A Gene-targeted Mouse Model for Familial Hypobetalipoproteinemia LOW LEVELS OF APOLIPOPROTEIN B mRNA IN ASSOCIATION WITH A NONSENSE MUTATION IN EXON 26 OF THE APOLIPOPROTEIN B GENE*

Familial hypobetalipoproteinemia, a syndrome characterized by abnormally low plasma levels of low density lipoprotein cholesterol, is caused by mutations in the apolipoprotein (apo) B gene that interfere with the synthesis of a full-length apoB100. In many cases of familial hypobetalipoproteinemia, nonsense or frameshift mutations result in the synthesis of a truncated apoB protein. To understand why these mutations result in low plasma cholesterol levels, we used gene targeting in mouse embryonic stem cells to introduce a nonsense mutation (N1785Stop) into exon 26 of the mouse Apob gene. The sole product of this mutant Apob allele was a truncated apoB, apoB39. Mice homozygous for this “apoB39-only” ( Apob 39 ) allele had low plasma levels of apoB39 and markedly reduced plasma levels of very low density lipoprotein and low density lipoprotein cholesterol when fed a high fat diet. Analysis of liver and intestinal RNA from heterozygous apoB39-only mice revealed that the Apob 39 mRNA levels were 60–70% lower than those from the wild-type allele. Interestingly, apoB39 was not cleared as rapidly from the plasma as apoB48. The apoB39-only mice provide new insights into the mechanisms of familial hypobetalipoproteinemia and the structural (TAA) and created a new Hin dIII site. To construct the targeting vector, a 6-kilo-base (kb) insert from the mutant 15A5 clone was excised with Sac I and Sal I and ligated into the Sal I and Sac I sites of p15A6. A neomycin resistance gene ( neo ) driven by the RNA polymerase II promoter was membrane, one would expect twice as many polyclonal antibodies to bind to apoB83 and expect the immunostaining to be twice as intense. According to this reasoning, the true ratio of apoB39 to apoB83 in the plasma of Apob 39/83 mice would probably be more than the 11:1 ratio that we observed with the polyclonal antiserum against mouse apoB100. In support of this reasoning, the apoB83 band in the plasma of Apob 39/83 mice was less intense (relative to apoB39) in Western blots performed with the antiserum against human apoB37 (data

Familial hypobetalipoproteinemia, a syndrome characterized by abnormally low plasma levels of low density lipoprotein cholesterol, is caused by mutations in the apolipoprotein (apo) B gene that interfere with the synthesis of a full-length apoB100. In many cases of familial hypobetalipoproteinemia, nonsense or frameshift mutations result in the synthesis of a truncated apoB protein. To understand why these mutations result in low plasma cholesterol levels, we used gene targeting in mouse embryonic stem cells to introduce a nonsense mutation (N1785Stop) into exon 26 of the mouse Apob gene. The sole product of this mutant Apob allele was a truncated apoB, apoB39. Mice homozygous for this "apoB39-only" (Apob 39 ) allele had low plasma levels of apoB39 and markedly reduced plasma levels of very low density lipoprotein and low density lipoprotein cholesterol when fed a high fat diet. Analysis of liver and intestinal RNA from heterozygous apoB39-only mice revealed that the Apob 39 mRNA levels were 60 -70% lower than those from the wild-type allele. Interestingly, apoB39 was not cleared as rapidly from the plasma as apoB48. The apoB39-only mice provide new insights into the mechanisms of familial hypobetalipoproteinemia and the structural features of apoB that are important for lipoprotein metabolism.
Familial hypobetalipoproteinemia (FH␤), 1 an autosomal codominant disorder characterized by low plasma cholesterol levels (1), is caused by mutations in the apolipoprotein (apo) B gene that interfere with the synthesis of a full-length apoB100. Most of the FH␤ mutations described to date have been nonsense or frameshift mutations within exon 26 of the apoB gene (1), which result in the synthesis of truncated forms of apoB. These truncated apoB proteins are invariably present in low concentrations in the plasma. Human FH␤ heterozygotes typically have moderately low levels of low density lipoprotein (LDL) cholesterol (Ͻ40 mg/dl) and are asymptomatic. FH␤ homozygotes have extremely low LDL cholesterol levels, typically less than 5 mg/dl, and a variable clinical phenotype, ranging from a complete absence of symptoms to a rather severe clinical phenotype with intestinal fat malabsorption, vitamin E deficiency, and neurological disease (1).
To understand metabolic mechanisms underlying FH␤, our laboratory has used gene targeting in mouse embryonic stem (ES) cells to introduce nonsense mutations into exon 26 of the mouse apoB gene (Apob). In a recent study, we reported the development of a mutant Apob allele (the "apoB83-only" allele, or Apob 83 allele) in which apoB100 amino acid 3798 was changed to a stop codon (2). The sole product of the Apob 83 allele was a truncated apoB, apoB83. The phenotype of heterozygous apoB83-only mice was strikingly similar to that of humans with FH␤ and an apoB83 mutation. In both human and mouse "apoB83 heterozygotes," the plasma concentrations of apoB83 were extremely low, and all of the apoB83 in the plasma was confined to the most buoyant very low density lipoprotein (VLDL) particles. An analysis of heterozygous apoB83-only mice revealed dual mechanisms for the "hypobeta" phenotype. First, compared with a wild-type Apob allele, the Apob 83 allele was associated with low levels of the apoB mRNA, which led to abnormally low amounts of apoB synthesis and secretion. Second, apoB83 was cleared from the plasma extremely rapidly, far more rapidly than the full-length apoB100.
