Profiling neural editomes reveals a molecular mechanism to regulate RNA editing during development

The C. elegans adult neural editome consists of nearly 600 A-to-I editing sites

To examine the C. elegans adult neural editome, synchronized larval-stage 1 (L1) worms were grown to adulthood for 50 h at 20°C before neural cell isolation. Neural cells were isolated from wild-type and adr-2(-) worms expressing green fluorescent protein (GFP) under the control of a pan-neuronal promoter, rab-3, as previously described (Spencer et al. 2014; Deffit et al. 2017; Kaletsky et al. 2018). Briefly, adult worms were subjected to chemomechanical disruption, and by using fluorescent-activated cell sorting (FACS), GFP-positive neural cells were selected from the live, single-cell population (Supplemental Fig. S1A,B). The percentage of adult neural cells isolated was similar to previously published studies using this technique (Kaletsky et al. 2018). Successful isolation of neural cells from nonneural cells was confirmed using real-time quantitative PCR (qPCR) to measure expression of both the neural-specific gene unc-64 (Saifee et al. 1998) and the muscle-specific gene myo-3 (Supplemental Fig. S1C; Ardizzi and Epstein 1987).

High-throughput sequencing of poly(A) selected RNA from the isolated adult neural cells was performed, and reads were uniquely mapped to the C. elegans reference genome using the Spliced Transcripts Alignment to a Reference (STAR) alignment program (Dobin et al. 2013). Pairwise comparisons of the replicates indicated high reproducibility (Supplemental Fig. S2A). We compared our adult neural transcriptome to the previously published adult C. elegans neural transcriptome, which was performed with a similar technique, except the neural GFP expression was driven by a different pan-neuronal promoter (Kaletsky et al. 2018). Comparison of the transcriptomes using DESeq2 revealed that a majority of genes showed no significant difference in expression (P-adj < 0.05) (Supplemental Fig. S2B; Supplemental Table S1).

A-to-I RNA editing sites were identified de novo from the adult neural RNA-seq data set using the Software for Accurately Identifying Locations Of RNA editing (SAILOR) program (Deffit et al. 2017). Each predicted site is reported with a confidence score that evaluates a number of parameters, including read coverage and percentage of editing at each site. Sites with a confidence score of ≥0.99 were used for downstream analysis. By using these stringent cut-offs, 574 candidate sites were predicted from the wild-type adult neural RNA-seq data set, and 17 sites were predicted for the adr-2(-) data set. As adr-2(-) worms lack A-to-I editing (Tonkin et al. 2002), the sites identified by SAILOR for the adr-2(-) worms are false positives. Subtraction of these false positives led to identification of 570 A-to-I editing sites in the wild-type adult neural editome, which were assigned to 66 different neural transcripts (Supplemental Table S2).

To determine whether this developmental- and tissue-specific approach identified novel editing sites, the adult neural editome was compared with a list of 203,934 editing sites (9444 transcripts) identified from previous C. elegans studies (Washburn et al. 2014; Whipple et al. 2015; Zhao et al. 2015; Deffit et al. 2017; Goldstein et al. 2017; Reich et al. 2018; Ganem et al. 2019). Except one (Deffit et al. 2017), the previous studies used whole-worm lysates to identify editing sites. Many of these studies performed control experiments, in which sites were also computationally predicted for adr-2(-) worms (that lack all A-to-I editing) and any false-positive sites were removed. This list also includes edited sites from all developmental stages. Nearly, 23% of the adult neural editing sites (130 out of 570) had not been previously identified (Fig. 1A), which suggests that these adenosines may be edited only or more highly in adult neurons. Nearly 80% of the novel sites (104 out of 130) mapped to different regions of 53 previously identified edited transcripts, whereas 10 neural transcripts were identified as novel edited mRNAs (Fig. 1A). The genic distribution of A-to-I editing sites indicates that the vast majority (>80%) of adult neural editing sites are located in the 3′ untranslated regions (UTRs) of transcripts (Fig. 1B). In addition, as the UTRs of many C. elegans genes are not well annotated and our previous studies showed that the intergenic regions primarily reside within 2000 bp upstream of or downstream from the annotated genes (Deffit et al. 2017), it is highly possible that some of the editing sites identified in intergenic regions also reside in 3′ UTRs.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1.

