Doubling Expression of the Low Density Lipoprotein Receptor by Truncation of the 3 * -Untranslated Region Sequence Ameliorates Type III Hyperlipoproteinemia in Mice Expressing the Human ApoE2 Isoform*

The primary receptor mediating clearance of apolipoprotein (apo)E- and apoB100-containing lipoproteins from the circulation is the low density lipoprotein (LDL) receptor. Reduced expression of the LDLR is believed to be a precipitating factor in the pathogenesis of type III hyperlipoproteinemia (HLP) in some humans homozygous for the apoE2 allele ( APOE*2 ). To test the effect of genetic changes in LDL receptor expression on the pathogenesis of type III HLP, we have generated a var-iant allele at the endogenous mouse Ldlr locus that ex-presses the human LDL receptor transcript. Transcription of the human LDLR minigene is regulated by the endogenous mouse promoter sequence, but a truncation of 3 * -untranslated region results in increased mRNA stability. Consequently, in liver of heterozygotes, steady state levels of mouse and human LDLR transcripts are 50 and 180% the levels of total transcript in wild type mice, respectively. Overall, the 2.3-fold normal level of LDLR message in heterozygotes completely ameliorates type III HLP caused by the homozygosity for the human APOE*2 allele, normalizing their plasma lipoprotein profile. We conclude that a modest increase in expression of the LDLR through message

The primary receptor mediating clearance of apolipoprotein (apo)E-and apoB100-containing lipoproteins from the circulation is the low density lipoprotein (LDL) receptor. Reduced expression of the LDLR is believed to be a precipitating factor in the pathogenesis of type III hyperlipoproteinemia (HLP) in some humans homozygous for the apoE2 allele (APOE*2). To test the effect of genetic changes in LDL receptor expression on the pathogenesis of type III HLP, we have generated a variant allele at the endogenous mouse Ldlr locus that expresses the human LDL receptor transcript. Transcription of the human LDLR minigene is regulated by the endogenous mouse promoter sequence, but a truncation of 3-untranslated region results in increased mRNA stability. Consequently, in liver of heterozygotes, steady state levels of mouse and human LDLR transcripts are 50 and 180% the levels of total transcript in wild type mice, respectively. Overall, the 2.3-fold normal level of LDLR message in heterozygotes completely ameliorates type III HLP caused by the homozygosity for the human APOE*2 allele, normalizing their plasma lipoprotein profile. We conclude that a modest increase in expression of the LDLR through message stabilization is sufficient to prevent precipitation of type III HLP in mice.
Type III hyperlipoproteinemia (type III HLP) 1 is a disorder of lipoprotein metabolism that leads to elevated plasma cholesterol and triglyceride concentrations due mainly to an increase of apoB-containing remnant lipoproteins. These particles are the product of lipolytic processing of chylomicrons and very low density lipoproteins (VLDL) derived from the intestine and liver, respectively. Like low density lipoproteins (LDL), these are atherogenic lipoproteins, and subjects with type III hyperlipoproteinemia are predisposed to atherosclerosis and premature death from myocardial infarction (1).
Type III HLP is a genetic disease associated with the expression of a metabolically impaired apoE protein (2,3) or apoE deficiency (4). It is most commonly associated with individuals homozygous for the APOE*2 allele, the product of which has decreased affinity for the LDL receptor compared with other common isoforms. However, homozygosity of APOE*2 is necessary but not sufficient for the common form of type III hyperlipoproteinemia, as only 5-10% of adult homozygotes develop this disorder. In fact, most APOE*2 homozygotes have mild hypocholesterolemia and reduced atherosclerosis risk (5).
It has long been appreciated that other genetic and environmental factors besides possession of two APOE*2 alleles are required for development of type III HLP. Type III HLP rarely manifests before adulthood, is more prevalent in men than women, and has an earlier age of onset in men than women (6). Women tend to express type III hyperlipoproteinemia only after menopause. Earlier onset is also associated with obesity, excessive alcohol consumption, diabetes mellitus, and hypothyroidism (1). The mechanism by which these conditions induce type III HLP is unclear. One possibility may be that these conditions result in down-regulation of hepatic LDL receptor activity, leading to reduced uptake of atherogenic lipoproteins containing apoE from plasma. For instance, estrogen has been shown to increase hepatic LDL receptor activity (7). The decrease in plasma estrogen levels in postmenopausal women would therefore be thought to lead to decreased hepatic LDL receptor activity and increased circulation time of remnant lipoproteins.
