A Cis-acting A-U Sequence Element Induces Kinetoplastid U-insertions*

A 34-nucleotide A-U sequence located immediately upstream of the editing sites of the Leishmania tarentolaecytochrome b mRNA induces a mitochondrial extract to insert U nucleotides independent of guide RNA. Insertions are localized to positions immediately 5′ and 3′ of the A-U sequence. When placed within an unedited mammalian transcript, the A-U sequence is sufficient to induce U-insertions. The sequence has a high degree of similarity with the templating nucleotides of a cytochrome b guide RNA and with a sequence adjacent to the editing sites in ND7 mRNA, the other characterized kinetoplastid mRNA supporting guide RNA-independent U-insertions. At least one protein specifically interacts with the A-U sequence. The reaction is consistent with a mechanism proposed for guide RNA-directed editing.

Several mitochondrial mRNAs of the kinetoplastid protozoa must be edited through the precise insertion and deletion of U nucleotides (reviewed by Refs. [1][2][3]. The editing has been proposed to proceed through a reaction that is initiated at an editing site by an endonucleolytic cleavage producing a 5Јfragment with a 3Ј-OH and a 3Ј-fragment with a 5Ј-monophosphate (4). U-addition by a terminal uridylyltransferase activity (TUTase) or U-deletion by an exonuclease activity at the 3Ј-OH followed by re-ligation of the 5Ј-and 3Ј-fragments would complete one editing cycle which could be reiterated at subsequent sites. Experimental evidence in support of the proposed mechanism has been obtained in vitro (5)(6)(7)(8).
Genetic information specifying the location of the U-insertions and deletions is carried on guide RNAs (gRNAs). 1 A specific gRNA is able to bind to its cognate mRNA immediately 3Ј of a block of editing sites and also contains a template complementary to a block of edited sequence (4). It has been demonstrated in vitro for certain Trypanosoma brucei and Leishmania tarentolae mRNAs that the appropriate gRNA directs the U-insertions and deletions at the editing sites; changing the templating sequence of the gRNA results in the corresponding change to the number of U-insertions and deletions (5)(6)(7)(8).
U-insertions within some L. tarentolae mRNAs have been reported to be catalyzed by a mitochondrial lysate independent of added gRNA. Initially it was suggested that gRNAs endog-enous to the mitochondrial lysate were directing the insertions (9). However, it was subsequently demonstrated that the Uinsertions were unaffected by mutations to the mRNA that should have inhibited interactions with endogenous gRNA (10). Characterization of the reaction was complicated by difficulty in solubilizing the majority of the U-insertion activity from the crude mitochondrial lysate, making subsequent enrichment inefficient and leaving the assays vulnerable to artifacts caused by contaminating activities. The low level of U-insertions also hindered any significant biochemical characterization and prevented the localization of the U-insertions within individual RNAs. As a result, the biological significance of the gRNAindependent reaction observed in vitro was unclear.
We describe here a novel assay that has facilitated the enrichment and characterization of the gRNA-independent Uinsertion activity. The majority of the U-insertions occurring independent of gRNA are adjacent to a 34-nucleotide A-U sequence element that is immediately 5Ј of the editing sites on the cytochrome b transcript. This element is both necessary for the cytochrome b gRNA-independent editing and also sufficient to induce gRNA-independent insertions within a mammalian transcript. The sequence of the A-U element is highly similar to the templating sequence of cytochrome b gRNA I and a sequence immediately upstream of editing sites of the ND7 mRNA, which is the other characterized mRNA that supports gRNA-independent U-insertions (8). The intermediates produced during the reaction are consistent with a mechanism proposed for gRNA-directed editing (4). Our results suggest that gRNA-independent editing may result from association of the editing components with the 34-nucleotide A-U element present on the cytochrome b mRNA.

EXPERIMENTAL PROCEDURES
Mitochondrial Extract Preparation-An old laboratory strain (UC strain) of L. tarentolae was grown to a density of 1 to 2 ϫ 10 8 cells/ml in 2-liter flasks containing 500 ml of BHI media (DIFCO) supplemented with 10 g/ml hemin. The flasks were loosely covered with foil and shaken at 100 rpm at 27°C in an Innova 4230 incubator (New Brunswick Scientific). After washing three times with SHE (2 mM EDTA, 20 mM Hepes 7.5 and 0.25 M sorbitol), cells were lysed in a Stansted Disrupter (Energy Service Co.) at 1200 p.s.i. as described previously (10). Mitochondria were purified by flotation in a Renograffin density gradient (11), and after washing three times with SHE, resuspended in 950 l of solubilization buffer (25 mM Hepes 7.5, 10 mM MgCl 2 , 1 mM EDTA, 20 mM KCl, 0.1 mM ATP, 1 mg/ml Pefabloc, and 10 g/ml leupeptin) for every 500 ml of starting culture. Triton X-100 was added to a final concentration of 0.5%, and the solution was gently mixed by inversion and left on ice for 5 min. The mitochondrial lysate was centrifuged at 11,000 ϫ g at 4°C for 5 min and the supernatant frozen on dry ice for storage at Ϫ80°C. The protein concentration of the unfractionated extract was 2.3 Ϯ 0.1 g/l. A 250-l aliquot of the extract was thawed on ice and loaded onto a linear 5-ml 10 -40% sucrose gradient and centrifuged in a Ti55 rotor at 100,000 ϫ g for 12 h at 4°C. Fractions (450 l) were collected from the top of the gradient.