De novo identification of editing sites in adult neurons. (A) The adult neural editing sites identified were cross-referenced with previously reported sites to identify novel editing sites. (B) Genomic distribution of adult neural sites. (CDS) Coding region; (UTR) untranslated region; (ncRNA) noncoding RNA. (C) A Two Sample Logo analysis of editing sites identified in adult neural cells compared with adenosines present in the same dsRNA region. Enriched and depleted nucleotides are shown above and below the axis, respectively. The level of conservation is represented by letter height. Logos were generated using a Student's t-test with P < 0.05 and no Bonferroni correction.

To determine the nearest neighbors of the 570 editing sites identified in adult neural cells, the percentage of occurrence for the 5′ and 3′ neighboring nucleotides were weighted. As the list of 570 sites contains editing sites mapped to different genomic locations, a background list was used to match the genomic distribution of these identified 570 edited sites. To determine whether or not the appearance of particular nucleotides in the nearest-neighbor positions was statistically significant (P < 0.05), a Two Sample Logo analysis (Vacic et al. 2006) was used to compare the edited adenosines to the sequence context of all adenosines in the ds region (Fig. 1C). Our data suggest that in adult neural cells, C/A and G/A are overrepresented whereas A/G and U/C are underrepresented at the 5′ and 3′ sides of editing sites, respectively. The identified 5′ neighboring nucleotides are consistent with previously published mammalian ADAR preferences (Polson and Bass 1994; Lehmann and Bass 2000). However, A is less preferred at the 3′ side by mammalian ADARs. In sum, our transcriptome-wide approach to identify tissue- and developmental-specific editing sites yielded 570 editing sites in adult neural cells, including 130 novel sites mapped to 63 individual transcripts.

The neural editome changes during development

Previous data from our laboratory reported 7361 editing sites in neural cells isolated from L1 worms (Deffit et al. 2017). This number is more than 10-fold higher than the number of adult neural editing sites identified herein. In addition, in contrast to the adult neural editome, nearly half of the previously identified L1 neural editing sites were annotated to intronic regions. It is possible that technical differences in sequencing and/or bioinformatic analysis could contribute to these differences. To rule out differences owing to variations in read length and bioinformatics analysis, the length of the L1 neural RNA-seq reads was shortened to 75 bases (see methods) and then reprocessed through the bioinformatics pipeline with the same parameters used for the adult neural RNA-seq reads. This reanalysis predicted 6213 candidate sites in wild-type and 28 false-positive sites in adr-2(-) L1 neural cells. After subtracting the false positives, 6202 A-to-I editing sites were identified in the L1 neural editome (Supplemental Table S3). Despite the slight reduction in number of L1 edited sites, these data still suggest a 10-fold higher number of editing sites in the C. elegans larval nervous system compared with the adult nervous system.

To test whether the quantity of editing sites between the life stages could be attributed to read coverage differences, the number of uniquely aligned reads were counted using SAMtools at various points in the computational analysis. Although the adult neural RNA-seq data set had twice as many aligned reads, the SAILOR analysis for the adult data set had 3.3 times fewer sequencing reads compared with the L1 data set (Fig. 2A). Before identifying editing sites, SAILOR discards duplicate reads (those that start and end with the same end nucleotide) and reads that contain more than one non A-to-G variant. Thus, the reads used for SAILOR identification had a depth of five reads covering 81.3% and 57.9% of the reference genome in the L1 and adult neural RNA-seq data sets, respectively. To simply test whether the difference in editing site number is caused by the differences in read coverage, the same number of reads (100,000 after removal of duplicated reads) was randomly chosen from both data sets and input into SAILOR. This approach identified 721 and 187 edited sites with a confidence score of ≥0.99 in the L1 and adult neural RNA-seq data sets, respectively. The fourfold larger number of L1 editing sites is consistent with our conclusion that the L1 neural editome is larger than the adult neural editome.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 2.

