OUP user menu

RNA editing

Axel Brennicke, Anita Marchfelder, Stefan Binder
DOI: http://dx.doi.org/10.1111/j.1574-6976.1999.tb00401.x 297-316 First published online: 1 June 1999


The term RNA editing describes those molecular processes in which the information content is altered in an RNA molecule. To date such changes have been observed in tRNA, rRNA and mRNA molecules of eukaryotes, but not prokaryotes. The demonstration of RNA editing in prokaryotes may only be a matter of time, considering the range of species in which the various RNA editing processes have been found. RNA editing occurs in the nucleus, as well as in mitochondria and plastids, which are thought to have evolved from prokaryotic-like endosymbionts. Most of the RNA editing processes, however, appear to be evolutionarily recent acquisitions that arose independently. The diversity of RNA editing mechanisms includes nucleoside modifications such as C to U and A to I deaminations, as well as non-templated nucleotide additions and insertions. RNA editing in mRNAs effectively alters the amino acid sequence of the encoded protein so that it differs from that predicted by the genomic DNA sequence.

1 Introduction – no border between RNA editing and RNA modification

RNA editing processes are confined to the world of eukaryotes and have not so far been recognized in microorganisms. Just as introns were eventually identified in bacterial genes including some in Escherichia coli, posttranscriptional changes in RNA sequences described as RNA editing may also occur in prokaryotes.

RNA editing covers a multitude of biochemically different processes, the term having been adopted in 1986 for the addition and deletion of uridine nucleotides to and from mRNAs in trypanosome mitochondria [1]. The observed changes in nucleotide identity or insertion/excision of a nucleoside described below are termed RNA editing rather than being classified as the classical (30 years older) ‘RNA modification’. The latter processes change (modify) the nucleoside in the RNA chain, sometimes only by a minor modification, for example by the addition a methyl group, at other times significantly (Fig. 1) [2]. Sometimes modification even alters the nucleotide structure to such an extent that base-pairing properties change significantly [3]. These modifications are most prevalent in tRNAs and rRNAs and are often essential for biological function. In some tRNAs a cytidine in the anticodon can be modified to function as a uridine in codon recognition to read different nucleotide triplets in the mRNA. In some of these cases modifications in both eukaryotes and bacteria change the information content of the tRNA so that it does not faithfully reflect the genomic blueprint. A further complication includes modification of some tRNAs by 3′ additions of oligoA, which however is commonly described as RNA editing.

Figure 1

Phylogenetic distribution of various types of nucleoside modifications in RNA. Reprinted from [55] with kind permission.

Here we adopt the definition of Price and Gray [4], which states that RNA editing describes processes of nucleotide alterations which result in different or additional nucleotides in the RNA which could as well have been encoded in the genomic sequence.

2 Different species – different genomes – different organelles – different processes

RNA editing is found in very diverse species of eukaryotes, in humans and trypanosomes, in slime molds and plants (Table 1). The evolutionary distances between the different organisms in which editing has been observed not only reflect the distribution of researchers’ favorite organisms, but also show the wide evolutionary distances between the different editing mechanisms. While one level of classification can thus be based upon a phylogenetic scheme, another grouping could reflect the mechanisms involved.

View this table:
Table 1

The different processes of RNA editingf

TypeaOrganism (genome)Transcript(s)bcis-acting element(s)trans-acting factor(s)Mechanism
Insertion/deletion editing
U insertion/deletionKinetoplastids (mt)mRNAsAnchoring sequencegRNAs, TUTase, RNA ligase, endonuclease, U exonuclease, other factorsCleavage, TUTase or U exonuclease action, and ligation
Mostly C insertion; also U, UA, AA, CU, GU, and GCPhysarum polycephalum (mt)mRNAs, tRNAs, rRNA??Linked to transcription
G insertionParamyxoviruses (v)P mRNASlippery sequenceViral polymerasePseudotemplated transcription
A insertionEbola viruses (v)GP mRNASlippery sequenceViral polymerasePseudotemplated transcription
3′-terminal A additionVertebrates (mt)mRNAsFlanking tRNA structureEndonuclease, TATaseCleavage and TATase action
3′-terminal A additiondMetazoan animals (mt)tRNAs3′-overlapping tRNA?Endonuclease, TATase?Cleavage and TATase action
C to A, A to G, U to G, and U to AAcanthamoeba castellanii, chytridiomycete fungi (mt)tRNAsInternal guide sequence?Endo- or exonuclease, nucleotidyltransferase?Replacement of first three 5′ nt
Modification editing
C to ULand plantsmRNAs (mt, cp), tRNAs (mt), rRNAs (mt)Flanking sequence?C deamination
Mammals (n)mRNAs, apoB (Gln→stop), NF1 (Arg→stop)Mooring sequence, efficiency and AU-rich elementsC deaminase (APOBEC-1), other factorsC deamination
Physarum polycephalum (mt)cox1 mRNA???
Marsupials (mt)tRNA (Gly→Asp anticodon)???
U to CLand plants (mt, cp)mRNAsFlanking sequence?U amination?
Mammals (n)WT1 mRNA (Leu→Pro)???
A to IMammals (n)mRNAs, GluR-B, -C, -D, -5, -6, 5-HT2CRdsRNA structuredsRNA A deaminases (ADAR1 and -2)A deamination
Human hepatitis deltaAntigenome (stop→Trp)dsRNA structureADAR1A deamination
A to I?eSquids (n)Kv2 K+ channel mRNA??A deamination?
Drosophila melanogaster (n)4f-rnp mRNA??A deamination?
G to AcMice (n)GPT mRNA (Cys→Tyr)??Base or nt replacement?
U to AHumans (n)α-Galactosidase mRNA (Phe→Tyr)??Base or nt replacement?
ADAR, a deaminase for RNA; apoB, apolipoprotein B; cox1, cytochrome c oxidase subunit 1; GluR, glutamate receptor; GP, glycoprotein; GPT, GlcNac-1-phosphate transferase; 5-HT2CR, serotonin 2C receptor; NF1, neurofibromatosis type 1 gene; TATase, terminal uridylyltransferase; WT1, Wilms’ susceptibility gene 1; cp, chloroplast; ds, double stranded; mt, mitochondrial; n, nuclear; nt, nucleotide; v, viral;?, uncertain or unknown.
This (putative) editing event has been reported for the NIH/SW mouse strain. Whether it occurs in other strains and species is unknown.
  • aInsertion/deletion editing is defined as an editing process during which phosphodiester bonds are made and/or broken, resulting in almost all cases in an edited RNA in which the number of nucleotides has changed. In modification editing, the identity of a nucleotide is changed by base conversion or substitution without disruption of the phosphodiester backbone. The status of G to A and U to A editing in two mammalian RNAs is unclear: it could be the result of either base substitution (in which case it would qualify as modification editing) or nucleotide replacement (in which case it should be considered insertion/deletion editing).

