Tissue-specific inhibition of apolipoprotein B mRNA editing in the liver by adenovirus-mediated transfer of a dominant negative mutant APOBEC-1 leads to increased low density lipoprotein in mice.

APOBEC-1 is a catalytic subunit of an apolipoprotein B (apoB) mRNA editing enzyme complex. In humans it is expressed only in the intestine, whereas in mice it is expressed in both the liver and intestine. APOBEC-1 exists as a spontaneous homodimer (Lau, P. P., Zhu, H.-J., Baldini, A., Charnsangavej, C., and Chan, L. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 8522-8526). We tested the editing activity and dimerization potential of three different mouse APOBEC-1 mutants using in vitro editing activity assay and immunoprecipitation in the presence of epitope-tagged APOBEC-1. One catalytically inactive mutant, mu1 (H61K/C93S/C96S), that retains its capacity to dimerize with wild-type APOBEC-1 was found to inhibit the editing activity of the latter and was thus a dominant negative mutant. Two other inactive mutants that dimerized poorly with APOBEC-1 failed to inhibit its activity. Intravenous injection of a mu1 adenovirus, Admu1, in C57BL/6J mice in vivo resulted in liver-specific expression of mu1 mRNA. On days 4 and 9 after virus injection, endogenous hepatic apoB mRNA editing was 23.3 ± 5.0 and 36.8 ± 5.7%, respectively, compared with 65.3 ± 11 and 71.3 ± 5.2%, respectively, for luciferase adenovirus-treated animals. Plasma apoB-100 accounted for 95 and 93% of total plasma apoB in Admu1 animals on days 4 and 9, respectively, compared with 78 and 72% in luciferase adenovirus animals. Plasma cholesterol on day 9 was 98 ± 17 mg/dl in the mu1-treated animals, substantially higher than phosphate-buffered saline-treated (57 ± 9 mg/dl) or luciferase-treated (71 ± 12 mg/dl) controls. Fast protein liquid chromatography analysis of mouse plasma showed that the intermediate density/low density lipoprotein fractions in the animals treated with the dominant negative mutant adenovirus were much higher than those in controls. We conclude that active APOBEC-1 functions as a dimer and its activity is inhibited by a dominant negative mutant. Furthermore, apoB mRNA editing determines the availability of apoB-100, which in turn limits the amount of intermediate density/low density lipoprotein that can be formed in mice. Liver-specific inhibition of apoB mRNA editing is an important component of any strategy to enhance the value of mice as a model for human lipoprotein metabolism.

The mouse is a useful animal model for lipoprotein metabolism and atherosclerosis (1,2). Its value as a model for human disease, however, is limited by the fact that there is a substantial difference in lipoprotein metabolism between the two species. One major difference is the presence of high levels of apolipoprotein B (apoB) 1 mRNA editing in the liver in mice but not in humans.
ApoB mRNA editing is a process by which apoB-100 mRNA is converted to apoB-48 mRNA (3,4) (reviewed in Ref. 5). It involves the conversion of the first base of the codon CAA, encoding glutamine 2153 in apoB-100, to UAA, a stop codon, in apoB-48 mRNA. In humans, editing occurs exclusively in the small intestine but not in the liver. Therefore, the human liver produces apoB-100 and the small intestine produces apoB-48. In mice, apoB mRNA editing occurs in both the small intestine and the liver. Therefore, the amount of apoB-100 produced by the liver is very small, and mice have very low levels of circulating apoB-100; consequently, they have low levels of intermediate density lipoprotein (IDL) and low density lipoprotein (LDL), atherogenic lipoproteins that require apoB-100 as an essential component.
ApoB mRNA editing is mediated by a multiprotein enzyme complex. APOBEC-1 is a cytidine deaminase-like protein that has been identified as a catalytic component of the complex (6 -10) that efficiently edits synthetic apoB mRNA in vitro in the presence of complementation factors. Lau et al. (10) showed that APOBEC-1 exists as a spontaneous homodimer. It shows sequence similarity to Escherichia coli cytidine deaminase, which also exists as a homodimer (11). In the case of APO-BEC-1, Lau et al. (10) postulated that dimerization may be mediated by hydrophobic interactions in a leucine-rich domain in the C-terminal third of the molecule. It is not known, however, if the active form of APOBEC-1 is in the form of a homodimer.
