Identification of Novel Alternative Splice Variants of APOBEC-1 Complementation Factor with Different Capacities to Support Apolipoprotein B mRNA Editing*

Two novel mRNA transcripts have been identified that result from species- and tissue-specific, alternative polyadenylation and splicing of the pre-mRNA encoding the apolipoprotein B (apoB) editing catalytic subunit 1 (APOBEC-1) complementation factor (ACF) family of related proteins. The alternatively processed mRNAs encode 43- and 45-kDa proteins that are components of the previously identified p44 cluster of apoB RNA binding, editosomal proteins. Recombinant ACF45 displaced ACF64 and ACF43 in mooring sequence RNA binding but did not demonstrate strong binding to APOBEC-1. In contrast, ACF43 bound strongly to APOBEC-1 but demonstrated weak binding to mooring sequence RNA. Consequently ACF45/43 complemented APOBEC-1 in apoB mRNA editing with less efficiency than full-length ACF64. These data, together with the finding that all ACF variants were co-expressed in rat liver nuclei (the site of apoB mRNA editing), suggested that ACF variants might compete with one another for APOBEC-1 and apoB mRNA binding and thereby contribute to the regulation of apoB mRNA editing. In support for this hypothesis, the ratio of nuclear ACF65/64 to ACF45/43 decreased when hepatic editing was inhibited by fasting and increased when editing was re-stimulated by refeeding. These findings suggested a new model for the regulation of apoB mRNA editing in which the catalytic potential of editosomes is modulated at the level of their assembly by alterations in the relative abundance of multiple related RNA-binding auxiliary proteins and the expression level of APOBEC-1.

Alternative mRNA polyadenylation, splicing, and mRNA editing are mechanisms of post-transcriptional RNA processing that introduce diversity in mammalian mRNAs, thereby regulating the number of structurally and functionally different proteins encoded by a single gene (reviewed in Refs. 1 and 2). Base modification RNA editing may involve the enzymatic deamination of single nucleotides within splice site consensus motifs (3) or in protein coding sequence, often with physiologically significant effects (4). Adenosine to inosine editing of mRNAs, such as that occurring on mRNAs encoding the glutamate-gated calcium channel and serotonin 2C receptor subunits, is mediated by a family of adenosine deaminases active on RNAs that function independently of cofactors to deaminate their cognate target within double-stranded RNA (reviewed in Ref. 5). Multiple cytidine deaminases active on RNA have been predicted through their homology with the zinc-dependent cytidine deaminase domain of the mRNA editing enzyme APO-BEC-1 1 (Ref. 6 and references therein); however, only APO-BEC-1 (7,8) and CDD1 (9) have been shown to deaminate cytidines within mRNA. Unlike adenosine deaminases active on RNA, cytidine deaminases active on RNA minimally require RNA-binding auxiliary proteins for their activity on RNA.
Editosomes isolated from rat hepatic or intestinal cell extracts are heterogeneous in protein composition and have a complexity of 27 S as active complexes in the nucleus and as inactive 60 S complexes in the cytoplasm (16,17,27,28). apoB RNA-binding proteins of 100 and 55 kDa and a cluster of RNA-binding proteins of 42-48 kDa referred to as p44 (16,29) are recovered with affinity-purified editosomes assembled in situ (17). Additional proteins with affinity for APOBEC-1 and/or apoB mRNA have been identified with varying effects on apoB mRNA editing in transfected cells or in vitro (15, 30 -36). However, none could replace the obligate requirement for ACF65/64. The mechanism by which auxiliary proteins modulate editing activity is unknown but, in some instances, may involve both chaperoning interactions leading to the nuclear accumulation of editing activity as well as interactions within the editosome resulting in APOBEC-1 catalytic activity modulation (17,37).
In this report, two novel cDNAs that encode 43-and 45-kDa proteins are identified and shown to arise from alternative splicing of acf pre-mRNA. The proteins, henceforth designated as ACF43 and ACF45, contain the NH 2 -terminal portion of ACF64 including all three RRMs and bind to both APOBEC-1 and apoB mRNA. The role these proteins may have in regulating hepatic apoB mRNA editing has been addressed in a metabolic model system. In fasted rats, a marked reduction in apoB mRNA editing activity was accompanied by an alteration in the relative abundance of ACF65/64 to ACF45/43 in the nucleus that was reversible upon restimulation of editing activity by refeeding. Based on their in vitro functional characteristics, it is proposed that the relative abundance of nuclear ACF65/64, ACF45, and ACF43, together with changes in APOBEC-1 abundance, modulate editosome assembly and activity.

EXPERIMENTAL PROCEDURES
cDNA Isolation, RACE, and DNA Manipulations-Plaque hybridization screening of a gt11-based rat liver 5Ј stretch cDNA library (Stratagene) with radioactively labeled probes (RTS Rad-Prime system; Invitrogen) and 3Ј-RACE on Marathon ready rat liver cDNAs (Clontech) were performed according to the recommendations from the manufacturer. Rat genomic DNA (Clontech) was amplified using Expand High Fidelity PCR System (Roche Molecular Biochemicals). Subcloned inserts, 3Ј-RACE, and genomic DNA PCR products were sequenced using Big-Dye version 3.0 Terminator Cycle sequencing and the products analyzed in the Core Nucleic Acid Laboratory of the University of Rochester Medical Center. For Southern blot analysis, 10 g of restriction enzyme-digested rat genomic DNA (Clontech) was resolved through 1% agarose and transferred to nylon membrane (Stratagene). Blots were hybridized to a 32 Plabeled PCR-amplified exon 11A-specific probe in ExpressHyb (Clontech) at 60°C, washed, and autoradiographed. Reverse transcript (RT)-PCR was performed using avian myeloblastosis virus reverse transcriptase (Promega) to synthesize first-strand cDNA from oligo(dT)-primed total cellular RNA prior to amplification of specific transcripts with Taq DNA polymerase (Promega).
