Tissue-specific patterns of RNA editing

Our study of temperature effects on editing in six tissues of zebrafish shows that a high level of tissue specificity occurs in total amounts of editing and in the sites that are edited. Both features of editing exhibited strong dependence on acclimation temperature, with reductions in temperature generally but not invariably leading to increased editing of a message, as shown in Fig. 6. Furthermore, temperature effects on editing differed among tissues. The causes and functional consequences of these temperature-dependent tissue-specific changes in RNA editing are largely unknown. In terms of causal relationships leading to tissue-specific editing, some of the observed differences in editing could simply be due to differences in the transcriptomes of the six tissues. In addition, differences might arise from the activities of different isoforms of ADAR that are active in different tissues. Vertebrates typically have three isoforms of ADAR, and these have different tissue distributions and preferences for editing sites [12]. ADAR1 is expressed in all tissues. ADAR2 is expressed most highly in brain but is found in other tissues as well. ADAR3 is expressed exclusively in brain. Because of gene duplication, zebrafish have four ADAR enzymes: ADAR1, ADAR2a, ADAR2b, and ADAR3 [13]. To our knowledge, comparisons of the temperature dependencies of activities or site specificities of the vertebrate ADAR isoforms have not been made. However, in Drosophila, which has only a single isoform of ADAR, editing activity at 35 °C was substantially reduced relative to activities at 15 °C [14]. A similar temperature dependence was reported in a later study of Drosophila [15]. This reduced activity at high temperature was due to combined effects of a reduced level of expression of the adar gene and temperature-dependent auto-editing of the enzyme that led to a reduced level of activity [14]. Temperature effects on ADAR activity could also result from alterations in RNA secondary structure, as discussed below. Further exploration of temperature-ADAR interactions is warranted in order to enhance our understanding of the mechanistic basis of temperature-dependent RNA editing.

The functional consequences of tissue-specific temperature-dependent editing are largely unknown. However, the categories of mRNAs edited in different tissues reflect in some measure effects related to tissue-related functions. Thus, in the highly edited brain, editing of mRNAs associated with synaptic assembly, axonal transport, and regulation of circadian rhythm are consistent with adaptive modification of neural function during acclimation. Studies of temperature acclimation in fish have revealed substantial changes in the biophysical properties of brain synaptic membranes that are correlated strongly with adaptive changes in behaviors like hyperexcitability and maintenance of equilibrium [34]. Studies of invertebrate species also have discovered relatively large amounts of editing in brain, and some of these are recoding events that foster temperature-adaptive changes in protein function. For example, comparisons of polar and tropical octopus species revealed temperature-adaptive recoding of the mRNA of a potassium voltage-gated ion channel that adaptively altered the channel’s function [1, 35]. Increased editing at low temperatures led to a protein with an intrinsically higher rate of channel closing, which would be compensatory for the effects of cold on rate of channel function. It is noteworthy that RNA editing by ADAR most frequently replaces a large side-chain (R group) with a relatively small one [1]. RNA editing thus leads to increased contents of glycine and alanine, for example, which would increase protein flexibility and thereby compensate for effects of reduced temperature on protein conformational stability. RNA editing led to only a relatively small amount of recoding in the cold-acclimated zebrafish (Fig. 4D). In keeping with earlier studies [1], glycine increased during cold acclimation, a change consistent with temperature-compensatory change in protein stability [1, 36]. The adaptive function, if any, of the substitution of arginine for lysine remains to be determined. Studies of temperature-dependent editing in Drosophila also have provided evidence for adaptive change in brain function mediated by ADAR-driven editing [4, 15, 17]. The best understood instances of RNA editing in nervous tissue are from studies of mammals, especially humans [1], but temperature effects have not been studied in these homeothermic animals. The many clear examples of how editing-induced changes in amino acid sequence modify the activities of neural proteins like ion channels in homeothermic mammals suggest that brain tissues of ectotherms also may benefit from recoding events that enable these key proteins to function well at different temperatures. In fact, it could be instructive to examine the ectothermic orthologs of mammalian brain proteins that are strongly edited. Such studies might reveal the taxonomic breadth of specific types of editing changes that can affect, for example, channel activity.