RNA Transcripts-RNAs were transcribed either from PCR products or directly from synthetic oligodeoxynucleotides using T7 RNA polymerase (12). The template for the parental transcript (wild type) contain-ing the 5Ј 178 nucleotides of pre-edited cytochrome b and a mutated gRNA-binding site was amplified from plasmid pNB2-S1 (9) using oligodeoxynucleotide primers 5692 and 10211. The same plasmid was used both for the amplification of the ⌬(3-10) template with primers 10513 and 10211, and the S(8A to 8C) template with 10703 and 10211. The ⌬(35-54) template was synthesized through reverse transcription of circular wild type cytochrome b RNA with primer 10076 followed by PCR amplification of the extension products using 10076 and 10132. The template for ⌬ (31)(32)(33)(34)(35)(36)(37)(38)(39) was similarly synthesized by RT-PCR using 10795 and 10796. The templates for the 5Ј-extended wild type and S(8A to 8C) cytochrome b RNAs were synthesized by RT-PCR of circular wild type cytochrome b and circular S(8A to 8C) RNAs, respectively, using primers 7845 and 10797. The template for the RNA transcript used to demonstrate the importance of a 5Ј-phosphate and 3Ј-OH to the ligation reaction was synthesized by RT-PCR of circular wild type cytochrome b using primers 9962 and 9964. This RNA has 5Ј and 3Ј termini that would be produced by endonucleolytic cleavage 3Ј of base 44, a location where several U insertions were detected (Fig. 2C). The template for the wild type 192-nucleotide human light chain ferritin transcript was amplified from ATCC clone 124280 using primers 10704 and 10705. The template for the ferritin transcript in which 34 internal nucleotides were substituted with the A-U element, was constructed in 3 steps. First, the 5Ј-fragment of the template was amplified with the forward PCR primer 10704 and reverse primer 12122 containing a 3Ј-extension complementary to the AU element. Second, the 3Ј-fragment was amplified with forward primer 12121, containing a 5Ј-extension with the sequence of the A-U element, and reverse primer 10705. Third, the products from the first two PCR amplifications were gel purified, mixed together, and PCR amplified using primers 10704 and 10705. The product was cloned into pBluescript II KS(ϩ) (Stratagene) and verified by sequencing. The template containing the 8A to 8C mutated A-U element was synthesized similarly except that primers 12333 and 12334 were used in place of 12121 and 12122. The RNAs for the gel shift assay were transcribed directly from synthetic oligodeoxynucleotides: 10899 for the wild type RNA, 12309 for the ferritin competitor transcript, and 12525 for the 8A to 8C mutant. The promoter portion of the oligonucleotide templates was annealed to oligonucleotide 5690 prior to initiating the reactions.
Transcription reactions (50 to 100 l) were carried out at 37°C with 1 mM each NTP, 0.5 mM dithiothreitol, 100 mM Tris, pH 7.7, 100 mM spermidine (basic), 1% Triton X-100, 13 mM MgCl 2 , and 50 units of T7 RNA polymerase (Life Technologies, Inc.). A 5-fold molar excess of GMP over GTP was included in the reactions for RNAs used in subsequent ligation reactions. RNAs for the gel-shift assay were synthesized to a specific activity of 2.4 ϫ 10 6 cpm/g by the addition of [␣-32 P]UTP under the conditions described above with the exception that the total MgCl 2 concentration was increased to 22 mM and the GTP, ATP, and CTP concentrations to 4 mM each. The transcript used to demonstrate the importance of 5Ј-phosphate and 3Ј-OH groups to the ligation reaction was primed with 1 mM ApG. ITP was substituted for GTP during this transcription to ensure that all transcripts were primed with a 5Ј-OH (12). The RNA was then either radiolabeled through kination of the 5Ј-OH with [␥-32 P]ATP or through ligation of [5Ј-32 P]Cp to the 3Ј-OH. All cytochrome b transcripts were purified on 6% polyacrylamide (acrylamide:bis-acrylamide weight ratio of 19:1), 8 M urea; the shorter RNAs used in the gel-shift assays were purified on 8% polyacrylamide, 8 M urea gels. RNAs were eluted and precipitated with ethanol.
Intramolecular circular RNAs were formed by incubating 2 to 5 M linear RNAs for 2 h at 37°C in 50 mM Tris, pH 7.8, 10 mM MgCl 2 , 10 mM dithiothreitol, 0.3 mg/ml bovine serum albumin, 1 mM ATP, 15% dimethyl sulfoxide, and 10 units/l T4 RNA ligase (New England Biolabs). The intramolecular circular ligation products were purified on 6% denaturing polyacrylamide gels, eluted, and precipitated with ethanol. Several criteria were used to identify and purify the circular RNAs from intermolecular ligation products potentially complicating the U-insertion assay. First, primer extension analysis was used to extend through the ligation junction. Extension of a circular RNA yields a cDNA product that is close to full-length whereas extension of an intermolecular dimer yields both this cDNA and, providing the RT primer-binding site is not at the extreme ends of the RNA, a second shorter product. Second, when radiolabeled GTP primed and unlabeled GMP primed RNAs are incubated together in a ligation reaction, the linear ligation products, identified by the primer extension, are the only predominant radiolabeled bands. Third, the electrophoretic mobility of circular RNAs relative to linear RNAs varies significantly with the percentage of acrylamide. Small quantities of linear contaminants co-migrating with circular RNAs on the 6% gel are readily detected when re-electrophoresed on higher percentage gels.
