Adaptation of A-to-I RNA editing in Drosophila

Abstract

Adenosine-to-inosine (A-to-I) editing is hypothesized to facilitate adaptive evolution by expanding proteomic diversity through an epigenetic approach. However, it is challenging to provide evidences to support this hypothesis at the whole editome level. In this study, we systematically characterized 2,114 A-to-I RNA editing sites in female and male brains of D. melanogaster, and nearly half of these sites had events evolutionarily conserved across Drosophila species. We detected strong signatures of positive selection on the nonsynonymous editing sites in Drosophila brains, and the beneficial editing sites were significantly enriched in genes related to chemical and electrical neurotransmission. The signal of adaptation was even more pronounced for the editing sites located in X chromosome or for those commonly observed across Drosophila species. We identified a set of gene candidates (termed "PSEB" genes) that had nonsynonymous editing events favored by natural selection. We presented evidence that editing preferentially increased mutation sequence space of evolutionarily conserved genes, which supported the adaptive evolution hypothesis of editing. We found prevalent nonsynonymous editing sites that were favored by natural selection in female and male adults from five strains of D. melanogaster. We showed that temperature played a more important role than gender effect in shaping the editing levels, although the effect of temperature is relatively weaker compared to that of species effect. We also explored the relevant factors that shape the selective patterns of the global editomes. Altogether we demonstrated that abundant nonsynonymous editing sites in Drosophila brains were adaptive and maintained by natural selection during evolution. Our results shed new light on the evolutionary principles and functional consequences of RNA editing.

Fig 1. The landscape of A-to-I editomes in D. melanogaster.

(A) A flowchart of A-to-I editing detection in brains of D. melanogaster. Editing sites are classified into five distinct classes based on the decreasing confidence of editing, sequencing coverage, and the number of libraries in which the editing events are detected. (B) Overlaps of the editing sites identified in this study and previous studies. (C) Boxplots of the editing levels of the common and novel sites in each brain library (P < 0.001 in each brain library, KS tests). (D) A summary of the editing sites with respect to their functional annotations. The numbers of high-confidence editing sites in each functional category in D. melanogaster are given in the top panel, and the proportion of editing sites is presented above the bars. mRNA-seq coverage (middle) and editing level (bottom) of editing sites in each category are also shown. For a site, the median value of coverage and editing level across all the libraries (if applicable) is used for the boxplots. (E) The observed numbers of editing sites located in stable hairpin structures and the expected numbers of sites (median and 95% confidence intervals) under randomness. ***, P< 0.001 revealed by simulations.

Fig 2. Conservation patterns of editing sites in brains of D. melanogaster and two other species.

“+”, the high-confidence editing sites were reliably detected in a species (Top). Bottom: possible gain and loss patterns of 87 sites that have a minimal editing level of 0.05 in D. melanogaster and have at least 200 raw reads in both D. simulans and D. pseudoobscura. “-”, the orthologous site is not edited with high probability [joint P(D₀) < 0.0002].

Fig 3. Evaluating the effect of detection bias on N/S ratio estimation.

(A) The simulated and observed N/S ratios with increasing cutoffs of sequencing coverage (C_min). The x-axis is the cutoff of coverage (C_min) and the y-axis is the simulated (median in black, and the range from 2.5% to 97.5% quantile is in blue) and observed (red) N/S ratio. The N/S ratio under neutral evolution (3.80) is indicated with dashed lines. l_min is 0.02 here. (B) The relative difference of the simulated vs. the observed N/S ratio with increasing C_min. Each plot is corresponding to the upper one in (A). (C) The simulated (median in black and the range from 2.5% to 97.5% quantile in blue) and observed (red) N/S ratios (the y-axis) when pooling all the brain libraries together and randomly sampling a fraction (f, from 0.05 to 1, the x-axis) of reads. Results with the cutoff of editing level, l_min = 0.01, 0.02 and 0.05 are presented. The N/S ratio under neutral expectation (3.80) is indicated with dashed lines. (D) The simulations of the editing sites in PSEB (upper panel) and non-PSEB genes (lower panel) as in (C).

Fig 4. The Gene Ontology (GO) enrichment analysis on the PSEB genes.

MF: molecular function; BP: biological process.

Fig 5. The effect of local nucleotide contexts on editing in brains of D. melanogaster.

(A) A 7-mer motif centered with the high-confidence editing sites. (B) The score cutoff that specified the top 90% quantile of the high-confidence editing sites. (-0.622) corresponds to the top 75.4% of all the 7-mer sequences centered with adenosine in the genes with editing events.

Fig 6. A-to-I editing increases mutation sequence space of evolutionarily conserved genes.

(A) The editing density in the N sites is significantly inversely correlated with the dN value (between D. melanogaster and D. simulans) of the host genes. The genes expressed in brains are ranked with increasing dN values and divided into 20 bins (the x-axis, and lower dN means evolutionarily more conserved). The left and right panel is for 1- to 5-day female (B2) and male (B6) brains of D. melanogaster, respectively (Table 1). In each bin, the editing density (y-axis) is calculated by dividing the observed number of editing sites with the total number of adenosine sites that cause amino acid changes if edited. (B) The editing density in the N sites is significantly positively correlated with the phyloP score of the sites. All the nonsynonymous adenosine sites (cause amino acid changes if edited; ≥ 5X sequencing coverage) are ranked with increasing phyloP scores and grouped into 20 bins (x-axis, and higher phyloP score means evolutionarily more conserved). (C) The editing density of the N sites is significantly lower in the non-conserved compared to conserved sites after controlling mRNA-Seq coverage. All the nonsynonymous adenosine sites (cause amino acid changes if edited; ≥ 5X) are ranked with increasing sequencing coverage and binned into 20 categories (x-axis). Within each bin, we further divided the sites into two equal-sized subgroups based on the phyloP scores. The y-axis is the editing density of the non-conserved relative to the conserved subgroup in each bin.

Fig 7. The effect of temperature on editing levels in brains of Drosophila.

(A) The changes of editing levels in N and silent (S and UTRs) sites in female and male brains under elevated temperature (stressed at 30°C for 48 hours). (Error bar represents the s.e. of the level changes for editing sites in each category). (B) The flanking sequences (100 nts at each side) have significantly lower MFE (Kcal/mol) for the N sites compared to the silent sites. (C) Clustering the brain libraries of D. melanogaster based on the editing levels of 391 high-confidence editing sites that have at least 20 raw reads in each brain library. Note flies of the same accommodation conditions always cluster together. (D) Clustering the brain libraries of D. melanogaster and D. simulans based on the editing levels of 289 high-confidence editing sites that have at least 20 raw reads in each brain library. Note species divergence plays a more important role than temperature in clustering the samples.

Fig 8. The expression profiles of PSEB genes and the expression pattern of Adar determine the overall N/S ratio of the editome in different developmental stages or tissues.

(A) The N/S ratios (x-axis) for the all the editing sites (dots), the editing sites in PSEB genes (orange) or non-PSEB genes (cyan) in different developmental stages of D. melanogaster in the modENCODE Project. Asterisks indicate different N/S ratios between PSEB and non-PSEB editing sites: *, P < 0.05; **, P < 0.01; ***, P < 0.001. The overall observed N/S ratios were compared to the expected N/S ratio under neutral evolution (3.80): red dots, positive selection; grey dots: neutral evolution; blue dots, purifying selection. (B) The proportions of N and S sites that are contributed by PSEB genes in different development stages. (C) The expression level of Adar (RPKM) and the number of PSEB genes expressed during the development of D. melanogaster.