C. elegans neural editome change during development. (A) Comparison of the RNA-seq data sets from L1 and adult neural cells. The number of input reads for each step of the SAILOR bioinformatic pipeline is listed along with the number of de novo identified edited sites. (B) Genomic distribution of the identified L1 neural editing sites was determined using the WormBase annotations (WS275). (C) A Two Sample Logo analysis of editing sites identified in L1 neural cells compared with adenosines present in the same dsRNA region. Enriched and depleted nucleotides are shown above and below the axis, respectively. The level of conservation is represented by letter height. Logos were generated using a Student's t-test with P < 0.05 and no Bonferroni correction. (D) qPCR was used to quantify adr-2 expression in neural cells (left) and in whole-worm lysate (right) for both L1 and adults. The average of three independent replicates is plotted; error bars, SEM. Statistical significance was calculated using a Student's t-test. (***) P < 0.001; (ns) P > 0.05.

In support of developmental differences in the neural editomes and consistent with the initial publication of the L1 analysis, the largest number of L1 neural editing sites (44.0%) localized to introns (Fig. 2B). It is possible that a higher number of intronic reads in the L1 data set could lead to a higher number of intronic editing sites identified. To address this possibility, the number of intronic reads in both RNA-seq data sets was counted using featureCounts. Both RNA-seq data sets contained a similar number of intronic reads (Fig. 2A). SAILOR analyzed 6.5-fold fewer intronic reads for the adult data set but identified 210-fold less intronic sites compared with the L1 data set. In sum, the C. elegans neurodevelopmental editomes differ in both editing site number and distribution, with a significantly higher number and intronic distribution of editing sites in L1 neural cells and a majority of 3′ UTR editing sites in adult neural cells. The nearest-neighbor preferences of the L1 editing sites suggest that U/A and G/A are overrepresented, whereas A/G and U/C are under-represented at the 5′ and 3′ sides, respectively (Fig. 2C). Compared with the preferences of the adult neural edited sites (Fig. 1C), L1 neural edited sites show similar neighbor preferences at the 3′ side, but at the 5′ side U is overrepresented specifically in L1 neural editing sites.

One possibility for the fewer editing sites identified in the adult nervous system could be that adr-2 expression decreases during development. Previous whole-worm analysis suggested similar, and perhaps slightly higher, adr-2 mRNA expression in adult worms compared with L1 worms (Hundley et al. 2008). To determine whether adr-2 expression is altered during neural development, adr-2 mRNA levels were quantified in L1 and adult neural cells. Absolute mRNA levels for adr-2 in neural cells and whole-worm lysates were measured for three independent biological replicates using qPCR. Our data indicate that neural adr-2 mRNA expression decreases during development, whereas global adr-2 mRNA expression remains unchanged (Fig. 2D). Together, these data suggest that the tissue-specific decrease in adr-2 mRNA expression in adult neural cells contributes to the smaller adult neural editome.

Differential expression of edited transcripts contributes to stage-specific neural editomes

The neural-specific decrease in adr-2 mRNA expression likely contributes to the decreased number of editing sites identified in adult neural cells. However, the distribution of editing sites identified within specific regions of transcripts also changes during development. To gain insight into whether editing has altered developmental contributions to neural transcriptome diversity and/or whether altered gene expression is affecting developmental editing site identification, variant detection at the identified sites was performed for both neural RNA-seq data sets using SAMtools (see Methods) (Li et al. 2009). First, a composite list of editing sites in the C. elegans nervous system was generated by merging the L1 and adult editing sites, which yielded a total of 6352 editing sites mapped to 453 unique transcripts. Transcripts were classified as edited if at least one variant (G) read was detected (Fig. 3A; Supplemental Table S4). Of the 6352 editing sites, 4642 and 68 sites were identified as edited only in the L1 or adult neural RNA-seq data sets, respectively (Fig. 3B). Among the 4642 edited sites in L1 neural cells, 1740 sites mapped to transcripts that were edited within a different region in adult neural cells and, thus, were excluded from the list of L1-specific edited transcripts (Fig. 3A,C). Likewise, among the 68 sites edited only in adult neural cells, more than half mapped to transcripts that were edited within a different region in the L1 neural RNA-seq data set (Fig. 3A). Thus, there were only a handful of neural sites (1.1%) edited specifically in adults (Fig. 3B). In contrast, the percentage of neural transcripts edited specifically in early development (73.1%) was significantly higher, and more than three-quarters of these transcripts are edited within introns (Fig. 3C).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 3.

