The mechanism of U insertion/deletion RNA editing in kinetoplastid mitochondria
The mechanism of U insertion/deletion RNA editing in kinetoplastid mitochondriaJuan D. Alfonzo, Otavio Thiemann and Larry Simpson*
Howard Hughes Medical Institute and Department of Molecular, Cell and Developmental Biology, University of California, Los Angeles, CA 90095-1662, USA
Received June 18, 1997;Revised and Accepted July 29, 1997
ABSTRACT
Recent advances in in vitro systems and identification of putative enzymatic activities have led to the acceptance of a modified `enzyme cascade' model for U insertion/deletion RNA editing in kinetoplastid mitochondria. Models involving the transfer of uridines (Us) from the 3'-end of gRNA to the editing site appear to be untenable. Two types of in vitro systems have been reported: (i) a gRNA-independent U insertion activity that is dependent on the secondary structure of the mRNA; (ii) a gRNA-dependent U insertion activity that requires addition of a gRNA that can form an anchor duplex with the pre-edited mRNA and which contains guiding A and G nucleotides to base pair with the added Us. In the case of the gRNA-mediated reaction, the precise site of cleavage is at the end of the gRNA-mRNA anchor duplex, as predicted by the original model. The model has been modified to include the addition of multiple Us to the 3'-end of the 5'-cleavage fragment, followed by the formation of base pairs with the guiding nucleotides and trimming back of the single-stranded oligo(U) 3'-overhang. The two fragments, which are held together by the gRNA `splint', are then ligated. Circumstantial in vitro evidence for involvement of an RNA ligase and an endoribonuclease, which are components of a 20S complex, was obtained. Efforts are underway in several laboratories to isolate and characterize specific components of the editing machinery.
INTRODUCTION
Kinetoplastid protozoa have a single mitochondrion which contains the mitochondrial (kinetoplast) DNA in the form of a nucleoid body situated within the kinetoplast portion of the mitochondrion adjacent to the basal body of the flagellum (1 ,2 ). There are two major known taxonomic groups within the kinetoplastids: the trypanosomatids and the bodonids/cryptobiids (3 ). Kinetoplast DNA (kDNA) has been extensively studied from several trypanosomatid species (Trypanosoma brucei, Leishmaniatarentolae, Trypanosoma cruzi and Crithidia fasciculata) and from one cryptobiid species (Trypanoplasma borreli).
Uridine (U) insertion/deletion RNA editing was first described in the kinetoplast-mitochondrion of T.brucei (4 ). This unusual RNA modification process (5 ) involves the insertion and, to a lesser extent, the deletion of U residues from transcripts of maxicircle `cryptogenes' (6 -11 ). The extent of editing varies from a few Us at a few adjacent sites to hundreds of Us at hundreds of sites over the entire gene (`pan-editing') (12 ). Editing corrects frameshifts, creates translation initiation codons and, in the case of pan-edited genes, converts the transcripts of unrecognizable cryptogenes into translatable mRNAs (13 ). Editing has been shown to be developmentally regulated in T.brucei (9 ,14 ,15 ).
The complementary sequence information for the specific insertion and deletion of U residues resides in a novel class of short 3'-oligo(U) RNAs, which can form `anchor' duplexes with mRNA just downstream of specific editing blocks (16 ). These `guide RNAs' (gRNAs) appear to specify the insertion and deletion of U residues by base pairing. In addition, the observed overall 3' -> 5' polarity of editing site selection within an editing domain results from the mediation of multiple overlapping gRNAs, in which upstream anchor sequences are created by downstream editing. gRNAs are transcribed from both the maxicircle and the minicircle components of the kDNA in trypanosomatids (16 -20 ) and from the 180 kb circles in the kDNA of T.borreli (21 ).
The initial hypothesis for the mechanism of RNA editing was the `enzyme cascade' model (16 ), in which the gRNA was mostly a passive carrier of the editing information and the act of information transfer was relegated to protein catalyzed cleavage-ligation reactions. The source of U residues was either UTP or, in a variant of this model, the 3'-oligo(U) tail of the gRNA itself (22 ). The evidence for this model was initially based on the existence in mitochondrial lysates of several enzymatic activities, including a terminal uridylyl transferase activity, RNA ligase activity (23 ,24 ) and pre-edited region-specific endoribonuclease activity (25 ,26 ). A second model was based on an analogy to Group I and Group II RNA catalyzed splicing reactions. In this `double transesterification' model the source of U residues is again either UTP (27 ) or the 3'-oligo(U) tail of the gRNA (28 ). The evidence for the latter type of transesterification model was the in vivo existence of gRNA-mRNA chimeric molecules (28 ), which were the predicted intermediates of the first transesterification. Chimeric gRNA-mRNA molecules could also be generated in vitro by incubation of synthetic RNAs with mitochondrial extracts or with glycerol gradient fractionated extracts (29 -32 ).