Direct Assay for U-insertion-The circular cytochrome b transcripts were heated at 65°C for 3 min in a 10-l volume containing 5 mM Tris, pH 8.3 (27°C), and 0.2 mM EDTA, pH 8.0. After the addition of 25 l of reaction mixture (40 mM KCl, 20 mM MgCl 2 , 6 mM potassium phosphate, pH 7.6, 40 mM dithiothreitol, 2 mM ATP, 2 mM GTP, 0.1 mM UTP, 5 Ci [␣-32 P]UTP, 1 mg/ml Pefabloc SC, and 10 g/ml leupeptin), the RNAs were incubated at 27°C for 10 min. Either 15 l of the fractionated or non-fractionated mitochondrial extracts were added to the RNAs and incubated at 27°C for an additional 50 min. Reactions with non-fractionated extract contained 1 pmol of substrate RNA, whereas those with the fractionated extract contained 0.5 pmol. Reactions were terminated by phenol/chloroform extraction followed by ethanol precipitation. Products were resolved on 9% denaturing polyacrylamide gels and quantified after PhosphorImager scanning (Molecular Dynamics).
Enrichment of RNAs with gRNA-independent U-insertions-Circular cytochrome b transcripts were extract treated under the direct assay conditions with 1 mM 4-thio-UTP (Amersham) substituted for both labeled and unlabeled UTP. Circular RNAs were purified on 6% denaturing polyacrylamide gels, eluted, and ethanol precipitated. RNAs were resuspended in a total volume of 50 l containing 0.5% SDS and 6 mM EDTA. After heating at 65°C for 3 min, the RNA solution was added to a 1.5-ml plastic tube containing 10 l of organomercuryderivatized agarose (Bio-Rad) that had been washed three times in 1 ϫ buffer (400 mM NaCl, 40 mM Hepes pH 7.5, 3 mM EDTA, and 0.5% SDS) and resuspended in 50 l of 2 ϫ buffer (800 mM NaCl and 80 mM Hepes pH 7.5). The solution was incubated at room temperature for 30 min on a rocker. The beads were pelleted for 5 s in a microcentrifuge and after removing the supernatant, resuspended in 600 l of 1 ϫ buffer and heated at 65°C for 5 min. The beads were re-pelleted, the supernatant discarded and the washing procedure repeated an additional 5 times. After the last wash, RNA was eluted from the beads by heating at 65°C in 1 ϫ buffer containing 100 mM dithiothreitol and 30 pmol of the DNA primer (7534 or 10040) used for subsequent cDNA synthesis. The RNA was ethanol precipitated using 5 g of glycogen (Boehringer-Mannheim) as a carrier, and the organomercury column enrichment procedure repeated an additional two times. The RNA from the 3rd cycle of enrichment was reverse transcribed into cDNA and PCR amplified for cloning using primers 7534 and 8494 or 10040 and 10041. Bio-X-ACT (Intermountain Scientific) was used to increase the fidelity during the amplification.
Other Assays-U-insertions within extract-treated linear RNAs were detected using an RNase H reaction (9). Linear RNAs were extracttreated as described for the circular substrates with the exception that the UTP concentration was at 0.5 M and the specific activity increased to 2.5 ϫ 10 5 Ci/mol. Heparin (6 g/ml) was included within the reactions to reduce the amount of 3Ј-end labeling relative to the internal Uadditions. The extract reactions were terminated by the addition of EDTA to a final concentration of 20 mM and incubation at 27°C for 3 min followed by the addition of SDS to 0.5%, proteinase K (50 g/ml), and a further 30-min incubation. The RNA was then phenol/chloroform extracted, gel purified, eluted, and ethanol precipitated. Oligodeoxynucleotide 10951 was annealed to the ferritin transcripts, 7622 to the 5Ј-extended cytochrome b RNAs, and 7534 to the wt cytochrome b RNAs for subsequent RNase H digestion.
The gel-shift assays were performed similar to a previously described procedure (13). After denaturation, 8 pmol of the RNAs were incubated for 50 min with extract under U-insertion conditions but in the absence of UTP (so that complications from TUTase activity would be avoided) and GTP. The concentration of dithiothreitol was reduced to 3 mM. Where indicated, the competitor ferritin transcript was added to the binding reactions at the same time as the A-U containing RNA. The binding reactions were resolved on a 6% nondenaturing polyacrylamide gel (acrylamide:bis-acrylamide weight ratio of 80:1). The assay for ligase activity through the gradient was performed as for U-insertion activity using 0.5 pmol of radiolabeled linear wild type cytochrome b transcript that had been GMP primed.
Analysis of Cytochrome b Cleavage Products-Circular cytochrome b RNA was reacted with the fractionated extract under U-insertion conditions either in the absence of NTP or presence of 1 mM each of 4-thio-UTP, GTP, and ATP. Full-length linear RNAs formed during the reactions were gel purified, eluted, and precipitated in ethanol. Linear RNAs isolated from the reaction containing 4-thio-UTP were enriched for those with insertions by one passage through the organomercury column matrix. RNAs were incubated with 4 units of poly(A) polymerase (Life Technologies, Inc.) at 37°C for 30 min in a reaction containing 50 mM Hepes pH 8.0, 250 mM NaCl, 10 mM MgCl 2 , 2.5 mM MnCl 2 , 1 mM CTP, and 40 units of RNasin (Promega). The poly(C) tailed RNAs were amplified for cloning and sequencing with primers 9209 and 9295.