Differential expression of many transcripts during neural development contributes to the stage-specific neural editomes. (A) Workflow for identifying stage-specific edited sites following the quantification of all the neural editing sites. (B) Distribution of stage-specific edited sites. (C) Genomic distribution of L1-specific edited transcripts based on the genomic location of identified edited sites. (CDS) Coding region; (UTR) untranslated region; (ncRNA) noncoding RNA. (D) Dots represent individual genes that are down-regulated (red; 4476, log₂fold < −0.5), up-regulated (blue; 5171, log₂fold > 0.5), or not significantly different (gray; 7697, P-adj > 0.05) between three biological replicates of wild-type neural L1 and adult RNA-seq data. (E) The average of three independent replicates of qPCR validation of neurodevelopmentally, differentially expressed genes is plotted; error bars, SEM. Statistical significance was calculated by Student's t-test. (*)P < 0.05; (***) P < 0.001. (F) Representation of the L1-specific edited transcripts based on differential expression during neural development. Dark blue represents transcripts differentially expressed in neural development and ADR-2 dependent (dotted) or independent (solid). Light blue represents transcripts that are not differentially expressed in neural development.

The above data suggest that editing impacts transcriptome diversity in the nervous system in a stage-specific manner. However, whether editing of these neural transcripts also leads to differential expression in the two life stages is unknown. To identify genes with altered expression during neural development and to identify the role of editing in any observed expression differences, differential gene expression analysis was performed on three biological replicates of wild-type neural RNA-seq data from both stages using DESeq2 (see Methods). Principle component analysis (PCA) indicates the three biological replicates of each sample cluster best, and the clusters of the two developmental stages were more distinct from one another than any individual samples (Supplemental Fig. S3A). The expression of 9620 genes were significantly altered (P-adj < 0.05) between L1 and adult neural cells isolated from wild-type animals (Fig. 3D; Supplemental Table S5). Compared with the neural L1 expression, slightly more than half of these genes (5171) significantly increased, whereas the other 4476 significantly decreased expression in adult neural cells (Fig. 3D). Independent qPCR validation of one up-regulated gene, clec-41, and one down-regulated gene, daf-2, confirmed the trends observed in the RNA-seq data (Fig. 3E). Gene Ontology (GO) analysis of the genes up-regulated (Supplemental Fig. S3B) and down-regulated (Supplemental Fig. S3B) in adult neural cells was distinct.

To determine whether the differential neural expression during development contributes to the L1-specific neural editome, the stage-specific edited transcripts were overlapped with the differentially expressed neural transcripts. More than two-thirds of the L1-specific edited transcripts (203 of 315) are down-regulated in adult neural cells (Fig. 3F). These data are consistent with previous studies (for review, see Pinto and Levanon 2019) and suggest that the decreased gene expression in adult neural cells affects the overall amount of editing observed.

Previous studies showed that ADR-2 regulates neural-specific gene expression in L1 worms (Deffit et al. 2017). To determine the extent to which ADR-2 contributes to neurodevelopmental gene expression, DESeq2 was performed on L1 and adult neural RNA-seq data sets from adr-2(-) animals (Supplemental Table S6; Supplemental Fig. S3C). The expression of 7088 genes was significantly altered (P-adj < 0.05) between L1 and adult neural cells from adr-2(-) worms (Supplemental Fig. S3C). Compared with L1 expression, nearly two-thirds of genes (4374) increased, whereas the other 2714 decreased, in expression in adult neural cells lacking adr-2(-). Compared with wild-type neural development, there is a large difference in the number of genes with decreased expression in adr-2(-) adult neural cells (4476 to 2714). These data suggest that ADR-2, and perhaps RNA editing, plays an important role in promoting the neural developmental down-regulation of several hundreds of genes.