  • bFor modification editing, the translational consequences of editing have been indicated for RNAs that contain a single editing site.

  • cFor other references, please check the corresponding chapters.

  • dIn platypuses, As and Cs are added.

  • eA comparison of genomic and cDNA sequences shows multiple apparent A to G changes, suggesting that the RNAs are edited by A to I conversion (see chapter 19 of [55]). The presence of Is in edited RNAs has not yet been established, however.

  • fReprinted from [55] with kind permission.

In ‘insertional editing’ nucleotides are inserted into (or deleted from) the single stranded RNA molecule. These insertions are not specified in the gene sequence from which the RNA has been transcribed. In the insertional editing processes found in mitochondria of the Trypanosoma parasites and of the slime mold Physarum, the mature mRNA will thus be longer than the respective primary transcript [58]. In trypanosomes some mature mRNAs have more nucleotides inserted than are encoded by the gene. In these cases the translated mRNA may thus be more than twice as long as the actual gene in the DNA.

Insertional RNA editing also includes the non-templated cotranscriptional insertion of nucleotides in certain viruses, which is specified by the surrounding sequence [9]. Other borderline cases are the 3′ polyA additions which complete UAA translational stop codons and some tRNAs in animal mitochondria.

A second class of RNA editing, ‘modification editing’, is more closely related to the RNA modification processes as reflected in the term. In this category the two most prevalent changes induce deaminations of C to U and A to I. Less widely distributed changes are U to C, G to A and U to A conversions, the latter being found in mammals such as mice and humans. To summarize the details of RNA editing processes from the clearly eukaryotic nuclear-cytoplasmic editing events to the RNA editing processes found in the organelles, we will require a few novel concepts developed in the investigation and description of RNA editing. One of these is the term ‘editosome’, which describes the usually unknown protein or protein-RNA complex responsible for the RNA editing reaction. Another concept is the ‘guide RNA’ molecule, which through base-pairing with the RNA molecule to be edited specifies the editing site. An editing site is the nucleotide actually altered and can include those nucleotides immediately adjacent or in the immediate vicinity.

3 RNA editing in the nucleus of animals

RNA editing of transcripts of nuclear genes in animals are very rare events and have been reported for only three or four specific genes. Examples in mammals are listed in Table 1 and include the single C to U changes in the apolipoprotein B and neurofibromatosis type 1 tumor suppressor mRNAs, a U to C alteration in Wilms’ tumor susceptibility mRNA, A to I deaminations in GluR channel mRNAs and in 5-HT2CR serotonin receptor mRNAs. A to I changes have also been described in Kv2 K+ channel mRNAs from squid, in the Drosophila 4f-rnp mRNA and in the human hepatitis delta virus. G to A changes have been observed in the mouse GlcNac-1 phosphate transferase mRNA, and U to A alterations in human α-galactosidase mRNAs. The nucleotide specificities and the enzymatic mechanisms involved in the tissue-specific C to U editing in the apolipoprotein B and the A to I deaminations in GluR channnel mRNAs have been characterized and will be considered in more detail [10,11].

3.1 RNA editing in vertebrate apolipoprotein B mRNAs

In vertebrates a single apolipoprotein B (apoB) gene (Fig. 2A) encodes two different proteins, one of 100 kDa (APOB100) and the other of 48 kDa (APOB48) [1118]. In humans APOB100 is synthesized in the liver, whereas APOB48 is made in the small intestine. Both proteins are involved in binding lipids to keep them in solution. APOB100 is secreted into the blood plasma, while APOB48 is involved in binding and resorbing fatty acids in the lining of the small intestine. In this tissue the apoB mRNA is modified at nucleotide 6666, where a cytosine is deaminated to a uridine. The CAA codon is converted to a UAA translational stop codon and as a result protein synthesis terminates to yield the smaller APOB48.

Figure 2

RNA editing in the nucleus of mammalia. A: RNA editing in the apoB transcript produces a stop codon by C deamination in a genomic CAA codon to a UAA codon in the intestine, but not in the liver. Two different proteins are synthesized from just one gene, one of 100 kDa as specified by the 29 exons of the genomic sequence (top part) and another of 48 kDa after editing in the intestine. B: Secondary structures folded between intron and exon sequences determines the A to I conversion sites in RNA editing of various glutamate-responsive channels (GluR), a serotonin receptor 5-HT2CR and the hepatitis delta virus (HDV). These A deamination editing sites are often found in double stranded regions of the primary transcript RNA. Reprinted from [55] with kind permission.