In the current study, we sought to determine whether these dual "hypobeta" mechanisms (low apoB mRNA levels and rapid clearance of truncated apoB-containing lipoproteins) are specific to the Apob 83 mutation or are general features of all Apob nonsense mutations and all truncated apoBs. To investigate that issue, we have used gene-targeting techniques to introduce a nonsense mutation into another site within exon 26 of the mouse Apob gene, generating "apoB39-only" mice. We also sought to further define the impact of Apob gene mutations on mouse development. Homozygous apoB83-only mice had severe neurodevelopmental abnormalities and died during embryonic development or shortly after birth (2). We wanted to determine whether these lethal developmental abnormalities were a general feature of Apob mutations that result in the synthesis of a truncated apoB. 39 Allele-A gene-targeting vector ( Fig. 1) designed to generate apoB39-only mice was constructed from mouse Apob clones 15A5 and 15A6 (strain B10.A) (3). Clone 15A5 begins 55 base pairs (bp) past the beginning of exon 26 and extends to a SacI site located 6540 bp into exon 26. A nonsense mutation was introduced into clone 15A5 by site-directed mutagenesis using oligonucleotide B39stop (see Table I for sequences of all oligonucleotides). The mutagenesis reaction, which involved changing two nucleotides, converted codon 1785 (AAT, specifying Asn) to a stop codon (TAA) and created a new HindIII site. To construct the targeting vector, a 6-kilobase (kb) insert from the mutant 15A5 clone was excised with SacI and SalI and ligated into the SalI and SacI sites of p15A6. A neomycin resistance gene (neo) driven by the RNA polymerase II promoter was then inserted into the polylinker SalI site. The vector was linearized at a unique SauI site (608 bp into mouse exon 26) and electroporated into RF8 embryonic stem cells (strain 129/Sv) (4). Targeted clones were identified by Southern blot analysis of HindIII-digested genomic DNA, using a probe located 5Ј to the sequences in the gene-targeting vector (2). The apoB39-only gene-targeting vector was designed to insert a subtle mutation into the Apob gene without disrupting the structure of the Apob gene or the apoB transcript. The apoB39-only vector is identical to vectors used to insert other subtle mutations into the Apob gene (2, 3) (e.g., the apoB100-only mutation and the apoB83-only mutation) except for the location of the point mutation.

Generation of Mice with an Apob
Targeted ES cells were microinjected into C57BL/6 blastocysts to create male chimeras (5), which were bred with C57BL/6 females to generate heterozygous apoB39-only mice (Apob 39/ϩ ). The Apob 39/ϩ mice were intercrossed to create homozygotes and also were bred with "apoB100-only" (Apob 100/100 ) and "apoB48-only" (Apob 48/48 ) mice to generate Apob 39/100 and Apob 39/48 mice. In addition, Apob 39/39 mice were mated with Apob 83/100 mice to generate Apob 39/83 offspring. All mice were weaned at 21 days of age, housed in a barrier facility with a 12-h light/dark cycle, and fed either a chow diet containing 4.5% fat (Ralston Purina, St. Louis, MO) or a high fat diet containing 18.45% butter, 1.25% cholesterol, and 0.5% cholate (ICN Biomedicals, Aurora, OH). All of the mice used in this study were of a mixed genetic background (ϳ50% C57BL/6 and ϳ50% 129/Sv). 2 Lipoprotein Fractionation, Western Blot Detection of ApoB, and VLDL Sizing-The distribution of lipids and apoB within the lipoprotein fractions was assessed by fractionating plasma on a fast-performance liquid chromatography (FPLC) column (6). To visualize apoB proteins in mouse plasma or lipoprotein fractions, we performed Western blots of 4% polyacrylamide/SDS gels, using a rabbit antiserum specific for mouse apoB100 (6). In some experiments, Western blots were performed with a rabbit antiserum against a truncated human apoB protein, apoB37 (7,8). The diameters of VLDL particles from Apob 39/39 , Apob 48/48 , and Apob 100/100 mice were determined by electron microscopy of negatively stained fractions (9,10). For these experiments, VLDL fractions were prepared from mouse plasma by ultracentrifugation (11).