Clones with the same mapped cleavage site were included only if the length of each poly(C) extension was unique indicating that they had originated from different RNAs. There is a 1 nucleotide uncertainty in mapping the cleavage sites adjacent to genomically encoded C nucleotides, and when UTP is included within the editing reaction, the mapping is also complicated by genomically encoded U nucleotides.
Northern Blot Analysis-RNA isolated from the sucrose fractions was electrophoresed in a 1.2% agarose/formaldehyde gel and transferred to Hybond N ϩ membrane (Amersham). Hybridization in Church-Gilbert buffer (0.5 M sodium phosphate, pH 7.2, 1 mM EDTA, 7% SDS, 1% bovine serum albumin) was performed at 65°C for the 12 S probe (oligodeoxynucleotide 10855) and at 55°C for the gRNA probe (oligodeoxynucleotide 10854). Oligodeoxynucleotide probes were 5Ј-end labeled and were added to the hybridizations at a final concentration of 1 pmol/ml and a specific activity of 1 ϫ 10 6 cpm/pmol.
Oligodeoxynucleotides-Oligodeoxynucleotides shown below were used in this study.

RESULTS
An Assay for gRNA-independent U-insertions-Cytochrome b mRNA contains 15 sites near its 5Ј-end that are edited in vivo through the insertion of a total of 39 U nucleotides. In vivo editing of cytochrome b mRNA has been proposed to be directed through the sequential action of 2 different gRNAs (4,10,14). Cytochrome b gRNA I contains the templating sequence to potentially guide the appropriate number of U-insertions into the first (3Ј-most) seven editing sites. Editing creates a gRNAbinding site for cytochrome b gRNA II which contains the template to complete editing at the remaining sites.
An RNA transcript containing the first 178 nucleotides from the 5Ј-end of the pre-edited cytochrome b mRNA of L. tarentolae was used as a substrate for an in vitro U-insertion assay.
The 178-nucleotide cytochrome b transcript was circularized with T4 RNA ligase, and then incubated with a solubilized L. tarentolae mitochondrial extract in the presence of [␣-32 P]UTP, salts, and the appropriate cofactors (Fig. 1A). Approximately 1% of the input circular cytochrome b transcripts are modified with U-insertions (assuming an average of 2 insertions per molecule). A major advantage of a circular substrate over a linear one is that the circularization prevents a TUTase activity from catalyzing U-addition to the 3Ј-end. Thus, all the U-additions to a circle, by definition, have to be internal. Circular cytochrome b RNA has a significantly different mobility on a denaturing gel than the corresponding linear molecules, and as a result the circular RNA is resolved from both the endogenous molecules that become intensely labeled during the reaction (Fig. 1A, ϪRNA) and from cytochrome b RNA degradation products that are also substrates for the TUTase activity (Fig. 1A, marker). In addition, circular RNAs have been demonstrated to be more stable to nucleolytic degradation than their linear counterparts, and the functional properties of many RNAs are not significantly altered as a result of the ligation (15)(16)(17).
The gRNA binding sequence (anchor) on the cytochrome b transcript was mutagenized (Fig. 1B) so as to inhibit interaction with the small amount of endogenous cytochrome b gRNA I. The mutagenized RNA is as good a substrate for U-insertions indicating that the reaction does not require endogenous gRNA to interact with the gRNA-binding site (Fig. 1A, Ϫanchor). This transcript is used in all subsequent assays and will be referred to as "wild type." The solubilized extract supporting the gRNA independent U-insertion activity can be fractionated over a linear 10 -40% sucrose gradient (Fig. 1A). Fractions were collected from the top of the gradient and assayed in the absence of added cognate-gRNA. The U-insertion activity is predominantly within fractions 6 and 7 and is resolved from the endogenous ribosomal RNA (fractions 10 and 11), identified by Northern analysis (Fig. 1A) and which also appears to be labeled by the TUTase activity (Fig. 1A, lanes 10 and 11, intensely labeled bands). TUTase activity appears to be present through most of the gradient since the endogenous RNA is labeled from fraction 3-11. The specific activity of the gRNA-independent U-insertion activity in fractions 6 and 7 is increased 50-fold over the unfractionated extract. Fractions 6 and 7 each contain approximately 10% of the total protein loaded onto the gradient. The remaining increase in specific activity is probably a result of fractionating away RNAs or proteins that are inhibitory to the reaction.
The endogenous cytochrome b gRNA I is spread throughout the gradient (Fig. 1A, Northern analysis). The wide distribution is suggestive that the gRNA is in heterogeneously sized complexes with other factors, in agreement with previous studies (18 -21). The fractions containing the peak U-insertion activity (6 and 7) each contain approximately 10% of the endogenous gRNA.
