Two efficiency elements flanking the editing site of cytidine 6666 in the apolipoprotein B mRNA support mooring-dependent editing.

Normally, apolipoprotein B (apoB) mRNA editing deaminates a single cytidine (C6666) in apoB mRNA. However, when the catalytic subunit of the editing enzyme complex, APOBEC-1, was overexpressed in transgenic mice and rabbits, numerous cytidines in the apoB mRNA and in a novel mRNA, NAT1, were aberrantly hyperedited, and the animals developed liver dysplasia and hepatocellular carcinomas. To identify the RNA motifs in the apoB mRNA that support physiological editing and those that support aberrant hyperediting, we constructed rabbit apoB RNA substrates and tested them in vitro for physiological editing and hyperediting. Three previously unrecognized RNA elements that are critical for efficient physiological editing at C6666 were identified. In concert with the mooring sequence (6671-6681), the 5' efficiency element (6609-6628), an A-rich region (6629-6640), and the 3' efficiency element (6717-6747) increased editing at C6666. The 5' efficiency element was the most potent, elevating physiological editing to wild-type levels in combination with the mooring sequence. The 3' efficiency element was somewhat less important but increased physiological editing to levels approaching wild type. These elements encompass 139 nucleotides on the apoB RNA transcript and are sufficient for editing with the efficiency of full-length apoB mRNA. Furthermore, a distal downstream apoB region (6747-6824) may function as a recognition element in the apoB mRNA. Hyperediting at C6802 in the rabbit apoB mRNA is mediated by RNA elements similar to those required for normal physiological editing at C6666. Similarly sized upstream and downstream flanking regions of C6802 are necessary for hyperediting in combination with a degenerate mooring sequence.

RNA editing is the alteration of the genetic information present in nascent RNA transcripts. One form of RNA editing, apolipoprotein B (apoB) 1 mRNA editing, deaminates a specific cytidine (C 6666 ) in the apoB mRNA, which generates a uridine (1-4) changing codon 2153 from a genomically encoded CAA (glutamine) to an in-frame stop codon (UAA) (5,6).
ApoB mRNA editing results in the formation of a truncated apoB protein (apoB48) that is about half the size of the fulllength genomically encoded apoB (apoB100) (reviewed in Refs. 7 and 8).
Under physiological conditions, apoB mRNA editing is an exquisitely precise process. No other RNAs are known that undergo this C 3 U deamination, and no other cytidines in the apoB mRNA are edited, except for a minor site in human intestinal apoB mRNA (9). Two nucleotide sequence elements have been identified that are necessary for editing: an 11nucleotide (nt) "mooring sequence" (6671-6681) (10 -12) and a 4-nt spacer sequence between the edited cytidine and the mooring sequence (10,11,13). Integration of 26 -63 nt of specific apoB RNA sequences encompassing the mooring sequence and the spacer sequence into heterologous mRNAs resulted in editing at levels substantially less than wild type, suggesting that the RNA environment is important and that other unidentified elements are necessary for efficient editing at C 6666 (14,15).
The apoB mRNA editing complex comprises a catalytic subunit designated APOBEC-1 (16) and other, as yet unidentified, auxiliary proteins (17)(18)(19). Overexpression of APOBEC-1 in the livers of mice and rabbits resulted in liver dysplasia and hepatocellular carcinomas (20). Overexpression of APOBEC-1 also resulted in the aberrant editing of several other cytidines in the apoB mRNA (21)(22)(23). The highest degrees of editing occurred at C 6738 , C 6743 , C 6762 , C 6782 , and C 6802 (21). Numerous cytidines were also edited in a novel mRNA, NAT1 (22). This aberrant editing was termed hyperediting to distinguish it from the physiological editing at C 6666 . Sequence analysis of the RNA surrounding the hyperedited cytidines showed that no exact mooring sequence was present within the correct distance to support this hyperediting (21). In contrast, other mRNAs containing a mooring sequence were not hyperedited (20), indicating that elements in addition to the mooring sequence are necessary for hyperediting.
Here we report results from in vitro studies of physiological editing at C 6666 and hyperediting at C 6802 in the apoB RNA. Our results show that the essential elements for physiological editing at C 6666 of the rabbit apoB mRNA encompass 139 nt, consisting of defined upstream efficiency elements, the mooring sequence, and the 3Ј efficiency element. Similar RNA features are necessary for the hyperediting of cytidines when APOBEC-1 is overexpressed.

