Nascent-seq indicates widespread cotranscriptional RNA editing in Drosophila

Abstract

The RNA editing enzyme ADAR chemically modifies adenosine (A) to inosine (I), which is interpreted by the ribosome as a guanosine. Here we assess cotranscriptional A-to-I editing in Drosophila by isolating nascent RNA from adult fly heads and subjecting samples to high throughput sequencing. There are a large number of edited sites within nascent exons. Nascent RNA from an ADAR-null strain was also sequenced, indicating that almost all A-to-I events require ADAR. Moreover, mRNA editing levels correlate with editing levels within the cognate nascent RNA sequence, indicating that the extent of editing is set cotranscriptionally. Surprisingly, the nascent data also identify an excess of intronic over exonic editing sites. These intronic sites occur preferentially within introns that are poorly spliced cotranscriptionally, suggesting a link between editing and splicing. We conclude that ADAR-mediated editing is more widespread than previously indicated and largely occurs cotranscriptionally.

Figure 1

Editing of mRNA edited sites is cotranscriptional. A. Overview of workflow used to detect cotranscriptional editing. Putative edited sites occurring in 10 of 12 replicates were filtered by 39X genomic sequence, resulting in 621 high ranking cotranscriptionally edited sites. B. Comparison between the replicate sample editing levels show that the editing levels are highly correlated (R=0.96). The percent editing levels per site are plotted. C. Distribution of editing levels is shown peaking at 15%. D. Overlap between the 621 high ranking cotranscriptional exon editing sites and the 276 high ranking mRNA exon editing sites. While a majority of mRNA sites (92%) were found to be cotranscriptionally edited, 41% of the nascent sites were found in the mRNA sites. Relaxing thresholds increases the mRNA overlap to 92% of the nascent sites.

Figure 2

mRNA editing level is set cotranscriptionally. A. Comparison of the percent editing levels between nascent and mRNA data show that editing levels are set cotranscriptionally. The 257 exon sites found in both nascent and mRNA datasets were sorted by the nascent level and the pooled editing levels were plotted. Editing levels slightly rose in the mRNA data by an average of 8% (2 tailed paired t-test, p = 1e-21). B. mRNA editing levels in a different background (Cs) are also similar to the nascent editing levels. C. Distribution of nascent/mRNA rank is shown for both edited (orange) and all genes (green). These two distributions are different (2 tailed t-test unequal variance, p = 1e-81). Log 2 rank ratios are plotted. (See text for more details.) D. Box plot of the same data as in C but binned as a function of gene size. Edited genes are still different from all genes (2 tailed t-test unequal variance, p < 0.005, Bonferroni corrected alpha).

Figure 3

A majority of edited sites occurs within introns. A. Introns were scanned for editing using the same pipeline resulting in 729 high ranking sites. Intron editing comprises 54% of all editing within the nascent data, in contrast to 46% within exons. B. Frequency of intron edited sites per gene follow a similar distribution as exon editing. Most genes contain one to four intron edited sites, while a few (<10%) contain more than 11. C. Overlap of exon edited genes and intron edited genes identify three classes of edited genes. 15% of edited genes are edited in both exon and intron. 36% of edited genes are edited only within introns. Lastly, 49% of all edited genes are edited only in the exons. D. Intron and exon edited events in syt1 and shaker genes occur on the same nascent transcripts. PCR amplicons were cloned into pGEM-T and individual clones were Sanger sequenced from the T7 and SP6 ends. The presence of a G in the chromatogram at the specified position is illustrated in red for exons and blue for introns. White represents the absence of editing, or an A in the Sanger sequencing. E. shaker is an example of a gene edited in both exon and intron. While 6 of 6 well known exon sites (red) are found cotranscriptionally edited, we observe 52 sites (blue) within introns. Editing sites are plotted by position and level at the shaker locus. Asterisk denotes exon and intron sites found to occur on the same transcripts in Fig. 3D. F. rdgA is an example of a gene edited only within the introns in which 26 cotranscriptionally edited sites are observed. G. CG10077 is edited only within exons. We identified 9 cotranscriptionally edited sites, 5 of which were previously identified in mRNA data from the modENCODE project.