The U-Insertions Are Not Random-The low efficiency of the gRNA-independent reaction had previously prevented the cloning and sequencing of individual RNAs with insertions (9). We enriched for RNAs containing inserted U nucleotides by exploiting the observation that 4-thiouridine could be incorporated by the mitochondrial extract into the cytochrome b transcript with 85% of the efficiency of uridine (data not shown). RNAs containing 4-thiouridine insertions form mercaptides with an organomercurial column matrix which permit their partitioning from the bulk non-modified RNA ( Fig. 2A). Those RNAs containing thiouridine insertions were eluted from the organomercurial matrix with dithiothreitol for RT-PCR, cloning, and sequencing.
After organomercury column enrichment, 28% of the sequenced clones contained inserted U nucleotides (Fig. 2, B and  C). As a result of the assay design, only 108 nucleotides of the substrate RNA could initially be screened for insertions (arrows in Fig. 2C); the remaining 70 nucleotides were used as primer-binding sites during the RT-PCR amplification and any U-insertions within this region would not have been detected. In a parallel experiment, however, primers binding to different regions of the circular cytochrome b RNA were used for the RT-PCR. This resulted in the generation of molecules with circularly permuted ends and permitted U-insertions to also be detected within the original primer-binding sites (circles in Fig.  2C). The U-insertions are predominantly localized either within an 8-nucleotide sequence corresponding to the 5Ј-end of the non-permuted cytochrome b transcript or within an 18nucleotide region that is normally edited in vivo (Fig. 2C, boxed  sequence). It should be emphasized that there are no insertions between nucleotides 9 and 34 and only rare insertions within the region 3Ј of the first editing site (nucleotides 58 -178). Thus, mutation of the gRNA-binding site still results in U-insertions within editing sites, although with less accuracy than occurs in vivo (Fig. 2C, bottom panel). All clones contained the mutagenized gRNA-binding site eliminating any possibility that they had been derived from endogenous mRNA. Insertions that occurred adjacent to genomically encoded U nucleotides could not be precisely localized (Fig. 2C, positions 8, 108, and 128).
The observed U-insertions are not the result of either a PCR or cloning artifact based on the following observations. First, clones containing inserted U nucleotides were only detected if the RNAs were enriched by passage through the organomercury column prior to RT-PCR and cloning. When the extracttreated RNAs were amplified and cloned without sufficient enrichment, none of 60 sequenced clones had insertions. Second, the insertions within all of the sequenced clones were restricted to U nucleotides. If the U-insertions had resulted from errors during the PCR amplification or cloning, the insertion of other nucleotides would have been expected. Nine of the cloned RNAs that were enriched for U-insertions also contain deletions. These deletions are preferentially located either at U-nucleotides or adjacent to inserted U-nucleotides (data not shown).
An A-U Sequence Element Immediately Adjacent to the 5Ј-Most Editing Site Is Necessary for the gRNA-independent Uinsertions-Since the editing sites of most pre-edited mRNAs are rich in purines, it was initially assumed that the purines may serve as a recognition site for the U-insertion machinery, accounting for the predominance of gRNA-independent insertions within this region (Fig. 2C, boxed sequence). However, deletion of the purine-rich pre-edited region (Fig. 3, ⌬(35-54)), rather than inhibiting U-insertions, stimulated the reaction by a factor of 1.7 Ϯ 0.3 (mean from six reactions). As already indicated (Fig. 1), modification of the gRNA-binding site has no significant effect on the reaction efficiency. Deletions made 3Ј of the gRNA-binding site also have no effect (data not shown).
The U-insertions, instead, are dependent on a 34-nucleotide A-U sequence that is within the untranslated region immediately 5Ј of the in vivo editing sites (Fig. 3A). Deletion of 7 nucleotides from the 5Ј-end of this sequence resulted in the loss of 82 Ϯ 8% (mean from six reactions) of the gRNA-independent insertions (Fig. 3B, lane 2), and deletion of 7 nucleotides from the 3Ј end of the A-U sequence resulted in an 84 Ϯ 4% (mean from six reactions) reduction (Fig. 3B, lane 3). Substitution of eight of the internal A nucleotides with C nucleotides likewise abolished 86 Ϯ 2% (mean from six reactions) of the insertions (Fig. 3B, lane 1). The decreased number of U-insertions within the mutated RNAs is not a result of a decreased stability within the mitochondrial extract; incubation of radiolabeled RNA with the mitochondrial extract in the absence of [␣-32 P]UTP indicated that the S(8A to 8C) mutated RNA is as stable as the wild type (Fig. 3B, lanes 7 and 8).
The A-U sequence is sufficient to induce U-insertions when placed within an unedited mammalian transcript. A 192-nucleotide linear transcript containing part of the coding sequence of mammalian ferritin was treated with the mitochondrial extract. After gel-purification, the extract-treated transcript was annealed with an oligodeoxynucleotide complementary to a sequence at the 3Ј-end. Digestion with RNase H followed by electrophoresis on a denaturing gel resolved the 5Ј-fragment containing internal U-insertions from the U nucleotides added at the 3Ј-OH by the TUTase (Fig. 4); the 3Ј-labeled fragments are too small to be visible in these reactions. The ferritin transcript has a low level of internal U-insertions as indicated by the labeling of the 5Ј-fragment (Fig. 4A, lanes 3  and 4). However, when the 34-nucleotide A-U element was substituted in place of nucleotides 51 to 85 of the ferritin transcript (Fig. 4A, lanes 5 and 6), the level of internal Uinsertions increased 16 Ϯ 1-fold (mean from six reactions). The majority of the U-additions are internal since the labeling of the 5Ј-RNase H fragment (Fig. 4A, lane 6) is 70 Ϯ 10% that of the non-digested transcript (Fig. 4A, lane 5). The difference (Ϸ30%) represents labeling at the 3Ј-end which also appears to be stimulated approximately 2-fold relative to the ferritin transcript. Substitution of 8A with 8C nucleotides within the 34 nucleotide element (S(8A to 8C), see also Fig. 3A) of the chimeric transcript inhibited the induction of internal U-insertions (Fig. 4A, lane 8) but did not inhibit the additions to the 3Ј-end (Fig. 4A, lane 7). This is consistent with the cytochrome b RNA mutagenesis (Fig. 3B, lane 1) and suggests that the U-insertions are induced directly by the A-U element rather than being an indirect consequence of the 34-nucleotide substitution. It is possible, however, that the stimulation of 3Ј-end labeling observed with both the wild type A-U and the 8A to 8C substitutions are the result of such an indirect effect.