This deamination reaction is catalyzed by a site-specific enzyme, the APOBEC-1 enzyme, which will also deaminate cytidines in vitro [19]. This is the catalytic as well as the sequence-specific subunit of an ‘editosome’ complex, which contains at least three other as yet unidentified proteins. These are required for the highly specific activity of apobec-1 in vitro, particularly to bind and access the apoB mRNA. In the yeast two-hybrid system an APOBEC-1 binding protein has been identified, which shows specific affinity for both the deaminase and the apoB mRNA. The tissue-specific expression of the deaminase APOBEC-1 coincides with the distribution of the apoB RNA editing in different vertebrates. In humans and rabbits RNA editing and APOBEC-1 activity only occur in the small intestine, while in rats and mice both activities are also present in the liver. In addition to liver and small intestine the APOBEC-1 activity is expressed in some tissues where the apoB gene is not activated. In these cells APOBEC-1 may play a role in the deamination of other mRNAs, e.g. the nf1 (neurofibromatosis type 1 tumor suppressor) and the nat1 (novel apobec-1 target) mRNAs.

The biological significance of the apoB mRNA editing and of the additional deaminations is as yet unclear, but may be revealed by experiments with apobec-1 overexpression and transient transfection experiments in knockout mice. These investigations have so far given equivocal results regarding the effects and importance of apoB editing. Transient expression of the APOBEC-1 deaminase in the liver by transfection reduces the amount of APOB100 with a corresponding increase of the APOB48 protein. After several days, however, the plasma concentration of the latter decreases. Overexpressing apobec-1 in mice in which the gene for the low density lipoprotein has been knocked out similarly decreases the concentration of APOB100 in the plasma and in effect also lowers the concentrations of the APOB100 containing lipoproteins. Elevated cholesterol concentration based on misregulation of these two carriers thus appears to be influenced by the increased editing efficiency. Transgenic mice overexpressing apobec-1 developed carcinomas apparently due to the deleterious over-editing of other transcripts such as nf1 and nat1.

In summary, RNA editing of a unique nucleotide in the apoB transcript results in the synthesis of two different proteins with different properties from one gene (Fig. 2A). In this case RNA editing increases product complexity from just one genomic coding region. However, considering the vast amount of non-coding DNA in the mammalian genome, encoding an additional gene does not seem to pose a problem. The presumed phenotypic and evolutionary advantage of this editing must thus be sought elsewhere, perhaps in the capability of a more rapid response to metabolic changes in the quality and quantity of fatty acids which are released depending on the type of food entering the small intestine.

3.2 RNA editing in glutamate-responsive ion channels

In mRNAs encoding some of the subunits of glutamate-responsive ion channels (GluR), nucleotides recognized as G-residues were found at some positions, where A nucleotides were encoded by the respective gene [10,12,20]. These nucleotides turned out to be inosines (I) derived by hydrolytic deamination of the purine ring in the specific adenosines. The I residues are recognized as G in almost all biological systems, including translation. This biochemical alteration of a nucleoside results in an RNA editing event leading to an altered information content which directs synthesis of one or more different proteins. Such A to I modifications are found in the different GluR subunit mRNAs and, depending on the subunit type, changes range from one to seven nucleotides (Fig. 2B). In addition, in the hepatitis delta virus RNA one and in a serotonin receptor transcript (5-HT2CR) four nucleotides are altered from A to I. Specificity in these RNA editing events is mediated by the preference of the single subunit deaminase (dsRAD or DRADA or ADAR1=adenosine deaminase acting on RNA) for double stranded RNA [19,21]. Backfolding into extensive hairpin loops generates double stranded regions surrounding the nucleotide to be altered. These secondary structures bind the enzyme and direct the activity to exposed adenosines. The first enzyme of the diverse family of dsRNA deaminases was identified by its destabilizing effect on RNA duplexes before RNA editing was recognized. Base-pairings are destabilized by changing adenosines to inosines, the RNA duplex unwinds and becomes single stranded.

In some gluR transcripts stable duplexes are only present in unspliced precursors, where intron and exon sequences can base-pair. This observation confirms that the A to I type of RNA editing occurs in the nucleus and not in the cytoplasm of animal cells, since it precedes intron splicing which only occurs in the nucleus. The physiological and phenotypic consequences of these editing events can be observed in the properties of the resulting ion channels. Depending on the respective proportion of edited and unedited subunits assembled in the multi-protein complexes which form the l-glutamate receptors, the properties of these main ion channels of excitatory synaptic signal transmission in the brain vary considerably. Coupled with the parallel alternative intron splicing of the precursor transcripts the varying extent of RNA editing in individual finally translated mRNAs generates a range of combinatory possibilities resulting in synaptic ion channels with different properties.

As in the case of the apoB mRNA editing, A to I editing in the nucleus of mammalia thus creates different protein products from a single gene. Again, an additional, more subtle level of regulation may be an underlying function of the posttranscriptional change in the resulting protein sequence. It is possible that RNA editing in cells of the nervous system can be influenced by developmental stimuli to yield a modulation of the ion channeling patterns during formation and maturation of the neuronal capacity of the brain.

4 RNA editing in kinetoplasts of trypanosomes

When Rob Benne and his coworkers in Amsterdam claimed in 1986 that the mRNA for subunit 2 of the cytochrome c oxidase (cox2) in mitochondria of trypanosomes contains four U nucleotides which are not encoded in the mitochondrial genome, the scientific community was very sceptical [1]. They called the unknown process RNA editing, by analogy to the proofreading performed by the editors of manuscripts. The extent of editing in a mitochondrial transcript in Trypanosoma brucei, the causal agent of sleeping sickness, can be dramatic where more than half of the mature mRNA is posttranscriptionally inserted. This is for example observed in the transcript for subunit 3 of the cytochrome c oxidase (cox3), where the genomic coding region is 463 nucleotides long, and to which 547 nucleotides (all Us) are added and from which 41 of the genome encoded Us are deleted. The resulting edited mRNA is 969 nucleotides long and is translated into a conserved COX3 polypeptide.