. Summary of U insertion and U deletion editing events
Maxicircle gene
Species
L.tarentolae
T.brucei
C.fasciculata
T.borreli
T.cruzi
ND1
0a-0b
0-0
5'-0
ND4
0-0
0-0
ND3
176-5
210-13
ND5
0-0
0-0
ND7
24-0
553-89
27-0
ND8
218-40
259-46
nd
ND9
334-41
345-20
nd
RPS12
117-32
132-28
135-7
133-32
A6
106-5
447-28
68-3
446-55
Cytb
39-0
34-0
39-0
5': 47-4
3': 144-40
CO1
0-0
0-0
0-0
5': 72-11
3': 177-13
CO2
4-0
4-0
4-0
0-0
4-0
CO3
29-15
547-41
32-2
0-0
MURF1
0-0
0-0
MURF2
28-4
26-4
30-0
MURF5
nd
nd
nd
G3
35-14
148-13
G4
326-5
325-40
12S rDNA
0-0
0-0
0-0
0-0
9S rDNA
0-0
0-0
0-0
0-0
aUridine additions. bUridine deletions. nd, not analyzed experimentally.
ORGANIZATION OF THE MITOCHONDRIAL GENOME IN KINETOPLASTIDS
kDNA is a unique biological structure, both physically and genetically. Physically, the kDNA of trypanosomatids consists of 20-50 catenated homoplasmic maxicircle molecules, 20-40 kb in size in different species, and ~5000-12 000 catenated minicircles, 0.8-2.5 kb in size in different species (33 ). Several small basic proteins from C.fasciculata have been isolated which could be cross-linked to kDNA by treatment of cells with formaldehyde (34 ). These proteins have limited homology with histone H1 sequences and have been proposed to be involved with condensation of the kDNA network in situ into the highly structured nucleoid body (35 ). Evidence has also been presented for a physical rotation of the nucleoid body within the mitochondrion during S phase in Leishmania and Crithidia, but not in Trypanosoma (36 ,37 ). Covalently closed minicircles are randomly removed from catenation in the network by a Type II topoisomerase and replication occurs within two replisomes situated at either end of the nucleoid body (38 ). We have previously speculated that rotation of the nucleoid body aids redistribution within the network of catenated minicircles encoding different gRNAs, to avoid loss of specific minicircle sequence classes upon segregation of the nucleoid body in daughter cells (7 ).
The kDNA of the cryptobiid T.borreli consists of two classes of large circular molecules, Component II, varying from 40 to 80 kb in two strains, and Component I, 180-200 kb in size, which do not appear to be catenated into a network (39 ,40 ).
Genetically, the kDNA genomes from both trypanosomatids and cryptobiids are similar in that they are composed of two separate but interacting genomes, one of which encodes rRNAs, structural genes and cryptogenes and the other of which encodes the gRNAs. In the trypanosomatids the maxicircle DNA encodes the cryptogenes (and a few gRNAs) and the minicircle DNA encodes the majority of the gRNAs. In T.borreli the 40-80 kb circles represent the maxicircle homologs and the 180 kb circles the minicircle homologs (21 ). The question of the evolutionary origin of the minicircle awaits investigation of additional bodonid/cryptobiid species, but the simplest scenario is that an ancestral kinetoplastid contained gRNA genes segregated on large circles which were excised, linked to autonomous replication elements and circularized to yield minicircles.
The location and polarity of the rRNA and structural genes in the maxicircle genome are conserved in all trypanosomatid species analyzed. This is shown in Figure 1 for the sequenced regions of the L.tarentolae, T.brucei and C.fasciculata maxicircles. The 9S and 12S rRNAs are not edited, but the RNA transcripts possess 3'-oligo(U) tails (41 ); the 3'-tail of the 12S rRNA is heterogeneous in length but that of the 9S rRNA has 11 Us. These non-encoded 3'-oligo(U) sequences are probably added by the known mitochondrial terminal uridylyl transferase (TUTase) activity, but whether this represents an adventitious by-product of the editing machinery or part of the normal maturation process is not known. gRNAs also have non-encoded 3'-oligo(U) tails ranging from 5 to 30 Us (42 ). mRNA transcripts of both genes and cryptogenes have non-encoded 3'-oligo(AU) tails, presumably resulting from interaction of the 3' polyadenylation machinery with TUTase activity (43 ).