Primer Extension Analysis of the Relative Amounts of ApoB Transcripts Arising from Different Apob Alleles-The relative amounts of Apob transcripts from different Apob alleles were assessed with a primer extension assay similar to the one that we described previously (2). In the current experiments, we assessed the relative proportions of the two different Apob transcripts in the tissues from Apob 39/ϩ , Apob 39/100 , Apob 39/48 , and Apob 39/83 mice. To obtain the template for the primer extension assay, reverse transcription-polymerase chain reaction (RT-PCR) was used to amplify a segment of the apoB transcript. Total RNA from the mouse tissues was isolated with the RNA STAT-60 kit (Tel-Test "B", Inc., Friendswood, TX). Then 5.0 g of RNA was treated with 2.0 units of DNase I (RNase-free; Boehringer Mannheim) for 30 min at 37°C. After the DNase was inactivated by heating for 10 min at 95°C, the first strand of cDNA was synthesized with an RT-PCR kit (Stratagene) and a specific primer (B39cDNA) corresponding to sequences downstream from the apoB39 mutation, within exon 26 of the Apob gene. After the first strand of cDNA synthesis was completed, we used Taq polymerase (Boehringer Mannheim) and primers B39pe1 and B39pe2 to amplify a 1400-bp segment of the Apob cDNA. That segment extended from the last 26 bp of exon 25 through the first 1374 bp of exon 26. The corresponding fragment of genomic DNA is ϳ2 kb in length.
After removal of the unincorporated nucleotides from the enzymatically amplified cDNAs, a 32 P-end-labeled oligonucleotide (corresponding to sequences immediately downstream from the apoB39 nonsense mutation (B39pe3)) was annealed to the cDNA template and extended with avian myeloblastosis virus reverse transcriptase (Life Sciences, St. Petersburg, FL) in the presence of dideoxy-CTP (2). Because of the nonsense mutation in the Apob 39 allele, the primer extension product from the Apob 39 cDNA was 33 bp in length versus 30 bp with the cDNAs from the Apob ϩ , Apob 48 , Apob 83 , or Apob 100 alleles. The different-sized primer extension products were resolved by electrophoresis on 8% polyacrylamide gels containing 7.5 M urea. After electrophoresis, the gels were dried and exposed to autoradiographic film. The radioactivity in each primer extension product was also quantified with the AMBIS radioanalytic imaging system (AMBIS, San Diego, CA). The levels of Apob 39 transcripts were expressed as percentages of the transcripts produced by Apob ϩ , Apob 48 , Apob 83 , or Apob 100 alleles. Data from different animals of the same genotype were expressed as mean Ϯ S.E.
Isolation and Metabolic Labeling of Mouse Primary Hepatocytes-Primary hepatocytes were prepared from Apob 39/48 mice as described previously (2). To label newly synthesized proteins, hepatocytes were incubated with Pro-mix (530 MBq/ml; Amersham Pharmacia Biotech), an L-[ 35 S]methionine and L-[ 35 S]cysteine metabolic labeling solution. The relative amounts of apoB39 and apoB48 in cells were determined by immunoprecipitating the apoB with a mouse apoB-specific antiserum (6) and size-fractionating the immune complexes on a 4% polyacrylamide/SDS gel. The radioactivity in each apoB band was quantified with a phosphor imager (Fuji Bio-Imaging Analyzer, BAS 1000 with MacBAS; Fuji Medical Systems USA, Stamford, CT).
Inhibition of Lipoprotein Clearance in Apob 39/ϩ Mice with Triton WR-1339 -Apob 39/ϩ mice (n ϭ 3) were anesthetized with avertin (0.02 ml/g of body weight) and injected with Triton WR-1339 (500 mg/kg of body weight) (12,13), a detergent that blocks the processing and uptake of apoB-containing lipoproteins (13). Each mouse was bled from the retroorbital plexus before and at several time points after the injection of Triton WR-1339. The relative levels of apoB39, apoB48, and apoB100 were assessed by Western blots of 4% polyacrylamide/SDS gels, using a rabbit antiserum against mouse apoB (6).

Gene-targeted Mice That Synthesize
ApoB39 -To generate a mouse model of human FH␤ associated with a truncated apoB, we used a gene-targeting vector ( Fig. 1) to insert a subtle mutation, N1785Stop, into exon 26 of the mouse Apob gene. After electroporation of the gene-targeting vector into ES cells, 768 colonies were picked and screened by Southern blots with a 5Ј-flanking probe. Three targeted clones were identified. The targeting frequency of ϳ1 in 250 was significantly lower than the ϳ1 in 50 frequency that we consistently observed with vectors designed to generate apoB48-only, apoB100-only, or apoB83-only mice (2, 3). 3 High percentage male chimeras were 2 From the perspective of the relative levels of apoB proteins in the plasma, mouse strains have been shown to be important. In 1987, Lusis et al. (42) reported that different strains of mice have different "apoB100:apoB48 ratios" in their plasma and that this ratio also varies with different diets. In this study, we analyzed the apoB proteins in mice with a mixed genetic background and were therefore concerned that the mixed genetic background might influence the interpretation of our results. However, in littermates that had been fasted for 4 h, we never observed significant variations in the relative amounts of different apoB proteins in the plasma of different animals. In addition, all of the experiments described in this paper were performed on multiple animals of the same genotype. In addition, some of the most important analyses in this paper did not involve comparisons of different animals of the same genotype but instead involved the comparison of two different proteins or two different mRNAs within the same animal. indicating that about one-half of the Apob 39/39 mice did not survive. The number of homozygous offspring was significantly fewer than predicted (p Ͻ 0.05 by 2 ). Although most of the surviving Apob 39/39 mice were healthy and fertile, three were hydrocephalic. The latter finding strongly suggested that the lethality in the Apob 39/39 mice was a consequence of the same developmental abnormalities previously documented in apoBdeficient mice (2, 14 -17) or in mice lacking microsomal triglyceride transfer protein (18).
