Fig 1.
Identification of editing sites and evolution of editing levels.
(A) A tree representing the schematic phylogeny of Drosophila species [43]. All 13 species were used for RNA editing site identification; the 6 species used for editing level evolution analyses are highlighted in orange. D. melanogaster, D.mel; D. simulans, D.sim; D. yakuba, D.yak; D. eugracilis, D.eug; D. takahashii, D.tak; D. ficusphila, D.fic; D. elegans, D.ele; D. kikkawai, D.kik; D. bipectinata, D.bip; D. ananassae, D.ana; D. pseudoobscura, D.pse; D. mojavensis, D.moj; D. virilis, D.vir. The divergence time between D.ele, D.eug, D.fic, D.kik, D.bip and others are unknown; the dotted lines in the tree only represent the topology. The data used for each species are indicated. (B) Venn diagram showing sites reported in this study and three previously published studies. (C) Numbers of sites used for pairwise comparisons between the 6 selected species derived from the RNA-seq, mmPCR-seq and combined sets. Sites covered by ≥50 reads were included. The average numbers of the male and female whole body data are shown. (D) Pairwise comparison of editing levels (male whole body data), with the Spearman’s rho values in red. Sites covered by ≥50 reads were included. A site that was not an ‘A’ at the DNA level was defined as having an editing level of 0%. (E) Editing level divergence distance values (1 –Spearman’s rho) plotted against estimated divergence time for all possible species pairs. The dashed line indicates the linear regression. (F) Neighbor-joining tree with branch lengths inferred using editing level (male whole body data) distance (1- Spearman’s rho) for all pairs of species. Only sites edited in both species were used for analysis.
Fig 2.
Cis regulatory features explain and predict the divergence of RNA editing levels.
(A) The difference in ADAR binding motif scores between sites that are edited (≥10%) in D.mel but not edited (≤1.5%) in other species (green) compared to the difference in sites that are edited (≥10%) in both species (purple). *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001 (one-tailed Mann-Whitney U test) (B) An example of the relationship between structure and editing level changes across species. The editing site is indicated by the purple arrows. (C) The difference in free energy, stem length, and paired bases between sites that are edited (≥10%) in D.mel but not edited (≤1.5%) in other species and control sites that are edited (≥10%) in both species. The numbers of editing sites are indicated above the bars. *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001 (one-tailed Mann-Whitney U test) (D) Relative importance of ADAR motif and dsRNA structural features to predict the establishment of editing (left) and the variation in editing levels (right). In the Presence/Absence editing scenario (left), given that a site was edited (≥10%) in one species, we predicted whether sites were edited (≥10%) or not edited (≤1.5%) in another species. In the Presence/Presence editing scenario (right), given the editing level in one species, we predicted the editing level in another species, requiring that the sites were edited (≥10%) in both species. The D.mel—D.vir comparison is shown. Features with positive and negative relationships with presence of editing (left) or increase of editing level (right) are marked in red and black, respectively. (E) Comparison of predictive accuracy of editing levels of two models (with or without motif and structural features). Given the editing levels in D.mel, editing levels in each of the other species were predicted using the random forests model. Sites with editing level ≥10% in at least one of the paired species were included in the analysis.
Fig 3.
Characterization of the evolution and function of edited genes.
(A-D) The relationship between age of editing sites and editing levels, the degree of clustering, and gene function. (A) Depiction of the species and age groups analyzed; numbers above the branches denote the numbers of D.mel editing sites in each age group and numbers in parentheses denote the numbers of genes in each age group. (B) The relationship between age of editing sites and their representative editing levels. Spearman's ρ is indicated. (C) The relationship between age of editing sites and the degree of clustering. We defined a site as clustered if the neighboring site is within a certain distance (30, 40, and 50bp). Error bars, standard deviation. (D) The relationship between age of editing sites and function of edited genes. CT: Cation Transport; R.: Regulation of; CA: Channel Activity; TA: Transporter Activity.
Fig 4.