EXPERIMENTAL PROCEDURES
Generation of ApoB RNAs-ApoB RNAs were produced by in vitro transcription from polymerase chain reaction (PCR) constructs, that included a T7 promoter. The apoB cDNA constructs were generated by PCR amplification from the plasmid pRabSK, a derivative of pRab-1 (24), encoding a 354-base pair rabbit apoB cDNA segment. The PCR products were purified by agarose gel electrophoresis and eluted by Qiaex II gel extraction (Qiagen, Santa Clarita, CA). The primers designed on the computer program OLIGO 4.0 were 18 -25-nt-long and were purchased from Life Technologies, Inc. The T7 promoter sequence * This work was supported in part by the National Institutes of Health Program Project Grant HL47660 and by Boehringer Ingelheim International, GmbH, as well as by fellowships from the Swiss National Science Foundation and Cancer Research Switzerland (to M. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
with GGA (5Ј-GGATCCTAATACGACTCACTATAGGGA-3Ј) added to achieve efficient expression was incorporated into the PCR constructs by a second PCR amplification. RNA was produced by the T7-MEGA shortscript in vitro transcription kit (Ambion, Austin, TX). The RNA was analyzed by agarose gels, treated with DNase 1, and purified by phenol-chloroform extraction and ethanol precipitation. The names of the RNAs refer to the first nucleotide of the RNA according to the apoB cDNA position (e.g. B6629 starts at nt 6629), and the length of the RNAs is given in nucleotides.
The DNA constructs for the P1 (GenBank accession number X62154) and the N-myc RNA (GenBank accession number M12731) were produced by reverse transcription-PCR from normal mouse liver RNA. The primers for the N-myc amplification were MU6879:5Ј-ACGGCCTGTATACTTTTG-TATG-3Ј and ML7053:5Ј-AACAAATACAGTAAACAAGGAA-3Ј. For the P1 amplification, the primers were P1U1279:5Ј-GCCCGCTGCAGTGTTT-TGG-3Ј and P1L1454:5Ј-CCCTGTATTGGTGCATCCTAA-3Ј. The N-myc and the P1-PCR fragments were TA subcloned (Invitrogen, San Diego, CA) and sequenced. Chimeric RNAs were produced from PCR-derived DNA constructs by recombinant PCR (25). The PCR products were purified, and the RNAs were produced as described for the apoB RNAs. Substitution mutagenesis was accomplished by incorporating the alterations in the primers used to produce the PCR constructs.
The The PCR templates for all the apoB/N-myc chimeric RNAs were sequenced.
In Vitro ApoB mRNA Editing Assay-For the in vitro apoB mRNA editing assay (26), 100 pg of synthetic RNA prepared as described above, 100 g of rabbit liver S100 extract, 5 g of recombinant APOBEC-1 (MBP-APOBEC-1) (21), 1 g of Escherichia coli tRNA, and 40 units of RNasin (Promega, Madison, WI) in buffer D containing 1 mM dithiothreitol in a reaction volume of 100 l were incubated at 30°C for the indicated times and then extracted as described (26). In control experiments, the RNAs were incubated in this assay mixture lacking rabbit liver S100 extract. All RNAs were amplified by reverse transcription-PCR, and the resulting single-band PCR products were purified over a microspin S-300 HR column (Amersham Pharmacia Biotech).

RESULTS
To investigate the influence of distal flanking sequences on editing of C 6666 and hyperediting of C 6802 in the apoB mRNA, we used an in vitro editing assay of small apoB RNAs. The RNAs were produced by in vitro transcription and incubated in the presence of rabbit liver extract (as a source of auxiliary proteins) and an excess of recombinant APOBEC-1. The percentage of editing of a specific cytidine was then determined by primer extension analysis.