A sequence extension on the 5Ј-end of the A-U element sig-nificantly enhances U-insertions into the linear cytochrome b transcript but is not absolutely required. The efficiency of Uinsertions within the linear 178-nucleotide cytochrome b transcript (Fig. 4B, lanes 3 and 4) is decreased approximately 15-fold relative to the corresponding circular RNA (data not shown). The 8A to 8C substitution within the linear transcript (Fig. 4B, lanes 5 and 6) reduced the U-insertions by a further factor of 2.0 Ϯ 0.4 (mean from 8 reactions), suggesting that approximately half the U-insertions within the linear transcript are still dependent on the A-U element (Fig. 4A, lanes 3  and 4). If 60 nucleotides are added to the 5Ј-end of linear cytochrome b transcript through circular permutation (22), the level of insertions approximates that of the corresponding circular RNA (Fig. 4A, lanes 7 and 8). Thus, the greater efficiency of U-insertion detected with the circular RNA is probably a consequence of adding nucleotides to the 5Ј-end of the A-U element and not an indirect effect of the circularization. The 8A to 8C substitution within the A-U element of the 5Ј-extended transcript inhibits the U-insertions (Fig. 4B, lanes 9 and 10). This is consistent with the effect of the same mutation on U-insertions within the circular RNA (Fig. 3B, lane 1).
There Is Specific Binding of the A-U Element to at Least One Protein within the Fractionated Mitochondrial Extract-Gelshift analysis was used to detect factors within the fractionated mitochondrial extract that specifically interact with the A-U sequence. Incubation of sucrose gradient fractions 6 and 7, the peak U-insertion fractions (Fig. 1A), with a 37-nucleotide transcript containing the A-U element results in formation of 6 complexes (Fig. 5, lane 2, A-F). Four of the complexes (labeled A, B, D, and E) are not detected when radiolabeled 8A to 8C substituted RNA is used in the assay (lane 1). These four bands are sensitive to pretreatment of the extract with proteinase K (data not shown). Since the same 8A to 8C substitutions also inhibited U-insertions within the cytochrome b transcripts (Fig. 3, lane 1), the missing complexes (A, B, D, and E) could be directly relevant to the U-insertion reaction. A greater amount of complex C forms with the 8A to 8C substituted RNA than with the wild type A-U transcript. This is suggestive that complex C could possibly be a precursor to the formation of complex E which is inhibited by the same mutation.
Three of the bands shifted by the A-U element can be competed by an unlabeled 41-nucleotide ferritin transcript (Fig. 5,  lanes 3-9). Whereas complexes A, B, and D are efficiently competed by the inclusion of low levels of the ferritin RNA during the binding reaction, bands C, E, and F are only competed approximately 50% by a 10-fold molar excess. The low level of U-insertion detected within the 192-nucleotide linear ferritin transcript (Fig. 4A, lanes 3 and 4) and in other transcripts not containing the A-U element (Fig. 3B, lanes 1-3) suggests that the U-insertion machinery can also interact relatively nonspecifically with RNA. It seems probable, however, that the induction of U-insertions when the A-U element is present involves complex E as it is both sensitive to the same A to C substitutions that inhibit U-insertions and is also relatively resistant to competition by the ferritin transcript. The K d for this complex is approximately 200 nM.
The gRNA-independent U-insertion Reaction Is Consistent with a Mechanism Proposed for gRNA-directed Editing-Previous studies have suggested that gRNA-directed editing proceeds through a mechanism involving sequential endonucleolytic cleavage at an editing site followed by the addition of uridine nucleotides to the 3Ј-OH group of the 5Ј-fragment and re-ligation of the fragments (23). For the in vitro editing of the T. brucei ATPase 6 mRNA, the best characterized in vitro kinetoplastid editing system, the U-additions and deletions are dependent upon the addition of gRNA (24,25). If the U-insertion components can assemble on some L. tarentolae pre-edited mRNAs independent of gRNA, an analogous mechanism could be mediating the gRNA-independent reaction. Such an assembly of U-insertion components on the pre-edited mRNA is supported by the gel-shift analysis (Fig. 5).
