ADAR RNA editing on antisense RNAs results in apparent U-to-C base changes on overlapping sense transcripts

Abstract

Despite hundreds of RNA modifications described to date, only RNA editing results in a change in the nucleotide sequence of RNA molecules compared to the genome. In mammals, two kinds of RNA editing have been described so far, adenosine to inosine (A-to-I) and cytidine to uridine (C-to-U) editing. Recent improvements in RNA sequencing technologies have led to the discovery of a continuously growing number of editing sites. These methods are powerful but not error-free, making routine validation of newly-described editing sites necessary. During one of these validations on DDX58 mRNA, along with A-to-I RNA editing sites, we encountered putative U-to-C editing. These U-to-C edits were present in several cell lines and appeared regulated in response to specific environmental stimuli. The same findings were also observed for the human long intergenic non-coding RNA p21 (hLincRNA-p21). A more in-depth analysis revealed that putative U-to-C edits result from A-to-I editing on overlapping antisense RNAs that are transcribed from the same loci. Such editing events, occurring on overlapping genes transcribed in opposite directions, have recently been demonstrated to be immunogenic and have been linked with autoimmune and immune-related diseases. Our findings, also confirmed by deep transcriptome data, demonstrate that such loci can be recognized simply through the presence of A-to-I and U-to-C mismatches within the same locus, reflective A-to-I editing both in the sense-oriented transcript and in the cis-natural antisense transcript (cis-NAT), implying that such clusters could be a mark of functionally relevant ADAR1 editing events.

Keywords: ADAR; DDX58/RIG-I; LINC-P21; MultiEditR; RNA editing; U-to-C.

FIGURE 1

A persistent U-to-C base change in DDX58 cDNA. (A) Upper: schematic representing the identification of 11 U-to-C (red bars) and 11 A-to-I (green bars) base changes within the 3′UTR of DDX58 in the B cell line RCK8. Primers used for PCR amplification are indicated as small grey arrows. Lower: representative image of Sanger traces showing that U-to-C base changes (inside the red rectangles) are only present in cDNA and not in genomic DNA (gDNA). (B) These RNA base changes are present at the same in other cell lines. (C,D) Quantification of the 11 U-to-C (C) and 11 A-to-I (D) sites within different cell lines. Quantification was performed directly from Sanger traces using MultiEditR (Kluesner et al., 2021). Center = mean and error bars = standard deviation, N = 3.

FIGURE 2

U-to-C base changes within DDX58 are dynamic. (A) Flowchart of the experiment. (B) Upper: schematic representing the 11 U-to-C (red bars) and 11 A-to-I (green bars) base changes within the 3′UTR of DDX58. Primers used for cDNA synthesis and PCR amplification are indicated as small grey arrows. Lower: Quantification of base changes within DDX58 3′UTR with and without interferon (IFN) treatment based on sequences from bacterial colonies (Supplementary Figure S2). U-to-Cs and A-to-Is are shown in red and green, respectively. (C,D) Dot plots showing the decrease in U-to-Cs and the increase A-to-Is upon IFN treatment. Each dot = one single site; line = mean. A two-tailed unpaired t-test was used to compare the differences (*p < .05).

FIGURE 3

U-to-C base changes are present and dynamic also in the long intergenic non-coding RNA hLincRNA-p21. (A) Flowchart of the experiment. (B) Upper: schematic representing the plasmid used for overexpression of hLincRNA-p21. This long non-coding RNA contains two inverted Alu elements (big orange arrows). Primers used for cDNA synthesis and PCR amplification are represented as small grey arrows. Lower: Quantification of base changes within hLincRNA-p21 sense Alu element with and without doxorubicin (doxo) treatment based on sequences from bacterial colonies (Supplementary Figure S3). U-to-Cs and A-to-Is are shown in red and green, respectively. (C,D) Dot plots showing the increase of both U-to-Cs and A-to-Is upon doxorubicin treatment. Each dot = one single site; line = mean. A two-tailed unpaired t-test was used to compare the differences (**p < 0.01; ****p < 0.0001).

FIGURE 4

U-to-C base changes originate from A-to-I RNA editing on antisense RNA. (A) Flowchart of the experiment. (B) Representative agarose gel of the amplification products of DDX58 and hLincRNA-p21 upon one-step PCR using different primers for cDNA synthesis. C- = negative control. (C) Dot plots showing the editing frequency in U-to-Cs and A-to-Is measured from Sanger sequencing dependent on the primer used for cDNA synthesis. Only sites with editing higher than 5% in at least one condition are plotted. Each dot = one single site; line = median; red dashed line represents the limit of detection of MultiEditR (Kluesner et al., 2021). U-to-Cs and A-to-Is are shown in red and green, respectively.

FIGURE 5

A-to-I antisense RNA editing in NGS data. Circular heatmap for a kidney RNA-seq sample. The external circle represents the chromosomes, and the two inner circles represent sense (intermediate circle) and antisense (internal circle) editing. A-to-I RNA editing locations are indicated with black lines connecting the heatmap to the cytoband context of the chromosomes (human genome assembly hg38). Editing levels are depicted in circular heatmaps using a color scale based on RPKM-like values.