The Apob 39/ϩ mice were mated with Apob 48/48 mice and Apob 100/100 mice to generate Apob 39/48 and Apob 39/100 mice, respectively. ApoB39 was easily detectable in the plasma of fasted Apob 39/ϩ , Apob 39/48 , and Apob 39/100 mice (Fig. 2). In contrast, apoB83 was virtually undetectable in the plasma of fasted Apob 83/100 mice ( Fig. 2A). With the antiserum to mouse apoB100, the apoB39 band in Apob 39/100 mice was only slightly less intense than the apoB100 band ( Fig. 2A). However, because that antiserum was developed against the full-length apoB100, we suspected that it contained many more antibodies against apoB100 than against apoB39 and might therefore stain the apoB100 band more intensely than the apoB39 band. To obtain a more accurate view of the relative amounts of apoB39 and apoB100 in the plasma, we performed Western blots with a rabbit antiserum against the amino-terminal 37% of the human apoB molecule, reasoning that that antiserum would be more likely to stain apoB100 and apoB39 equally. With that antiserum, apoB39 was present in low concentrations in the plasma, relative to apoB100 (Fig. 2B). These results were further confirmed by Coomassie Blue-stained SDS-polyacrylamide gels. In the d Ͻ 1.21 g/ml lipoprotein fraction, the apoB100 band was 9-fold more intense than the apoB39 band, as judged by densitometry (data not shown).
Distribution of ApoB39 and ApoB100 within the Plasma Lipoproteins of Chow-fed Mice-To examine the distribution of apoB39 within the plasma lipoproteins, plasma samples from chow-fed Apob 39/39 (n ϭ 6) mice were pooled and size-fractionated on an FPLC column. Western blot analysis of the FPLC fractions revealed that apoB39 was distributed across a broad range of lipoprotein sizes from the VLDL (fractions 19 and 20) to the high density lipoproteins (HDL) (fractions 29 -34); the most intense bands were located in small LDL particles (fractions 25-28) (Fig. 3). In identical experiments performed with Apob 100/100 mice, the apoB100 peak has always eluted slightly earlier (fractions 23-26) (3). That discrepancy led us to suspect that apoB100-containing LDL were larger than apoB39-containing LDL. To directly compare the distribution of apoB100 and apoB39 within the LDL fractions, we size-fractionated pooled plasma from chow-fed Apob 39/100 mice (n ϭ 5) on an FPLC column and performed a Western blot analysis on fractions 23-32. These studies revealed that apoB100-containing particles were larger than apoB39-containing lipoproteins (Fig.  4). Once again, the most intense apoB39 band was located in fractions 27 and 28, whereas the most intense apoB100 band was located in larger particles, fractions 23-26 (Fig. 4). The fact that apoB39 was located in denser lipoproteins, compared with tors relates to the different distances between the targeted nonsense mutations and the site where the vectors were linearized. In the case of the apoB39-only targeting vector, that distance was 829 bp, whereas it was 6868 bp in the apoB83-only vector. A shorter distance between the site of linearization and the targeted mutation can increase the frequency of gene-conversion events (43). Gene-conversion events can eliminate the targeted mutation as well as the restriction site used to detect the gene-targeting event, thereby reducing the frequency of recognized gene-targeting events.  (2), the concentration of apoB83 in Apob 83/100 mice was so low that it was undetectable. In contrast, apoB39 was easily detectable in the plasma of Apob 39/100 mice. B, Western blot of plasma samples from Apob 39/100 mice (n ϭ 3) with the antiserum to human apoB37 (7,8). apoB100, was corroborated by Coomassie Blue-stained SDSpolyacrylamide gels. As judged by densitometry, the apoB100 band was 9-fold more intense in the d Ͻ 1.21 g/ml fraction but was 20 -30-fold more intense than the apoB39 band in the d Ͻ 1.70 g/ml fraction (data not shown).
Response of Apob 39/39 Mice to a High Fat Diet-In Apob 39/39 mice fed a high fat diet, little cholesterol accumulated in the VLDL-and LDL-sized lipoproteins (Fig. 5). In parallel experiments, we documented a striking accumulation of cholesterol in the VLDL/LDL fractions of Apob 48/48 mice when fed a high fat diet (Fig. 5). The size of the VLDL/LDL cholesterol peak in the Apob 48/48 mice was similar to that of Apob 100/100 and wildtype mice fed a high fat diet (3). We also analyzed the distribution of apoB proteins within the plasma lipoproteins of Apob 39/39 and Apob 39/100 mice fed the high fat diet. In the Apob 39/39 mice, apoB39 was widely distributed within the different lipoprotein fractions, but most of the apoB39 was in small LDL-sized lipoproteins (data not shown). In the Apob 39/100 mice, most of the apoB100 and apoB39 were in the LDL, with the apoB39 peak eluting on slightly smaller particles (Fig. 6).