Antibody Generation and Immunoprecipitations-Polyclonal antibodies were raised in rabbits against peptides corresponding to amino acids 4 -18 (NH 2 -terminal) (28) and 508 -522 (COOH-terminal) of ACF64 and affinity-purified (Bethyl Laboratories). Rat liver nuclear extracts were UV cross-linked to apoB RNA (38) and the complexes immunoprecipitated with NH 2 -or COOH-terminal specific antibody at 4°C overnight. The complexes were bound to Protein A-agarose (Oncogene) for 2 h, washed extensively by vortexing in phosphate buffer containing 1 M NaCl and eluted with 3 M sodium thiocyanate, acetone-precipitated, resolved by 10.5% SDS-PAGE, and identified by autoradiography.
Rat Fasting/Refeeding Regime-Male Sprague-Dawley rats (250 -300 g; Taconic Farms) were housed in metabolic cages and control rats fed regular chow. Experimental rats were fasted for 48 h but allowed access to drinking water or fasted for 48 h and then fed on high sucrose diet (Low EFA Purified diet; Purina) for 48 h prior to sacrifice.
Cytoplasmic and Nuclear S100 Extract Preparation-Cytoplasm and purified nuclei were isolated from perfused rat livers as described previously (28,39).
Cell Culture and Transfection-Cell lines were obtained from ATCC (Manassas, VA) and maintained as recommended. Transfections were performed using FuGENE 6 (Roche Molecular Biochemicals), and total cellular RNA or protein was isolated as previously described (40). APO-BEC-1, ACF43, ACF45, and ACF64 proteins were expressed with HA and/or V5 epitope tags at their amino and carboxyl termini, respectively, from a modified pcDNAIII (Invitrogen). Drosophila S2 cells (Invitrogen) were maintained in Schneider's modified media (Invitrogen) as recommended. 3 ϫ 10 6 cells were transfected with a total of 2 g of DNA in six-well cluster plates using FuGENE 6 and harvested 48 h after DNA addition. PCR-amplified acf cDNA variants, apobec-1, and apoB DNA (nucleotides 6413-6802) were subcloned into pAc 5.1-V5-His (Invitrogen).
RNA Editing Assays-apoB RNA editing efficiency was determined upon RT-PCR-amplified transfected human apoB reporter RNA transcripts by poisoned primer extension analysis (40) and quantified by PhosphorImager densitometry.
Yeast Two-hybrid Analysis-cDNAs for acf43, acf45, acf64, and apobec-1 were subcloned into pAS2-1 and pACT2 (Matchmaker System; Clontech) and two-hybrid analyses conducted in yeast strain Y190. All yeast manipulations, ␤-galactosidase filter assays, and histidine prototrophy screens were performed according to the recommendations from the manufacturer.
Protein Expression and Purification-cDNAs for acf43, acf45, and acf64 were subcloned into a modified pET28a Escherichia coli expression vector (Novagen) encoding epitope tags and hexahistidine motifs such that proteins could be expressed under isopropyl-1-thio-␤-D-galactopyranoside induction and purified by nickel affinity chromatography (Qiagen) (41).
RNA Binding Assay-Ultraviolet cross-linking assays were performed as previously described using 498-nucleotide in vitro transcripts of apoB RNA (38), NF-1 RNA (42), or WT-1 (a control RNA lacking mooring sequence motifs) (17). For protein excess competition, purified recombinant ACF64, ACF45, and/or ACF43 were combined in editing buffer prior to the addition of radiolabeled apoB RNA and UV cross-linking.

Identification and Molecular Cloning of Rat acf45 and acf43
cDNAs-Screening of a rat liver cDNA library with a fulllength human acf cDNA together with 3Ј-RACE techniques identified two clones of 1394 and 1916 bp. Both were 100% identical to the 5Ј 1240 bp of full-length rat acf64 (28) but diverged at the 3Ј end, where both sequences were polyadenylated. The cDNAs encoded 405-and 383-amino acid proteins, respectively, for which the theoretical molecular masses were 45 and 43 kDa and which were therefore referred to as ACF45 and ACF43.
The possibility that acf45 and acf43 mRNAs resulted from alternative splicing of rat acf pre-mRNA was suggested by the recent finding of multiple alternatively spliced variants of human acf pre-mRNA (23,26). To investigate this, the genomic region encompassing exons 11 and 12 (by comparison to the human acf gene structure) was PCR-amplified, subcloned, and sequenced. BLAST analyses of the resultant 2277-bp fragment with the two cDNAs indicated the exon structure shown in Fig.  1 with the corresponding COOH-terminal amino acid sequences of the translation products indicated below. The exons are numbered with respect to the human acf gene, although it remains to be determined whether the two genes share identical exon structure 5Ј of this region.
acf45 cDNA appeared to be encoded by splicing of exon 11 to exon 11A and utilization of a polyadenylation signal in intron 11A; the stop codon for the resulting translation product being located in the 3Ј end of exon 11A (Fig. 1, Table I). acf43 cDNA appeared to be derived by inclusion of intron 11 and exon 11A and the utilization of the same polyadenylation sequence as acf45. Although acf43 mRNA is longer than acf45 mRNA, the inclusion of the intron introduced a stop codon within the 5Ј end of intron 11, thereby giving rise to the shorter translation product (Fig. 1, Table I). Rat acf65 and acf64 cDNAs appeared to result from exon 11A skipping and alternative splicing of exon 11 to include or exclude, respectively, the 5Ј-most 24 nucleotides of exon 12 in a manner analogous to human acf pre-mRNA ( Fig. 1, Table I, and Ref. 23). Each of the splice sites and the polyadenylation signal conformed to their respective consensus motifs (Table I) (43,44).