In the flagellated protozoa of the order Kinetoplastida, this type of RNA editing which inserts and deletes uridines is prevalent, and at the same time restricted to this class of organisms [57, 2227]. These protozoa are characterized by a single mitochondrion in the cell, the kinetoplast, which is located at the base of the flagella and thus presumably in an ideal position to provide the chemical energy for kinetic propulsion. In the kinetoplastid mitochondrial genome a classic set of genes encodes some of the hydrophobic subunits of the respiratory electron transport chain (Fig. 3). These genes are sometimes hardly identifiable from their primary sequence conservation in the genome, particularly in those instances where extensive editing is required for the conserved open reading frame. In the nad9 transcript, encoding subunit 9 of the respiratory chain NADH dehydrogenase in T. brucei, 345 uridine nucleotides are added to create the 649 nucleotides long mRNA. Other primary transcripts such as nad4 and nad5 are not edited at all, the first identified RNA editing in cox2 (encoding subunit 2 of the cytochrome c oxidase) involves editing at three sites, where altogether four uridines are inserted. The final length of the mRNA encoding ATP6 (subunit 6 of the F0-ATPase) of 811 nucleotides contains more uridines added by RNA editing (448) than the 392 nucleotides encoded by the mitochondrial genome. Trypanosomes have evolved a complex system to ensure the fidelity of the number of nucleotides which have to be inserted as well as the precise location of the editing events in the transcript.

Figure 3

RNA editing in trypanosome mitochondria involves the insertion and deletion of uridines into and from the primary transcript to create the mature, translated mRNA. A: Genes are encoded in the maxicircle molecules and require varying numbers of changes by RNA editing. The sites of U insertion and their number are specified by the guide RNAs (gRNA) encoded in the minicircle DNAs and transcribed from their own promoters located within 18-bp repeats. B: The guide RNAs base-pair with the transcript before editing downstream of a site of U insertions and anchor the guide to the transcript. The antisense sequence of the guide determines where and how many uridines are to be inserted or deleted. The mature edited mRNA can perfectly base-pair with the guide RNA. Reprinted from [55] with kind permission.

In trypanosomatid RNA editing the location of the correct insertion site and the precise number of inserted Us are specified by the so-called ‘guide RNAs’. These are small RNA molecules, which contain a short antisense sequence which can base-pair with the unedited primary transcript downstream of a site where uridines are to be inserted (Fig. 3B). As a result of such base-pairing the guide RNA is anchored to the transcript, the anchor forming a perfect match of 10–15 nucleotides between guide RNA and transcript RNA. This double stranded region and the specific structure of the guide RNA are enveloped in a large multi-protein complex, the ‘editosome’, which performs the actual editing. This editosome contains several enzymatic activities, which begin to open the transcript RNA chain at the first mismatched nucleotide and start inserting uridines. These will base-pair with the next nucleotides in the guide RNA and U insertion in the transcript will continue as long as A (or G) in the guide RNA is matched. Insertion of U into the transcript RNA molecule will stop when C or U is encountered in the guide, which can base-pair with the next original nucleotide in the transcript RNA. This mechanism requires certain properties of the editing process. Firstly, insertion or deletion of Us progresses from the 3′ to the 5′ end of the primary RNA molecule. Secondly, a single guide RNA will not cover all of the editing sites in an extensively altered primary transcript and several guide molecules will be required. Thirdly, in frequently edited transcripts an anchor site for a guide RNA in the 5′ region of the transcript will sometimes only be created by the downstream editing events, thus resulting in a chain of guiding events proceeding in the 3′ to 5′ direction along the primary transcript eventually generating the mature mRNA.

The guide RNA molecules are usually encoded in the so-called minicircle DNA (Fig. 3A), specific molecules of the mitochondrial genome, of which thousands are present in a kinetoplast. They are mostly interconnected from unresolved replication catenates and form a type of three-dimensional ‘chain-mail’ structure in the organelle. The guide RNA genes are characterized by a specific sequence, which is present as an 18 nucleotides long imperfect repeat upstream of each guide RNA gene. These sequences possibly act as transcription promoters and may also play a role in the guide RNA evolution by duplication. The 18 bp long imperfect repeats are potentially involved in signaling individual transcript initiation, since most of the guide RNA molecules contain the di- or triphosphates characteristic for primary 5′ termini. At their 3′ ends guide RNAs have a tract of about 15 uridines which are added posttranscriptionally by a terminal uridylyl transferase. It has been proposed that these U tails act as potential donors of the Us inserted into the RNA during editing, but experimental evidence suggests that UTP is utilized for the mRNA editing, possibly catalyzed by the same enzyme which adds the 3′ oligo-U tails to the guide RNAs.

The mechanism of U insertion and deletion involves an endonucleolytic cut at the mismatch between guide RNA and unedited transcript. The newly opened ends of the transcribed molecule are held in place within the editosome complex by several of the polypeptides contained therein. One of the enzymes in the editosome is a terminal U-transferase which adds Us from UTP in the 5′ to 3′ direction at the 3′ end of the upstream part of the mRNA. Another enzyme, an RNA ligase, joins the two parts of the mRNA molecule after the gRNA-specified number of Us has been inserted. Probably during this alignment step the superfluous added Us as well as excessive genomically encoded Us are excised from the 3′ end of the upstream part of the mRNA. This is where deletion of Us is templated by the overriding antisense sequence information of the guide RNA molecule.

RNA editing of at least some of the mitochondrial mRNAs in these sometimes parasitic protozoa is developmentally regulated. Transcripts for cytochrome b and for subunit 2 of the cytochrome c oxidase (cox2) of T. brucei and T. congolese exhibit little or no editing in the bloodstream forms in the infected human or animal, where the respiratory chain is not functional in the parasites. These transcripts are only fully edited in the intermediate developmental forms, which are initiated in the midgut of the insect vector such as the tsetse fly, where the parasites depend on the ATP generated by the respiratory chain. RNA editing may thus confer a regulatory advantage if a clear chain of cause and consequence can indeed be established in this case. Minimizing the size of the mitochondrial genome cannot be the raison d’etre of RNA editing in trypanosomes, since the numerous guide RNAs take up considerably more genomic sequence than a normal protein coding gene.