Editing at C 6666 -To determine the impact of downstream elements on physiological editing of C 6666 , we first generated RNAs with 3Ј sequences of different lengths (Fig. 1). Editing at C 6666 was determined in an in vitro editing assay after 1, 2, 4, and 16 h. The parental RNA B6507 (354 nt) was highly edited (82%) after 16 h (Fig. 1). Editing of smaller RNAs was similar; even B6629⌬-77 (119 nt) was edited to 77% after 16 h of incubation. However, removal of an additional 31 nt at the 3Ј end (6717-6747) decreased editing by almost half. We termed this 31-nt spanning region located 51-81 nt downstream of C 6666 the "3Ј efficiency element." That the loss of editing efficiency was not simply due to the length of the RNA was shown by using the N-myc chimeric RNA MS-M (197 nt) and the P1 chimeric RNA MS-P1 (196 nt) in which sequences 3Ј of the mooring sequence were replaced with N-myc or P1 sequences. The AU-rich apoB/N-myc RNA and B6629⌬-118 were edited to a similar extent. The GC-rich apoB/P1 RNA was marginally edited (MS-P1, Fig. 1).
To investigate the influence of upstream flanking regions on editing of C 6666 , we transcribed small RNAs lacking the 3Ј efficiency element. RNAs starting at nt 6609 (e.g. B6609 (78 nt) and B6609L (98 nt)) were edited with the same efficiency as the parental construct B6507, resulting in 80 -90% editing within 4 h of incubation (Fig. 2). However, deletion of 20 nt in the 5Ј end resulted in a 50% decrease in editing after 16 h of incubation ( Fig. 2, B6629⌬-108 (88 nt) and B6629⌬-118 (78 nt)). The size of the RNA was not responsible for the difference in editing, because B6629⌬-118 and B6609 are the same length (78 nt). Thus, the difference in editing efficiency was due to the RNA element 6609 -6628, located 38 -57 nt upstream of C 6666 , which we termed the "5Ј efficiency element." Because RNAs lacking the 5Ј efficiency element were still edited to about 50% of the wild-type level after 16 h, we deleted more 5Ј sequences to define closer regions that promote editing at C 6666 . Additional deletion of 12 nt from the 5Ј end resulted in an RNA, B6641 (46 nt), that was edited only to 7% (n ϭ 3) after 16 h of incubation (data not shown), indicating the importance of a second region 26 -37 nt upstream of C 6666 for efficient editing at C 6666 . This second region is A-rich (AAAUGAAAAA), and mutations in this A-rich region (6629 -6640) revealed that three A 3 U and three A 3 G substitutions decreased editing at C 6666 (data not shown). However, introduction of a mooring sequence (UGAUCAGUAUA) returned editing efficiency to high levels (65% after 16 h (data not shown)).
To distinguish further whether the 5Ј and 3Ј efficiency elements are part of the minimal flanking requirement for efficient physiological editing at C 6666 or whether they are specific enhancer elements for physiological editing, we made chimeric apoB/N-myc RNAs (Fig. 3). Previous deletion and mutagenesis experiments indicated that the 26-nt apoB editing cassette, including C 6666 and the mooring sequence, is sufficient for low level editing in certain heterologous RNA contexts (10,12). We therefore exchanged sequences flanking this apoB-editing cassette with N-myc sequences, resulting in 20 Ϯ 14% (M-MOOR-M) editing at C 6666 after 16 h. This represents about 25% of the editing activity detected in the parental apoB RNA (B6609C). However, chimeric RNAs containing the apoB 5Ј efficiency element were edited to wild-type levels (EFF-MOOR-M and EFF-MOOR-EFF), emphasizing the importance of the 5Ј efficiency element (6609 -6628) for efficient mooringdependent editing. Adding the 3Ј efficiency element (6717-6747) did not further increase editing efficiency when the 5Ј efficiency element was present (EFF-MOOR-EFF), but in the absence of the 5Ј efficiency element, the 3Ј efficiency element increased editing 2-fold (M-MOOR-EFF). These data indicate that in combination with the apoB editing cassette, the 5Ј efficiency element and to a lesser extend the 3Ј efficiency element restored editing at C 6666 to wild-type levels.