An alternative origin for acf45 and acf43 mRNAs would be if an acf gene duplication and modification had occurred. To investigate this possibility, rat genomic DNA was analyzed by Southern blotting (Fig. 2) using diagnostic restriction sites located within exons 11 and 11A and their flanking intronic sequences. An exon 11A-specific probe hybridized only to the predicted single EcoRI, StyI, and 0.56-kb BamHI fragments. Importantly, the probe hybridized only to the predicted 1.0and 0.77-kb fragments in the RsaI digest. The lack of probe hybridization to other restriction fragments indicated that there was only one gene that encodes acf65, acf64, acf45, and acf43.
acf65, acf64, acf45, and acf43 mRNAs Are Co-expressed in a Species-and Tissue-specific Manner-If ACF45 and ACF43 are translation products of alternatively spliced acf transcripts, then mRNAs encoding the identified cDNAs should exist. RNA species shorter than those encoding ACF64/65, however, were not detected in normal rat liver poly(A) ϩ RNA by Northern blot analysis (25,28). Similarly, acf43/45-specific RNAs were below the detection limits of ribonuclease protection assays. Therefore, reverse transcription polymerase chain reaction (RT-PCR) amplification of total cellular RNA from rat liver was performed using primers to co-amplify acf65 and acf64 (primers 64 and 65 in Fig. 1) or acf45 and acf43 (primers 43 and 45 in Fig.  1). Products of the predicted size for both acf45 and acf43 were observed (Fig. 3A, Liver ϩ lane). A transcript that includes intron 11 and is therefore represented by the larger PCR product encodes ACF43. ACF45 is encoded by splicing of intron 11,  so the smaller PCR product had to have arisen from RNA and not genomic DNA. Therefore, acf45 and acf43 transcripts were expressed in rat liver, and they were expressed simultaneously with acf64/65 as alternative spliced mRNA variants.
The acf gene in both human and rats encodes the alternative 5Ј end of exon 12 that is included in the alternative splicing of acf65 mRNA (23,26). However, exon 11A of the rat gene had no similar DNA sequence or amino acid coding potential within the equivalently sized intron 11 of the human gene (23). Therefore, the ability to carry out acf43/45 alternative polyadenylation and splicing was species-specific.
Total cellular RNA isolated from several rat tissues was evaluated for the alternative processing of acf64/65 and acf43/45 mRNAs using the discriminating primer pairs. All RNA variants were amplified from the same cDNA synthesis reaction; therefore, the identification of acf64/65 amplicons serves as the positive control for the cDNA synthesis reactions in which acf43/45 mRNAs were not detected. Consistent with previous findings (23)(24)(25), acf64/65 mRNAs were expressed in multiple tissues (Fig. 3B). However, acf43/45 mRNAs were only detected in liver and small intestine (Fig. 3A). These data demonstrate tissue-specific expression of acf43/45 mRNA in the two tissues that support apoB mRNA editing and, together with ACF43/45 protein binding to apoB mRNA (see Figs. 4 and 6; Refs. 16 and 45), further implicate a role for these proteins in apoB mRNA editing.
ACF45 and ACF43 Are Members of the p44 Cluster of Editosomal ApoB RNA-binding Proteins and Are Not Proteolytic Products of ACF64 -Translation of the alternatively spliced acf43 and acf45 mRNAs is predicted to produce two proteins that lack the COOH terminus of ACF65/64 but contain the amino-terminal three identified RRMs. Consequently, ACF45 and ACF43 are predicted to bind to apoB RNA and react with an ACF NH 2 -terminal specific antibody but not a COOH-terminal specific antibody.
UV cross-linking of rat liver nuclear extract to apoB RNA identifies ACF64/65 (formerly p66 (Ref. 28)) and the p44 cluster of proteins (Fig. 4). Immunoprecipitation of UV cross-linked nuclear extract with the ACF64 carboxyl-terminal (␣C) peptide-specific antibody resulted in a marked reduction in the recovery of the radiolabeled p44 cluster of proteins relative to the recovery of radiolabeled ACF64/65; i.e. only 9% of the total apoB RNA-binding protein was recovered within the p44 cluster compared with 60% recovery in the unfractionated nuclear extract. In contrast, the ACF amino-terminal (␣N) peptidespecific antibody yielded a similar recovery of cross-linked p44 (57%), as seen in the unfractionated nuclear extract. The data demonstrated that members of the p44 cluster of proteins have the amino-terminal, but not the carboxyl-terminal, domain of ACF64 and that the acf43 and acf45 mRNAs likely encode proteins in this profile. The recovery of low levels of p44 by the COOH-terminal antibody is most likely the result of the coimmunoprecipitation of ACF45/43 with ACF64/65 in the context of residual editosomal complexes that were not dissociated by the rigorous immunoprecipitation conditions.
It has been speculated previously that the p44 cluster of proteins with apoB RNA UV cross-linking activity (16,29) were proteolytic products of the p66 RNA-binding protein now identified as ACF64/65. In fact, motif analysis predicts potential thrombin recognition sites at amino acids 380 or 390. To evaluate potential proteolytic cleavage and to determine whether the rate of cleavage might be linked to the level of editing activity, ACF64 containing an NH 2 -terminal HA and COOHterminal V5 epitope tag was overexpressed alone in McArdle rat hepatoma cells or together with HA-tagged APOBEC-1 (which increased editing efficiency from wild type levels of 15% to Ն60% (Ref. 46)).
Western blots of whole cell extracts 48 h after transfection (Fig. 5) reacted with either anti-HA or anti-V5 antibodies demonstrated only full-length ACF64 and APOBEC-1 in the appropriate extracts. Proteolysis of ACF64 to ACF45/43 therefore was not significant. The possibility of a low level of ACF64 proteolysis or that proteolysis of both NH 2 -and COOH-terminal tags took place (thereby negating our ability to detect the cleavage product) cannot be ruled out. However, the role of proteolysis in generating ACF43/45 appears unlikely, given the transcription and translation of mRNAs encoding both proteins.
