Engineered alterations in RNA editing modulate complex behavior in Drosophila: regulatory diversity of adenosine deaminase acting on RNA (ADAR) targets

Abstract

Select proteins involved in electrical and chemical neurotransmission are re-coded at the RNA level via the deamination of particular adenosines to inosine by adenosine deaminases acting on RNA (ADARs). It has been hypothesized that this process, termed RNA editing, acts to "fine-tune" neurophysiological properties in animals and potentially downstream behavioral outputs. However, the extreme phenotypes resulting from deletions of adar loci have precluded investigations into the relationship between ADAR levels, target transcripts, and complex behaviors. Here, we engineer Drosophila hypomorphic for ADAR expression using homologous recombination. A substantial reduction in ADAR activity (>80%) leads to altered circadian motor patterns and abnormal male courtship, although surprisingly, general locomotor coordination is spared. The altered phenotypic landscape in our adar hypomorph is paralleled by an unexpected dichotomous response of ADAR target transcripts, i.e. certain adenosines are minimally affected by dramatic ADAR reduction, whereas editing of others is severely curtailed. Furthermore, we use a novel reporter to map RNA editing activity across the nervous system, and we demonstrate that knockdown of editing in fruitless-expressing neurons is sufficient to modify the male courtship song. Our data demonstrate that network-wide temporal and spatial regulation of ADAR activity can tune the complex system of RNA-editing sites and modulate multiple ethologically relevant behavioral modalities.

FIGURE 1.

Visualization of dADAR expression using ends-out homologous recombination. A, schematic representation of the targeting construct used to insert an HA epitope tag at the 3′ of the dAdar locus. B, representative Western blot showing HA-positive bands in two independent lines lacking the white⁺ mini-gene. Actin was used as a loading control. *, nonspecific labeling. This is likely to be a head/brain-specific cross-reaction because it is not observed when using whole fly tissue (see Fig. 4B). C, quantification of relative dADAR-HA levels (normalized to actin) before and after Cre expression. Values are expressed relative to the mean of each post-Cre dAdar^HA line (n = 6 Western blots, three independent samples). Error bars, S.E. values. D, lamin and dADAR-HA staining in the male brain and thoracic ganglion. Scale bar, 10 μm. E, dADAR-HA co-localizes with DAPI-stained nuclei and Elav, but not Repo, in the male brain. Scale bar, 20 μm.

FIGURE 2.

Molecular reporter of RNA editing reveals neuron-specific patterns of dADAR activity. A, design of the reporter, termed syt-T. Exons 8–10 of syt-1, along with the intervening introns, were cloned into the pUAS expression vector. Upon transcription, the E1 and E2 elements form a pseudo-knot structure by base pairing with coding sequences in exon 9 (3), leading to formation of a dADAR substrate and editing of sites 3 and 4. B, example electropherograms showing editing of sites 3 and 4 in three genetically distinct cell types as follows: mushroom body γ neurons (201y), fruitless-positive (fru), and glutamatergic (ok371) neurons. Average editing of site 3 and 4 in 21 classes of neurons, defined by distinct Gal4 drivers, is shown in C and D. Each value is the mean of 4–6 RT-PCRs derived from males carrying each Gal4 driver and one of two independent insertions of syt-T. Error bars, S.E. values. E, dADAR expression in glutamatergic neurons. dADAR-HA and the nuclear red fluorescent protein red-stinger (21) driven by ok371-Gal4 are shown in the male brain and thoracic ganglion (upper panel), and at higher magnification in the central brain (middle) and thoracic ganglion (lower panel). F, dADAR-HA expression in Dachshund-positive Kenyon cells is clearly reduced relative to the surrounding nuclei. Scale bars, 20 μm.

FIGURE 3.

Varied impacts on mRNA re-coding following reduction of dADAR expression. A, heat map representation of editing levels at 68 sites in dAdar^WTLoxP heads and thoraxes, and the corresponding reduction of editing in the same tissue from dAdar^hyp males. Each editing site is represented by a four-symbol code (see supplemental Table 3 for details). B and C, editing levels in female heads for 8 LE sites (B) and 10 HE sites (C). The homozygotic and heterozygotic backgrounds containing various dAdar alleles are noted below each graph. Each value is the mean of ≥3 RT-PCRs. Error bars, S.E. values.

FIGURE 4.

Dynamic control of dADAR expression underlies developmental patterns of editing at low efficiency sites. A, dADAR and Elav expression in the L3 ventral nerve chord (VNC) and the adult thoracic ganglion (TG). Scale bar, 20 μm. B, LE sites are subject to strong developmental regulation. Editing at eight LE sites was examined at two larval stages (L1 and L3) and in adult males (A). Inset, representative example of n = 4 Western blots showing a strong increase in dADAR expression at the male adult-stage relative to L1 and L3. The two dADAR-HA bands likely represent dADAR proteins containing or lacking the alternatively spliced 3a exon, which is included at a higher level in the abdomen relative to the head and thorax (43). C, developmental profiles of 10 HE sites in L1, L3, and adult (A) males. Error bars, S.E. values.