5 RNA editing in mitochondria of Acanthamoeba

RNA editing in the amoeboid protozoon Acanthamoeba castellanii only affects tRNAs, where in addition a plethora of posttranscriptional modifications change numerous nucleoside structure(s) (Figs. 1 and 4). Following the definition of RNA editing given above as nucleotide alterations which could just as well have been encoded in the genomic sequence, the sequence alterations in tRNAs of Acanthamoeba classify as bona fide editing [28,29]. During sequence analysis of the mitochondrial genome in this amoeba 13 of the 16 tRNAs encoded in the mitochondrial DNA were predicted to contain mismatches in the aminoacyl acceptor stem when folded according to the classical structure model. The mismatches occur exclusively in the first three base pairs, involving nucleotides 1–3 and the complementing nucleotides 70–72 respectively. These base-pairings are present in all tRNAs and are essential for the functional folding of the molecule in space and in some cases are crucial for aminoacylation. Direct sequence analysis of the mature tRNA molecules revealed that the nucleotides in positions 1–3 differed from those predicted by the genomic sequence and that all of the mismatches in base-pairing to nucleotides 70–72 had been resolved. These posttranscriptional changes are limited to the first three nucleotides in all 13 tRNAs. G-U pairings are corrected to A-U pairs at nucleotide positions 1, 2 or 3, while a U-G low affinity pair involving nucleotide 4 is not altered.

Figure 4

RNA editing in tRNAs. A: In Acanthamoeba nucleotides 1–3 (filled circles) often do not pair with the respective opposite nucleotides in the aminoacyl stem of many mitochondrial tRNAs. These nucleotides are removed and replaced with the correctly pairing nucleotides. B: In Physarum mitochondria nucleotides are inserted at various different positions in the tRNA precursor molecules by an as yet unknown specificity and mechanism. C: In metazoan animals missing or not base-pairing nucleotides at the 3′ termini of tRNAs are added by the general mitochondrial polyA polymerase and the tRNA is completed at this end by the CCA adding enzyme. In marsupials a C nucleoside in the anticodon of tRNA-Asp is deaminated to a U, changing its identity to a tRNA-Gly. D: RNA editing in land plants alters tRNA sequences by C to U deamination mostly in base-pairing regions and improves the secondary structure. The folding opposite strand may be involved in specifying these editing events. Reprinted from [55] with kind permission.

To change one or all of the first three nucleotides, replacement is the most parsimonious explanation, since a single biochemical modification reaction cannot result in the many different changes of nucleotide identity. One of the two enzymes required would be a tRNA-specific 5′ to 3′ exonuclease, the other a 3′ to 5′ nucleotidyltransferase, which uses the 3′-terminal nucleotides of the tRNA as template to add only complementing nucleotides. The enzymes most likely to be involved have, however, not yet been identified or purified.

The same pattern of tRNA editing as in Acanthamoeba mitochondria is also found in the mitochondria of some of the chytridiomycete fungi such as Spizellomyces punctatus[30]. These fungi are phylogenetically very distant and unrelated to Acanthamoeba, making a common evolutionary origin of this type of tRNA editing very unlikely. Thus apparently independent events have established the same patterns of RNA editing in amoebas and these fungi, suggesting that pre-existing enzymes could fairly easily be adapted and modified during evolution.

6 RNA editing in mitochondria of Physarum

In the myxomycete slime mold Physarum polycephalum RNA editing appears to be the most complex described to date and involves both types of RNA editing, the insertional and the substitutional types [8,31,32] (Fig. 5). Any nucleotide, A, C, G or U, can be inserted as a monomer or as a dinucleotide, and in addition, C to U exchanges are observed. Nucleotide insertions are found in almost all of the RNA transcripts analyzed, all in all about 1000 editing sites are expected in the tRNA, rRNA and protein coding transcripts of the about 60-kb large mitochondrial genome.

Figure 5

RNA editing in mitochondria of the slime mold Physarum. A continuous open reading frame, here for the cox1 subunit of the cytochrome c oxidase, is only created by the insertion of various nucleotides, mostly C, into the primary transcript. In addition several C to U deaminations are observed, which improve the conservation of the encoded protein with its homologues in other organisms. The depicted three reading frames show the resulting polypeptide sequence (underlined) to be derived as a continuous frame only in the edited mRNA. Reprinted from [55] with kind permission.

The insertion of nucleotides into transcripts in Physarum mitochondria is closely linked to the transcription processivity and is apparently very efficient, whereas substitutional editing occurs more slowly and by an unrelated mechanisms. Possibly the latter C to U transition is catalyzed by a deamination reaction analogous to the apoB RNA editing, since deletion/excision of a nucleotide has never been observed in this system.

Insertional maturation of the newly synthesized RNA molecule is already complete 14–20 nucleotides away from the site of chain elongation by the RNA polymerase. Insertion probably utilizes nucleotide triphosphates, since specific concentrations of CTP are required for the insertion of Cs at the proper sites. The hypothetical ‘editosome’ is most likely linked to the transcription complex and only acts on nascent RNAs. Partial in vitro editing systems cannot insert the missing nucleotides posttranscriptionally into an unedited RNA molecule. The close coupling to transcription results in a 5′ to 3′ processivity of RNA editing, and thus contrasts with the 3′ to 5′ progression of the U insertions in trypanosome mitochondria.