Our data indicate that efficient physiological editing at C 6666 is enhanced by both the 5Ј efficiency element (38 -57 nt upstream) and by the 3Ј efficiency element (51-81 nt downstream). If these apoB elements enhance editing at C 6666 , they may promote hyperediting in N-myc RNA containing a mooring-like motif. To test this possibility, we exchanged the 26-nt apoB editing cassette for the equivalent N-myc sequences (Fig.  4). The N-myc RNA (MY) was not hyperedited, although it contains mooring-like motifs and its AU content is equivalent to that of apoB RNA. When the apoB 5Ј and 3Ј efficiency elements were added to the N-myc RNA, no hyperediting was detected (Fig. 4, EFF-M-EFF). Thus, the apoB 5Ј and 3Ј efficiency elements promote editing in combination with the apoB editing cassette (Fig. 3). However, providing the entire apoB upstream flanking region (6609 -6661), including the 5Ј efficiency element (F-M) or the entire apoB downstream flanking region (6688 -6747), including the 3Ј efficiency element (M-F), resulted in hyperediting of one (C 6914 ) or two cytidines (C 6914 and C 6909 ) in the N-myc RNA. The flanking regions of the editing cassette in the apoB RNA therefore act as recognition elements for hyperediting of a cytidine in the N-myc RNA context. Because we hypothesized that several elements in the apoB mRNA could function as recognition elements and could support mooring-dependent editing (22), we examined the influence of the more downstream apoB RNA element, B6747, on hyperediting of the N-myc RNA (Fig. 4). The apoB sequence applied in these chimeric RNAs does not cover sequences relevant for efficient editing of C 6666 , but B6747 starts directly downstream of the 3Ј efficiency element mapped above. Again, this longer N-myc RNA (M) was not hyperedited (Fig. 4). In M-B6747, however, the substitution of 78 nt of apoB RNA (B6747) in the N-myc RNA caused editing of two cytidines 5Ј of the N-myc mooring-like motif (Fig. 4). These results indicate that an element in the apoB RNA more than 81 nt downstream of C 6666 influences editing.
Hyperediting-Our data indicate that both upstream and downstream flanking regions contribute to editing of C 6666 . To compare these findings to hyperediting, we first mapped in vitro the minimal upstream sequence required for hyperediting of C 6802 (Fig. 5). B6629 includes C 6666 , the mooring sequence downstream, and the region with the clustered cytidines that are hyperedited in the apoB mRNA from transgenic mice overexpressing APOBEC-1. All cytidines investigated in B6629 (196 nt) were hyperedited in vitro in a pattern similar to that observed in vivo (21). At C 6802 the level of hyperediting was 4.9 Ϯ 1.7% (n ϭ 3). Deletion of the mooring sequence (Fig. 5, B6687, 138 nt) abolished hyperediting at C 6743 and increased hyperediting 2-fold at C 6802 (11.8 Ϯ 4%, n ϭ 3). Deletion of an additional 60 nt of 5Ј sequences (Fig. 5, B6747, 78 nt) resulted in the loss of combined hyperediting of C 6782 and C 6783 (C 6782/3 ), but C 6802 was still hyperedited (4.2 Ϯ 1.5%, n ϭ 3). However, deletion of an additional 13 nt at the 5Ј end abolished hyperediting at C 6802 (Fig. 5, B6760, 65 nt), indicating that sequences 42-55 nt upstream are essential for hyperediting at this site.
Intriguingly, the sequence 42-55 nt upstream of C 6802 is AU-rich (AAAAAUUAAAA), similar to the A-rich region (AAAUGAAAAA) that contributes to physiological editing of C 6666 . Furthermore, the AU-rich sequence contains a cryptic polyadenylation signal (AUUAAA). We tested the possibility that the cryptic polyadenylation motif (6752-6757) influences hyperediting of C 6802 , because we and others have shown a link between editing and activation of this cryptic polyadenylation signal in the apoB mRNA (5,6,15,27). The 78-nt apoB RNA, B6747, contains the cryptic polyadenylation signal (Fig. 6). Mutation of the cryptic polyadenylation signal to AUUAUA by the substitution of a U for A did not diminish hyperediting of C 6802 in UB6747 (Fig. 6). Moreover, in the MB6747 RNA, the replacement of the cryptic polyadenylation signal with a mooring sequence (UGAUCAGUAUA) had no effect on hyperediting of C 6802 . Therefore, the cryptic polyadenylation signal (AUUAAA) itself is not necessary for hyperediting at C 6802 . However, three A 3 G substitutions in this region abolished hyperediting at C 6802 (Fig. 6, GB6747). Thus, there appears to be a preference, if not a necessity, for an AU-rich sequence in this 5Ј-flanking region 42-55 nt upstream of C 6802 .