Consistent with a cleavage/U-addition/ligation mechanism, an RNA ligase is present within the same sucrose gradient  5 and 6). The 8A to 8C substitution within the A-U element of the chimeric transcript inhibits this induction (S (8A to 8C), lanes 7 and 8). B, nucleotides 5Ј of the A-U element influence the U-insertions. The RNase H assay was also used to detect U-insertions within the linear wild type cytochrome b RNA containing the mutated gRNA-binding site (lanes 3 and 4). The insertions are induced 15-fold when 60 nucleotides are added to the 5Ј-end of the A-U element (5Ј-extension, lanes 7 and 8). The U-insertions within both the linear and 5Ј-extended RNAs are inhibited by the 8A to 8C substitution (lanes 5, 6, 9, and 10).
fractions as the gRNA-independent U-insertion activity (Fig.  6A, lanes 1-11). Fractions 6 and 7, the peak U-insertion fractions (Fig. 1A), efficiently ligate radiolabeled linear cytochrome b RNAs containing a 3Ј-OH and 5Ј-monophosphate in a similar manner to T4 RNA ligase; linear dimers and intramolecular circles are formed in both reactions (Fig. 6A, lanes 6 and 7). The relative concentration of dimer and intramolecular ligation products is dependent upon the RNA substrate concentration and the sequence of the RNA used in the reaction. The reactions were performed in the presence of ATP and UTP under the same conditions in which the U-insertions occur. RNAs with a 3Ј-monophosphate and a 5Ј-OH are not substrates for the mitochondrial ligase activity (Fig. 6B), suggesting that if the reaction does proceed through a cleavage/ligation mechanism, it would require that a 3Ј-OH be produced during the endonucleolytic cleavage. The peak fraction (Fig. 6A, lane 6) has a low level of ligase activity in the absence of added ATP which is probably the result of ligases that had been adenylated prior to the fractionation (Fig. 6A, lane 12). Similar to some other RNA ligases (26), the kinetoplast ligase can also utilize UTP (Fig. 6A, lane 13). The presence of both ATP and UTP stimulates the reaction 6-fold relative to UTP alone (Fig. 6A,  compare lanes 6 and 13). Kinetoplastid RNA ligases have previously been described (27)(28)(29)(30).
Intermediates were isolated from the gRNA-independent reaction that are consistent with a cleavage/U-addition/ligation reaction mechanism. Endonucleolytic cleavage of the circular cytochrome b transcript during the U-insertion reaction produces full-length linear RNAs. The full-length linear molecules formed during the U-insertion reaction were gel purified and a poly(C) extension was added by poly(A) polymerase. The extension products were amplified for cloning by RT-PCR using a poly(G) primer and a primer containing the sequence of the mutated gRNA-binding site. The position at which the circular RNA was cleaved within the extract could then be ascertained by determining the sequence immediately adjacent to the poly(C) extension (Fig. 6C). Under the U-insertion reaction conditions, the sites of RNA cleavage do not correspond well with the sites of U-insertion (filled circles, Fig. 6C). Since the gRNA-independent U-insertion machinery appears to be part of a complex (Fig. 1A), it was reasoned that those RNAs cleaved within the extract as part of the editing reaction would be rapidly re-ligated back to circular RNAs, whereas those that are cleaved by contaminating nucleases would accumulate at a faster rate. In fact, when NTPs are excluded from the reaction so as to inhibit the ligase (Fig. 6A, ϪNTP), the sites of gRNAindependent cleavage in 50% of the sequenced RNAs now correspond with the sites of the gRNA-independent U-insertions (compare Fig. 2C with the open circles in Fig. 6C). In the presence of UTP, 41% of the linearized RNAs also had 1 to 3 U nucleotides added at the 3Ј-end (data not shown). Thus, the intermediates of the gRNA-independent reaction are suggestive of an endonuclease/U-addition/ligase-mediated reaction. FIG. 6. The gRNA-independent U-insertion reaction is consistent with the mechanism proposed for gRNA-directed editing. A, an RNA ligase co-fractionates with the U-insertion activity. Linear 32 P-labeled cytochrome b transcript containing a 5Ј-monophosphate and a 3Ј-OH group was incubated with the sucrose gradient fractions. There is little ligation in the absence of added NTP (ϪNTP, lane 12). UTP can support the reaction (ϩUTP, lane 13). B, although linear cytochrome b RNAs containing a 3Ј-OH and a 5Ј-monophosphate are chased into circular products by the fractionated mitochondria extract, RNAs with a 3Ј-monophosphate and a 5Ј-OH are not. C, under the U-insertion reaction conditions, the sites of endonucleolytic cleavage (filled circles) are not within the in vivo edited region (boxed) and do not correspond well with the sites of gRNA-independent U-insertion (Fig. 2C). If the RNA ligase is inhibited through the exclusion of NTP from the reaction, the sites of cleavage (open circles) in 50% of the sequenced molecules now correspond to the sites of U-addition.

FIG. 7. A biological role for the A-U element.
A, the L. tarentolae maxicircle was searched for sequences containing the A-U element. There is a 20-nucleotide match of the A-U element within a 22-nucleotide sequence of cytochrome b gRNA I (nucleotides 16760 -16781 in LEIKPMAX) and a 16-nucleotide match within an 18-nucleotide sequence immediately 5Ј of the editing sites in the mRNA encoding ND7 (nucleotides 3282-3299 in LEIKPMAX). Vertical lines indicate identical bases. The location of the 8A to 8C substitutions that inhibit over 80% of the U-insertions are also indicated. Cytochrome b gRNA I is shown below base pairing with the pre-edited mRNA (bottom). Lowercase letters represent the templating nucleotides. The sequence of the gRNA that matches that of the A-U element is shaded and the first 9 editing sites are boxed. B, A model for the function of the A-U element. Editing factors assemble on the A-U element of the pre-edited mRNA independent of gRNA. Accurate editing would ensue after the binding of gRNA and the transfer of the editing factors to its A-U element. editing components would be predicted to be detrimental even if the potential to guide a functional coding sequence within the pre-edited mRNA was maintained. As a result, there could be two different evolutionary forces conserving this sequence and limiting the opportunity for genetic drift.
