Widespread RNA editing decreases at elevated temperature
To investigate the effect of temperature on editing, we limited our scope to wild type Canton-S Drosophila melanogaster, in which the editosome has been well characterized [9]. Instead of by-hand data processing, as we have done previously [8,9,28], we created a batch processing script to remove bias, eliminate human error and increase feasibility of such a large-scale endeavor. This program does not improve upon the by-hand method, but is able to quickly process prepared files en masse using the same strategy as the by-hand method. We investigated 54 editing sites in fourteen different transcripts (see Additional file 1: Figure S1), representing the sites most highly conserved and best characterized of the over 3,000 adenosines known to comprise the Drosophila editosome [29]. The results from each individual site are compiled in Figures 1A, Additional file 2: Figure S2 and Additional file 3: Table S1.
We grew engineered D. melanogaster at 25°C before transferring newly-ecclosed male animals to equivalent incubators held at 10°C, 20°C and 30°C. We used only males to control for possible sex-specific differences. Drosophila eggs are most viable at 20°C [22]; therefore, the 20°C temperature range chosen is biologically relevant. After 72 hours of temperature acclimation, the animals were snap frozen at −80°C. We then performed RNA editing analysis using previously published methods [8].
Thermo-sensitive editing at individual sites occupies a range of possible patterns. For example, editing at sites 2 to 4 in the calcium sensor synaptotagmin-1 transcript is universally insensitive to temperature (Figure 1B). However, editing of paralytic at sites 1 to 3 decreases significantly at 30°C (Figure 1C). Contrary to the overall trend (Figure S2A), editing of site 6 in the potassium channel shab transcript is potentiated by temperature, significantly increasing in editing level over the temperature range (Figure 1D). Notably, shab site 7, located adjacent to site 6, is edited at nearly 100%, and is temperature-insensitive within the 20°C range studied (Figure 1D).
As there are several examples of multiple editing sites within a single transcript that respond differently to temperature (see Additional file 4: Figure S3), it is unlikely that the observed changes in editing are due to fluctuating transcript levels. Thus, while the general trend is for many sites to decrease editing level at 30°C, there are notable examples that are temperature invariant, as well as those whose temperature profile is counter to the overall trend.
Thermo-sensitive editing patterns are largely conserved
In order to determine whether the observed temperature-dependent editing responses are conserved in other species of Drosophila, we studied editing of select transcripts in five closely related Drosophilidae species (Figure 2, Additional file 5: Figure S4, Additional file 6: Table S2), representing over 10 million years of evolutionary divergence (Figure 2D) [30]. We observed that the thermo-responsive editing patterns at sites at which editing is insensitive to temperature, such as sites 2 to 4 in the synaptotagmin-1 transcript (Figure 1B) and site 7 in the shab transcript (Figure 1D), are also highly conserved between the species investigated (Figure 2A and C).
The quite diverse thermo-responsive editing patterns at paralytic sites 1 to 3 (Figure 2B), shab site 6 (Figure 2C), sites 1 to 3 in the complexin SNARE-binding protein transcript (see Additional file 5: Figure S4A), and dadar auto-editing (see Additional file 5: Figure S4B) are generally highly conserved, although absolute editing varies slightly and slope also varies significantly between species (see Additional file 6: Table S2). However, editing at the single site in uncoordinated-13, which encodes a protein involved in calmodulin binding in the larval central nervous system, is not conserved, either in absolute levels or in response to temperature (see Additional file 5: Figure S4C, Additional file 6: Table S2).
It is possible that the slight but significant changes in thermo-responsiveness of editing between species are a result of altered RNA structures, perhaps due to single nucleotide changes or polymorphisms within the genes of the Drosophila species studied here. As most structures that direct editing are formed between exonic and intronic sequences, changes in primary sequence often occur in intronic cis elements, leading to alterations in RNA structure and a corresponding change in editing. However, it is notable that editing at both dadar [31] (see Additional file 5: Figure S4B) and unc-13 (RA Reenan, personal observation, Additional file 5: Figure S4C) is directed by an entirely exonic secondary structure, whereas synaptotagmin-1 [7] and paralytic [8] are sites directed by structures comprised of paired exon and intron sequences. One would, therefore, expect the structures of dadar and unc-13 to be under higher sequence—and, therefore, structural—conservation, leading to highly similar editing at these sites across Drosophilidae. While auto-editing of the dadar transcript satisfies this prediction, editing in unc-13 is highly variable (see Additional file 5: Figure S4B and C), suggesting that either the unc-13 RNA structure or the specificity of editing machinery has evolved in the species studied, leading to the observed variability.