The close physical and temporal coupling of RNA polymerase activity and RNA editing in this system suggests an enzymatic involvement of the RNA polymerase itself in the editing process. Mitochondrial RNA polymerases are evolutionarily related to the prokaryotic phage T7 single subunit polymerase type, which has been found to add faithfully those nucleotides that are inserted during RNA editing in the Ebola virus transcript (see below). Thus some conserved sequence motif structure and additional specificity factors may cause the RNA polymerase to pause or even to ‘back up’ and by re-reading the last transcribed nucleotide to insert the correct additional nucleotide, the polymerase thus stuttering. In this model the RNA polymerase has to reattach to the DNA template to resume faithful transcription and to then copy correctly at least up to the next editing site.

The signal motif(s) for specificity in Physarum mitochondrial editing are as yet unclear, but clearly do not involve stuttering of the RNA polymerase. Firstly the site of nucleotide insertion has to be marked and secondly the type of nucleotide inserted must be identified. The physical connection of nascent transcript elongation and RNA editing suggests that in the elongating RNA the sequences 3′ to the adjacent 10 nucleotides cannot be involved in determining the editing site, since they are synthesized only after the relevant RNA editing event is complete. Sequences 5′ to an editing site could, on the other hand, interact by backfolding, although this may create steric problems as the RNA polymerase complex covers several nucleotides of the emerging DNA-RNA heteroduplex in the transcription bubble.

Certain nucleotide motifs could trigger the insertions, but have so far not been detected in the sequences analyzed to date. It is possible that trans-acting guide molecules confer specificity, but as yet such molecules have not been identified either in the mitochondrial genome or as imported RNAs encoded in the nucleus. Availability of the complete mitochondrial genome sequence of Physarum may help resolve this question.

7 RNA editing in mitochondria and plastids of plants

Mitochondria and plastids of almost all of the land plants examined contain examples of posttranscriptional conversion of C to U and U to C in the sequences of many transcripts [3345]. In plastids including chloroplasts only between 5 and 20 Cs are found to be deaminated to Us in total (Fig. 6). In mitochondria, by extrapolation from the RNAs analyzed to date, probably up to 500 in some plants up to 1000 C to U changes are expected. In the majority of plant species U to C conversions are very rare events, amounting to only about two or three in the total mitochondrial RNA of any one flowering plant. RNA editing sites are mostly found in coding regions of mRNAs and less frequently in introns and other non-translated regions. In some cases RNA editing in tRNA molecules restores essential base-pairings (Fig. 4). In these instances only complete RNA editing allows correct folding and further maturation and processing of the tRNA precursors. The ribosomal RNAs in plant organelles appear to undergo very little RNA editing.

Figure 6

RNA editing in mitochondria and plastids of plants alters C to U and less frequently U to C. The enzyme(s) responsible deaminate C to U at specific sites in mRNAs and tRNAs. It is as yet unclear how the editing sites are recognized and which enzymes are involved. For the reaction from U to C transaminases and respective cofactors may be involved. Reprinted from [55] with kind permission.

Since no nucleotide insertions or other changes of nucleotide identities have been observed in plants, the most parsimonious explanation of the mechanism of RNA editing would be to propose the existence, and to invoke the activity, of an RNA-specific C-deaminase analogous to the APOBEC-1 enzyme described in the apoB editing of mammalia for example. All the evidence gathered to date does indeed point to a deamination reaction being involved in the C to U conversions. Such deaminases are, however, usually from energetic considerations unlikely to catalyze the reverse reaction of adding an amino group to a U residue to generate a C.

How a given C in an RNA molecule is identified for conversion to a U is at present unclear. Given the sheer numbers of editing sites that need to be specified in mitochondrial transcripts, a sequence-specific recognition by proteins alone, as in the apoB editing, is extremely unlikely. Current thinking thus favors the coinvolvement of RNA sequences guiding the, as yet hypothetical, editosome to specific sites. Such a guide function may be provided by cis- or trans-acting RNA molecules, but cannot reside in the sequence vicinity of the editing sites alone. No common sequence motifs or related groups of such can be identified around the different C to U conversion sites beyond a low percentage of Gs in the nucleotide position preceding the edited Cs. In both mitochondria and plastids downstream nucleotides appear to have little to do with the editing site specifications, while the upstream sequence context is apparently critical. In both types of organelle the essential upstream region varies between different editing sites, sometimes only 25–30 nucleotides are sufficient while at other sites 200 nucleotides may not be enough to specify a C to be edited. In mitochondria, sequence duplications have been found, in which RNA editing is correctly maintained as long as sufficient upstream sequences are present. These differ between individual sites as found in experiments with transgenic plastids, in which various lengths of upstream and downstream sequence insertions were tested in vivo.

RNA editing in plant organelles is a posttranscriptional process as suggested by the identification of potential RNA editing intermediates. In these partially edited transcripts some Cs have been changed to Us, while at other potential editing sites the Cs encoded by the genome are present. In these partially edited RNA molecules there is no order of editing along the RNA molecule suggesting that the editosome complex does not scan the RNA molecule along its linear length, but rather randomly selects certain sites or regions and releases the transcript again after editing.

Partially edited mRNAs are found in polysomal fractions of plant mitochondria and are apparently translated into a family of variant proteins. However, only one type of protein sequence appears to be incorporated into the polypeptide complexes of the respiratory chain. This polypeptide sequence corresponds generally to the polypeptide sequence best conserved with the homologues in other organisms and may be selected by its physiological and biochemical functionality. Indeed it is unlikely that polypeptides synthesized from unedited RNAs would function properly and such proteins would thus impair the efficiency of mitochondrial respiration.