THE MECHANISM OF U INSERTION/DELETION RNA EDITING
Several variations of the two original models for RNA editing have been progressively made as new features of the RNA and protein components have accrued (59 -62 ). All editing models share a common need to accommodate gRNA and mRNA interactions as a prelude to information transfer. The models diverge in regard to the underlying chemistry of the process and the roles played by protein components in catalyzing information transfer.
In the double transesterification model either the 3'-end of the gRNA (28 ) or free UTP (27 ) acts as the nucleophile that attacks the phosphodiester bond between the last nucleotide of the gRNA-mRNA anchor and the nucleotide following the first editing site on the mRNA (Fig. 2 A). This nucleophilic attack leads to transesterification of the attacking group onto the mRNA and the production of a free mRNA 5'-fragment. The 3'-end of the newly produced 5'-fragment of the mRNA will subsequently act as the nucleophile in the second transesterification reaction, leading to regeneration of a partially edited mRNA (Fig. 2 A). The existence of the predicted gRNA-mRNA chimeric molecules and the similarity to Group I and Group II intron RNA catalyzed splicing reactions provided the appeal of this model.
THE ACTIVITIES
Evidence for a protein-mediated editing mechanism, albeit circumstantial, has also been provided by the detection of various enzymatic activities in mitochondrial extracts from trypanosomatids. Initially, a TUTase (23 ), an endoribonuclease (25 ,26 ) and an RNA ligase activity (23 ) were reported. All of these activities seem to possess some sort of specificity for pre-edited mRNA and/or gRNA and this specificity suggested some involvement of these activities in RNA editing. In fact, the presence of these activities provided the initial stimulus for the development of the enzyme cascade model of editing.
These activities have been fractionated to different extents. Initially, Pollard et al. (69 ) found that the three activities of T.brucei extracts as well as the mRNA and gRNA could be separated into two distinct peaks by glycerol gradient sedimentation. One peak sedimenting at 19S contained TUTase, RNA ligase and gRNAs, while mRNA and also gRNA were found in a second peak in the 35-40S region of the gradient. Since then, co-fractionating putative editing activities have been obtained in several laboratories, both with T.brucei and with L.tarentolae extracts (70 -74 ). Peris et al. (75 ) recently reported the separation of potential editing activities from both species into two peaks; a major peak sedimenting at 10S, which contains TUTase and gRNAs, and a minor peak sedimenting at 20S, which contains the RNA ligase and some TUTase. The TUTase-containing and ligase-containing complexes could be visualized on native gels and were termed the T-IV complex and the ligase complex. In addition, a heterodisperse series of gRNA- and mRNA-containing complexes were found throughout the gradient. The main difference between the T.brucei and L.tarentolae extracts is that the 20S peak of T.brucei contains as much TUTase activity as the 10S peak. Both the gRNA-independent U insertion activity in L.tarentolae (70 ) and the gRNA-dependent U deletion and U insertion activities from T.brucei (67 ,71 ) were found to sediment at ~20S, together with the ligase complex, suggesting that this complex may represent a minimal editing enzymatic unit (76 ).
Although the data from various laboratories seems to differ in the number and/or size of the putative editing complexes, this might simply reflect subtle differences due to variations in the fractionation procedures and/or the species. However, given the high hydrophobicity of many mitochondrial proteins, it is probable that the observed apparent sizes of the editing complexes may be influenced by interaction with proteins that bind editing complexes non-specifically. Thus the actual size of the editing complexes under non-denaturing conditions could easily be overestimated. However, from the glycerol gradient fractionation data it is clear that RNA editing does take place within the realm of a multiprotein complex. It is expected that for complex assembly not only catalytic components of editing will be required, but also a set of structural proteins to provide a scaffold within which the actual catalysis occurs. The ultimate answer to the question of the size and/or nature of the editing complex will be through the establishment of a highly sensitive and quantitative in vitro editing assay, purification of the individual components and reconstitution of the active editing complex from the individual parts.