Sizes of ApoB39-and ApoB100-containing VLDL-The FPLC fractionation of plasma from Apob 39/39 mice demonstrated that apoB39-containing LDL were smaller than apoB100-containing LDL. To determine whether the apoB39containing VLDL were also smaller, we isolated VLDL from chow-fed Apob 39/39 and Apob 100/100 mice and measured particle diameters by electron microscopy. The VLDL were significantly smaller in Apob 39/39 mice than in Apob 100/100 mice (mean of 33.9 Ϯ 5.9 versus 44.4 Ϯ 13.4 nm, p Ͻ 0.0001). This reduction in diameter corresponds to a ϳ55% reduction in particle volume. The electron micrographs also revealed that the VLDL from Apob 39/39 mice were homogeneous in size and almost entirely devoid of large lipoprotein particles. In the Apob 39/39 mice, more than 92% of the VLDL particles were between 22 and 39 nm. In contrast, only 50% of the VLDL particles in Apob 100/100 mice were in this size range. Less than 1% of the VLDL particles from Apob 39/39 mice were larger than 50 nm, whereas 24% of the particles in Apob 100/100 mice were larger than 50 nm. We also compared the sizes of VLDL from Apob 39 Although small, this reduction in diameter is predicted to correspond to a 45% reduction in particle volume. Virtually identical differences in VLDL size were observed in chow-fed Apob 39/39 and Apob 48/48 mice. With both diets, there were fewer large particles (Ͼ50 nm) in the Apob 39/39 mice than in Apob 48/48 mice.
The Apob 39 Allele Is Associated with Low Levels of ApoB mRNA-In a prior study (2), we demonstrated that the levels of apoB mRNA from an Apob 83 allele were 75-80% lower than those from Apob 100 or Apob ϩ alleles (2). To examine if the Apob 39 allele also was associated with low mRNA levels, we used primer extension assays to assess the relative amounts of Apob 39 and Apob 100 transcripts in the liver, intestine, and heart 4 of Apob 39/100 mice (n ϭ 3). Enzymatic amplification of the RNA yielded the expected "cDNA-sized" band of 1400 bp (and none of the ϳ2-kb "genomic DNA-sized" band) (Fig. 7A).
Primer extension experiments revealed that the steady-state levels of Apob 39 mRNA were 60 Ϯ 1, 66 Ϯ 3, and 71 Ϯ 6% lower in the liver, intestine, and heart, respectively, than the levels of the Apob 100 mRNA (Fig. 7B). In the livers of Apob 39/48 mice, there was an identical reduction in Apob 39 transcripts, relative to Apob 48 transcripts (Fig. 7C).
Synthesis and Secretion of ApoB39 -The decreased Apob 39 mRNA levels in Apob 39/48 mice suggested that apoB39 synthesis and secretion rates might also be low, relative to those for apoB48. To test this idea, we prepared primary hepatocytes from Apob 39/48 mice, incubated them with [ 35 S]methionine and [ 35 S]cysteine, and analyzed apoB39 and apoB48 secretion in metabolic labeling/immunoprecipitation experiments. As judged by phosphor imager analysis, the amount of apoB39 in Apob 39/48 hepatocyte lysates was reduced by 60%, and apoB39 secretion into the cell culture medium was reduced by 71 Ϯ 2%, compared with apoB48 (Fig. 8).
Generation and Analysis of Apob 39/83 Mice-To compare the steady-state levels of the Apob 39 and Apob 83 mRNAs and to gain insights into differences in the metabolism of apoB39-and apoB83-containing lipoproteins, we bred and analyzed Apob 39/83 mice. Since we detected an ϳ4:1 ratio of Apob 100 and Apob 83 transcripts in the livers of Apob 83/100 mice (2) and an ϳ2:1 ratio of Apob 100 and Apob 39 transcripts in Apob 39/100 mice (Fig. 7), we suspected that we would observe an ϳ2:1 ratio of the Apob 39 and Apob 83 transcripts in Apob 39/83 mice. This suspicion was confirmed; primer extension analysis of the liver RNA from an Apob 39/83 mouse revealed that the ratio of apoB39 mRNA to apoB83 mRNA ratio was ϳ1.8:1 (Fig. 9).
Although the ratio of Apob 39 transcripts to Apob 83 transcripts was ϳ1.8:1, the ratio of Apob 39 and Apob 83 transcripts in the plasma was quite different. As judged by Western blot analysis, the ratio of apoB39 to apoB83 in the plasma of Apob 39/83 mice was 11:1 (Fig. 10). Given the relatively modest difference in the amounts of Apob 39 and Apob 83 transcripts in the liver, we suspected that the marked discrepancy in the plasma levels of apoB39 and apoB100 was probably due, in large part, to differences in the rates at which these two proteins were cleared from the plasma. In our previous studies of Apob 83/100 mice, we demonstrated that apoB83-containing lipoproteins were cleared extremely rapidly from the plasma, accounting in large part for the high ratio of apoB100 to apoB83 in the plasma of Apob 83/100 mice (2). Accordingly, we suspected that the high ratio of apoB39 to apoB83 in Apob 39/83 mice was due to the fact that apoB83-containing lipoproteins are cleared from the plasma much more rapidly than apoB39-containing lipoproteins.
