RNA Editing of Apolipoprotein B mRNA SEQUENCE SPECIFICITY DETERMINED BY IN VITRO COUPLED TRANSCRIPTION EDITING*

(apo) is produced by in viva RNA editing which involves a C+U conversion of the first base of the codon CAA for Gln-2153, changing it to UAA, an in-frame stop codon. We have reproduced the editing reaction in vitro using nuclear extracts. Efficient RNA editing was demonstrated by using apoB segments as substrate or in a coupled transcrip-tion-editing reaction using apoB minigenes as tem- plate. ApoB minigenes constructed by ligating late a the for the defined

RNA editing is a molecular biological phenomenon whereby the primary structure of an RNA transcript is altered by mechanisms other than splicing (l-4). The first description of this process involved the addition of nongenomically encoded uridine (U) residues to mitochondrial mRNAs in the kinetoplastid protozoa (5). Subsequently, U residues have been found to be removed from some transcripts (6, 7). Recently, Thomas et al. (8) described the addition of two nontemplated G residues in Paramyxovirus SV5 which joined two open reading frames to produce the P protein. An analogous situation was also described in measles virus where variable numbers of Gs were inserted into the molecule (9). The only putative RNA editing described in mammals involves the conversion of a C to a U residue in apoliprotein (apo) B mRNA in the small intestine (10, 11); editing was also demonstrated recently in cells transfected with apoB gene constructs (12,13).
ApoB is a major protein constituent of plasma lipoproteins. The plasma concentration of apoB shows a strong direct correlation with the development of coronary artery disease (14,15). Plasma apoB exists in two major forms (16), apoB-* This work was supported by National Institutes of Health Grants HL 27341 (to L. C.) and AR 38858 (to W. S. L. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "uduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 100 and apoB-48, which contain 4536 and 2152 amino acid residues, respectively (10,11,(17)(18)(19)(20)(21). In humans, apoB-100 is the product of a 14-kb' mRNA in the liver. ApoB-48 is the product of an intestinal mRNA of adult humans which is identical in structure to apoB-100 mRNA except for a single C+U base substitution involving C-6666, the first base of the codon CAA for Gln-2153, changing it to UAA, a stop codon (10, 11).
The mechanism behind the C-U conversion is unknown. In this study, we have reproduced the RNA editing in vitro using a nuclear extract and a template consisting of either a segment of apoB DNA (i.e. coupled transcription editing) or apoB mRNA (i.e. direct RNA editing). By altering the nucleotide sequence of the template by site-specific mutagenesis, we have defined the sequence specificity of the editing reaction. The sequence specificity of the reaction was found to be relatively lax, which has important implications for the potential role of RNA editing in the regulation of gene expression.

Primer Extension
Assay for RNA Editing-Primer extension in the presence of dideoxy GTP was used as a rapid assay for the absence (i.e. unedited template) or presence (i.e. edited template) of primerextended product extending beyond C-6666. An oligonucleotide with the sequence 5'-AATCATGTAAATCATAACTATCTTTAATA-TACTG-3' (34-mer) was used as primer. Primer extension was performed essentially as described by Driscoll et al. (22). The primer extension products were separated by electrophoresis on a 12% polyacrylamide sequencing gel and exposed to Kodak XAR-5 x-ray film for varying periods of time.
Rut Liver Nuclear Extract Preparation-All manipulations were performed in the cold, and all solutions, tubes, and centrifuges were chilled to 0 "C. Rat liver nuclear extracts were prepared essentially as described by Gorski et al. (23). We normally obtained approxi-

AND DISCUSSION
Editing of ApoB RNA in Vitro-In our initial testing of RNA editing in uitro, we incubated T7 transcripts of cloned apoB-100 cDNA segments containing C-6666 with a nuclear extract from rat liver, a tissue known to contain large amounts of edited apoB mRNA (28, 29). The RNA was then purified and analyzed by two different assays: (i) the PCR cloninghybridization method described under "Materials and Methods" and (ii) a primer extension assay similar to the one described recently by Driscoll et al. (22). The primer extension method is rapid and moderately sensitive, detecting down to 1% of C+U conversion in standard RNA mixtures of 0.1 pg. Unfortunately it is not linear in the lower concentration range, and its sensitivity is also limited. The PCR cloninghybridization procedure was much more sensitive. For example, while editing was demonstrated with 100 and 10 ng of apoB RNA transcripts in the reaction, the primer extension failed to detect the editing of an in vitro apoB transcript when only 1 ng was used in the reaction (Fig. 1). In contrast, when the PCR cloning-hybridization assay was used, editing was detected with all three concentrations of input RNA (Table  I, A). It was more efficient in the reaction using 10 ng than those using higher (100 ng) or lower (1 ng) amounts of input RNA. The latter two RNA concentrations were edited to about the same extent. The fact that editing was detected in the lOO-and lo-ng but not in the l-ng reaction in the primer extension assay indicates that the B-Stop signal was a function of the total amount of B-Stop sequences generated, which tsil -c r-54-mer is in turn dependent on the total input RNA in the assay. Therefore, it is not an accurate assay for the extent of editing, and its sensitivity is much inferior to the PCR cloninghybridization assay, which is independent of the input RNA concentration. Thus, the latter assay alone was used in all subsequent experiments. Both nuclear extract and apoB mRNA were needed for the reaction. Incubation of apoB mRNA with buffer alone or with heat-inactivated nuclear extract failed to show any C+U conversion. The authenticity of the edited product was confirmed by direct sequencing. Of all the Cs in the B-mRNA,kt,, only one (C-6666 of the codon CAA for Gln-2153) was converted to a U; none of the other Cs were changed. Also, this base substitution is the only one observed; by direct sequencing, there is no other type of base changes in the edited RNA.