RNA Binding Activities of ACF45 and ACF43-The cDNAs encoded by the acf45 and acf43 alternatively spliced mRNA variants encode all three RRMs of ACF64/65. To directly evaluate RNA binding activity, recombinant His 6 -tagged proteins were expressed in E. coli, purified, and assayed by UV crosslinking. The analyses employed equivalent lengths of in vitro transcribed apoB, neurofibromin (NF-1), and NAT-1 RNAs, all three of which under appropriate conditions could be edited by APOBEC-1 in a mooring sequence-dependent manner (47,48). NF1 and NAT1 mRNA were evaluated as they deviate from the tripartite editing motif of apoB mRNA in enhancer, spacer, and mooring sequence elements such that they support reduced (NF1) and no (NAT1) editing activity in rat liver cells (42,47,49). Differences in editing activity on these mRNAs has been proposed to be the result of the favorable AT-rich context of the apoB mooring sequence-and tissue-specific differences in auxiliary proteins (45,49). An equivalent length of the Wilms' tumor mRNA (WT-1) that lacks mooring sequence similarities has been used in prior studies as a control for general or nonspecific RNA binding (17).  Fig. 1) were used in RT-PCR analyses to detect mRNA transcripts for the indicated acf splice variants. ϩ and Ϫ indicate the presence or absence, respectively of reverse transcriptase in the cDNA synthesis reaction. The lack of a product in the Ϫ lane, from which reverse transcriptase was omitted in the cDNA synthesis reaction, indicates that PCR products, especially acf43, were amplified from mRNA and not genomic DNA. Representative RT-PCR products were verified by DNA sequencing .   FIG. 4. acf43 and acf45 encode proteins that are members of the p44 apoB RNA binding cluster of proteins. Rat liver nuclear extracts were UV cross-linked to radiolabeled apoB RNA, treated with RNase and the complexes immunoprecipitated with ACF COOH-(␣-C) or NH 2 -terminal (␣-N) specific antibody at 4°C overnight. Complexes were bound to Protein A-agarose (Oncogene) for 2 h, resolved by 10.5% SDS-PAGE, identified by autoradiography, and quantified by Phospho-rImager scanning densitometry. NE is non-immunoprecipitated nuclear extract cross-linked to apoB RNA and represents a lighter exposure of the same gel than is shown for the immunoprecipitations.
Equal moles of each recombinant protein were combined with an equivalent specific activity of each radiolabeled RNA under in vitro editosome assembly conditions, UV cross-linked, RNase-digested, and analyzed by SDS-PAGE. ACF64 UV cross-linked to apoB and NF-1 RNAs with similar efficiency but did not bind well to NAT-1 or WT-1 RNA (Fig. 6A). The efficiency of ACF45 and ACF43 cross-linking to apoB mRNA was ϳ3-fold higher than that observed with ACF64. ACF45/43 demonstrated the greatest yield of cross-linked complexes with apoB mRNA but also bound with decreasing order of yield to NF-1 and NAT-1 RNAs and, to a limited extent, to WT-1 RNA. Therefore, all of the ACF protein variants had autonomous RNA binding ability. The differences in UV cross-linking efficiencies suggested that ACF45/43 had greater apoB-specific binding and a broader range of RNA binding activity than ACF64. The data also suggest that, although the mooring sequence is essential for ACF64/65 and ACF43/45 binding, flanking sequence context is also important.
The co-expression of ACF64, ACF45, and ACF43 in liver cells raises the possibility that the proteins may compete for apoB mRNA binding. Therefore, competition analyses were performed with pairs of recombinant proteins under the conditions of in vitro editosome assembly using apoB RNA UV crosslinking efficiency as the end point. ACF64 cross-linking was reduced ϳ60% by the addition of ACF45 (at a molar ratio of ACF45:ACF64 of 1:5) relative to that seen with ACF64 alone (Fig. 6B). Increasing the amount of ACF45 in the reaction resulted in increasing yields of ACF45 cross-linking to apoB RNA but virtually eliminated ACF64 cross-linking at molar ratios of ACF45:ACF64 of 1:1 and 5:1. These data suggested that ACF45 is an effective competitor with ACF64 for mooring sequence RNA binding.
In contrast, ACF43 cross-linking to apoB RNA was virtually undetectable at a molar ratio of ACF43:ACF64 of 1:5 (Fig. 6B). Increasing the amount of ACF43 in the reaction relative to ACF64 (molar ratios of ACF43:ACF64 or 1:1 and 5:1) increased ACF43 cross-linking efficiency but also enhanced ACF64 crosslinking to apoB RNA. The data suggested that ACF43 is not an effective competitor with ACF64 but in fact may, at an appropriate concentration, facilitate ACF64 binding to the mooring sequence. In control analyses, apoB RNA binding by any of the three recombinant proteins was not affected by a 10-fold molar excess of BSA, demonstrating that the observed effects of mixing ACF variants on their apoB RNA binding activity were specific (Fig. 6B).
The effect of ACF43/45 on ACF64 binding to apoB RNA raised the possibility that ACF45 and ACF43 may also affect each other's ability to bind to apoB RNA. To evaluate this, apoB RNA UV cross-linking competition analyses were performed employing recombinant ACF45 and ACF43. Titration of increasing amounts of ACF43 did not significantly affect crosslinking efficiency of ACF45 (at molar ratios of ACF43:ACF45 of 1:5, 1:1, and 5:1). ACF43 only demonstrated efficient crosslinking to apoB RNA at the highest concentration used, which was equivalent to that which resulted in significant UV crosslinking activity in the ACF64 competition studies (Fig. 6B) (molar ratio of ACF43:ACF45 of 5:1, Fig. 6C). These data suggest that ACF43 also could not compete with ACF45 for mooring sequence RNA binding.
Competition analyses in which ACF43 remained constant and ACF45 was titrated also suggested that ACF45 had higher affinity for apoB RNA (Fig. 6C). ACF43 UV cross-linking was reduced but detectable in reactions where the molar ratio of ACF43:ACF45 was 5:1. Increasing the molar ratio of ACF43: ACF45 from 1:1 to 1:5 resulted in increasing cross-linking of ACF45 but loss of detectable ACF43 cross-linking activity. Taken together, the data suggested that ACF protein variants have different affinities for apoB RNA and that their co-expression in liver (and intestine) is likely to affect the binding efficiency of each protein to the apoB mooring sequence and possibly affect the dynamics of editosome assembly and disassembly and thereby editing activity.
Interactions of ACF45 and ACF43 with APOBEC-1-The interaction of ACF64 with APOBEC-1 and complementation of apoB mRNA editing was observed with COOH-terminally truncated forms of the protein (22,50). Given that ACF43/45 comprise the NH 2 -terminal 379 amino acids of ACF64 plus unique COOH-terminal sequences, the ability of these proteins to bind APOBEC-1 was a possibility. To evaluate this, the ACF vari- ants and APOBEC-1 were co-expressed in yeast as bait and prey fusion proteins and both ␤-galactosidase activity and histidine prototrophy screens were used to evaluate potential interactions (Fig. 7).
