Pleiotropic constraints, expression level, and the evolution of miRNA sequences

Abstract

Post-transcriptional gene regulation mediated by microRNAs (miRNAs) plays critical roles during development by modulating gene expression and conferring robustness to stochastic errors. Phylogenetic analyses suggest that miRNA acquisition could play a role in phenotypic innovation. Moreover, miRNA-induced regulation strongly impacts genome evolution, increasing selective constraints on 3'UTRs, protein sequences, and expression level divergence. Thus, it is essential to understand the factors governing sequence evolution for this important class of regulatory molecules. Investigation of the patterns of molecular evolution at miRNA loci have been limited in Caenorhabditis elegans because of the lack of a close outgroup. Instead, I used Caenorhabditis briggsae as the focus point of this study because of its close relationship to Caenorhabditis sp. 9. I also corroborated the patterns of sequence evolution in Caenorhabditis using published orthologous relationships among miRNAs in Drosophila. In nematodes and in flies, miRNA sequence divergence is not influenced by the genomic neighborhood (i.e., intronic or intergenic) but is nevertheless affected by the genomic context because X-linked miRNAs evolve faster than autosomal miRNAs. However, this effect of chromosomal linkage can be explained by differential expression levels rather than a fast-X effect. The results presented here support a universal negative relationship between rates of molecular evolution and expression level, and suggest that mutations in highly expressed miRNAs are more likely to be deleterious because they potentially affect a larger number of target genes. Finally, I show that many single family member miRNAs evolve faster than miRNAs from multigene families and have limited functional scope, suggesting that they are not strongly integrated in gene regulatory networks.

Fig. 1

A. Conservation matrix among miRNAs in three nematode species. The name of the Caenorhabditis briggsae miRNA used as query against the genome assemblies of Caenorhabditis sp. 9 (Csp9) and Caenorhabditis sp. 5 (Csp5) is shown in left column. Dark grey boxes indicate orthologs with conserved mature sequences. Light grey boxes indicate orthologous miRNAs with substitutions in the mature sequence. C. briggsae miRNAs with no orthologs are shown with white boxes. Crosses illustrate ambiguous orthology status due to missing data in the genome assembly. Note that orthology is tentatively assigned for members of the mir-35 family between C. briggsae and C. sp. 5 but that it cannot be excluded that some of the mir-35 genes are paralogs rather than orthologs (Supplementary Text and Supplementary Figure 1). B. The number of orthologous miRNAs in each species. C. The number of miRNA families defined by distinct seed motif

Fig. 2

A. miRNAs experience strong purifying selection, as evident from comparisons of mean nucleotide divergence between the mature sequence, the backbone (the hairpin minus the mature miRNA) and their flanking sequence. Wilcoxon two-sample tests: *** P < 0.001. Error bars indicate ± 1 standard error of the mean (SEM). B. Nucleotide substitutions between orthologous miRNAs (or perhaps between paralogous miRNAs for Csp5-mir-35 genes, see text for details) in the seed motif predict functional divergence and the regulation of distinct sets of target genes. The seed sequence is in bold and dots represent identical nucleotides relative to the top miRNA. Shaded boxes highlight substitutions between orthologs in the seed. Dashes indicate indels. Cbr: C. briggsae; C9: C. sp. 9; C5: C. sp. 5

Fig. 3

No pattern of co-evolution between divergent miRNAs in C. sp. 5 and C. sp. 9 and the targets of their orthologs (or paralogs) in C. briggsae. The frequency of target sites corresponding to the novel seed is low and is only slightly increased in species having a substitution in the seed. Sequences of miRNA seeds are shown on the left. The frequency of the target sites for the ancestral seed (dark grey) and the derived divergent seed (white) are plotted on the left

Fig. 4

X-linked miRNAs evolve faster than autosomal miRNAs in Caenorhabditis (A) and in Drosophila (B). Comparison of mean nucleotide divergence between miRNAs located on the X chromosome and on the autosomes in C. briggsae and in D. melanogaster. Wilcoxon two-sample tests: ** P < 0.01, *** P < 0.001, NS: not significant. Error bars indicate ± 1 SEM

Fig. 5

The average level of nucleotide divergence in the miRNA haipin, the backbone and the mature sequence decreases with increasing levels of miRNA expression. Error bars indicate ± 1 SEM. ANOVA: hairpin: F(2, 121) = 4.89, P = 0.010; backbone: F(2, 121) = 4.88, P = 0.009; mature: F(2, 121) = 3.85, P = 0.024