Studies on the individual components of the editing complexes are beginning to accumulate. An earlier report of a putative editing endoribonuclease failed to show a gRNA dependence of cleavage. The first experimental evidence for gRNA-dependent cleavage came in the in vitro system described by Seiwert et al. (61 ), Cruz-Reyes et al. (67 ) and Piller et al. (77 ) for U deletion editing. Piller et al. (77 ) were able to resolve three endoribonuclease activities from the T.brucei mitochondrial extract, by sensitivity to DTT, cleavage specificity and a requirement for gRNA. One activity cleaves Cyb mRNA within the pre-edited region and the specificity is unaffected by addition of cognate gRNA. A second activity is single-strand specific and cleaves Cyb mRNA in the 3'-end of the editing domain. A third activity has all the features of the predicted gRNA-dependent editing nuclease. It cleaves immediately 5' of a duplex but is also specific for the gRNA-directed editing site; digestion yields a 5'-P and a 3'-OH, consistent with the cleavage fragments being substrates for RNA ligase. This activity also sediments at ~20S. Isolation and characterization of purified nucleases, however, is necessary to confirm these assignments of activities to different enzymes. Alfonzo and Simpson (unpublished results) have recently purified to homogeneity, cloned and expressed a 22 kDa nuclease from L.tarentolae extracts with some specificity towards pre-edited RNAs and which appears to be gRNA independent; this probably corresponds to the nuclease activity previously described by Simpson et al. (25 ).
Two proteins of 50 and 45 kDa (reported as 57 and 50 kDa from T.brucei) have been identified as putative RNA ligase intermediates, since they can be covalently charged with AMP (24 ,75 ). The involvement of a particular RNA ligase in editing will require further purification and characterization. So far, the only evidence that this RNA ligase might be involved in editing is the finding that it co-fractionates in glycerol gradients with other putative editing components.
If indeed editing happens within a ribonucleoprotein complex, the process will require structural components that will not be directly involved in catalysis. It is presumed these structural factors should have specificity and high affinity for either the mRNA and/or gRNA. It is appealing to think that some interactions between editing components will help in assembly of an active complex and also help stabilize the complex once formed. The best candidate so far for a structural editing factor is a gBP21 protein purified from T.brucei by Köller et al. (78 ). gBP21 was identified by UV crosslinking to gRNA. Further analysis showed that gBP21 could bind gRNAs strongly in vitro (Kd = 5-10 nM). Although no biological significance has been proven for the strong interaction between gRNA and gBP21, in view of the low dissociation constant it may play some role in RNA editing. However, the specific role of gBP21 in editing remains an open question.
Another activity which was predicted to explain opening of the duplex formed by the 5'-most gRNA in an editing domain and the fully edited mRNA, to allow interaction with ribosomes, is an RNA helicase. This activity could also be invoked, in addition to breathing of the less thermodynamically stable portion of the duplex (51 ), to explain the melting of downstream gRNA-edited mRNA duplexes to allow formation of the anchor duplex with the adjacent upstream gRNA. A putative RNA helicase of the DEAD box family was isolated from mitochondrial extracts of T.brucei (79 ,80 ). Again, the specific role if any of this helicase in RNA editing remains an open question.
CONCLUSIONS
The overall evidence strongly indicates that U insertion/deletion RNA editing in trypanosomatids involves RNA-protein complexes and involves a series of successive specific gRNA-directed cleavages, 3'-terminal U addition, 3'-exonuclease trimming and re-ligation, almost exactly as predicted by the original enzyme cascade model (16 ). The field has progressed to the stage of biochemical characterization of individual components and reconstitution of editing activities. Major advances have been made in obtaining in vitro systems in which a complete cycle of U deletion and U insertion editing occurs, but the systems are extremely inefficient and do not progress past a single cycle of editing. There is clearly a need to optimize and reconstitute the in vitro systems with recombinant proteins and obtain progressive editing in multiple sites. Antisera against purified putative editing components should be generated and tested for immunodepletion activity and immunoprecipitation of editing complexes. Finally, gene disruption data must be obtained to confirm a role of each component in the editing process.
7 Simpson,L. and Thiemann,O.H. (1995) Cell, 81, 837-840.MEDLINE Abstract
8 Simpson,L., Maslov,D.A. and Blum,B. (1993) In Benne,R. (ed.), RNA Editing-The Alteration of Protein Coding Sequences of RNA. Ellis Horwood, New York, NY, pp. 53-85.
9 Stuart,K. (1993) In Benne,R. (ed.), RNA Editing-The Alteration of Protein Coding Sequences of RNA. Ellis Horwood, New York, NY, Vol. 1, pp. 25-52.