In both assays, APOBEC-1 interacted strongly with ACF43 and ACF64 with growth and ␤-galactosidase activity comparable to the positive control interaction between p53 and SV40 large T antigen (Fig. 7). A positive but weaker interaction was observed in histidine prototrophy and, to a lesser extent, ␤-galactosidase activity for the interaction between ACF45 and APOBEC-1. There was no significant interaction between lamin C, a recognized control for false positive interactions, and APOBEC-1, ACF43, and ACF45 in either assay.
ACF45 or ACF43 Can Serve as the Minimal APOBEC-1 Complementation Factor-Given that ACF45 and ACF43 bind to both apoB RNA and to APOBEC-1, it was likely that they might alone be sufficient for complementing APOBEC-1 in apoB mRNA editing. Ideally, complementation would be performed under defined in vitro conditions in the complete absence of ACF64 of ACF65 using recombinant APOBEC-1. However, soluble APOBEC-1 could not be isolated from E. coli or yeast in high enough yield and in vitro translated APOBEC-1 proved to be contaminated with complementing activity from the reticulocyte lysate. Conducting these analyses in mammalian cells was also problematic, in that all mammalian cell lines tested, including HeLa, COS, and Chinese hamster ovary, expressed proteins that cross-reacted with the ACF NH 2 -terminal polyclonal antibody (28). More significantly, these cell lines supported low to moderate levels of editing when transfected with APOBEC-1 and apoB cDNAs, suggesting that complementation activity is present in a wide range of tissue types. This activity is probably caused by ACF64/65 expression that is detected in a broad tissue range (Fig. 3B and Ref. 25).
Drosophila S2 cells provided a null background for the complementation analysis in that they had no endogenous immuno-cross-reactivity to the ACF NH 2 -terminal polyclonal antibody (data not shown) and did not support editing when transfected with apobec-1 and apoB cDNAs (Fig. 8B). Expression of similar amounts of ACF43, ACF45, or ACF64 with similar amounts of APOBEC-1 (Fig. 8A) together with apoB RNA resulted in complementation in apoB RNA editing (Fig. 8B). These findings define the minimal editosome as comprising APOBEC-1 plus either ACF45 or ACF43. The data also suggest that ACF64 is more effective in complementing APOBEC-1 than ACF43, which in turn is more effective than ACF45.

The Expression of ACF Splice Variants Responds to Metabolic Perturbation-The rat metabolic model of fasting and
refeeding is an ideal model in which to study both increases and decreases of hepatic apoB mRNA editing in adult animals. In the metabolic model, fasting decreased hepatic apoB mRNA editing ϳ2-fold and refeeding with a low fat, high carbohydrate diet stimulated editing activity to levels higher than that measured in non-fasted rats fed normal chow (51) (Fig. 9A). Blood insulin levels change dramatically during this dietary regimen, and, in this regard, insulin alone stimulated apobec-1 mRNA expression and apoB mRNA editing activity (52,53). Insulin treatment of hepatocytes also increased the recovery of ACF64/65 with isolated nuclei (28), suggesting an increased affinity of these proteins for nuclear binding sites and/or a change in their nuclear/cytoplasmic equilibrium.
To evaluate possible regulation of auxiliary proteins in a whole animal metabolic model, rats were subjected to a fasting and/or refeeding regimen and equal amounts of hepatic cytoplasmic and nuclear S100 extracts were resolved by SDS-  7. Protein-protein interactions of ACF43, ACF45, ACF64, and APOBEC-1. Yeast two-hybrid analyses using both histidine prototrophy (His-) and ␤-galactosidase (␤-Gal) activity assays were used to assay for interactions between the indicated proteins. A representative clone of three separate transformants for each interaction is shown. A comparison of the histidine prototrophy assays performed in the presence of 10 and 100 mM 3-aminotriazole (3AT) indicated the relative affinities of APOBEC-1 with the ACF alternative splice variants. Western blot analysis of total yeast protein extracts verified similar expression levels of all ACF fusion proteins (data not shown). PAGE (Fig. 9B) and Western-blotted with ACF65/64 NH 2 -terminal peptide-specific polyclonal antibodies (28) (Fig. 9C). Consistent with previous analyses (28,54), ACF65/64 immunoreactivity was more abundant (on a per microgram of protein basis) in the nuclear extract of control rat liver compared with the cytoplasmic extract. In contrast, the amount of ACF65/64 in the cytoplasmic extract from fasted rat liver markedly increased, such that the signal was equal to or greater than that observed in the nuclear extract. Refeeding rats resulted in the re-establishment of ACF64/65 reactivity in nuclear and cytoplasmic extracts to levels similar to that observed in control rat liver. Taking into account the amount of protein analyzed and the total extract volumes, the total nuclear and cytoplasmic ACF64/65 immunoreactivity was calculated (28) to be approximately equal in all three metabolic states, suggesting little or no de novo synthesis of ACF64/65.
The alterations in ACF64/65 recovery in nuclear and cytoplasmic extracts associated with metabolic perturbation were paralleled in the recovery of their apoB RNA UV cross-linking activity (Fig. 9D). PhosphorImager scanning densitometry demonstrated that, relative to control rat liver extracts, ACF64/65 UV cross-linking activity increased 3-fold (Ϯ 0.6 S.E., n ϭ 3) in cytoplasmic extract from fasted rats commensurate with a 1.6-fold (Ϯ 0.17 S.E., n ϭ 3) decrease in the nuclear extract. Refeeding was associated with a recovery of cytoplasmic and nuclear ACF64/65 UV cross-linking activity, the abundance and relative distribution of which were similar to that seen in control liver extracts. These data suggested that metabolic regulation of editing activity was associated with alterations of the availability and/or activity of ACF64/65 in the nucleus for apoB RNA binding and editosome assembly.