In muscle, a considerable amount of editing was in mRNAs encoding proteins involved in muscle cell differentiation, including striated muscle differentiation. Temperature acclimation in fish commonly leads to alterations in capacities for swimming and muscle phenotype, as noted in a study of zebrafish [31]. Thus, some of the editing in classes of RNA like “calcium ion trans-membrane transport” and “myofibril assembly” could reflect temperature-compensatory changes in muscle functional capacities to support swimming and for supporting muscle fiber differentiation, respectively. In gill tissue, editing of mRNAs associated with “regulation of intracellular pH” could reflect adaptive change in properties of systems of acid–base regulation, which might be important in establishing the appropriate temperature-dependent values of extra- and intracellular pH [36]. In mammals, there is evidence for regulation of editing by intracellular acidification [37]. Further investigations of interactions between temperature, RNA editing, and pH regulation are warranted. Likewise, the editing categorized as “response to temperature stimulus” and “response to heat” that occurred in some tissues points to additional lines of future study.

Similarities in editing activities in different tissues

Whereas there was marked variation among tissues in editing patterns and temperature sensitivity of editing, there were also cases where similarities among tissues were striking. The editing of mRNAs related with “ER to Golgi vesicle–mediated transport” and/or “response to endoplasmic reticulum stress” occurred in all tissues. The results are consistent with previous studies that showed that cold enhanced endoplasmic reticulum stress [38]. Such stress may reflect the extreme sensitivities of membrane biophysical state (static order and propensity to form non-lamellar structures) and, hence, membrane function to changes in temperature [36].

Temperature effects on RNA structure and amounts of editing

There are a number of reasons to expect that temperature will have a pronounced effect on RNA editing processes. First, the secondary and tertiary structures of RNA are thermally labile, such that even small changes in temperature are likely to perturb these structures [19, 39]. Second, and following from the previous point, because RNA editing by ADAR requires that RNA be double-stranded, temperature changes that either stabilize (lower temperatures) or disrupt (higher temperature) dsRNA could alter the availability of suitable substrates for ADAR activity. To a first approximation then, reduced temperature would favor increased editing due to the availability of more regions of double-stranded structure. We usually found that the amount of editing rose as acclimation temperature was reduced, but as discussed above, there were exceptions to this general pattern.

The most common pattern of temperature dependence of editing was the increase in edited sites that accompanied acclimation to reduced temperature. For example, in brain the number of edited sites rose from 24,678 at 27 °C to 32,471 at 18 °C and to 34,400 at 13 °C. The enhanced level of editing at reduced temperature is consistent with expectations based on the need of ADAR activity for dsRNA. As noted above, the levels of dsRNA would be expected to rise as temperature falls due to more stable base pairing at low temperature. However, this direct effect of temperature on levels of dsRNA, while likely to play some role in determining the temperature dependence of editing, cannot fully account for inter-tissue differences in editing levels or for the variation in editing sites among tissues (Fig. 3).

In the principal components analysis of RNA editing (Fig. 3), the missing values are imputed using missForest [40]. However, it is important to note that the editing levels exhibit a heavy right-skew, which means that imputation of missing values could potentially introduce bias.