Our observations suggest that, in general, patterns of editing responsiveness to temperature are highly conserved and gene-specific in nature. The fact that insensitivity to temperature, temperature-sensitivity and temperature-potentiated site patterns are conserved is significant, and further suggests that while RNA structures that direct editing at certain sites are highly thermodynamically stable, others may be inherently more temperature labile, allowing editing levels to track with temperature based upon an as yet unknown interaction between temperature and RNA structure.
ADAR protein expression level is temperature-sensitive
Changes in editing levels may manifest due to effects at the RNA level (RNA folding, interactions with transcription/splicing machinery, negative influence of RNA chaperones) and/or directly via changes in dADAR protein concentration. We assayed dADAR protein in extracts from the heads of male flies reared at the experimental temperatures, as described above. We used engineered animals in which Jepson et al. tagged the endogenous dadar gene with an HA (hemagglutinin) epitope sequence using homologous recombination (HR) [32]. The appropriate control for mutations generated via HR is an allele in which no targeted mutations have been introduced through the engineering process, but which contains a loxP sequence, the only remnant from the HR procedure, within an intron [33]. The intermediate allele generated during the HR process contains a 5.5 kb mini-white gene, inserted within the same intron in the dadar locus, which produces a viable dADAR hypomorph with approximately 20% of the protein found in the loxP ADAR+ control [32].
Drosophila ADAR protein levels are stable between 10°C and 20°C, but decrease significantly at 30°C (Figure 3A-B, Additional file 7: Figure S5), generating a hypomorphic state. However, the amount of dADAR protein produced from a control loxP allele at 30°C is still significantly greater than that produced from the hypomorphic allele at any temperature. Moreover, the dadar hypomorphic allele is also sensitive to temperature, appears to experience reductions of protein levels between 10°C and 20°C and to be nearly undetectable at 30°C (Figure 3A-B). Although dADAR is completely undetectable in western blot analysis from engineered hypomorphic animals at 30°C, we observed no behavioral or phenotypic consequences. This is surprising given that dADAR null Drosophila display striking behavioral defects in motor control [13]. Although dADAR levels decrease when animals are kept at 30°C, editing at individual sites responds differently and independently to changes in temperature (Figure 1A), suggesting that dADAR concentration accounts for only part of the temperature responsiveness of editing sites.
Temperature-dependent editing patterns in loxP control and dadar hypomorphic animals are different and diverse even within a transcript (see Additional file 8: Figure S6), suggesting that dADAR levels are not fully responsible for temperature responsiveness. Finally, when editing in the hypomorph is compared to editing at different temperatures (see Additional file 9: Figure S7), it is clear that the dADAR level accounts for some, but not all, of the observed temperature dependent editing patterns (R = 0.48). Previously, we defined editing sites as ‘low-efficiency’ or ‘high-efficiency’, depending on their responsiveness to the hypomorphic dadar allele. Low-efficiency sites are sensitive to decreased dADAR concentrations, while high-efficiency sites are largely insensitive and are edited at a fairly constant level regardless of dADAR concentration [32]. In the present study, we manipulate the environment (30°C) to mimic the genetic hypomorph. However, the pattern of low-efficiency and high-efficiency sites is different between environmental (30°C) and genetic (hypomorph) lowered dADAR states (Figure 3 and Additional file 9: Figure S7), suggesting that the dADAR concentration is not the only factor driving editing levels and another temperature-sensitive factor, such as RNA folding, also directs editing.
Thermo-sensitive editing response is partially reversible
Although dADAR concentration decreases at 30°C compared to cooler temperatures (Figure 3), protein levels begin to recover after the animals are shifted to 20°C for just 24 hours, but do not fully recover even after 72 hours (Figures 4A and B, Additional file 10: Figure S8). Editing at specific sites displays varying degrees of recovery, most of which are statistically different from editing at 30°C and in the direction of editing at 20°C (Figure 4C). Editing at sites 2 to 4 of the synaptotagmin-1 transcript is largely resistant to temperature (Figures 1B, 4C), while editing of the paralytic transcript at sites 1 to 3 decreases significantly at 30°C (Figure 1C); all three sites recover slowly (Figure 4C) toward 20°C levels. Editing at shab site 6, which increases substantially at elevated temperatures (Figure 1D), actually appears to overshoot the 20°C level and is decreased to the 10°C level after 24 hours, and stabilizes to the 20°C level after 72 hours (Figure 4C). This suggests that editing of certain sites is driven more by dADAR level, and that these sites take longer to recover from elevated temperature, while editing at other sites is driven by another factor, such as the more rapid response of RNA structure to temperature.