Analyzing ApoB Metabolism in Apob 39 A, RT-PCR reactions demonstrating the enzymatic amplification of a 1400-bp apoB cDNA fragment from the liver, intestine, and heart from three Apob 39/100 mice. The cDNA was amplified using Taq polymerase and oligonucleotides corresponding to sequences in exon 25 and exon 26 of the Apob gene. Amplification of genomic DNA with the same oligonucleotides yielded a ϳ2-kb DNA fragment. B, a primer extension assay was used to assess the relative proportions of Apob 39 and Apob 100 transcripts in the tissues from three Apob 39/100 mice, using enzymatically amplified cDNA fragments as templates (see "Materials and Methods"). With the Apob 100 allele (or with the Apob 48 or Apob ϩ alleles), the product of the primer extension assay was 33 bp; because of the targeted mutation, the product of the primer extension assay in the Apob 39 allele was 30 bp. C, primer extension assay performed with template cDNA amplified from hepatic RNA of an Apob 39/48 mice. Identical results were obtained with samples from two other Apob 39/48 mice. Also shown are primer extension experiments with template cDNAs amplified from the hepatic RNA of Apob 100/ϩ , Apob 39/39 , and Apob 39/ϩ mice. l) was size-fractionated on a 4% polyacrylamide/SDS gel; the apoB proteins were then analyzed with Western blots with an antiserum specific for mouse apoB100. By scanning densitometric analysis, the ratio of apoB39 to apoB83 band intensity was 11:1. Identical results were observed repeatedly in Apob 39/83 mice. Although the observed apoB39:apoB83 ratio was 11:1, the true ratio was probably even higher. The rabbit antiserum against mouse apoB100 would be expected to contain approximately twice as many antibodies against apoB83 than against apoB39 (simply because apoB83 is twice as large as apoB39 and would be expected to bind twice as many antibodies) (46). Thus, if equal numbers of apoB83 and apoB39 molecules were immobilized on a nitrocellulose membrane, one would expect twice as many polyclonal antibodies to bind to apoB83 and expect the immunostaining to be twice as intense. According to this reasoning, the true ratio of apoB39 to apoB83 in the plasma of Apob 39/83 mice would probably be more than the 11:1 ratio that we observed with the polyclonal antiserum against mouse apoB100. In support of this reasoning, the apoB83 band in the plasma of Apob 39/83 mice was less intense (relative to apoB39) in Western blots performed with the antiserum against human apoB37 (data not shown). metabolism of apoB39, relative to that of apoB48 and apoB100, we assessed the relative levels of apoB39, apoB48, and apoB100 in the plasma of Apob 39/ϩ mice, both at base line and at several time points after the injection of Triton WR-1339, a detergent that inhibits the lipolytic processing and clearance of apoB-containing lipoproteins. After injection of the detergent, apoB48 accumulated in the plasma, relative to apoB100 (Fig.  11). This finding is consistent with the data of Li and coworkers (13). The accumulation of apoB48 in the setting of diminished lipoprotein clearance indicates that, under normal circumstances (i.e. the absence of Triton WR-1339), apoB48 is cleared more rapidly than apoB100 (13). Interestingly, the ratio of apoB39 to apoB100 in the plasma of Apob 39/ϩ mice was not perturbed after the injection of Triton, indicating that, under normal circumstances, apoB39 and apoB100 are cleared at similar rates (Fig. 11). These results with apoB39 are quite different from those that we observed previously with apoB83 (2). After injection of Apob 83/100 mice with Triton, there was a striking accumulation of apoB83 relative to apoB100, indicating that apoB83 is normally cleared much more rapidly than apoB100.
Apob 39/39 Mice Do Not Exhibit "Fasting Chylomicronemia"-By plasma electrophoresis, the plasma of some humans with FH␤ contains chylomicron-sized particles, even after a prolonged fast ("fasting chylomicronemia") (1,19). Similarly, fasting chylomicronemia has been observed in the plasma of apoB70 mice (14), which have low apoB mRNA levels as a result of an insertional mutation in the Apob gene. In those mice, an examination of the VLDL by electron microscopy revealed increased numbers of large VLDL particles (Ͼ70 nm) (14). Fasting chylomicronemia is thought to be caused by the large lipoproteins that are synthesized in the setting of decreased apoB synthesis (i.e. reduced amounts of apoB synthesis in the setting of normal amounts of neutral lipid synthesis) (1). Of note, the Apob 39/39 mice showed no fasting chylomicronemia (data not shown) despite their low levels of apoB mRNA and decreased apoB synthesis rates. The absence of fasting chylomicronemia in the Apob 39/39 mice is consistent with the distinct absence of large VLDL particles as determined by electron microscopy.