Coupled Transcription
Editing of ApoB-100 Minigene-The nuclear extract (23) used for RNA editing had been found earlier to be active in transcribing a number of mammalian genes (including the albumin and serum amyloid A genes, data not shown). We tested its ability to direct the coupled transcription editing of an apoB-100 minigene. An apoB-100 minigene was constructed by ligating the adenovirus major late promoter to the 5' end of an apoB-100 DNA segment corresponding to different lengths of apoB mRNA ( Fig. 2A). The AdB-100 minigene was incubated with the nuclear extract. The RNA product was purified and assayed for the relative proportions of B-Stop versus B-Gln transcripts by the PCR cloning-hybridization assay. Authenticity of the products was further confirmed by direct sequencing. It is evident that the nuclear extract can transcribe the AdB-100 minigene constructs as well as edit the transcripts under these conditions (Table I, B). The nuclear extract was somewhat more effective in the coupled transcription-RNA editing of AdR-100260 than the two longer constructs. Furthermore, both the supercoiled and linearized minigenes were transcribed and edited although the efficiency may be slightly higher with the supercoiled substrate. As in the case of RNA substrates, the only base change detected in these coupled transcription-RNA editing reactions was C-6666. In both the direct RNA-editing and the coupled transcription-editing experiments, misincorporation in the PCR as the cause of the C-&J substitution was excluded by the following: (i) in the direct editing experiments, incubation of the RNA in buffer alone did not produce any B-Stop sequences assayed by the PCR cloning-hybridization technique; furthermore, primer extension assay of the RNA products without PCR amplification confirmed the editing reaction; (ii) in the coupled transcription-editing experiments, the PCR cloning-hybridization assay performed directly on the DNA template (i.e. AdB-1002& showed 100% B-Gln sequences; and (iii) direct sequencing of multiple cloned B-Gln or B-Stop PCR products revealed no other base substitutions.
Comparison of the direct RNA-editing and coupled transcription-editing experiments indicates that the two reactions were comparable in efficiency. The fact that the editing reaction occurred under cell-free conditions using nuclear extracts suggests that RNA editing in uiuo may occur in a similar manner, i.e. as a coupled reaction in the nucleus. Therefore, we used the coupled transcription editing of supercoiled AdB-100 minigenes to analyze the tissue and sequence specificity of the reaction.
In mammals, there is species-and tissue-specific variation in the efficiency of RNA editing. We tested the ability of nuclear extracts from three different tissues to perform the coupled transcription-RNA editing (Table I, B). All three extracts efficiently transcribed the apoB minigene. However, only nuclear extracts from rat liver edited the transcript. The activity of the extracts from HeLa or Hep3B cells was undetectable. This observation correlates with the presence of substantial amounts of apoB-48 mRNA in rat liver RNA and its absence in Hep3B RNA (data not shown); HeLa cells normally do not produce any apoB mRNA.
Sequence Specificity of ApoB mRNA Editing in Vitro-The sequence specificity of the coupled transcription-editing reaction was examined by site-directed mutagenesis of the AdB-1OO26o DNA minigene (Fig. 2B). We constructed 22 different mutant apoB-100 minigenes, including constructs that contain single or multiple base substitutions in the bases immediately flanking C-6666, as well as one construct that contains a single-base insertion, and another, a single base deletion. The efficiency of editing was quantified by the PCR cloninghybridization assay and compared with the editing efficiency of wild type Ad13-100Z60. The most obvious conclusion that can be drawn from these in uitro mutagenesis experiments is that the reaction is promiscuous with respect to the sequence of the RNA substrate; of the '22 mutant RNA sequences, all except two were edited by the nuclear extract in vitro (Table  II). The results obtained in Table II were all confirmed by direct sequence analysis (data not shown).
In mutants u-n, the three bases flanking the 5' and 3' sides of C-6666 were mutated individually, producing transition mutants (a+) or transversion mutants (g+n). These constructs were edited in vitro with varying efficiency. For transition mutants, enhanced editing efficiency was observed for the mutant containing CGA instead of CAA (mutant a, which had a a-fold increase in efficiency); mutant T-6669-C (mutant c) was also edited somewhat more efficiently (-1.5fold) than wild type. Mutant A-6665+G (mutant d) was edited with normal efficiency. The two other constructs containing transition mutations (mutants e and f) were edited with an efficiency of 50% of wild type. Three of the transversion mutants, h, g, and k, were edited with markedly, moderately, and slightly reduced efficiency, respectively. The other transversion mutants were edited with the same (mutants i, j, and 1) or higher efficiency (mutants m and n) compared with wild type. As noted below, introduction of single Cs in the proximity of C-6666 appears to stimulate editing in some instances. This complete inhibition of editing should be contrasted with the roughly 50% editing efficiency of the corresponding double transition mutant (0). Complete inhibition of editing was also observed for the single deletion mutant, s, where one of the two As following C-6666 was deleted. In the insertion mutant, t, where an extra C was inserted next to C-6666, both Cs were edited with reduced efficiency.