To identify differentially expressed neural genes that are dependent on ADR-2, the genes with altered expression in wild-type adult neural cells were overlapped with those with altered expression in adr-2(-) adult neural cells (Supplemental Table S7). This analysis identified 1941 genes that were dependent on adr-2 for proper neurodevelopmental gene expression. These genes were enriched for gene ontologies including response to stimulus and determination of adult lifespan (Supplemental Fig. S3D), which is consistent with previous studies showing that worms lacking editing show chemotaxis defects and altered lifespan (Tonkin et al. 2002; Sebastiani et al. 2009; Ganem et al. 2019). It is also possible that these 1941 genes include some genes that are already differentially expressed in L1 neural cells isolated from adr-2(-) animals compared with wild-type L1 neural cells. To address this concern, DESeq2 was also performed between wild-type and adr-2(-) L1 neural cells. This analysis identified five (C39H7.4, rncs-1, asah-1, nep-17, and hpo-15) of the 1941 genes are already down-regulated in the adr-2(-) neural cells at the L1 stage. qPCR of one of these five genes (C39H7.4) confirmed the differential expression (Supplemental Fig. S3E). Three genes (kgb-1, nhr-122, and srw-17) from the remaining 1936 genes were randomly selected, and qPCR confirmed that the gene expression for all three genes was developmentally regulated and ADR-2 dependent (Supplemental Fig. S3F). Thus, providing additional support that ADR-2 is important for regulating neurodevelopmental gene expression for approximately 1900 genes.

To determine if editing could contribute to the ADR-2-dependent neurodevelopmental gene regulation, the list of developmentally regulated neural genes was overlapped with the known stage-specific edited transcripts. Nearly 30% of the differentially expressed and L1-specific edited genes were dependent on ADR-2 for proper gene expression during neural development (Fig. 3F). Together, these data indicate that differential expression contributes to the stage-specific neural editomes, and ADR-2 regulates neurodevelopmental gene expression in both an editing-dependent and an editing-independent manner.

A subset of neural transcripts shows increased editing during development

Although our above analysis indicates that stage-specific editing of many transcripts occurs owing to differential gene expression, nearly one-quarter of edited transcripts showed editing in both life stages (Supplemental Fig. S4A; Supplemental Table S4). Of the 119 transcripts edited in both stages, 10% show editing within a specific dsRNA region in a development-specific manner (Supplemental Fig. S4B). In these 12 transcripts, the intronic edited sites are L1 specific. Although 3′ UTR editing is observed in both stages, the editing level in the 3′ UTR increases in adults. The majority of transcripts (90%) are edited within the same dsRNA region throughout development. In these transcripts, even though the same dsRNA region is edited, both the number of sites edited and the editing level vary between L1 and adult neural cells.

As read coverage influences the ability to accurately quantify editing at a given site, to quantitively examine differences in editing levels during development, a threshold of five reads covering the edited nucleotide and one A-to-G variant in both data sets was required. This stringency reduced the number of edited sites found in both developmental stages to 901 (Supplemental Table S8). Despite the decrease in adr-2 expression during neural development (Fig. 2C), ∼80% of these sites showed increased editing in adult neural cells (Fig. 4A). This is consistent with previous observations of increased A-to-I editing during mammalian brain development and with the fact that ADAR expression does not directly correlate with the changes in editing levels (Wahlstedt et al. 2009; Ekdahl et al. 2012; Shtrichman et al. 2012; Zaidan et al. 2018).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 4.

Editing level increases in majority of the sites during neural development. (A) Heatmap representing the percentage of editing in L1 and adult neural cells at the sites (901) with more than five reads in neural RNA-seq from both life stages. (B) The 837 sites mapped to annotated transcripts were categorized into five groups depending on differential editing between L1 and adult neural cells. Bar graph represents the number of editing sites in each group. The number of unique transcripts in each group are listed above each bar.

To begin to understand how neurodevelopmental editing is regulated, the sites with changes in editing were annotated to genomic regions. For the adenosines edited in both life stages, 64 editing sites mapped to unannotated regions and were removed from downstream analysis. The remaining 837 editing sites mapped to 67 transcripts. The editing sites were classified into five groups depending on the percentage of change in editing observed in the two life stages (Fig. 4B). Together, the two classes with increased editing held the majority of sites analyzed, indicating a prevalent higher editing efficiency in adult neural cells.