To investigate the importance of the mooring-like motif (6809 -6819) downstream of C 6802 , we used the constructs shown in Fig. 7. This mooring-like motif is located five nucleotides from the 3Ј end of all short apoB RNAs investigated and is the only mooring-like motif downstream of C 6802 . In rabbit apoB RNA, this mooring-like motif matches only six out of the 11 nucleotides in the mooring sequence and has a 6-nt instead of a 4-nt spacer region (Fig. 7). In the 138-nt B6687 RNA template, all of the cytidines investigated were hyperedited except for C 6743 (Fig. 7). Scrambling the mooring-like motif at 6809 -6819 (B6687S) decreased editing at C 6802 to 33% of the level of the parental RNA (B6687). Furthermore, scrambling of the mooring-like motif (B6687S) and deletion of this region (B6687⌬-16) abolished hyperediting of C 6782/3 . In contrast, both mutations only slightly decreased hyperediting at C 6762 and at other hyperedited cytidines. For example, C 6762 was hyperedited to 9 -15% in the mutated RNAs, which is slightly less than in the parental RNA (19%). Therefore, the mooring-like motif at 6809 -6819 influenced hyperediting of cytidines up to 27 nt upstream, but hyperediting of C 6762 , located 47 nt upstream, was unaffected by this mooring-like motif.
To investigate this mooring-like motif at 6809 -6819 without the influence of distal elements, we concentrated on the smallest apoB RNA template, B6747, that was still hyperedited at C 6802 . In contrast to the 138-nt B6687S RNA (Fig. 7), where scrambling of the mooring-like motif decreased hyperediting of C 6802 , in the smaller 78-nt RNA B6747S, the same mutations abolished hyperediting at C 6802 (Fig. 8). Changing the mooringlike motif to an exact 11-nt mooring sequence (B6747M) motif enhanced editing of C 6802 2-fold. Furthermore, restoring the pattern of the mooring-like motif with an extra A in the spacer region (B6747R) between the mooring-like motif and C 6802 had no effect on hyperediting of C 6802 . These results indicate that an AU-rich sequence is not sufficient to support hyperediting at C 6802 unless it resembles the mooring sequence.
An AU-rich element and the mooring-like motif support hyperediting of C 6802 , but downstream elements also influence hyperediting. B6760 (65 nt) was not hyperedited at C 6802 because it lacks the AU-rich element (Fig. 9). When constructs were generated that transcribed the B6760 RNA with either 28 or 58 nt of additional 3Ј sequences, only the longer transcript was hyperedited at C 6802 . Thus, if the AU-rich element is not present, hyperediting at C 6802 requires that the 3Ј-flanking region include sequences 51-80 nt downstream (6853-6882) of the hyperedited site. DISCUSSION Previous studies indicated that several sequence elements are needed for efficient editing of apoB mRNA. For example, small apoB RNAs and chimeric RNAs with only 26 -63 nt of specific apoB sequence were edited inefficiently (11,14,15,28). Specifically, in the GC-rich apoE RNA, 63 nt of specific apoB mRNA were not edited (15). In contrast, when 354 nt of specific apoB sequence were translocated into the same locus of the apoE mRNA, wild-type levels of editing were detected, indicating that all sequence requirements for efficient editing at C 6666 are present within this segment (15). In this in vitro study, we have defined essential RNA sequence elements in the apoB mRNA necessary for its physiological editing and for its aberrant hyperediting. Physiological Editing of ApoB mRNA-We found that for normal physiological editing, an apoB RNA fragment of 139 nt consisting of the upstream 5Ј efficiency element (6609 -6628), an upstream 5Ј A-rich element (6629 -6640), the proximal efficiency sequence defined by Driscoll et al. (6648 -6661) (11), the mooring sequence (6671-6681) (10 -12), and the 3Ј efficiency element (6717-6747) were sufficient for editing with an efficiency equal to that of full-length apoB mRNA. We also found evidence that sequences even further downstream (6747-6824) contributed to editing at C 6666 (Fig. 10).
