal tracks of limb, brain, and liver are shown. d Snake-diverged CNEs are broadly distributed in the limb patterning network. An asterisk marks the genes which are associated with at least one snake-diverged CNE that overlaps a limb regulatory element

Fig. 3

CNEs diverged in subterranean mammals are associated with eye genes and eye regulatory elements. a Phylogenetic tree of species in our multiple genome alignment. Subterranean mammals are in orange, mammals without degenerated eyes are in black, and outgroup species are in gray. b Difference in the mean percent sequence identity between 52 CNEs overlapping the promoter regions of 64 eye-specific genes, which are diverged or lost in subterranean mammals, and 18,033 CNEs overlapping promoters of other genes. Compared to other mammals, the subterranean mammals have substantially higher divergence in the CNEs overlapping the 64 eye-specific gene promoters. Promoters are defined as ±1.5 Kb around the transcription start site. c Three examples of diverged CNEs located in the promoter of diverged lens-specific genes. d Diverged CNEs are enriched near genes that lead to lens defects in a mouse knockout (top 10 enrichments are shown here; Supplementary Table 14 lists all enrichments). Lens-related knockout phenotypes are shown in orange. Left: Bars depict Benjamini & Hochberg adjusted p-values derived by a one-sided Fisher’s exact test. Middle: Observed (orange vertical bar) and expected number (gray violin plots, based on 10,000 random subsets sampled from all CNEs) of diverged CNEs associated with phenotypes in each set (the thick box inside the violin plot indicates the first quartile, the median and the third quartile). The Z-score measuring the number of standard deviations that the observed number is above the random expectation is indicated. Right panel: Many of the diverged CNEs (dots) are far away from the transcription start site of genes in these sets. CNEs associated with the same gene have the same color. e Diverged CNEs significantly overlap regulatory elements active in whole eye, retina, and lens. Orange bars indicate eye regulatory datasets. ATAC-seq datasets generated in this study are highlighted in green. Remaining visual representation as in d. All animal illustrations were taken from phylopic.org

Fig. 4

Sequence divergence results in widespread loss of transcription factor (TF) binding sites. a Difference in the median motif scores of limb-related TFs between 933 snake-diverged CNEs that overlap limb regulatory data and all other CNEs (excess score, x-axis). A positive excess score as observed for limbed species reflects a preference to preserve motifs of limb-related TFs. P-values comparing the distribution of motif scores of diverged and all other CNEs were computed with a one-sided Wilcoxon rank sum test and corrected for multiple testing. b Mutations in snakes in a CNE overlapping a limb-specific ATAC-seq peak weaken or destroy binding sites for key limb transcription factors. The binding preference of the TFs is visualized as a sequence logo, motif scores were computed with SWAN. c As panel a, but using motifs of eye-related TFs to compare CNEs that are diverged in subterranean mammals and overlap eye regulatory data with all other CNEs. d Mutations in two subterranean mammals in a CNE overlapping a lens-specific ATAC-seq peak weaken a binding site for c-Maf, a crystallin-inducing TF that is required for lens fiber cell differentiation