Editing as a means of influencing efficiency of translation and re-organizing the proteome

The discovery that editing affects the efficiency of translation and does so in ways that differ among types of mRNAs (Fig. 6) has major implications for a potential role of editing in re-organizing the proteome during acclimation to temperature. It has been clear for many years that the relative as well as the absolute activities of different biochemical pathways may shift during thermal acclimation in ectotherms [41, 42]. Whereas these shifts may result from temperature-acclimatory alterations in patterns of transcription of new mRNAs [42], our finding that editing of 3′UTRs leads to significant changes in the efficiency of translation and that these effects differ among classes of mRNA points to another mechanism for acclimation of the proteome and metabolic re-organization. Importantly, the finding that one of the five mRNAs examined (srsf5b) exhibited greater editing at high (18 °C) than low (13 °C) temperatures illustrates the potential versatility of ADAR editing in orchestrating re-organization of the proteome, whether acclimation temperature is increased or decreased. The role of editing on the 5′UTR sequence could also be important in governing translation, even though much less editing occurred in this region relative to the 3′UTR. Because the 5′ region is key for initiation of translation, even small changes in base composition that lead to alterations in conformational stability could be instrumental in adjusting rates of initiation [43].

The mechanisms by which editing of the 3′UTR altered the efficiency of translation of these five mRNAs were not examined in our study, but there is evidence to support the conjecture that editing of UTRs that alters their secondary structures might account for these effects. Studies in Drosophila have shown that high fractions of edited sites (at 18 °C) are in regions of secondary structure [15]. Moreover, editing at a single base has been shown to be sufficient to alter RNA secondary structure [15]. When such editing occurs in UTRs, it could be important in governing different events in the process of translation. Thus, the secondary structures of 5′ and 3′UTRs are known to play important roles in modulating key steps of translation, including rates of initiation and elongation. For example, reduced secondary structure in the 5′UTR region enhances initiation, whereas increased secondary structure in the CDS and 3′UTR favors high protein expression [44]. Our finding that increased editing of the 3′UTRs in 5 different mRNAs led to higher efficiency of translation might be a reflection of increased secondary structure in these regions due to editing, increases that occur in response to either a rise or fall in acclimation temperature, depending on the mRNA in question.

The time courses of editing-mediated changes in the transcriptome and proteome merit further consideration. Rapid changes in the transcriptome in response to an acute change in temperature may involve differential effects on the 3′UTRs of different classes of mRNAs. For example, a rapid and adaptive re-organization of the transcriptome in response to acute change in temperature was demonstrated by Su et al. [45]. Their work (with rice) showed that disruption of 3′UTR structure by acute increases in temperature fostered degradation of “housekeeping” mRNAs by making the RNAs susceptible to RNA degradation. This effect was not found in mRNAs that encode stress-related proteins [20]. The roles of such temperature effects on 3′UTR regions during longer-term acclimatory processes remain to be determined. However, these findings suggest that RNA editing that occurs in 3′UTR sequences and influence mRNA stability and turnover could be important in adjusting the composition of the transcriptome and, therefore, the proteome during temperature acclimation. Proteome re-organization mediated through editing-governed translational efficiency would likely be more quickly achieved than effects requiring transcription because editing effects could arise from rapid alterations of translational activities involving pre-existing mRNAs. Thus, editing of existing transcripts might contribute importantly to at least the initial shifts in metabolic organization that occur during acclimation. Although the effects of acute temperature changes on 3′UTR secondary structures are well established, the roles of ADAR editing in adjusting the stabilities of 3′UTRs over longer acclimation periods such as the 30-day acclimation period of the present study remain to be demonstrated.

Although the work performed with the dual luciferase system involved only five types of mRNAs, the consistency in the findings—translation was stronger in the edited mRNAs in all cases—points to the possibility of a general role of mRNA editing in altering translational patterns in the face of changes in temperature. These ADAR-mediated effects could be important in responses to both decreases and increases in temperature, as shown by the data in Fig. 6. This discovery raises interesting questions about differences among RNAs in temperature sensitivities in their ADAR sites that determine whether editing is favored by increase or decrease of temperature. Thus, do ADAR sites evolve to have specific responses to increases or decreases in temperature that lead to adaptive changes in gene (protein) expression, as mediated through translational efficiency? Do the effects of editing modify the secondary or tertiary structures of the mRNAs in manners that influence initiation of translation, rates of translation, and mRNA half-life? Answers to these questions would shed important light on the mechanisms by which ectotherms re-organize their metabolic pathways during thermal acclimation.