Histology of Liver and Intestine from Apob 39/39 Mice-The reduced apoB secretion in the setting of the Apob 39 allele, along with the smaller size of apoB39-VLDL, led us to consider the possibility that the Apob 39/39 mice might develop fatty pathology in the intestine and liver (as a result of an abnormally low capacity to export neutral lipids). Oil Red O and hematoxylin and eosin staining showed no fatty pathology in the intestines or livers of chow-fed Apob 39/39 mice (data not shown). On the high fat diet, the Apob 39/39 mice developed fatty pathology in the liver, but that pathology is characteristic of the high fat diet in wild-type mice (20) and was also observed in Apob 48/48 and Apob 100/100 mice. There was no fatty pathology in the intestines of Apob 39/39 mice fed the high fat diet (data not shown). DISCUSSION The vast majority of the FH␤ mutations described to date result in the synthesis and secretion of a truncated form of apoB. While it is clear that the properties of the apoB gene mutations and the truncated apoB proteins must underlie the low plasma cholesterol levels in humans with FH␤, only limited information exists regarding the cellular and molecular mechanisms underlying this disorder (1). Gaining insights into basic mechanisms of FH␤ from human studies poses logistic and ethical difficulties because those experiments would require liver and intestinal biopsies from healthy and asymptomatic human subjects. In the current study, to further define the basic mechanisms of FH␤, we used gene targeting in mouse ES cells to introduce a nonsense mutation into exon 26 gene of the mouse Apob gene, generating an authentic FH␤ allele that yielded exclusively apoB39. The apoB39-only mice manifested several phenotypes characteristic of hypobetalipoproteinemia (1,16). First, the plasma concentration of apoB39 in Apob 39/100 mice was low, compared with that of apoB100, and second, the plasma of Apob 39/39 mice fed a high fat diet contained extremely low levels of VLDL/LDL cholesterol. There are several likely causes for the low plasma apoB39 levels and low plasma cholesterol levels in apoB39-only mice. One is that the apoB39 nonsense mutation causes a 60 -70% reduction in apoB mRNA levels, leading to a proportionate decrease in the synthesis and secretion of apoB-containing lipoproteins. Another factor, particularly in the setting of the high fat diet, might be a diminished capacity of apoB39 to assemble a lipoprotein with a large core of neutral lipids. Electron microscopy revealed the VLDL of Apob 39/39 mice were small (ϳ34 nm). That finding, along with the absence of fasting chylomicronemia, suggests that the apoB39 molecule may simply be too short to assemble very large, lipid-rich VLDL particles.
Low levels of RNA in association with nonsense mutations have been documented in several human genetic diseases (21)(22)(23)(24)(25). The extent to which nonsense mutations reduce mRNA levels can vary substantially, depending on the location of the mutation within the gene. For example, in the case of the ␤-globin gene, some nonsense mutations markedly reduce mRNA levels, whereas others have no effect on mRNA levels (22). We suspect that the same situation applies to the apoB gene. We previously introduced a nonsense mutation into a "physiologically normal" site in the mouse apoB gene, codon 2153, the same codon that is converted to a stop codon by the intestinal mRNA editing machinery (26,27). Of note, that mutation had no effect on apoB mRNA levels. More recently, we generated apoB83-only mice by introducing a nonsense mutation into codon 3798 and found that that mutation caused a 75-80% reduction in apoB mRNA levels. In the setting of the apoB83-only allele, the reduced mRNA levels likely reflected abnormal mRNA processing or stability, since the levels of apoB83 pre-mRNA were not decreased (2). In the current study, we documented a 60 -70% decrease in apoB mRNA levels in association with a nonsense mutation at codon 1785. The decrease in RNA was, however, smaller than with the apoB83 allele, since apoB39 transcripts were nearly twice as abundant as apoB83 transcripts in liver tissue from Apob 39/83 mice.
It is intriguing that a nonsense mutation in a "physiologically normal" site (i.e. codon 2153, the mRNA-editing codon) had no effect on mRNA levels, while mutations at codons 3798 and 1785 lowered mRNA levels. We suspect that the stability of apoB transcripts containing a nonsense mutation in codon 2153 has been "safeguarded" by millions of years of mammalian evolution, whereas this would not be the case for nonsense mutations at "abnormal" sites within the apoB gene. The clearance of lipoproteins from the plasma of Apob 39/ϩ mice was inhibited with an intravenous injection of Triton WR-1339. Blood was collected at base line and at several time points after the injection of Triton WR-1339. As judged by Western blot analysis, the intensity of apoB48 increased significantly, relative to apoB100. The intensity of apoB39 did not increase significantly, relative to apoB100. Results from a single Apob 39/ϩ mouse are shown. Identical results were observed in two other Apob 39/ϩ mice.
During the final phases of our investigation of the Apob 39 allele in mice, we fortuitously obtained a duodenal biopsy from a human "apoB37 heterozygote" (7, 28) (subject 18 from Ref. 28) during an elective cholecystectomy. Interestingly, primer extension studies of the duodenal RNA from that subject revealed that the apoB37 mRNA was reduced by 60%, relative to the apoB mRNA from the wild-type allele. 5 Those results are quite consistent with the results that we obtained with the Apob 39 allele in mice.