Previously we have generated mice that express human apoE2 in a physiologically regulated manner by replacing the coding sequences of the endogenous mouse Apoe gene with the human APOE*2 allele (Apoe 2/2 mice). These mice exhibit full penetrance of the type III HLP phenotype, regardless of age or gender in the presence of normal murine LDL receptor expression (8). This suggested to us that genetic factors that trigger the type III HLP phenotype in some humans are already present in mice. One possible factor could be a relatively low hepatic level of the LDL receptor in mice. Others have developed transgenic mice expressing human apoE2 that exhibit features of type III HLP when endogenous apoE is absent (9,10). Unregulated overexpression of LDLR gene by adenovirus-mediated gene transfer has been shown to be effective in normalizing the plasma lipid profiles in these mutants (11). Here we report that moderate and controlled overexpression of the LDL receptor completely ameliorates the type III hyperlipoproteinemia phenotype of the Apoe 2/2 mice. This increase in LDLR expression was achieved through enhanced stability of the LDLR mRNA in mice engineered to express human LDL receptor in place of the mouse LDL receptor.

EXPERIMENTAL PROCEDURES
Construction of the Human LDLR Replacement Targeting Vector-As a 5Ј region of homology, a 7-kb fragment containing intron 1 of the mouse Ldlr gene was isolated from plasmid pSIT1. Plasmid pSIT1 was used for a targeted disruption of the murine Ldlr gene by Ishibashi et al. (12) and was kindly provided by Dr. Shun Ishibashi, University of Tokyo, Japan. A 12-kb SalI/NotI fragment containing the human LDLR minigene was isolated from plasmid pMY3, which was used for generating LDLR transgenic mice by Yokode et al. (13) and was kindly provided by Dr. Masahiro Yokode, University of Kyoto, Japan. The human minigene consists of a 7.5-kb fragment of a SalI site in intron 1 through an EcoRI site in exon 5 of the human genomic DNA, a 2.2-kb fragment of an EcoRI site in exon 5 through a SmaI site in exon 18 of the human LDLR cDNA (14), and a 0.7-kb fragment of human growth hormone poly(A) addition signal sequence. As a 3Ј region of homology, a 1.4-kb fragment from a SalI site in exon 4 through SacI site in intron 4 of the mouse gene was amplified from mouse genomic DNA by PCR using primers designed from published cDNA sequence (15). These fragments were inserted into a TKϪ Neo vector to yield the targeting construct shown in Fig. 1A below. The neomycin phosphotransferase (Neo) gene was placed in the same transcriptional orientation as the LDLR gene. The human minigene and the Neo gene replaces ϳ8 kb of DNA from an XhoI through a SalI site in exon 4 of the mouse Ldlr gene.
Replacement of the Murine Ldlr Gene with the Truncated Human LDLR Minigene-The targeting vector was linearized and introduced into embryonic stem cells, TC-1, and cultured under selection media as previously described (16). Cells resistant to both G418 and ganciclovir were clonally isolated and screened by PCR using a Neo-specific primer 5Ј-GCT TCC TCG TGC TTT ACG GT-3Ј for the targeting construct and a primer 5Ј-GCA AGA TGG CTC AGC AAG CA-3Ј corresponding to intron 4 sequence. Correct targeting was confirmed by Southern blot analysis using a probe derived from exon 4 of the mouse gene.
Chimeras were generated from targeted ES cells and bred with C57BL/6 mice to obtain germ line transmission of the modified chromosome. The genotype of the modified allele (h) in the animals was determined by the presence of a 300-bp PCR fragment produced by using the Neo-specific primer above and a 3Ј exon 4-specific primer, 5Ј-GCA GTG CTC CTC ATC TGA C-3Ј. The wild type mouse Ldlr allele (ϩ) was detected as a 380-bp PCR fragment produced by a 5Ј exon 4-specific primer 5Ј-CTC CCA GGA TGA CTT CCG AT-3Ј and the 3Ј exon 4-specific primer above.