The importance of the 5Ј efficiency element for physiological editing was demonstrated by using short apoB RNA constructs in which the 5Ј efficiency element was present or absent (Fig. 2) and by testing chimeric apoB/N-myc RNAs (Fig. 3). In these chimeric RNAs, the presence of the apoB 5Ј efficiency element with the apoB editing cassette (6662-6687) was sufficient to give wild-type levels of editing at C 6666 . How the two elements increased editing of C 6666 to wild-type levels is unknown. One possibility is the formation of Watson-Crick base pairing between the two elements. However, according to RNA folding data obtained with the Mfold program (29,30), no secondary structures are favored between the 5Ј efficiency element and the editing cassette (data not shown).
Our data also indicate that a second apoB RNA element supported physiological editing of C 6666 . This A-rich element is located at position 6629 -6640, directly downstream of the 5Ј efficiency element. Deletion and mutagenesis studies in this A-rich element influenced editing at C 6666 severalfold (data not shown). Furthermore, a third proximal efficiency sequence at positions 6648 -6661 was reported to promote editing in combination with the mooring sequence (11).
Independent of other sequences, the 3Ј efficiency element of the apoB mRNA increased editing at C 6666 . A previous study indicated that 3Ј-flanking sequences of the apoB mRNA increase editing at C 6666 (14), but no efficiency element was defined. Our results indicate that the 3Ј efficiency element located within a 31-nt fragment (6717-6747) was necessary for efficient physiological editing in the apoB RNA (Fig. 1). It also increased editing in combination with the apoB editing cassette (Fig. 3), although to a lesser extent than did the 5Ј efficiency element. Furthermore, in the presence of the 5Ј efficiency element, the 3Ј efficiency element did not further increase editing (Fig. 3), indicating that the 5Ј efficiency element had a stronger impact on editing of C 6666 in small apoB RNA at least in vitro.
Physiological editing at C 6666 may also be enhanced by elements more than 81 nt from the editing site. Although these sequences were not required for efficient editing, the apoB RNA region encompassing C 6802 may function as a recognition element for the editing complex (Figs. 4 and 10). According to our two-step model for apoB mRNA editing, such recognition could bring the editing complex into closer proximity to the editing site at C 6666 (22). In a second step, the enzyme complex may then bind with a much higher affinity to the editing site at C 6666 than to the sequence elements that compose the binding site for the C 6802 (Fig. 5). RNAs were incubated for 16 h in an in vitro editing assay, and the editing of each cytidine was determined by primer extension analysis. B6687, which includes the mooring-like motif (striped bar), was used as parental apoB RNA for mutagenesis studies. B6687S has the mooring-like motif at 6809 -6819 scrambled (S). B6687⌬-16 lacks the last 16 nt of B6687, including the mooring-like motif. The numbers in the top part of the figure refer to the apoB cDNA position. The cytidines investigated for hyperediting are indicated below. Underlined letters indicate nucleotides that match the mooring sequence. B6687 was hyperedited at C 6802 (11.8 Ϯ 4%, n ϭ 3), C 6782/3 (2 Ϯ 0.1%, n ϭ 2), and C 6762 (19%). B6687S was hyperedited at C 6802 (3.9%), C 6782/3 (Ͻ1%), and C 6762 (9%). B6687⌬-16 was hyperedited at C 6782/3 (Ͻ1%) and C 6762 (15%).
FIG. 8. Scrambling the mooring-like motif at 6809 -6819 in the 78-nt apoB RNA B6747 abolished hyperediting at C 6802 . RNAs were incubated for 16 h in an in vitro editing assay, and the editing of C 6802 was determined by primer extension analysis (n ϭ 2). The smallest apoB RNA that was hyperedited at C 6802 , B6747, was used for a scrambling study of the mooring-like motif at 6809 -6819 (striped bar). B6747S has the 6 nt that match the mooring sequence scrambled (S). B6747M contains the exact 11-nt mooring sequence (dark bar), and B6747R has a reconstituted mooring-like motif in a different frame (R). Underlined letters indicate nucleotides that match the mooring sequence. The numbers in the top part of the figure refer to the apoB cDNA position. Hyperediting of ApoB mRNA-When APOBEC-1 is overexpressed, multiple cytidines are hyperedited (21)(22)(23). Detailed examination of the editing of one of these cytidines, C 6802 , in small apoB RNAs showed that sequences similar to those important in physiological apoB mRNA editing are also involved in the hyperediting of C 6802 . Hyperediting at C 6802 was supported by three RNA regions in the apoB mRNA: a proximal 5Ј-flanking region (6747-6759), a mooring-like motif (6809 -6819), and a 3Ј-flanking region (6853-6882) (Fig. 10).
