Complexity and conservation of regulatory landscapes underlie evolutionary resilience of mammalian gene expression

Abstract

To gain insight into how mammalian gene expression is controlled by rapidly evolving regulatory elements, we jointly analysed promoter and enhancer activity with downstream transcription levels in liver samples from 15 species. Genes associated with complex regulatory landscapes generally exhibit high expression levels that remain evolutionarily stable. While the number of regulatory elements is the key driver of transcriptional output and resilience, regulatory conservation matters: elements active across mammals most effectively stabilize gene expression. In contrast, recently evolved enhancers typically contribute weakly, consistent with their high evolutionary plasticity. These effects are observed across the entire mammalian clade and are robust to potential confounders, such as the gene expression level. Using liver as a representative somatic tissue, our results illuminate how the evolutionary stability of gene expression is profoundly entwined with both the number and conservation of surrounding promoters and enhancers.

Figure 1. Liver gene expression levels are highly conserved across 25 mammalian species

(a) Pairwise correlations of normalized expression levels for 17,475 one-to-one orthologous genes in livers isolated from 25 mammalian species show high conservation of gene expression. Shading of individual tiles in the heatmap depict pairwise correlation coefficients between species (Spearman’s Rho). Known phylogenetic relationships and species divergences are represented by an evolutionary tree (left of Y-axis), which includes twenty-three placental species (in four orders) and two marsupial species (in two orders). In bolded blue: species with higher-quality reference genomes; in grey: species with either draft or proxy reference genomes (excluded from analysis, Methods and Figure S2). (b-c) Housekeeping and core liver genes show slower expression divergence, compared to controls matched for gene expression levels. Pairwise correlation values were plotted against evolutionary distance for housekeeping (gold, b; 3,612 genes) and core liver genes (brown, c; 2,224 genes), and compared to the correlation values of control genes with the same distribution of mean expression levels across species (grey). Control genes were matched in expression to either housekeeping (b) or core liver genes (c), and are thus different sets for the two panels. Lines correspond to linear regression trends (after log transform of the time axis), with 95% confidence intervals in grey shading, and were added for visualisation purposes. Regression R² are reported in Table S2.

Figure 2. The number of promoters and enhancers corresponds with gene expression stability across evolution

(a) Genes are associated with all active regulatory elements sitting between their TSS and the TSS of the next gene on either side, within a limit of 1Mb. Regulatory elements sitting directly on the TSS of a gene (max. 5kb upstream and 1kb downstream) were exclusively associated to that gene (darker shading, exclusive TSS proximal). The cartoon example illustrates this procedure for three genes R, S, and T and their regulatory association domains ρ, σ and τ. (b) Numbers of promoters and enhancers associated to a gene are correlated across species. Shading of individual tiles corresponds to pairwise tie-corrected Spearman correlation coefficients for numbers of promoters and enhancers associated to orthologous genes across 15 mammalian species. (c) Examples of genes with simple (EIF1) and complex (APOB) regulatory landscapes in liver. Regulatory complexity was measured as the median number of promoters and enhancers associated to each gene across species. Histone modification ChIP-seq fold enrichments are shown in blue (H3K4me3) and orange (H3K27ac), and RNA-seq reads in green, for three representative species (human, mouse and dog). Numbers in blue and orange: maximum fold enrichment; numbers in green: gene expression values (TPM, normalised across species). (d) Expression distributions (mean expression across species) are shown for genes associated with increasing numbers of active promoters (purple) or enhancers (orange) in an average mammal. Active enhancers associated to a gene have an additive effect, whereas promoters show a more switch-like effect on gene expression levels. Classes containing fewer than thirty genes are greyed. (e) The number of associated promoters and enhancers contributes to evolutionary stability of gene expression. Grey insets: Expression divergence across species is compared between (i) genes associated to multiple promoters or enhancers (top) and (ii) control genes with the same expression level but associated to few promoters or enhancers (one or none, bottom). Plots: Pairwise Spearman correlation coefficients of expression levels between species were plotted against evolutionary distance for genes associated with multiple promoters (left; 1,688 genes) or enhancers (right; 1,479 genes), and compared to control gene sets. In both cases the number of associated promoters or enhancers corresponds to the median number across species. Lines are as described in Figure 1b-c.