Longer exposures of the ACF64/65 Western blots (Fig. 9C) revealed a profile of immunoreactive proteins ranging from 42 to 48 kDa that correspond to the p44 cluster. p44 immunoreactivity and the coincident apoB RNA UV cross-linking activities were only detected in nuclear extracts, with little or no enhanced recovery of these proteins apparent in the cytoplasmic extract from fasted rats. The data suggested that the relative abundance of members of the nuclear p44 cluster changed when rats were fasted (Fig. 9C), which was partially reversed upon refeeding. PhosphorImager scanning densitometry demonstrated similar levels of p44 apoB RNA UV crosslinking activity in hepatic nuclear extracts from control and refed rats but a 1.7-fold (Ϯ 0.23 S.E., n ϭ 3) increase in fasted rat extracts (Fig. 9D). These data suggested that decreased hepatic apoB mRNA editing in fasted rats was associated with an increase in nuclear ACF43/45 apoB mRNA binding activity at a time when there was a relative loss of nuclear ACF64/65 to the cytoplasm.
Expression of acf mRNA Splice Variants during Metabolic Regulation of ApoB mRNA Editing-The possibility that changes in the relative abundance of acf mRNA spliced variants contributed to observed differential expression of proteins in the p44 cluster was evaluated by semiquantitative RT-PCR on total liver RNA isolated from rats that were conditioned by fasting and refeeding. Initial calibration experiments (Fig.  10A) revealed that the ratio of PCR products within an amplification remained constant throughout the exponential phase FIG . 10. Expression of acf65, acf64, acf45, and acf43 mRNAs during metabolic perturbation. A, semiquantitative RT-PCR was calibrated by the amplification of different amounts of input cDNA (1 or 5 units) and the ratio of PCR products assayed throughout the amplification at the cycle numbers indicated across the top. PCR product ratios were determined from fluorescent intensities of multiple exposures and calculated using NIH Image version 1.63. All values were determined to be in the linear range by comparison to a DNA mass standard. The ratios were constant throughout the exponential phase (cycles 22-26) of the amplification and independent of the amount of input cDNA. The expression of hepatic acf64/65 mRNAs and acf43/45 mRNAs during the fasting and refeeding regime was determined using primer pairs 64/65 (panel B) and 43/45 (panel C) (see Fig. 1) using semiquantitative RT-PCR under the conditions determined in panel A. ϩ and Ϫ indicates the presence or absence, respectively, of reverse transcriptase in the cDNA synthesis reaction. The PCR product gels are representative data from a single rat for each metabolic state, whereas the ratios listed below the figure are an average from duplicate rats (B; standard deviations are as follows: control ϭ 0.032, fasted ϭ 0.027, and refed ϭ 0.011.) or single determinations from two separate rats (C).
FIG. 9. The relative abundance of hepatic of ACF protein variants is metabolically regulated. A, poisoned primer extension assays of RT-PCR-amplified hepatic apoB RNA from the indicated metabolically conditioned rat are shown. Products corresponding to unedited (CAA) and edited (UAA) apoB are indicated. Editing efficiencies, an average of duplicate determinations from separate rats, was determined by PhosphorImager densitometry as described in Fig. 8. Standard deviations were as follows: control (5.3), fasted (2.2), and refed (3.6). B, Coomassie Blue-stained SDS-PAGE upon which 25 g of protein from cytoplasmic (C) and nuclear (N) S100 liver extracts from rats conditioned as indicated were resolved. C, Western blot of gel as in B reacted with anti-ACF NH 2 -terminal peptide-specific polyclonal antibody (28) and developed using chemiluminescence. The panels for ACF64/65 and ACF43/45 are from the same blot at film exposure times of 5 and 30 s, respectively. D, ACF64/65 (p66) and ACF43/45 (p44) apoB RNA UV cross-linking activities in 25 g of protein from liver cytoplasmic and nuclear extracts of animals conditioned as indicated. of amplification (cycles [22][23][24][25][26] and was independent of the amount of input cDNA (1 or 5 l of a reverse transcription reaction). The ratio of acf65 to acf64 RT-PCR products remained relatively constant between treatment groups (Fig.  10B). The ratio of acf43 to acf45 RT-PCR products increased from ϳ1:1 in normal rat liver to ϳ3:1 in fasted rat liver (Fig.  10C). Refeeding fasted rats decreased the ratio of acf43 to acf45 RT-PCR products to below that in normal chow fed rats. These data are consistent with the possibility that regulated alternative splicing of acf pre-mRNA during fasting leads to a relative increase in the abundance of ACF43.

DISCUSSION
Two novel RNA-binding proteins of 45 and 43 kDa (ACF45 and ACF43) have been identified as expressed specifically in tissues that edit apoB mRNA and as each being sufficient in complementation of APOBEC-1 in apoB mRNA editing. These proteins together with ACF65 and ACF64 are expressed from alternatively spliced acf pre-mRNA. ACF45 and ACF43 are proposed to be members of a previously ill defined group of apoB RNA-binding proteins, formerly referred to as the p44 cluster that encompasses a molecular mass range of 42-48 kDa.
A Model for the Regulation of ApoB mRNA Editing-Kinetic analyses of editosome assembly suggested that both p44 and ACF65/64 (p66) are components of the apoB mRNA recognition complexes, assembly of which precedes formation of catalytically active 27 S editosomes (16,17,38). The interactions of ACF variants with apoB mRNA and APOBEC-1 described here support the recognition complex hypothesis (16) and implicate a role for ACF45/43 in regulating the protein composition of the editosome and its subsequent editing activity. The most salient functional characteristics of ACF variants suggested from the results were that: (i) all four variants were co-expressed in liver, (ii) editing activity was maximal in ACF64-APOBEC-1 editosomes, (iii) ACF64 and ACF43 had approximately equivalent abilities to interact with APOBEC-1 and that these interactions were stronger than those between ACF45 and APO-BEC-1, (iv) ACF43 could facilitate ACF64 binding to apoB mRNA (and to a lesser extent ACF45 binding to apoB mRNA), and (v) ACF45 interaction with apoB mRNA was stronger than that observed with either ACF64 or ACF43 and, in fact, ACF45 could displace ACF64/43 from apoB mRNA.