Divergence of editing levels and DNA sequences surrounding nonsynonymous, synonymous, and 3’UTR editing sites.
(A) The comparison of Omega (ω) between genes with nonsynonymous editing sites, genes with synonymous editing sites only, and unedited genes. Unedited genes are genes without any exonic editing sites. P-values were calculated using the Mann-Whitney U test. (B) The comparison of average ω between edited and unedited genes for various Gene Ontology (GO) terms. GO terms with at least 20 edited genes were used for the analysis. P-value was calculated using the paired-sample Wilcoxon test. (C) The relationship between ω of edited genes and the normalized number of nonsynonymous editing sites per gene. The number of nonsynonymous editing sites is normalized by the number of nonsynonymous sites between D.mel and D.sim in the gene. Spearman's ρ is indicated. (D) The comparison of ω for genes with younger or older D.mel editing sites. “Genes with younger sites” refer to genes with editing sites conserved in D.mel or D.sim only. “Genes with older sites” refer to genes with editing sites conserved beyond D.sim. P-values were calculated using Mann-Whitney U test by comparing these genes with unedited genes. (E) The pairwise comparisons of normalized editing level differences between D.mel strains for nonsynonymous, synonymous, and 3’UTR sites. Sites edited at least 20% in one or both strains in the pairwise comparisons were used. For each editing site, the normalized difference is the ratio of editing difference to mean editing level. P-value was calculated using Paired-sample Wilcoxon test. The un-normalized editing level difference is shown in S9A Fig. (F) Conservation of DNA sequences surrounding D.mel editing sites. Left: the PhastCons score distributions for the 60 bp regions flanking the editing sites are plotted separately for nonsynonymous, synonymous and 3’UTR sites using a 30 bp window size (Materials and Methods). Regions that are significantly different (Kolmogorov-Smirnov Tests, fdr corrected p value ≤ 0.01 and D (Kolmogorov–Smirnov statistic, i.e. the distance between two conservation score distributions) ≥ 0.05) from the editing loci (centered, colored red) are colored gray. Right: normalized conservation score, i.e. minus D statistic of the flanking regions of editing loci relative to the region with the highest conservation score. The dashed line indicates the editing site. (G) The comparison of editing levels between highly, moderately constrained and unconstrained D.mel sites. For this analysis, we used the representative editing level of each editing site, which is the maximum editing level across 59 D.mel RNA-seq datasets, including 30 developmental stages and 29 different tissues (Materials and Methods). The editing levels of nonsynonymous and 3’UTR sites were significantly higher than that of synonymous sites (p < 8.7Х10−12 and p < 6.2Х10−11, Mann-Whitney U Test). The editing levels of constrained sites were significantly higher than that of unconstrained sites (p = 6Х10−8). The numbers of editing sites are indicated above the bars.
Fig 5.
3’UTR RNA editing is associated with gene expression changes.
(A-B) Gene expression difference between EA mutant and wild type (wt) flies. 3’UTR, genes with 3’UTR editing sites; 3’UTR (1 site), genes with one 3’UTR editing site; 3’UTR (>1 site), genes with more than one 3’UTR editing site; control, genes with 5’UTR or coding editing sites only. Only sites with editing level >5% in the wild type flies were used for analysis. P-value was calculated using the Mann-Whitney U Test. (C) Gene expression difference between EA and wt flies for genes with highly edited (≥50% editing level) or lowly edited (<50%) 3’UTR sites (D) Gene expression difference between EA and wt flies for genes with constrained or unconstrained 3’UTR sites. (E) Editing level difference between nascent and polyA RNA-seq for 3’UTR sites and coding sites. P-value was calculated using the Mann-Whitney U Test. Only sites with read coverage ≥20 in both datasets and editing level >5% in the nascent RNA-seq data were used for analysis. (F) Editing creates putative miRNA target sites. Editing sites are highlighted in red. miRNA seed regions are underlined. Gene expression difference between EA and wt flies is listed in parentheses.