To begin to mechanistically dissect how increased editing occurs during neural development, we focused on sites that showed a >20% increase in editing in adult neural cells. These editing sites map to 23 unique transcripts (Table 1). In contrast to previously identified editing targets that have multiple (five to 125) editing sites in noncoding regions of the C. elegans transcriptome (Tonkin et al. 2002; Zhao et al. 2015; Deffit et al. 2017; Goldstein et al. 2017), nearly two-thirds (65%) of these transcripts have only one or two specific editing sites, and 70% of these transcripts are edited within the 3′ UTR. In sum, despite decreased adr-2 expression in adult neural cells, increased editing at a majority of sites within neural transcripts that are constitutively expressed during development was observed, thus suggesting the presence of a neural-specific molecular mechanism to regulate RNA editing during development.

The inhibitory role of ADR-1 on Y75B8A.8 editing is development-specific

A major regulator of RNA editing in C. elegans is the deaminase-deficient ADAR protein, ADR-1 (Washburn et al. 2014; Rajendren et al. 2018). Initial studies of adr-1(-) worms identified both sites that show increased and decreased editing compared with wild-type worms (Tonkin et al. 2002). In previous studies, ADR-1 was shown to promote ADR-2 binding to transcripts, which leads to increased editing (Washburn et al. 2014; Rajendren et al. 2018). However, ADR-1 has also been shown to inhibit editing of a specific transcript, Y75B8A.8, when transgenically expressed in neurons, and this repression is largely masked in studies of RNA isolated from whole worms (Washburn and Hundley 2016b). The genes identified in this study, which show increased editing in adult neural cells include Y75B8A.8 (Table 1).

From the neural RNA-seq data sets, editing of endogenous Y75B8A.8 increased during neural development (Fig. 5A), suggesting ADR-1 repression of editing is most efficient at the L1 stage. To directly test this hypothesis, developmental editing of the Y75B8A.8 3′ UTR was examined in neural cells in the presence and absence of adr-1. Transgenic worms expressing the Y75B8A.8 3′ UTR under the control of the rab-3 pan-neuronal promoter were synchronized at the L1 and adult stages. After RNA isolation, reverse transcription was performed using a reporter-specific primer. Editing at two specific sites (referred to as sites 227 and 228 owing to the nucleotide numbers within the 3′ UTR) within the Y75B8A.8 3′ UTR was apparent in both stages and genotypes (Fig. 5B). Editing at these sites drastically increased (∼40% in both 227 and 228) in adr-1(-) worms at the L1 stage (Fig. 5C, left), whereas a slight increase (∼20% at 228) in one of the sites was observed in adr-1(-) worms at the adult stage (Fig. 5C, right).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 5.

The inhibitory role of adr-1 on editing of neural transcripts is developmentally regulated. (A) Representation of neural RNA-seq reads covering the Y75B8A.8 transcript. Reads from adr-2(-) are negative controls. Green represents adenosine and brown represents guanosine at the marked chromosomal positions. (B) Sanger sequencing chromatograms of cDNA amplified from Y75B8A.8 reporter RNA. Editing sites are listed below the chromatogram. Sites 227 and 228 correspond to chromosomal positions of Chr III: 12,171,074 and 12,171,075. The nucleotides at each position are represented with a different color (green, adenosine; black, guanosine; blue, cytidine; red, thymidine). (C,D) The average of three independent replicates was plotted; error bars, SEM. One-way ANOVA was performed to determine the statistical significance. (****) P < 0.0001; (*) P < 0.05; (ns) P > 0.05. (C) Editing levels of the Y75B8A.8 neural reporter were measured in wild-type and adr-1(-) worms in L1 (left) and adult (right) stages using Sanger sequencing. (D) qPCR was performed to measure the absolute mRNA levels of adr-2 relative to adr-1 in both life stages. (E) Neural RNA-seq data were used to plot the L1 editing levels of representative sites from the group of neural RNAs that showed increased editing in development (P < 0.01, pairwise comparison using two-way ANOVA).

