INTRODUCTION

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 APOBEC-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 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.

MATERIALS AND METHODS

Mouse APOBEC-1 mutants (mu1, H61K/C93S/C96S; mu2, APOBEC-1 1-172 that misses the C-terminal 57 residues containing the leucine-rich domain; mu3, L182R/L187R) were produced by the polymerase chain reaction (PCR). In brief, pBluescript II KS (Stratagene) containing the wild-type mouse APOBEC-1 cDNA (7) was amplified using the reverse primer (5-GGAAACAGCTATGACCATG-3), forward primer (5-GTAAAACGACGGCCAGT-3), and specific mutagenesis primers. The specific mutagenesis oligonucleotides used for PCR are listed below. The underlined nucleotides represent restriction sites for cloning purposes. The mutations are in boldface. H61K (5-GAAGTTGACTTCAACTTTGTTGCTGGTGTTTTG-3); C93S/C96S (5-GCCCTGGAGCTCTCCCCGCTGGGACTCCAG-3); L182R/L187R (5-CTGTATGTACTGGAGCGCTACTGCATCATTAGAGGACTTCCACCCTGT-3); mu2 was generated by PCR using forward primer and 3 downstream primer (5-CGCGGTCAATGGGGGTACCTTGGCCAAT-3); wild-type APOBEC-1 tagged with a 9-amino acid Haemophilus influenzae hemagglutinin epitope (HA-1) preceded by a 3-alanine residue spacer (10) was generated by PCR using forward primer and 3 downstream primer (5-CGCGGTCAAGCGTAATCTGGAACATCGTATGGGTAGGCTGCGGCTTTCAACCCTGTAGCCCAAAGG-3). After PCR, the products were purified by electrophoresis on 1.5% agarose gels followed by extraction using a QIAquick gel extraction kit (Qiagen). The purified products were digested with BamHI and EcoRI and subcloned into the BamHI/EcoRI sites of pBKCMV (Stratagene) or pBluescript KS. DNA sequences were verified by double-stranded DNA sequencing.

RNAs were transcribed from cDNAs subcloned into pBKCMV and capped with T3 RNA polymerase using the mMESSAGE mMACHINE kit (Ambion). They were translated in nuclease-treated rabbit reticulocyte lysate (Promega) in the presence of [³H]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 80 °C. The relative intensities of the fluorographic images were determined with a photometric image scanner (BioImage).

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).

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 × 10⁹ plaque-forming units; particle/plaque-forming unit ratio ~30) was injected into the external jugular vein of 3-month-old (~30 g) C57BL/6J mice that were maintained on laboratory chow (Teklad 4% mouse/rat diet 7001).

Total cellular RNA was prepared by using the Ultraspec RNA Isolation System (Biotecx Laboratory). 20 µg of RNA was electrophoresed on a 1.0% agarose, 6% formaldehyde gel, transferred to a Hybond-N+ membrane (Amersham Corp.), and hybridized to a ³²P-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 ³²P-end-labeled mouse antisense apoB DNA was annealed overnight at 45 °C with 20 µg of total RNA. The annealed products were precipitated with ethanol and reconstituted in 10 µl 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).

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% NaN₃ (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 < 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).

RESULTS AND DISCUSSION

We examined the capacity of various APOBEC-1 mutants to dimerize with epitope-tagged wild-type APOBEC-1 (10). mu1 (H61K/C93S/C96S) efficiently formed a heterodimer with epitope-tagged wild-type APOBEC-1 (Fig. 1, lane 4). However, a mutant that lacks the leucine-rich region (mu2) or one with mutations in this region (mu3) showed a markedly decreased ability to dimerize with epitope-tagged wild-type APOBEC-1 (Fig. 1, lanes 6 and 8). These results indicate that the leucine-rich region is important for dimer formation.

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 >95% of its activity in the in vitro editing assay (Fig. 2). Both mu2 and mu3 also had essentially no editing activity by themselves (<5% 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 APOBEC-1 functions as a dimer.

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 ± 5.0% (n = 4) and that from AdLuc animals contained 65.3 ± 11% (n = 4) edited mRNA. On day 9, the proportions of edited mRNA were 36.8 ± 5.7% for Admu1 animals (n = 3) and 71.3 ± 5.2% for AdLuc animals (n = 4). The ratio for PBS-treated animals was 69.3 ± 4.3% (n = 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 ± 3.5% (n = 4) and 88.4 ± 2.2% (n = 4) for days 4 and 9, respectively, compared with AdLuc values of 92.7 ± 2.2% (n = 4) and 94.7 ± 0.48% (n = 4) for these two days; it was 93.8 ± 0.7% (n = 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 ± 4.0 mg/dl in Admu1, 2.84 ± 0.5 mg/dl in Adluc, and 3.63 ± 2.1 mg/dl in PBS animals, p < 0.0005, on day 9) in Admu1-treated animals.

The ratio of apoB-100 and apoB-48 protein was examined in mouse plasma by SDS gel analysis (Fig. 3C). Pooled plasma from four AdLuc-treated 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).

Table I. Plasma lipids (mean ± S.D.) in adenovirus-treated and PBS-injected (control) mice Day 4 Day 9  Control (8)a Admu1 (15) AdLuc (14) Control (8) Admu1 (13) AdLuc (11) mg/dl mg/dl Cholesterol 61  ± 6 80  ± 17b,c 59  ± 6 57  ± 9 98  ± 17c,d 71  ± 12 Triglyceride 63  ± 20 143  ± 30c,d 93  ± 31 60  ± 8 131  ± 28d,e 104  ± 25f a  The number in parentheses indicates the number of animals. b  p < 0.005 versus control. c  p < 0.0005 versus AdLuc. d  p < 0.0001 versus control. e  p < 0.05 versus AdLuc. f  p < 0.005 versus control.

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 PBS-treated control samples. However, compared with controls the 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 dimerization-incompetent 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 APOBEC-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 liver-specific 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.