The crystallographic structure of E. coli cytidine deaminase suggests that the active enzyme functions as a dimer (11). We reasoned that active APOBEC-1 might also function as a dimer and enzymatically inactive APOBEC-1 mutants (e.g. those that have substitutions in the zinc-coordinating residue in the active site (11)) might under appropriate conditions inactivate the wild-type enzyme by forming an inactive heterodimer with * This work was supported by National Institutes of Health Grants HL51586 and HL56668. 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. the latter. In contrast, mutants involving the leucine-rich domain have a high likelihood of having impaired dimerization potential (6,10). In this communication, we have examined the capacity of APOBEC-1 mutants to dimerize with wild-type APOBEC-1 and correlated this property with their ability to inhibit the editing activity of the latter. We then made use of the technique of adenovirus-mediated gene transfer to deliver a dominant negative mutant APOBEC-1 to the mouse liver in vivo. We found that mice treated with the dominant negative mutant APOBEC-1 developed significant hypercholesterolemia because of a marked increase in IDL/LDL cholesterol. Our observation is consistent with hepatic apoB mRNA editing being a limiting factor in the production of IDL/LDL in mice.
In Vitro Translation-RNAs were transcribed from cDNAs subcloned into pBKCMV and capped with T3 RNA polymerase using the mMES-SAGE mMACHINE kit (Ambion). They were translated in nucleasetreated rabbit reticulocyte lysate (Promega) in the presence of [ 3 H]leucine (Amersham Corp.) for 2 h at 30°C. Immunoprecipitation was performed as described (10). The immunoprecipitated products were separated on an SDS-12% polyacrylamide gel. Gels were fixed, soaked in Amplify (Amersham Corp.), and dried before they were exposed to x-ray film at Ϫ80°C. The relative intensities of the fluorographic images were determined with a photometric image scanner (BioImage).
In Vitro ApoB mRNA Editing Assay-The in vitro editing assay was carried out as described previously (12) using 8 fmol of synthetic substrate in the presence of chicken enterocyte S-100 extracts. Primer extension products were fractionated on a 6% polyacrylamide, 8 M urea gel and quantified by exposure to a phosphor screen (Molecular Dynamics).
Recombinant Adenoviral Vectors-The cDNA for mu1 was excised from the KS vector by BamHI/ClaI digestion and subcloned into the BglII/ClaI sites of pAvCvSv shuttle vector (13). The recombinant adenovirus was prepared by cotransfection of pAvCvSv containing mu1 cDNA and pJM17 (14) into 293 cells (15). Infection, propagation, screening, and large scale production of high titer recombinant adenovirus were carried out as described (13,16). AdLuc (17) contained luciferase cDNA instead of mu1 cDNA. Recombinant adenovirus stock was diluted with phosphate-buffered saline (PBS) to the appropriate concentrations, and 0.2 ml of diluted recombinant adenovirus (2 ϫ 10 9 plaqueforming units; particle/plaque-forming unit ratio ϳ30) was injected into the external jugular vein of 3-month-old (ϳ30 g) C57BL/6J mice that were maintained on laboratory chow (Teklad 4% mouse/rat diet 7001).
Northern Blot Analysis and Primary Extension Assay-Total cellular RNA was prepared by using the Ultraspec RNA Isolation System (Biotecx Laboratory). 20 g of RNA was electrophoresed on a 1.0% agarose, 6% formaldehyde gel, transferred to a Hybond-Nϩ membrane (Amersham Corp.), and hybridized to a 32 P-labeled 0.7-kilobase pair mu1 cDNA probe. The same membrane was then hybridized to a mouse glyceraldehyde-3-phosphate dehydrogenase cDNA probe (Ambion). Direct primer extension was carried out as described (16). A 32 P-endlabeled mouse antisense apoB DNA was annealed overnight at 45°C with 20 g of total RNA. The annealed products were precipitated with ethanol and reconstituted in 10 l of first strand synthesis buffer (Life Technologies, Inc.) containing 0.5 mM each dATP, dCTP, dTTP, 0.5 mM ddGTP, 10 mM dithiothreitol, and 20 units of Superscript II reverse transcriptase (Life Technologies, Inc.). The extension was carried out for 1 h at 45°C. The extension products were resolved by 6% polyacrylamide, 8 M urea gel electrophoresis, and the radiolabeled products were quantitated by PhosphorImager (Molecular Dynamics).
Plasma Lipoprotein and ApoB Analysis-Animals were fasted for 6 h and blood was collected into tubes containing EDTA. Plasma (0.2 ml) collected from individual mice was fractionated by fast protein liquid chromatography (FPLC) using two Superose 6 columns (Pharmacia Biotech, Inc.) connected in series (17). Fifty 0.5-ml fractions were collected using the eluent 1 mM EDTA, 154 mM NaCl, and 0.02% NaN 3 (pH 8.2) (18). Cholesterol and triglyceride contents of each FPLC fraction and of plasma were determined enzymatically (Sigma kits 352-50 and 336-10). For analysis of plasma apoBs, lipoproteins of d Ͻ 1.21 prepared by ultracentrifugation from plasma pooled from four individual animals were fractionated on 4 -15% SDS gel and stained by Coomassie Blue. The apoB-100 and apoB-48 bands were quantified by densitometer measurements (13).