Figure 4

Comparison of previously identified mRNA sites with cotranscriptionally edited sites. A. Comparison between the cotranscriptionally edited sites and the modENCODE mRNA edited sites shows that a majority (63%) of the (665) mRNA edited sites are cotranscriptionally edited. B. Comparison between the cotranscriptionally edited sites and a pooled set of experimentally verified sites also shows that a majority (92%) of the (75) edited sites are cotranscriptionally edited. C. nAcRalpha-34E (dalpha5) contains 10 cotranscriptional edited sites within exons 6 were identical to the 8 previously identified in the pooled dataset. 2 novel sites are found located within an alternatively spliced exon (red) which also have significant intron signal. We also identify 4 intronic sites for a total of 14 sites. The locations and editing levels are plotted for exon and intron sites. Plotted in blue are the sequenced nascent reads per base pair at the nAcRalpha-34E locus. A negative slope and intron signal are observed, indicative of nascent RNA. D. Comparison of nascent and genomic reads at the alternatively spliced exon locus of nAcRalpha-34E, containing 2 novel sites. Both sites are present in close proximity to each other, with the first one being slightly higher edited. G nucleotides in red, A nucleotides in yellow. Editing levels for each site are plotted for novel exon (red circle), previously characterized exon (blue diamond) and intron (green triangle) edited sites.

Figure 5

Most exon and intron sites are poorly edited in ADAR0 mutant flies. Comparison of editing levels in Nascent-seq data from ADAR0 and FM7a flies reveals that most exon and intron sites have a reduced editing level in the ADAR0 mutant flies. Editing levels of the 1350 editing intron and exon sites were determined in ADAR0 and control groups. Only edited sites observed in both replicates of control groups were considered resulting in 609 total sites (e.g., FM7A and yw flies; See Experimental Procedures). A false positive rate of 4.5% and 5.1% was observed for exon and intron sites respectively. A. Exon editing levels are shown for each of the 374 sites in ADAR0 and FM7a samples. B. Intron editing levels are shown for each of the 235 sites in ADAR0 and FM7a samples.

Figure 6

syt1 intron editing is conserved and a link between editing and splicing. A. syt1 exon and ECS pairing contains two dsRNA domains DI and DII (Reenan, 2005). These domains consist of an exon and intron pairing, and are important for substrate definition. Cotranscriptionally edited sites were identified in syt1 exon and intron regions. Exon sites illustrated by black arrows, intron sites illustrated by blue arrows. Edited sites are found on the opposite strand of paired dsRNA structures for both DI and DII. Structures were folded with mfold. B. syt1 intron sites are edited in different Drosophila species. Sanger sequencing chromatograms are shown for total RNA and genomic DNA. Validation was performed for each species from two independent replicates of total RNA. C. syt1 intron signal from high throughput sequencing is higher in ADAR0 mutant flies compared to WT controls yw and FM7a. yw replicates in blue, ADAR0 replicates in orange, and FM7a replicates in green. IGB genome illustration shown with genome annotation in black. All samples normalized by the number of uniquely mapped reads. Location of exon sites illustrated by black arrows. Intron sites illustrated by blue arrows. D. Intron retention from the high throughput sequencing data is quantified for each replicate and sample. Intron retention is defined as the average base pair signal within the intron divided by the average base pair signal across all exons. E. qRT-PCR validation from total RNA of the increased intron signal in syt1 between ADAR0 and WT control FM7a. The bar graph plots the ratio of a PCR amplicon within the intron and a PCR amplicon within the neighboring exon (See Experimental Procedures). We observe a significant difference in intron and exon ratio between the ADAR0 and FM7a (n=3, 2 tailed t-test unequal variance, p<0.01). Asterisk denotes p < 0.05. F. Intron retention of introns neighboring exon only edited genes is significantly lower than all introns in transcribed genes (median upstream intron: 0.21, median downstream intron: 0.185, population median of 0.268; Kruskal-Wallis, p < 1e-5). Intron retention of intron only edited genes and all intron edited genes is significantly higher than all introns in transcribed genes (medians of 0.487, and 0.468 respectively, population median of 0.268, Kruskal-Wallis, p < 1e-15). Three asterisks denote p < 0.001.