One possibility for the reduced ADR-1 inhibition of editing during neural development is that the relative expression of adr-1 and adr-2 changes during neural development, thus allowing ADR-2 to overcome inhibition by ADR-1. To test for this possibility, the absolute adr-1 mRNA levels in neural cells and whole-worm lysates were measured using qPCR. Our data indicate that neural adr-1 expression decreased during development, whereas adr-1 expression in whole-worm lysates remained unchanged (Supplemental Fig. S5). Absolute mRNA counts for adr-2 (Fig. 2D) were divided by the corresponding life stage of adr-1 mRNA counts, and the observed ratio of adr-2 to adr-1 mRNA expression was found to significantly increase from L1 to adult neural cells (Fig. 5D). In contrast, the adr-2 to adr-1 ratio in whole-worm lysate remained unchanged during development. Consistent with previous studies, adr-1 mRNA levels are higher than those of adr-2 throughout development (Hundley et al. 2008). Thus, these data suggest a tissue-specific regulation of editing during development, in which the levels of an ADAR enzyme (ADR-2) exceed that of a repressive factor (ADR-1) to allow for increased editing during development.

To determine whether the inhibitory role of ADR-1 was specific to Y75B8A.8 or extended to other transcripts that show increased editing in adult neural cells (Table 1), editing levels were quantified for adr-1(-) neural RNA-seq data from both stages (Supplemental Table S9). Editing sites within transcripts that show increased editing during neural development also show increased editing in the absence of adr-1 at the L1 stage (Fig. 5E). This increase is significant across all sites analyzed in Figure 5E (P < 0.01, pairwise comparison). In contrast, a randomly selected group of a similar number of L1 edited sites did not show any significant increase in editing in the absence of adr-1 (Supplemental Fig. S6A). In addition, editing at the majority of the sites showed in Figure 5E did not significantly increase in the absence of adr-1 in the adult stage (Supplemental Fig. S6B). Together, these data suggest that the inhibitory role of adr-1 in neural cells is developmentally regulated. These data are also consistent with previous data indicating an embryo-specific enrichment of editing sites in the absence of adr-1 compared with later (L4) developmental stages (Ganem et al. 2019). Together, these studies suggest that ADR-1 negatively regulates editing primarily during early C. elegans development.

In the absence of adr-1, ADR-2 binding to neural mRNAs increases early in development

Our sequencing data suggest a regulatory mechanism in which adr-2 expression increases relative to adr-1 expression during neural development to result in increased editing of specific neural transcripts. To begin to understand the mechanism of how ADR-1 represses editing of these transcripts during early development, we tested whether ADR-1 binding to transcripts was required for inhibition of editing. Editing of the Y75B8A.8 3′ UTR neural reporter was examined in the presence of either wild-type ADR-1 or an RNA binding–deficient ADR-1 mutant, where mutations (KKxxK to EAxxA) within the first dsRNA binding domain (dsRBD1) abolished ADR-1 binding to mRNA in vivo (Rajendren et al. 2018). Transgenic worms expressing FLAG-tagged wild-type ADR-1 or the ADR-1 RNA binding mutant driven by the endogenous adr-1 promoter along with Y75B8A.8 3′ UTR under the control of the rab-3 pan-neuronal promoter were generated. Synchronized L1 worms were harvested, and total RNA was used to reverse transcribe the neural Y75B8A.8 reporter. By using Sanger sequencing, editing of the neural Y75B8A.8 reporter in the presence of the wild-type ADR-1 transgene (Fig. 6A) was similar to that observed for the neural Y75B8A.8 reporter from wild-type L1 worms (Fig. 5C). However, in the presence of an RNA binding–deficient ADR-1, editing levels significantly increased (Fig. 6B). These data suggest that ADR-1 competes with ADR-2 to bind to Y75B8A.8 in L1 neural cells. In addition, overlap of our previous studies of ADR-1-bound RNAs from whole-worm lysates (Ganem et al. 2019) and the neural transcripts with neurodevelopmentally increased editing identified above indicates that 80% of these transcripts are bound by ADR-1 (Table 1). This suggests that ADR-1 inhibition of editing by binding to neural RNAs could be a widespread mechanism to regulate RNA editing in a tissue- and development-specific manner in C. elegans.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 6.