Capacity of Mutant APOBEC-1s to Inhibit the Editing Activity of Wild-type APOBEC-1s-In order to study the effect of heterodimer formation on apoB mRNA editing activity, individual mutant and wild-type APOBEC-1s were cotranslated in vitro and the editing activity of the translation products was determined. By itself, mu1, in which the 3 zinc-coordinating residues were mutated, displayed no detectable editing activity in vitro. Increasing the ratio of mu1 mRNA in the presence of a constant amount of wild-type APOBEC-1 mRNA produced an APOBEC-1 product that had lost Ͼ95% of its activity in the in FIG. 1. Immunoprecipitation of APOBEC-1 and mutants cotranslated in vitro. Translation products before and after immunoprecipitation with anti-HA-1 antibody were analyzed as described under "Materials and Methods." wt, wild-type APOBEC-1; wt*, HA-1 epitope-tagged wild-type APOBEC-1; mu1, (H61K/C93S/C96S) mutant; mu2, APOBEC-1 1-172; mu3, (L182R/L187R) APOBEC-1. In each lane, the upper band represents the epitope-tagged APOBEC-1 and the lower band represents the untagged wild-type or mutant APOBEC-1. Since the antibody is only against the HA-1 epitope, the coprecipitation of the untagged smaller APOBEC-1 band indicates that the tagged and untagged proteins existed as heterodimers (10).
vitro editing assay (Fig. 2). Both mu2 and mu3 also had essentially no editing activity by themselves (Ͻ5% editing activity compared to that of wild type). The presence of increasing amounts of mu2 or mu3 translation products did not affect the editing activity of wild-type APOBEC-1 (Fig. 2). These results suggest that only a mutant that can efficiently dimerize with the wild-type APOBEC-1 is able to inhibit the editing activity of the latter. This supports the hypothesis that active APO-BEC-1 functions as a dimer.
Effect of mu1 Gene Transfer-Normal C57BL/6J mice were treated with an intravenous injection of Admu1, an adenoviral vector containing mu1, or AdLuc, an adenovirus containing luciferase cDNA. Liver and blood samples were obtained at days 4 and 9 after treatment. The highest level of mu1 mRNA was detected in the liver on day 4. It decreased substantially but was still readily detectable on day 9 (Fig. 3A). In contrast, mu1 mRNA was not detectable in the small intestine either on day 4 or day 9. This agrees with our previous experience (13,16) and that of others (reviewed in Ref. 19) that intravenous administration of adenoviral vectors in mice in vivo targets the transgene to the liver almost exclusively. To determine whether the Admu1 effectively inhibited endogenous apoB mRNA editing, we measured the relative amounts of apoB-100 and apoB-48 mRNAs by primer extension (Fig. 3B). On day 4 following adenovirus administration, the hepatic apoB mRNA from Admu1 animals contained 23.3 Ϯ 5.0% (n ϭ 4) and that from AdLuc animals contained 65.3 Ϯ 11% (n ϭ 4) edited mRNA. On day 9, the proportions of edited mRNA were 36.8 Ϯ 5.7% for Admu1 animals (n ϭ 3) and 71.3 Ϯ 5.2% for AdLuc animals (n ϭ 4). The ratio for PBS-treated animals was 69.3 Ϯ 4.3% (n ϭ 4). Therefore, Admu1 but not AdLuc administration markedly reduced the amount of edited apoB mRNA in the liver. The amount of edited apoB mRNA in the small intestine in Admu1-treated animals was 89.5 Ϯ 3.5% (n ϭ 4) and 88.4 Ϯ 2.2% (n ϭ 4) for days 4 and 9, respectively, compared with AdLuc values of 92.7 Ϯ 2.2% (n ϭ 4) and 94.7 Ϯ 0.48% (n ϭ 4) for these two days; it was 93.8 Ϯ 0.7% (n ϭ 4) for PBS-injected controls. Thus, adenovirus-mediated transfer of a dominant negative mutant APOBEC-1 in vivo selectively inhibited apoB mRNA editing in the liver. Furthermore, this maneuver resulted in a significant increase in apoB-100 level (15.2 Ϯ 4.0 mg/dl in Admu1, 2.84 Ϯ 0.5 mg/dl in Adluc, and 3.63 Ϯ 2.1 mg/dl in PBS animals, p Ͻ 0.0005, on day 9) in Admu1-treated animals.