The metabolic study suggested that apoB mRNA editing is also regulated by reversible alterations in the relative abundance of nuclear ACF64/65 to ACF43/45. These alterations are proposed to affect the competitive binding equilibrium of ACF variants for APOBEC-1 and apoB RNA, thereby influencing which editosome assemblies predominate, and the level of editing activity in the hepatocyte. In this regard, the ACF NH 2terminal specific antibody was effective for assessing the relative abundance of ACF64/65 to ACF43/45 within subcellular fractions and for determining how that ratio was altered during metabolic regulation. It was apparent that, although ACF43/45 proteins were less abundant in nuclear extracts than ACF64/65 ( Fig. 9), their specific activity in binding to apoB mRNA (as determined by UV cross-linking efficiencies, Figs. 6 and 9) was equal to or greater than that observed for ACF64/65.
In this regard, ACF64/65 (p66) UV cross-linking in the nuclear extract from fasted rats was reduced compared with control rats in part as a result of the reduction in nuclear ACF64 abundance (because of its export to the cytoplasm or its failure to import to the nucleus). This may have established a competitive RNA binding advantage for ACF45 and a competitive APOBEC-1 binding advantage for ACF43. In fact, UV crosslinking activity of the p44 cluster increased. The Drosophila cell transfection data predict that APOBEC-ACF43 or APO-BEC-ACF45 editosomes have reduced editing efficiency. Moreover, both APOBEC-ACF43 and APOBEC-ACF64 complexes are not predicted to interact effectively with apoB mRNA because of competition from ACF45. Thus, the combined actions of ACF45 and ACF43 in the fasted state are proposed to favor net disassembly of ACF65/64-containing editosomes and to reduce editing efficiency.
Refeeding enhanced editing activity and was accompanied by a reversal of the trends in acf variant mRNA processing and ACF isoform subcellular distributions seen in fasted rat liver. The increased abundance of nuclear ACF65/64 may have provided these proteins with a competitive advantage over ACF43/45 for apoB mRNA binding and APOBEC-1 interaction. We therefore propose that, in addition to the insulin-dependent increase in APOBEC-1 abundance that occurs upon refeeding (52,53,55,56), apoB mRNA editing activity was stimulated in the livers of refed rats through the establishment of a new nuclear equilibrium of ACF variants that promoted assembly of more active editosomes.
APOBEC-1 and ACF65/64 contain nuclear localization signals (NLS) (17,37,(57)(58)(59). However, both proteins have nuclear and cytoplasmic distributions (28,57). Recent evidence is conflicted as to whether the NLS in APOBEC-1 (58) or ACF64/65 (59) determines the nuclear trafficking of each respective protein, as well as APOBEC-1-ACF64/65 multimers, all studies have suggested that co-expression of ACF64/65 and APOBEC-1 impacts the subcellular distribution of each other. In this regard, increased ACF64/65 abundance in the hepatocyte cytoplasm of the fasted rat may have facilitated nuclear import of APOBEC-1 subsequent to its elevated expression upon refeeding, leading to editing factor accumulation in the nucleus.
Proteins within the p44 cluster appeared to be differentially regulated, as the ACF NH 2 -terminal specific antibody immunoreactivity suggested an increased abundance of a select protein within the p44 cluster from liver nuclear extracts of fasted rats that decreased upon refeeding. Similarly, RT-PCR data demonstrated that the abundance of hepatic acf43 mRNA increased relative to that of acf45 mRNA in fasted animals and returned to normal hepatic levels upon refeeding. This suggested that the predominant immunoreactive band in the p44 cluster from nuclear extract of fasted rats corresponded to ACF43, although there were subtle changes in the relative abundance of other proteins within the p44 cluster as well as an increased UV cross-linking activity. If ACF43 expression was increased in fasting rat liver nuclei, the model predicts that this would reduce apoB mRNA editing because ACF43 would compete with ACF65 for APOBEC-1 binding but not bind efficiently to apoB mRNA in the presence of ACF64/65 and ACF45.
The p44 cluster has always appeared heterogeneous in electrophoretic mobility. UV cross-linking, and immunoreactivity with the ACF NH 2 -terminal specific antibody suggested that these proteins are all related. Consistent with the possibility of post-translational modification, there are several high probability phosphorylation sites predicted by NetPhos (www.cbs. dtu.dk/services/NetPhos) and Prosite (us.expasy.org/prosite) within the amino-terminal 379 amino acids of ACF. The COOH terminus of ACF43 contains no candidate post-translational modification sites, whereas ACF45 COOH terminus includes a potential protein kinase A site and two calcium/calmodulin kinase II sites. There are no other predicted motifs in either COOH-terminal extension. We cannot exclude the possibility that other alternatively processed acf pre-mRNAs encoded additional proteins in the p44 cluster.
Expression of ACF Variants-The relevance of ACF43/ACF45 to apoB mRNA editing is underscored by their restricted expression to tissues that supported apoB mRNA editing (liver and intestine). acf45/acf43 mRNA alternative splicing occurred in rats but not humans as a result of the absence of exon 11A in the human acf gene. This contrasts with the use of alternative 3Ј splice sites in exon 12 that generates acf65 or acf64 mRNAs in both rats and humans and was detectable in all tissues and cell lines examined. This supports the proposed regulatory role of ACF43/45 in apoB mRNA editing, but also suggests that ACF64/65 may have other cellular functions.
Human intestinal cells edit all of the apoB mRNA they express, whereas human liver does not edit apoB mRNA owing to a lack of APOBEC-1 expression (60). It is therefore conceivable that humans lost acf exon 11A because there is no editing to be regulated in liver but preserved the ability to express ACF64/65 to support highly efficient constitutive apoB mRNA editing in the intestine. Human acf gene encoded numerous alternatively spliced isoforms, most of which have as yet no known function, and the relative roles of ACF64 and ACF65 in editing remains unresolved as they supported editing to equivalent levels (23,26).