Figure 3. Conserved regulatory activity is associated with both high and stable gene expression levels

(a) Example of gene expression and regulatory landscapes around the PROX1 gene in livers from ten placental mammals. Each row shows PROX1 expression (left, green background) and activity of promoters and enhancers around the PROX1 locus in one species (H3K4me3 (blue) and H3K27ac (orange) ChIP-seq signals, grey background; as described in Figure 2C). A placental-conserved promoter and two placental-conserved enhancers at this locus are highlighted. (b) Genes associated with placental-conserved promoters and enhancers show high expression levels. Grey inset: The contribution of placental-conserved regulatory activity to gene expression was evaluated using control genes associated with the same number of active promoters or enhancers, none of which are placental-conserved. Boxplots show the distribution of mean expression levels across species for all 1-to-1 orthologs (all genes); for genes associated with placental-conserved elements (dark purple for promoters, 2,384 genes; dark orange for enhancers, 387 genes); and for control genes (pale purple for promoters, pale orange for enhancers). ***: p < 0.001, **: p < 0.01, Wilcoxon rank sum test. (c) Genes associated to placental-conserved promoters and enhancers exhibit slow expression divergence across species. Grey inset: The contribution of placental-conserved regulatory activity to gene expression conservation was evaluated using control genes with similar expression levels and associated with the same number of active promoters or enhancers, none of which are placental-conserved. Plots: Pairwise Spearman correlation coefficients of expression levels between species were plotted against evolutionary distance, for genes associated with placental-conserved promoter(s) (purple) or enhancer(s) (orange) and control gene sets. Lines are as described in Figure 1b-c.

Figure 4. Recently-evolved enhancer activities weakly contribute to gene expression levels

(a) The contribution of recently-evolved regulatory elements (active in a single study species, here human) to gene expression was analysed. Genes with recently-evolved regulatory elements are typically also associated with shared regulatory elements (active in two or more species). (b) When compared to control genes with the same number of shared regulatory elements, human genes associated with additional recently-evolved promoter(s) or enhancer(s) exhibit significantly higher expression levels (***: p < 0.001, Wilcoxon rank sum test; promoters: 995 matched genes; enhancers: 5,173 matched genes). (c) When compared to control genes with the same total number of regulatory elements, human genes associated with recently-evolved enhancer(s) (orange) are expressed at lower levels (***: p < 0.001, Wilcoxon rank sum test; 3,054 matched genes). Recently evolved promoters are as active as shared ones (purple; 995 matched genes).

Figure 5. Recurrent recently-evolved regulatory elements contribute to gene expression stability

(a) Example of recurrent association of a gene with recently-evolved enhancers in multiple species. Genomic tracks show the regulatory landscape around the liver-specific gene CEBPA in human, mouse and dog (H3K4me3 (blue) and H3K27ac (orange) ChIP-seq signals; as described in Figure 2C). Recently-evolved enhancer activity in the three species is delineated with orange boxes and arrowheads. An orthologous enhancer with conserved activity across species is highlighted with orange shading. (b-c) Genes associated with recently-evolved regulatory activity significantly overlap across three reference species (b: promoters; c: enhancers; ***: p < 0.001, Chi-squared test). Numbers in Venn diagrams correspond to the number of genes with recently-evolved elements in all three species (center) and restricted to a single species. Overlaps between pairs of species are not shown. (d) Genes recurrently associated with recently-evolved elements across species exhibit high conservation of expression. Pairwise Spearman correlation coefficients of expression levels between species were plotted against evolutionary distance for genes recurrently associated with recently-evolved promoters (purple; 1,208 matched genes) or enhancers (orange; 729 matched genes) across multiple species, and control genes with similar mean expression levels across species. Lines are as described in Figure 1b-c. (e) Compared to control genes with similar expression levels and regulatory complexity, genes associated with recurrent recently-acquired promoter activity in multiple species diverge more slowly in expression (purple; 1,207 matched genes). Recently-evolved enhancers however are weaker at stabilising gene expression evolution: genes recurrently associated with recently-evolved enhancers across species exhibit higher divergence than control genes with similar expression levels and number of enhancers (orange; 207 matched genes). Plots as above.

Figure 6. An integrated summary of the evolution of mammalian regulatory complexity

(a) Representative example of the reference-free approach to connect promoter and enhancer activity with gene expression across species. Tracks in each of twenty species show an indicative landscape of active promoters and enhancers around the ONECUT1 gene, with orthologous regions linked across species by vertical lines. The reference-free mapping of these regulatory elements across species results in a meta-gene regulatory landscape that includes a single meta-promoter and 28 meta-enhancers (bottom barplot, x-axis). For each meta-element, evolutionary conservation is recorded as the number of species where promoter or enhancer activity is detected (y-axis). (b-c) The number of meta-promoters (b, purple) or meta-enhancers (c, orange) in the meta-gene landscape correlates with increased expression levels (x-axis) and expression stability (y-axis). Meta-genes were categorised according to the number of meta-promoters or meta-enhancers in their regulatory landscape. For each category, the median gene expression level is plotted against the median expression stability (1-CV, where CV—coefficient of variation across species). Insets in each plot show the spread of the distributions (interquartile ranges). Classes containing fewer than 30 meta-genes are not shown. (d-e) The evolutionary conservation of meta-promoters (d, purple) and meta-enhancers (e, orange) correlates with increasingly high and stable gene expression. Individual meta-promoters or meta-enhancers were classified according to the number of species where their activity is detected, and the median expression levels and expression stability of their putative target meta-genes were plotted as above. Insets as above. Classes containing fewer than 30 meta-genes are not plotted.