FIGURE 5.

Global reduction in dADAR activity leads to altered patterns of locomotor activity. A–C, mean activity profile of dAdar^WTLoxP (A, n = 30), dAdar^hyp (B, n = 30), and dAdar^5g1 males (C, n = 16) under 12-h light-dark cycles (white and black bars). Each bar is an average value of time points from three consecutive days. In dAdar^WTLoxP, but not dAdar^hyp or dAdar^5g1 males, anticipation of morning and evening can be observed under both environmental conditions (arrowheads). D, quantification of morning anticipation in the three experimental genotypes. E, mean locomotor activity in male and female dAdar allelic backgrounds. Error bars, S.E. values. **, p < 0.005; ***, p < 0.0005; not significant (ns): p > 0.05 (Mann-Whitney U test).

FIGURE 6.

RNA editing is required for appropriate male courtship. A, time taken to initiate courtship (latency) is significantly higher in dAdar^hyp males (n = 20) relative to dAdar^WTLoxP controls (n = 15), yet the total time spent courting virgin females over a 10-min period (courtship index, CI) is not significantly different between either genotype (B and C). Courtship index was either calculated over the whole 10 min (B) or following initiation of courtship (C). Examples of three separate song trains are shown from a single dAdar^WTLoxP (D) or dAdar^hyp male (E). Note that although the trains from the dAdar^WTLoxP male are highly stereotyped, trains from even a single dAdar^hyp male show striking variability in waveform pattern. Scale bar, 10 ms. F–H, song parameters in dAdar^WTLoxP (n = 26 songs, 5 males) and dAdar^hyp (n = 44 songs, 9 males). Error bars, S.E. values. *, p < 0.05; ***, p < 0.0005; not significant (ns): p > 0.05 (Mann-Whitney U test).

FIGURE 7.

Knockdown of dADAR in fruitless-expressing neurons alters the male courtship song. A, example of electropherograms showing editing of syt-T site 3 and 4 expressed in fruitless-positive (fru) neurons within the male and female head or thorax. B, quantification of editing of two independent insertions of syt-T (n = 6–7 RT-PCRs for each value). C, dADAR expression was examined specifically in fru neurons by expressing a nuclear red fluorescent protein (23) using the fru-Gal4 driver line, in a dAdar^HA background. Nuclei of fru neurons can be detected throughout the brain and thoracic ganglion (upper panel). Examples of dADAR expression in fru neurons in the dorsal anterior segment and pars intercerebralis (middle panels) and meso-thoracic ganglion (lower panel) are shown at higher magnification below. D and E, example of song trains from control males heterozygous for driver (w⁺; +; fru-Gal4/+, n = 26 song trains, 10 males) or RNAi transgenes (w⁺; adr-IR1/+; adr-IR2/+, n = 30 song trains, 10 males). Note the similarity in waveform between song trains shown in D and E compared with those from dAdar^WTLoxP males (Fig. 6D). F, example of song trains from males with reduced dADAR expression in fru neurons (w⁺; adr-IR1/+; fru-Gal4/adr-IR2) (n = 27 song trains, 11 males). Note the extra spike in the first pulse and the polycyclic waveform in the last pulse. Scale bar, 10 ms. Error bars, S.E. values. *, p < 0.05; **, p < 0.005; not significant (ns): p > 0.05 (Mann-Whitney U test).

FIGURE 8.

Model for neuron to neuron variation in editing levels within the Drosophila nervous system. Top panel shows a graphical representation of the change in editing of one HE site (shab site 4; shb4) and two LE sites (ard site 2; ard2, and unc-13; unc1). Shab site 4 is edited at almost wild-type levels even in genotypes with very low dADAR expression, as is the case for all HE sites (Fig. 3). Thus, editing at this, and similar sites, is unlikely to vary widely from neuron to neuron, even though dADAR activity is highly variable in different neuronal populations (Fig. 2). In contrast, editing at LE sites is likely to vary substantially in neurons with differing levels of dADAR expression. Certain LE sites only required 50% of wild-type dADAR expression for achieving wild-type editing levels, while others required more robust dADAR expression (Fig. 3). The bottom panel shows a diagrammatic representation of three distinct neuronal subtypes (derived from Fig. 2), with low, medium (med), and high relative expression of dADAR. In neurons with low dADAR activity (such as mushroom body neurons), only HE sites such as shab site 4 are likely to be strongly edited. At slightly higher levels (for example, fru neurons), both shab site 4 and ard site 2 (i.e. the “higher efficiency” LE sites) will show editing but not weak LE sites such as unc-13. Finally, in neurons with high dADAR expression (such as photo-receptors; supplemental Table 2), all subclasses may be open to robust editing.