Alternative acf Pre-mRNA Processing-Recent studies implicated SRp40 in regulating alternative splicing of exon 12 to generate either acf64 or acf65 mRNA (23). acf45 and acf43 arose from alternative splicing of exon 11 to exon 11A or intron 11 retention, respectively, and polyadenylation at the end of exon 11A. Analysis of exon 11A (ESEFinder; www.cshl.org/ esefinder) suggested multiple high score SR protein binding motifs but unlike the specific effect of SRp40 on exon 12 alternative splicing, all SR proteins tested promoted acf43 mRNA splicing over acf45 mRNA (data not shown). It is possible that other factors contributed to the regulation of alternative splicing of acf45/acf43 mRNAs in the intact animal.
In this regard, the patterns of alternative splicing that generate the various ACF variants encompassed many known examples of alternative splicing including exon skipping (ACF64 or ACF65), alternative 3Ј splice site utilization (ACF64 versus ACF65), use of an alternative exon (ACF45), intron retention (ACF43), and alternative polyadenylation (ACF43 and ACF45) (reviewed in Ref. 61). Although common in viral transcripts, intron retention is rare in higher eukaryotes, but the utilization of a translation stop codon within a retained intron has been documented in the Drosophila P element wherein it regulates tissue-specific restriction of P element transposition (62).
Intronic elements and their cognate trans-acting factor can regulate tissue-specific or developmentally cued alternative splicing (63). These include ETR3, binding of which to conserved elements within introns flanking cardiac troponin T exon 5 results in its inclusion only in embryonic muscle and KSRP that directs inclusion of the N1 exon of the c-Src pre-mRNA in neural cell types (64,65). Indeed, KSRP has been postulated to modulate editing efficiency by interacting with ACF (24). Therefore, an intronic element, perhaps in concert with polypyrimidine tractbinding protein that represses certain exons (64, 66 -69), may have dictated the tissue-specific utilization of exon 11A and/or the retention of intron 11. Assuming the use of a common promoter for all acf splice variants, the relative abundance of each variant could be altered by a number of mechanisms including synthesis, phosphorylation, and/or change in subcellular localization of the factors responsible for directing splicing of each mRNA. Interestingly, physiological stimuli of apoB RNA editing, including insulin and ethanol, also affect alternative splicing of other transcripts (70).
Regulation of ApoB mRNA Editing through Auxiliary Proteins-Alterations in auxiliary protein activity have been pro-posed to mediate developmental (56,71), hormonal (72), and ethanol-regulated (73) apoB mRNA editing. Furthermore, a regulatory role for ACF64 as a molecular chaperone that increases the nuclear accumulation of editing activity or that facilitates interactions within the editosome has been suggested, based on its co-localization with APOBEC-1 in both the nucleus and cytoplasm (28,59).
Heterokaryon analyses (58,59) suggested that APOBEC-1 and ACF65/64 shuttle between the nucleus and cytoplasm, supporting the possibility that editing factors are in a dynamic equilibrium. Trafficking of ACF64-APOBEC-1 complexes between active nuclear 27 S complexes and inactive 60 S complexes would provide a rapid and reversible mechanism for regulating apoB mRNA editing that requires little or no de novo protein synthesis. In this regard, the regulation of apoB mRNA editing by ethanol is independent of continued protein or RNA synthesis (74). Alterations in the relative abundance of nuclear and cytoplasmic localized ACF64 have been documented for four metabolic conditions: insulin-or ethanoltreated rat primary hepatocytes (28), thyroid hormone administration (72), and in liver of fasted and refed rats (this report). It is conceivable that metabolic perturbation induced posttranslational modifications of ACF64 that changed its nuclear binding affinity, resulting in altered nuclear leaching during biochemical fractionation. The consistently high nuclear recovery of ACF45/43 argues against the trivial explanation of random loss of nuclear proteins during fractionation. Recent evidence indicated that the NLS of ACF 64/65 is confined to amino acids 360 -401 (59). ACF43/45 contain only the first 20 amino acids of the predicted NLS. The predominant nuclear localization of ACF43/45 suggests either that this portion of the NLS together with the additional amino acids unique to ACF43/45 were sufficient for nuclear localization and/or that the COOH terminus of ACF64/65 that is missing on ACF43/45 was necessary for shuttling and cytoplasmic retention. Additional complexity is derived from the fact that APOBEC-1 contains an NLS (58) and itself maybe capable of trafficking ACF64/65/43 or ACF45 to the nucleus under the appropriate conditions. Additional studies are clearly warranted.
Rethinking the Complexity of the C to U mRNA Editing Editosome-APOBEC-1 cannot edit apoB mRNA without binding to a mooring sequence-selective RNA-binding protein. The currently accepted minimal in vitro editosome consists of a homodimer of APOBEC-1 together with ACF65 or ACF64 (23)(24)(25). Our data suggest a new operational definition for the minimal editosome wherein either ACF45 or ACF43 are independently sufficient to complement APOBEC-1 in editing of apoB mRNA. These data corroborate mutational analyses of human ACF64 wherein amino acids 1-390 were sufficient for APOBEC-1 complementation, albeit at reduced efficiencies compared with full-length ACF64 (22,50). The in situ editosome, however, may be more complex, given four ACF variants that reside in the nucleus simultaneously and the numerous auxiliary proteins that have been implicated in the editing process as activators, inhibitors, and/or chaperones (Ref. 28 and references therein). In this regard, the macromolecular composition of the active nuclear editosome has an aggregate size of 27 S, whereas inactive complexes in the cytoplasm have a sedimentation of 60 S (28). Each of these biological isolates of editosomes was significantly more complex than any of the minimal editosomes described to date. The proposed role of ACF variants and APOBEC-1 abundance in regulating editing activity has implications for the mechanism of catalytic turnover of editosomes and may explain the co-expression of the four ACF variants in normal rat liver and intestine. We hasten to add that this is a working model, and, although it implies heterogeneity in the ACF variant composition of editosomes, we cannot rule out that individual editosomes contain more than one ACF variant. In summary our data suggest that the site-specific deamination of a single cytidine in apoB mRNA is regulated by the expression of ACF variants through alternative splicing and polyadenylation and that their activity in the assembly of active editosomes is regulated through changes in their relative abundance, which in turn determines their interaction with apoB mRNA and APOBEC-1 in the cell nucleus.