8 RNA editing in mitochondria of animals

The first descriptions of RNA editing in mammalian mitochondria were not recognized as such, since they were identified before the term RNA editing was coined and were more readily explained as being due to the well-established mechanism of polyA addition [46]. Many of the mRNAs in animal mitochondria derive their 3′ termini by processing at the 5′ ends of downstream tRNAs followed by limited polyadenylation of the mRNA molecules. Some of the nascent mRNA 3′ termini do not contain a translational stop codon, but end with for example a U after the last in-frame codon. This U is posttranscriptionally extended to generate a UAA stop codon by 3′ addition of a polyA tract. It is not clear whether translation by the mitochondrial ribosome could effectively produce a full-length polypeptide including the last complete codon prior to polyadenylation at the solitary U. Certainly translation would be effectively terminated if the transcript template is incomplete and the ribosome would dissociate. The non-templated completion of the transcript information by polyadenylation may be required for faithful protein biosynthesis or may alternatively be involved in RNA metabolism.

Addition of a polyA tract by polyA polymerase is also the underlying mechanism of RNA editing of tRNAs in mitochondria of many metazoan animals [29,47,48]. In the mitochondrial genomes of platypus, chicken and many snails, genes for tRNAs sometimes overlap by one or two nucleotides. The generally correct processing at the respective 5′ termini of the tRNAs by RNase P leaves the upstream tRNA lacking the 3′ terminal nucleotides. These are added as a short polyA sequence by a polyA polymerase and subsequently trimmed back to generate the mature 3′ terminus by 3′ processing catalyzed by RNase Z (Fig. 4).

RNA editing in animal mitochondria is thus template-independent and utilizes the non-discriminating polyA polymerase, which will accept 3′ ends of any RNA molecule as template, be it mRNA, tRNA or rRNA. A template dependence might however be involved in completing the 3′ terminus of the tRNA-Ser in the platypus, where the four nucleotides 5′-CCCA-3′ have to be added. In postulating that in this instance the polyA polymerase may also catalyze CTP addition to the truncated tRNA, there would be no need for yet another enzymatic activity. However, this simple speculation has yet to be confirmed.

A different mechanism of tRNA editing may be operative in marsupial tRNAs, where a C nucleoside in the anticodon of tRNA-Asp is deaminated to a U at least in the American opossum. This deamination is apparently part of the tRNA modification processes and occurs after 5′ and 3′ trimming of the initially synthesized tRNA precursor. The enzyme responsible may be a modified CTP-deaminase or an adapted modification enzyme the nature of which has not yet been clarified. The biological consequence of this editing is the creation of two distinct tRNA species from just one gene: the unedited tRNA-Asp (GCC) and the edited tRNA-Gly (GUC) are present in about equal proportions in the marsupial mitochondria. The sole difference between the two tRNAs is this C to U change in the anticodon, which specifies the difference in codon recognition as well as altering the specificity elements for cognate amino acylation. In this case RNA editing results in two gene products being synthesized from just one coding region and amplifies the biological diversity beyond the genomic capacity. This is consistent with the very compact animal mitochondrial genomes, the size of which appears to be constrained in evolution.

9 Cotranscriptional RNA editing of virus RNAs

During transcription of several RNA viruses into mRNA additional nucleotides are incorporated which are not specified by the viral genome [4951]. In the paramyxoviruses such as measles, Sendai, parainfluenza and mumps viruses one or two and even up to eight or 10 G residues are inserted at specific sites (Fig. 7). In the Ebola viruses additional As are incorporated during transcription, but not replication, as mentioned earlier.

Figure 7

Cotranscriptional RNA editing of RNA virus transcripts inserts non-templated nucleotides by stuttering of the RNA-dependent RNA polymerase. The frame shifts induced by the insertion of one or two nucleotides extend the open reading frames most often of the P gene and the edited mRNA now encodes a much larger protein, usually the V protein. Specific sequence motifs in the viral RNA induce the RNA polymerase complex to stutter and to repeat the last transcribed nucleotide of the template RNA before resuming faithful transcription. Reprinted from [55] with kind permission.

All of these RNA viruses, including the paramyxoviruses and the Ebola type, propagate via antisense RNA molecules, which are then transcribed by a virus-encoded RNA-dependent RNA polymerase. This polymerase is prone to pausing and stuttering at certain nucleotide combinations, mostly mono- or oligonucleotide tracts. At the nascent mRNA 3′ ends up to several hundred As are added by the same polymerase, although they are not templated. These additional As stabilize the mRNA in a fashion analogous to the polyA tails added by specific polyA polymerases in other systems. While the mRNA polymerase (complex) pauses at these positions, the RNA replicase (complex) faithfully synthesizes the replication intermediate RNA from the virion RNA. Most likely the same RNA polymerase is differentially influenced by additional cofactors with replication taking place sheltered in the virion, while transcription to mRNA usually occurs in the cytoplasm.

This virus-encoded RNA-dependent RNA polymerase also pauses and incorporates non-templated nucleotides by ‘stuttering’ near the genomic stop codon of the first open reading frame in the unedited mRNA. Incorporation of one or two Gs or As upstream of the translational stop codon shifts the reading frame, which upon translation generates different proteins with divergent carboxy-terminal sequences. In the mumps and some other viruses the additional amino acid sequence can more than double the size of the genomically predicted first open reading frame.

The pausing and subsequent ‘stuttering’ of the RNA polymerase is compared to the sequence-specified pausing and transcription termination of the E. coli RNA polymerase. Processively unstable transcription complexes are induced in a similar fashion by certain sequences at the template sites. Pyrimidine-rich sequences with long U stretches before several Cs cause the viral RNA polymerase to slow down, slip back one nucleotide on the template and incorporate another G nucleotide opposite the same C again. This stuttering progression alters a sequence of, for example, three Gs predicted by the genomic RNA to four Gs in the edited mRNA. In the RNA viruses this dissociation and realignment of the polymerase-product complex to the previously transcribed nucleotide on the RNA template is explained by the specific rate constants of dissociation and RNA-protein binding. These and the calculated free energy levels are very similar in the different alignments even when including G-C as well as G-U pairings.