ADR-1 inhibits editing at transcripts that show increased editing in neural development by binding. (A) Sanger sequencing chromatograms of cDNA amplified from Y75B8A.8 reporter RNA isolated from the indicated worms at the L1 stage. (B) Editing in A was quantitatively measured. The average of two independent replicates is plotted; error bars, SEM. Two-way ANOVA was used to determine statistical significance. (***) P < 0.001. (C) Lysates from the indicated worm strains were incubated with FLAG magnetic beads. A portion of the lysates before immunoprecipitation and the IPs was subjected to immunoblotting with a FLAG antibody (Sigma-Aldrich F1804). (D) Bar graph represents the fold enrichment of cDNA present in the IPs divided by cDNA present in the input lysates from the indicated worm strains and normalized to the same ratio in adr-1; adr-2(-) worms. The average of three (Y75B8A.8, WO7G4.3 and lam-2) and two (daam-1 and uba-2) independent replicates were plotted; error bars, SEM. One-way ANOVA was used to determine statistical significance. (*) P < 0.05; (ns) P > 0.05. (E) Proposed model for editing regulation of neural transcripts that show increased editing during development.

To further test the model that ADR-1 represses editing by outcompeting ADR-2 binding to transcripts, we sought to determine whether ADR-2 binding to neural targets is affected by loss of adr-1. To specifically examine ADR-2 binding to neural mRNAs, transgenic worms coexpressing 3X FLAG-tagged adr-2 under control of the rab-3 promoter and the neural Y75B8A.8 reporter were generated in both adr-2(-) and adr-1(-);adr-2(-) backgrounds. Reporter editing indicated the expression of functional ADR-2 in these transgenic lines (Supplemental Fig. S7). To examine ADR-2 binding to neural transcripts, a FLAG RNA immunoprecipitation (RIP) was performed in synchronized L1 and adult worms that were UV crosslinked. Total RNA from IPs and input lysates was used to reverse transcribe specific genes, and qPCR was performed. Equal expression and efficient immunoprecipitation of ADR-2 were observed in the presence and absence of adr-1 for both life stages (Fig. 6C). As expected, FLAG–ADR-2 was not immunoprecipitated from adr-2(-) worms (Fig. 6C). The qPCR data indicate that, on average, sevenfold more Y75B8A.8 neural reporter mRNA was present in ADR-2 IPs from L1 worms lacking adr-1 compared with IPs in the presence of adr-1 (Fig. 6D). In contrast, the enrichment of neural Y75B8A.8 mRNA in ADR-2 IPs from adult worms was similar in the presence and absence of adr-1. To examine the competition of ADR-1 and ADR-2 to bind other transcripts with increased editing during neural development, ADR-2 binding to W07G4.3, daam-1, and uba-2 was examined. Similar to Y75B8A.8, increased ADR-2 binding to all three transcripts was observed in adr-1(-) L1 worms (Fig. 6D), and binding of neural ADR-2 to both W07G4.3 and daam-1 in adult worms was not significantly different in the presence or absence of adr-1. A small (1.7-fold), but significant (P = 0.04) increase in neural ADR-2 binding to uba-2 was detected in adr-1(-) adult worms. In contrast, ADR-2 binding to lam-2, an mRNA in which ADR-2 is known to require ADR-1 for binding (Rajendren et al. 2018), was reduced in both L1 and adult adr-1(-) worms.

In sum, our model suggests that ADR-1 inhibits ADR-2 binding to a specific group of neural transcripts in early development (Fig. 6E). Neural-specific changes in gene expression result in adr-2 increased levels compared with adr-1 in adult worms. The neurodevelopmental gene expression changes allow ADR-2 to bind more efficiently to select neural transcripts, resulting in increased editing. Together, these data reveal a mechanism for neural- and development-specific regulation of RNA editing and provide new insights into molecular mechanisms that regulate ADAR substrate recognition and editing efficiency.