Edited dADAR isoforms are not differentially sensitive to temperature
The dADAR protein auto-edits its own transcript, causing an amino acid substitution at position 458 in the deaminase domain. Unedited dadar transcripts produce a peptide encoding a serine at position 458 (dADARS), while edited transcripts produce the glycine isoform (dADARG). We previously reported that the dADARS and dADARG isoforms target the same editing sites, but edit them to slightly different levels, with the dADARG isoform displaying less activity in vivo [9]. Auto-editing is stable between 10°C and 20°C at about 55%, although dadar editing decreases to 37% at 30°C (see Additional file 2: Figure S2B, Additional file 3: Table S1), increasing the proportion of the unedited dADARS isoform. We, therefore, suspected that our observed overall temperature-dependent editing patterns (Figure 1A) might be a result of differential stability of the dADARS and dADARG isoforms.
We previously used HR to generate engineered animals in which the dADARS and dADARG isoforms were permanently hardwired into the HA-tagged dadar locus. Although behavior was altered in both the dADARS and dADARG lines [9], the isoforms do not have significant differential temperature stability (Figure 5, Additional file 11: Figure S9) compared to each other and to wild type mixed isoform dADAR (Figure 3). Therefore, it is unlikely that differential temperature stability of the dADARS and dADARG isoforms is responsible for the overall changes in RNA editing apparent in Figure 1A.
Structural RNA mutations affect temperature-dependent patterns
If RNA structures directing editing are partially responsible for the thermo-sensitivity of editing at specific sites, we reasoned that as temperature changes, the specific structures required for RNA editing may be more or less common leading to altered editing levels.
We tested the editing thermo-sensitivity profile of knock-in RNA structural mutations in the paralytic voltage-gated sodium channel transcript, engineered into the endogenous locus using HR [8,33]. The RNA structure directing editing at paralytic sites 1 to 3 is a complex tertiary pseudoknot (Figure 6A). The intronic editing site complementary sequence (ECS) is required for editing at all three sites, while the ‘donor site complementary sequence’ (DCS), which sequesters the splice donor in secondary structure, titrates the level of editing at all three sites. Finally, the tertiary pseudoknot interaction, formed between an intronic hairpin loop and a ‘docking site’ 3′ to the ECS, selectively directs editing at site 1 [8]. We decided to use mutant lines, in which the structure of the paralytic transcript is altered in multiple directions, to assay the effect of temperature via RNA structure on editing levels because these sites are among the most temperature sensitive in our study (Figure 1A). We have previously published a series of mutations, engineered via HR into the endogenous paralytic locus. These mutations, which are described below, alter editing at paralytic sites 1 to 3 by perturbing the local and long-range primary and secondary RNA structure around the edited adenosines [8]. As with other instances of HR, the appropriate control for HR-engineered mutations is a loxP control allele.
The ‘DCS delete’ mutation, which is predicted to decrease secondary structure (Figure 6B), decreases editing at all three paralytic editing sites, while the ‘DCS zip’ mutation (Figure 6C), which increases secondary structure, is known to increase editing at all three sites [8]. The thermo-responsive editing profile of these structural mutations appears to be a product of both temperature and structure. The ‘DCS delete’ mutation displays decreased editing at all three sites compared to the loxP control, as expected, yet the thermo-responsive editing pattern, measured as slope, is significantly different at all three sites than that seen in the loxP control (Figure 6D, Additional file 12: Table S3). Similarly, the ‘DCS zip’ mutation shows decreased thermo-responsive editing patterns at all three sites compared to the loxP control (Figure 6D), in addition to increasing editing in all conditions, as expected [8]. The altered thermo-sensitivity of these mutations, defined here as slope, is very subtle, yet is still significant (see Additional file 12: Table S3).
We also tested the thermo-responsiveness of structural mutations in the paralytic tertiary pseudoknot structure (see Additional file 12: Table S3, Additional file 13: Figure S10 A-D). The individual pseudoknot mutations, ‘Loop > α’ and ‘Dock > α’, both disrupt the tertiary RNA interaction, and selectively abolish editing at site 1, while preserving editing at sites 2 and 3. The double mutation, ‘Loop/Dock > α/α’, combines these mutations and is predicted to restore the tertiary pseudoknot structure, rescues editing at site 1 and increases editing at all three sites [8]. These mutations reveal temperature sensitivity patterns (slope) distinct from that seen in the loxP control (see Additional file 12: Table S3, Additional file 13: Figure S10E), suggesting that engineered RNA structures are affected differently by temperature, which manifests in the thermo-sensitive editing response of paralytic editing sites.