10 RNA editing and its consequences and advantages

RNA editing is required for gene expression in many of the systems discussed above. Often the genomic information encoding an open reading frame or a tRNA is cryptic or incomplete and will not yield a functional product. Thus the genetic system involved is dependent on RNA editing for its biological optimization and eventual survival. In the mitochondrial and plastid systems this process is particularly important for the synthesis of functional proteins which after editing exhibit closer sequence conservation with their homologues in other systems. In addition in these organelles RNA editing ensures that tRNAs can fold correctly.

Nevertheless, from first impressions RNA editing is rather extravagant and costly, since without exception all of the RNA editing events described would be rendered unnecessary if the sequence of the mature mRNA was encoded in the genome, as it is in bacteria. We thus have to look for or produce a hypothesis to explain the biological significance of RNA editing. This could include regulatory control of gene expression at the posttranscriptional level, where the rapid introduction of an AUG translational start codon could rapidly select an RNA for translation without requiring the complete de novo synthesis of the transcript. This may be reflected for example in the observed developmental differences in RNA editing efficiency in trypanosomes.

In addition, another advantage of RNA editing could be the potential of synthesizing two or more distinct products from a single gene as exemplified by the viral editing systems or in the tissue-specific apoB editing, where two lipid carrier proteins with different properties are derived from one and the same genomic coding region. Editing in the GluR receptor channels results in increasing the family of variant proteins available for the combinatory assembly of the channels and allows fine-tuning of the chemo-electric potential transmission properties. Here the available variation of a gene family is increased without increasing the number of genes. This presumed advantage may only be illusory as additional polypeptides are required for the editing process. RNA editing thus appears to be rather a biologically selfish process, which has become established and perpetuated.

11 Origin and evolution of RNA editing

Considering the diverse types and characteristics of RNA editing (Table 1) and the multitude of different biochemical and enzymatic processes involved it is obvious that there are several independently established mechanisms which have been grouped under the terminological umbrella of RNA editing. In some types of editing nucleoside conversion is apparently the result of an evolutionary recruitment of modification enzymes to act on RNA, while in the case of the insertional editings other preexisting enzymes have become assembled into complexes with modified specificities [6,52].

The animal nuclear RNA editing systems have apparently evolved starting from mononucleotide deaminases, which gave rise to larger gene families including the apobec-1 and adar genes. These exhibit clear evolutionary relationships with the bacterial deaminases involved in nucleotide metabolism. However, the adenosine nucleotide deaminase of E. coli cannot deaminate a nucleoside in the RNA since the reaction pocket in the enzyme is too small to accommodate the RNA strand. This internal active site has been widened by amino acid changes in the APOBEC-1 and ADAR proteins as modelled in the active site access structures [19,53].

The insertional editing in trypanosome mitochondria has enzymatically nothing in common with the nucleoside conversion processes. The enzymes involved, an endonuclease, terminal U transferase and RNA ligase, have apparently been recruited and adapted from different sources and contexts totally unrelated to nucleoside modification. However, the specificity determinants of nucleotide insertion in trypanosome mitochondria, as mediated by the interaction of double stranded RNA regions formed of mRNA and guide RNA, are analogous to the tRNA editing processes in animal mitochondria as well as in Acanthamoeba mitochondria and the GluR editing in mammalia. Since all of these RNA editing processes have evolved independently, (partially) base-pairing double stranded RNAs in cis or trans can apparently readily be recruited as specificity determinants. Such antisense guide sequences may also be involved in the still rather mysterious types of RNA editing such as those in Physarum mitochondria and in plant organelles. Antisense guide RNA molecules also offer another connection between RNA editing and modification, since in eukaryotes ribose methylations of rRNAs are also specified in trans by guide RNA molecules, the small nucleolar RNAs [54].

The presence of RNA editing in its various guises and appearances suggests that most if not all of these processes have evolved in specific lineages of speciation. None of the diverse RNA editing mechanisms can be convincingly linked to any of the processes assumed to have existed in an ancient primary RNA world. RNA editing mechanisms appear to have evolved much later to compensate for gene sequences gone awry or to increase evolutionary variation.

12 Conclusion and recommendation

If you still want to know more about RNA editing, which we hope to have convinced you is a fascinating subject, we recommend a book recently published by the American Society for Microbiology. It is entitled ‘Modification and Editing of RNA’, edited by Henri Grosjean and Rob Benne [55]. This treatise provides in a collection of essays an up-to-date and comprehensive coverage of the different processes of editing and modification.

The remaining question as to whether RNA editing occurs in bacteria is up to the microbiologist to resolve. Clues that RNA editing may exist could include ‘sequence errors’, unusual amino acid or codon alignments and the like, anything that could be explained by a modification of a nucleoside or by the addition of a nucleotide in the RNA molecule.

One candidate case has been identified in the tRNA-Ser in E. coli, which exists in two forms containing either a C or a dihydrouridine at nucleotide position 20 [29,56]. Attempts to set up in vitro systems have so far met with mixed success, this event is closely coupled to modification and/or transcription and may or may not be a genuine case of RNA editing.


We gratefully thank C.J. Leaver, Oxford, for his kind and critical comments on the text and for having defused many awkward phrases. Work in the authors’ laboratory is supported by grants from the Deutsche Forschungsgemeinschaft, the Bundesministerium für Forschung and the Fonds der Chemischen Industrie.


  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
  22. [22].
  23. [23].
  24. [24].
  25. [25].
  26. [26].
  27. [27].
  28. [28].
  29. [29].
  30. [30].
  31. [31].
  32. [32].
  33. [33].
  34. [34].
  35. [35].
  36. [36].
  37. [37].
  38. [38].
  39. [39].
  40. [40].
  41. [41].
  42. [42].
  43. [43].
  44. [44].
  45. [45].
  46. [46].
  47. [47].
  48. [48].
  49. [49].
  50. [50].
  51. [51].
  52. [52].
  53. [53].
  54. [54].
  55. [55].
  56. [56].
View Abstract