Effect of Inhibiting Hepatic ApoB mRNA Editing on Plasma ApoB-100 and ApoB-48 and Plasma Lipids-The ratio of apoB-100 and apoB-48 protein was examined in mouse plasma by SDS gel analysis (Fig. 3C). Pooled plasma from four AdLuctreated mice on days 4 and 9 contained 78 and 72%, respectively, of the total apoB as apoB-100. The proportion of apoB-100 increased to 95% at day 4 and remained at 93% at day 9 in Admu1-treated animals.
There were significant differences in plasma lipids among the three groups following adenovirus treatment. In Admu1 animals, plasma cholesterol (day 9) and triglyceride (both day 4 and day 9) were almost double the values in the other two groups (Table I).
In order to determine which lipoprotein fractions were affected by Admu1 injection, plasma lipoproteins from day 9 samples were fractionated by FPLC and their cholesterol content determined (Fig. 4). It is evident that there is no significant difference in the very low density or high density lipoprotein fractions between Admu1-treated and the AdLuc-or PBStreated control samples. However, compared with controls the FIG. 2. Effect of the presence of excess mutant APOBEC-1s on editing activity of wild-type APOBEC-1. Capped wild-type APO-BEC-1 cRNA and mutant cRNAs were cotranslated in vitro using a rabbit reticulocyte lysate. Translation products were analyzed for apoB mRNA editing activity in vitro. ApoB RNA editing activity of wild-type APOBEC-1 in the absence of mutant APOBEC-1 was 74 Ϯ 4% (mean Ϯ S.D., n ϭ 4) of input RNA in this assay and this activity is set to 100%. All assays were done at a constant amount of wild-type APOBEC-1 in the presence of increasing amounts of mutant APOBEC-1s. IDL/LDL fractions were markedly elevated in the animals treated with the dominant negative mutant APOBEC-1 vector. SDS-polyacrylamide gel analysis indicates that the apoB-100 band was substantially higher in the Admu1 compared with the AdLuc1 samples (data not shown).
The experiments above lead us to the following conclusions. (i) Active APOBEC-1 works as a dimer and can be inhibited by a dominant negative mutant. Here we show that a mutant APOBEC-1 (H61C/C93S/C96S) with its zinc-coordinating residues mutated is totally inactive. This is consistent with previous observations (8,20,21) and with the catalytic role of the homologous residues in E. coli cytidine deaminase (11). The other two mutants, mu2 and mu3, which had mutations in the leucine-rich region, were also enzymatically inactive. The loss of catalytic activity in these mutants is probably related to their inability to efficiently form dimers, the active form of the enzyme. The fact that the mutations in these dimerizationincompetent APOBEC-1s involve residues in the leucine-rich domain indicates that this region is important for dimer formation. In any case, only mu1, which is competent in dimerization, acts as a dominant negative mutant when coexpressed with wild-type APOBEC-1 (Fig. 2). Therefore, these experiments support the contention that the active form of APO-BEC-1 is a homodimer (10).
(ii) Liver-specific inhibition of APOBEC-1 action leads to elevated IDL/LDL cholesterol. We hypothesized that the low IDL/LDL cholesterol observed in mice compared to humans is partly the result of the diversion of apoB mRNA to the production of predominantly apoB-48, severely limiting the amount of apoB-100, which is essential for the formation of IDL/LDL. Here we showed that hepatic APOBEC-1 action in mice was inhibited by adenovirus-mediated transfer of a dominant negative mutant. As a result of the decrease in apoB mRNA editing, the plasma apoB-100/apoB-48 ratio was greatly increased and the plasma IDL/LDL cholesterol was much higher in these animals than in the luciferase-or PBS-treated controls. We have repeated the adenovirus experiments and consistently observed the IDL/LDL elevation (data not shown). We conclude that the availability of apoB-100 (indirectly determined by apoB mRNA editing efficiency in the liver) limits the amount of IDL/LDL that can be formed in mice. It is interesting that in APOBEC-1 knockout mice that do not edit apoB mRNA either in the liver or small intestine there is little (22) or no change (23,24) in plasma IDL/LDL. The difference in phenotype between the animals reported in this study and the complete knockout mice may be related to the fact that 1) the liverspecific nature of APOBEC-1 inactivation and the normal editing in the small intestine may in some way contribute to enhanced very low density lipoprotein and subsequently IDL/ LDL production, and/or more likely 2) the acute inhibition of APOBEC-1 action by adenovirus-mediated gene transfer does not allow sufficient time for the animals to marshal a compensatory response that may play a role in minimizing the lipoprotein changes in the APOBEC-1 gene knockout animals. In any case, future experiments aimed at "humanizing" the mouse should include the permanent liver-specific inactivation of apoB mRNA editing, which would greatly enhance the value of this animal as a model for human lipoprotein metabolism and atherosclerosis.