In Apob 83/100 mice, lipoproteins containing apoB83 were cleared far more rapidly from the plasma than apoB100 (2). The rapid clearance of apoB83-containing lipoproteins was demonstrated in experiments in which the uptake of lipoproteins in Apob 83/100 mice was blocked by an intravenous injection of the detergent Triton WR-1339. After the Triton injection, the amount of apoB83 in the plasma increased sharply, relative to apoB100. That result indicated that apoB83 is cleared much more rapidly than apoB100 under normal circumstances (i.e. in the absence of detergent). Other experiments indicated that the enhanced clearance of apoB83 was dependent both on the LDL receptor and on apoE (2). In generating the apoB39-only mice, one of our principal goals was to determine whether rapid clearance was a general property of all truncated apoBs or whether there were intrinsic differences in the metabolic properties of apoB39-and apoB83-containing lipoproteins. Several lines of evidence strongly favor the latter scenario. First, in the plasma of Apob 39/83 mice, the concentration of apoB39 was at least 11-fold greater than that of apoB83, although the amount of the apoB39 mRNA was less than 2-fold greater. Second, when we injected Triton WR-1339 into an Apob 39/ϩ mouse, apoB39 did not accumulate relative to apoB100. However, apoB48 did accumulate, relative to both apoB100 and apoB39, indicating that the apoB48-containing lipoproteins are normally cleared more rapidly than those containing either apoB39 or apoB100.
The portion of the apoB100 molecule that binds to the LDL receptor is located near amino acids 3300 -3500, far downstream from the carboxyl terminus of apoB48 (29 -31). Consequently, the clearance of apoB48-containing lipoproteins from the plasma is entirely dependent on the binding of apoE on the surface of the lipoprotein to either the LDL receptor or the LDL receptor-related protein (32)(33)(34). In the case of the LDL receptor-related protein, binding of lipoproteins appears to require that the surface of the lipoprotein be enriched with a supplemental "dose" of apoE (35,36). The fact that apoB48 accumulated relative to apoB39 in the Triton WR-1339 experiments suggests that apoE-mediated clearance may be less efficient for apoB39-containing lipoproteins. In view of the small size of apoB39-containing VLDL, it is tempting to speculate that apoB39-containing particles cannot accommodate as many molecules of apoE on their surface, compared with apoB48containing lipoproteins. Of course, apoB83 does contain the portion of the apoB molecule that interacts with the LDL receptor. Thus, we suspect that the extremely low levels of apoB83 in the plasma of Apob 39/83 mice relate in large part to the ability of apoB83 to bind to the LDL receptor (2).
ApoB39 was distributed widely within the VLDL, intermediate density lipoproteins, LDL, and HDL fractions of apoB39only mice, in a manner virtually identical to that described for similarly sized truncated apoBs in humans with FH␤ (1,37). Within the LDL fractions, apoB39-containing particles were smaller than those containing apoB100, presumably because of a decreased ability of apoB39 to bind lipids. The distribution of apoB39 within the plasma lipoproteins was markedly different from that of apoB83, which was confined to the largest particles within the VLDL fraction, both in humans and in mice (2,38). The absence of apoB83 from the LDL probably reflects the extraordinarily rapid clearance of apoB83-containing VLDL by lipoprotein receptors in the liver (2).
Approximately one-half of the predicted number of Apob 39/39 mice were viable, and about 10% of those developed hydrocephalus. This phenotype is similar to that reported for homozygous apoB70 mice, which have low apoB mRNA levels as a result of an insertional mutation within the Apob gene (14). The Apob 83/83 mice were more severely affected, all of them dying before birth or shortly after birth and the majority manifesting severe neurodevelopmental abnormalities. Farese et al. (39) reported data suggesting that the developmental abnormalities in apoB knockout mice were due to an absence of lipoprotein production by the visceral endoderm of the yolk sac, causing impaired delivery of lipid nutrients to the developing embryo. With that concept in mind, we believe that the most parsimonious explanation for the more severe developmental abnormalities in the Apob 83/83 mice (than in Apob 39/39 mice) is that the lower level of apoB83 mRNA resulted in a more severe nutritional deficit in the developing mouse embryos. However, other factors could contribute to the more severe phenotype in the Apob 83/83 mice. As noted earlier, apoB83 is taken up very rapidly by cellular lipoprotein receptors, far more rapidly than apoB39 or apoB100 (2). We can imagine that the peculiar properties of apoB83 might cause the yolk sac lipoproteins to be taken up and metabolized by the "wrong" cell types within the developing Apob 83/83 embryo, thereby preventing them from reaching the critical cell types within the developing central nervous system.
In the past, we (7,28,40) and others (41) have documented that truncated apoB proteins shorter than apoB48 or apoB100 can assemble large and buoyant lipoproteins. Those observations have occasionally led us to wonder why the extraordinary size of the apoB proteins has been conserved during tens of millions of years of mammalian evolution. In other words, why has evolution not "whittled away" at the length of the apoB molecule? The current studies suggest several likely explanations. First, the embryonic lethality in the homozygous apoB83only and apoB39-only mice suggests that nonsense mutations within the apoB transcript may not always be well tolerated, both from the standpoint of normal embryonic development and from the standpoint of maintaining normal levels of the apoB mRNA. Second, the apoB39-only mice have shown that whittling away 358 amino acids from apoB48 results in plasma lipoproteins that are small and have distinctly different metabolic properties. Given the central role of the apoB-containing lipoproteins in delivering lipids, antioxidant vitamins, and fuel to cells, we suspect that mutations affecting the intrinsic metabolic properties of lipoproteins might be poorly tolerated by evolution.