The intermediates produced during the gRNA-independent reaction are consistent with the endonuclease/U-addition/ligation mechanism previously proposed for gRNA-directed editing based on the following observations. First, there is a ligase activity within the fractionated mitochondrial extract that cofractionates with the gRNA-independent U-insertion activity. Linear molecules containing 3Ј-OH and 5Ј-phosphate groups, the same groups produced during gRNA-directed editing (24,25), are chased into products by the co-fractionating ligase activity whereas RNAs containing a 3Ј-phosphate and 5Ј-OH are not. Second, there is significant accumulation of linear RNAs corresponding to cleavage of the circular cytochrome b RNA at sites of U-insertion when the ligase activity is inhibited through the exclusion of NTPs from the reaction. Third, in the presence of UTP, U nucleotides are added to the 3Ј-end of the linear molecules. These apparent similarities in mechanism are consistent with the possibility that the gRNA-independent reaction exploits some of the same editing components used during gRNA-directed editing.
The A-U element could, thus, be serving as an assembly point for the editing machinery on those pre-edited mRNAs for which it is especially important to be efficiently edited (Fig.  7B). The original model for gRNA-directed editing, in fact, postulated that editing components may interact with the preedited mRNA independent of gRNA (4). The pre-assembled editing factors would be able to insert U nucleotides within both the 5Ј and 3Ј adjacent sequence (Fig. 7B, top). This would explain the localization of the majority of the in vitro gRNAindependent U-insertions (Fig. 2C). U-insertion components could be transferred to the A-U element of the gRNA upon its binding to the pre-edited mRNA (Fig. 7B, middle). Cytochrome b gRNA I was previously shown to specifically inhibit the independent U-insertion reaction (9,10) and this, in part, could be a result of the machinery being transferred from the A-U element of the mRNA to that of the gRNA. Accurate editing would occur after the transfer of editing factors to the gRNA (Fig. 7B, bottom). Additional experiments are being performed to futher test this model.
Although the A-U element has only been identified within cytochrome b gRNA I, there are several other kinetoplastid gRNAs that are very A-U rich. Selection-amplification (33,34) is being used to identify the sequence permutations of the A-U element that are able to support the internal U-insertions. Knowing the permissible permutations will be informative in determining whether related elements are found on other gRNAs. It is also possible that there are gRNA and/or mRNA-specific factors required for editing. The latter possibility is likely since there are life cycle stages in T. brucei during which specific mRNAs are not edited, even though both the pre-edited mRNA and appropriate gRNA are present (35)(36)(37).
Although the A-U element is able to support the U-insertion reaction and form specific complexes with the mitochondrial extract, the reaction is significantly enhanced by the addition of 5Ј-nucleotides either as a result of the intramolecular ligation or an artificial extension (Fig. 4B). This additional requirement for efficient U-insertion could also limit inappropriate U-insertions by the editing components assembled on the mRNA in vivo. The enhancement could be occurring in vivo, however, if the cytochrome b pre-edited mRNA, like the mitochondrial genome of some eucaryotes (see Ref. 38; reviewed in Ref. 39) is transcribed as part of a polycistronic precursor.
Alternatively, an enhancement could possibly result from interactions between the 5Ј-and 3Ј-ends of the mRNA during processing and translation (40,41). There are examples of in vivo misediting of cytochrome b mRNA (42), and it is possible that some of these events are a consequence of having editing components assembled on the pre-edited mRNA independent of gRNA.
The enhancement of the gRNA-independent U-insertions by an artificial 5Ј-extension on the cytochrome b mRNA is in agreement with previous results from a primer extensionbased assay (10). However, the results from the primer extension assay also suggested that the 5Ј-extension was required to be base paired, and no such requirement was detected in this study. The discrepancy probably is a result of the primer extension assay being limited to the detection of U-insertions within the first 2 editing sites. Subtle changes in the structure of the RNA could be influencing the sites at which the Uinsertions are occurring. Although the A-U element is a major determinant of the reaction, we cannot eliminate the possibility that the reaction is also influenced by other sequence and/or structural elements. Changes in structure may account for the increased level of U-insertion observed with the deletion of the pre-edited region (Fig. 3B, lane 4).
A-U rich sequences are found within the 5Ј-untranslated region of several mitochondrial mRNAs. The L. tarentolae A-U element is not completely conserved within the cytochrome b mRNA of T. brucei and Crithidia fasciculata, but both of these related genera of trypanosomes have similar sequences immediately upstream of the editing sites (43). An A-U-rich sequence present within the 5Ј-untranslated region of the Saccharomyces cerevisiae cytochrome b mRNA has been implicated in an interaction with the translation machinery (44). This raises the intriguing possibility that the editing components interacting with the L. tarentolae A-U element could either be associated with the translation machinery or perhaps evolutionarily related.