A final point to consider in the context of how mRNA editing might lead to metabolic re-organization during temperature acclimation is the finding that acclimation temperature has an effect on mRNA splicing. Healy and Schulte [46] discovered that cold acclimation affected splicing patterns in hundreds of genes in four species of fishes. A number of the genes for which temperature-dependent splicing was observed were genes that exhibited upregulation in the cold. It is not known if RNA editing was the mechanism involved in governing alternative splicing, but this possibility merits investigation [9]. Thus, it is possible that mRNA editing could alter the proteome by changing splicing patterns as well as by modulating translational efficiencies of mRNAs.

CDS editing: roles of synonymous base changes

Whereas the importance of recoding editing in modifying the thermal responses of proteins has been demonstrated [1], the roles of synonymous changes in the CDS remain to be analyzed in the context of thermal acclimation. Because changes between synonymous codons can affect rates of translation [47] and the final folded state of a protein [48], even the small fraction of synonymous changes noted in the CDS could have physiological significance through affecting the amounts of protein synthesized and the final conformational and functional states of the proteins.

Conservation of corresponding states of stability in RNA secondary and tertiary structures—a major role for RNA editing in ectotherms?

Lastly, we wish to consider issues that reflect major differences between editing in homeothermic endotherms like mammals and ectothermic species like fishes. We conjecture that ADAR editing may play quite different roles in these two broad groups of animals. To approach these differences, we first consider another perspective on the significance of temperature-dependent stability of the base pairing that establishes RNA secondary and tertiary structures. This analysis is based on considerations of the need to conserve the abilities of RNAs to undergo reversible conformational changes that are essential for function at different body temperatures. Reversible formation of secondary structural elements is key to several aspects of mRNA function, so the ability to retain a capacity for undergoing these reversible changes in conformation at different temperatures seems critical. This type of conservation of conformational stability is found in orthologous proteins of species adapted to different temperatures [19, 49, 50] and is termed the retention of “corresponding states.” The recent discovery of evolved temperature-adaptive differences in stability of mRNA secondary structure suggests that mRNA structure, like protein structure, evolves to maintain the correct balance between conformational stability and flexibility, i.e., corresponding states of conformational stability, at normal cell temperatures [21].

We conjecture that a similar temperature-adaptive pattern in RNA stability could arise from RNA editing and could potentially conserve the right balance between stability and flexibility in all components of mRNA (CDS, 3′UTR, introns, etc.). Increased strength of base pairing as temperature is reduced could lead to perturbation of RNA function if, for example, secondary structures become too rigid to allow needed changes in mRNA conformation during translation, splicing reactions, or RNA turnover. Editing might lead to reduced amounts of base pairing at low temperature, an effect that could compensate for the strengthening of base pairing due to reduced thermal energy. This type of adaptive change in RNA structural stability across different genomic regions during acclimation may be the most general role of temperature-dependent RNA editing in ectotherms. Conversely, the fact that the preponderance of RNA editing in mammals is focused on repetitive sequences, e.g., Alu elements, and is less common in other genomic regions, could be a sign of a lack of need for genome-wide temperature-compensating changes in RNA stability in homeotherms. In this context, it merits mentioning recent work on hibernating (heterothermic) ground squirrels in which RNA editing increased during periods of low body temperature [7]. Few changes in CDS were observed; most editing was in interspersed repeats (SINE elements) that engage in dsRNA formation and can trigger the innate immune response [51]. Enhanced RNA editing during hibernation may be a mechanism for minimizing inflammation during periods at low, near-zero body temperatures [7], rather than a mechanism for ensuring corresponding states of RNA structural stability to allow maintenance of all categories of RNA function.