The 5Ј-flanking region has a greater influence than the 3Јflanking region on the hyperediting of C 6802 . A proximal 5Јflanking region (6747-6759) and 5Ј sequences upstream from that element plus the mooring-like motif were all that was necessary for high level (11.8%) hyperediting of C 6802 (Fig. 5). Like the 5Ј A-rich element that enhanced physiological apoB mRNA editing, the proximal 5Ј-flanking region (6747-6759) that enhanced the hyperediting of C 6802 is AU-rich (Fig. 6). Three A 3 G substitutions in this element abolished hyperediting (Fig. 6). The importance of the 3Ј-flanking region for the hyperediting of C 6802 was demonstrated in the experiment shown in Fig. 9. An apoB RNA substrate lacking the proximal 5Ј-flanking region was not hyperedited at C 6802 until the 3Јflanking region (6853-6882) was added. Intriguingly, the minimal size of flanking regions required for hyperediting at C 6802 was comparable to that required for efficient physiological editing at C 6666 . Hyperediting and physiological editing required 55-57 nt of upstream flanking sequences and approximately 80 nt of downstream flanking sequences (Fig. 10). The similar size of the required flanking sequences and the importance of the mooring-like motif for hyperediting at C 6802 (Fig. 8) suggest that hyperediting depends on loose recognition of RNA features that define the physiological editing site at C 6666 .
Besides the flanking sequences, a mooring-like motif supports hyperediting at C 6802 . Several mutagenesis studies demonstrated that physiological apoB mRNA editing depends on the highly conserved mooring sequence and the spacer region (10 -13). Single substitution mutations in the mooring sequence either drastically decreased or abolished editing (10,12), and alterations in the 4-nt spacer decreased editing severalfold (10,13). However, the mooring-like motif downstream of C 6802 has only 6 of the 11 nt in the mooring sequence and has a spacer region of 6 rather than 4 nt. Furthermore, it consists of seven Us and four As and lacks the TGAT motif previously suggested to support hyperediting (23). Nevertheless, this mooring-like motif supported hyperediting of C 6802 , and scrambling the mooring-like pattern abolished hyperediting of C 6802 in short apoB RNAs and reduced hyperediting in longer apoB RNAs (Figs. 7 and 8). Furthermore, the introduction of an exact mooring motif (B6747M) increased hyperediting at this site 2-fold (Fig. 8). These results emphasize that the mooring-like pattern is essential for hyperediting at C 6802 and that not every AU-rich sequence supports hyperediting at C 6802 , although APOBEC-1 has been shown to bind to AU-rich RNA (31,32).
The mooring-like motif at C 6802 supports hyperediting of Cs located up to 27 nt upstream. When the mooring-like motif downstream of C 6802 was mutated, hyperediting of C 6802 was diminished or abolished ( Figs. 7 and 8). Furthermore, both mutations abolished hyperediting at C 6782/3 (Fig. 7), indicating that hyperediting of C 6782/3 depends also on the mooring-like motif located 26 -27 nt farther downstream. Hence, hyperediting has a relaxed spacer constraint compared with physiological editing, which was abolished by increasing the spacer to 12 nt (10,11,33).
Our in vitro study shows that the essential elements for physiological editing at C 6666 of the apoB mRNA encompass 139 nt, consisting of defined upstream efficiency elements, the mooring sequence, and the 3Ј efficiency element. Further investigations on the secondary structure of the apoB mRNA editing locus are needed to define the interaction of these elements. Our study also shows that similar RNA features are necessary for the hyperediting of cytidines when APOBEC-1 is overexpressed. FIG. 10. Schematic of elements that influence physiological editing at C 6666 and hyperediting at C 6802 . The apoB mRNA encompassing C 6666 and C 6802 is drawn linearly, and the positions according to the cDNA are given below. Elements and RNA regions that influence efficient physiological editing and hyperediting in vitro are marked (shaded bars). The distal downstream apoB region is marked as a hatched bar.