The fact that the vast majority of mutations do not seriously impair editing efficiency suggests that the sequence requirement of this reaction is not very stringent.
To test this hypothesis, we constructed two 6-base substitution mutants where the two codons flanking the CAA are mutated, being replaced completely by transitional substitutions (mutant u) or five transitional substitutions and a single transversional substitution (mutant v). Transcripts from both constructs were edited with an efficiency approximately one-third (for mutant u) and two-thirds (for mutant v), respectively, of that of wild type.
In a number of the mutants that we studied, one or more Cs were introduced in the sequence in close proximity to C-6666. These constructs include mutants m, n, r, and t. Like C-6666, most of the neighboring Cs were also found to be edited in vitro. It is interesting that introduction of a single C immediately next to C-6666 stimulated editing when the relative position of C-6666 was not altered (mutant m). Both Cs in mutant m were simultaneously edited with high efficiency. In mutant t, where there was an insertion, there was moderate impairment of editing efficiency, the 3' C being affected more than the 5' C. In comparing the relative efficiency with which each C was edited in individual mutants, there appears to be some general pattern. In three instances where the relative position of C-6666 was not changed (mutants m, n, and r), the original C-6666 was edited much more efficiently than the Cs at the neighboring positions. Mutant r contains a substitution of the triplet CCC for CAA. In this case, the first C-6666 was found to be edited with normal efficiency, whereas the two 3' Cs, C-6667 and C-6668, were edited poorly. Simultaneous editing of the two 5' Cs was also observed, though infrequently.
These results suggest that there may be a distance effect between the edited C and some flanking DNA sequence. Results with mutants m, n, and r suggest that a minimal distance between C-6666 and some 3' sequence or a maximal distance between this base and some 5' sequence is preferred for efficient editing. The slightly more efficient editing of the 5' C of the insertion mutant t is difficult to interpret because there is significant impairment of editing of both Cs.
Implications of Sequence Specificity of ApoB mRNA Editing-We have found that a rat liver nuclear extract not only can edit synthetic apoB RNA in vitro (Table I, A), as was demonstrated recently by Driscoll et al. (22), but also could perform coupled transcription editing of a minigene construct in which the in vitro nascent apoB transcript was edited with fidelity (Table I, B). RNA editing is likely a nuclear event in vivo. The system described here may mimic the in vivo situation better than one that uses a synthetic RNA as substrate. We have tested synthetic RNAs corresponding to each of the mutant constructs in direct editing experiments.
All of them, including mutants q and s, were edited in vitro, which suggests that coupled transcription editing may have a more stringent sequence requirement than direct RNA editing. The ability to introduce multiple mutations to apoB RNA without seriously reducing the editing efficiency indicates that specific base pairing or hydrogen bonding involving the bases immediately flanking C-6666 is not required for recognition by the putative cytidine deaminase enzyme. The relative laxity of the structural requirement is also supported by the fact that when additional C residues were introduced to the vicinity of C-6666, they were edited often to the same extent as C-6666. On the other hand, since mutants q and s, which contained double transversions and a single deletion, respectively, were consistently not edited in the coupled transcription editing reaction, there is definitely a preferred RNA structure recognized by the enzyme.
Although most of the mutant apoB RNA sequences were edited with normal or reduced efficiency, two of the constructs (mutants a and p) were consistently edited at an efficiency approximately 3-fold that of wild type. Synthetic transcripts of both mutants that were fed into the nuclear extracts were also edited efficiently (data not shown). It is possible that both apoB RNA mutants are deleterious to the organism.
Mutant a, which contains CGA in place of CAA, would be edited so efficiently in the liver that insufficient apoB-100, a physiologically important protein, would be produced. In contrast, editing of mutant p, which contains CGG instead of CAA, would produce an Arg+Trp substitution and not a stop codon, and no apoB-48 would be produced in the intestine. For these reasons, we speculate that both mutations, if they occurred, would be eliminated from the population. Recently, an analogous RNA-editing mechanism involving C-U conversions has been described as a common phenomenon in plant mitochondria (30,31). In our current analysis, the relatively loose stringency of the editing reaction suggests that other RNAs may also be edited in vivo. Thus, in mammals as in plant mitochondria, RNA editing may not be a unique biologic phenomenon confined to apoB.