Mice heterozygous for the targeted Ldlr gene (Ldlr h/ϩ ) were bred with Apoe 2/2 mice that were homozygous for targeted replacement of the mouse Apoe gene with the human APOE*2 gene (8). Doubly heterozygous mice were crossed to Apoe 2/2 mice again to generate mice homozygous for the human APOE*2 gene and heterozygous for the targeted Ldlr locus (Apoe 2/2 Ldlr h/ϩ mice). These mice were further crossed to Apoe 2/2 mice to generate littermates homozygous for the human APOE*2 gene with wild type mouse Ldlr (Apoe 2/2 Ldlr ϩ/ϩ mice) or homozygous for the human APOE*2 gene and heterozygous for the human LDLR gene (Apoe 2/2 Ldlr h/ϩ mice) for characterization. All mice used for characterization were a mix of strains 129 and C57BL6/J and were fasted for 4 h before plasma lipid analysis. Mice were fed normal chow (Prolab RMH 3000, number 5P76, St. Louis, MO) or western-style diet (0.15% (w/w) cholesterol and 21% (w/w) fat, TK 88137, Teklad Premier, Madison, WI) ad libitum.
Hepatic LDLR mRNA Analysis-Mice were sacrificed with an overdose of 2-2-2 tribromoethanol, and livers were harvested, flash frozen in liquid nitrogen, and stored at Ϫ70°C. RNA was prepared from the tissue using Trizol reagent following standard protocols (17). For Northern blot analysis, a 380-bp fragment of mouse exon 4 and a 350-bp fragment of human exon 4 were used as probes specific for the mouse and human genes, respectively. A primer extension assay was designed to quantitate both human and mouse message simultaneously using a primer 5Ј-GGA GCA CGT CTT GGG GGG ACA GCC T-3Ј corresponding to a shared sequence within exon 3 of the two LDLR genes. Dideoxy-CTP terminates extension of the primer, resulting in murine message extension by 5 bp and human message extension by 3 bp. The 30-and 28-bp fragments were separated in a denaturing 20% polyacrylamide gel. A primer 5Ј-GCA GCG AAC TTT ATT GAT GGT ATT 3Ј was included in each reaction to determine the expression of the glyceraldehyde-3-phosphate dehydrogenase (Gapdh) gene as an internal control. Primer extension analysis was performed on 50 g of total mRNA as previously described (18). Image densitometry was used to determine relative amounts of message (FLA-2000, Fuji Photo Film USA Inc., Elmsford, NY). To determine the decay rate of the LDLR mRNA, heterozygous animals were injected via tail vein with actinomycin D (150 g/100 g body weight) and ␣-amanitin (50 g/100 g body weight). Animals were sacrificed at various time points after injection, and liver mRNA was prepared using standard protocols for primer extension analysis.
Plasma Lipid Analysis-Plasma was isolated and total cholesterol, HDL cholesterol, and triglycerides were measured as described previously (19). Equal volumes of plasma from at least five sex-and agematched animals were pooled, and lipoproteins were separated by ultracentrifugation or by fast protein liquid chromatography (FPLC) as described (20). Lipoprotein fractions were dialyzed against phosphatebuffered saline, pH 7.4, and subjected to SDS-PAGE analysis as described (20).
In Vivo Clearance of VLDL and LDL-VLDL and LDL fractions were isolated from plasma of apoE-deficient mice (21) by ultracentrifugation for use as tracers for the turnover studies. The VLDL and LDL were radioiodinated with 125 I and 131 I, respectively, using previously published procedures (22,23). The external jugular vein of recipient mice was exposed and cannulated with a sterile tapered microrenathane catheter, and 3 Ci (specific activity ϳ10 5 cpm/g protein) of each radiolabeled ligand were injected. Immediately after dose injection, the catheter was removed, and the skin incision was closed using interrupted sutures. Blood samples were collected by retro-orbital sampling of anesthetized mice at various time points, and radiolabel remaining in the plasma was quantified with a gamma counter.

Replacement of the Mouse Ldlr Gene with the Truncated
Human LDLR Minigene-The targeting strategy used to disrupt the mouse Ldlr gene and replace it with exons 2-18 of the truncated human LDLR minigene is illustrated in Fig. 1. Homologous recombination between the endogenous locus ( Fig.  1A) and the targeting construct ( Fig. 1B) results in a hybrid gene in which the human LDLR minigene is expressed under the control of the endogenous promoter (Fig. 1C). All 5Ј-regulatory sequences and exon 1 of the endogenous mouse locus are intact, but ϳ8 kb of mouse DNA spanning 3Ј of intron 1 through exon 4 is replaced with the human LDLR minigene. The 3Ј-UTR of the minigene is shortened by a deletion of two of the three 3Ј "AU-rich elements" that destabilize the mRNA transcript (24,25). In addition, the human LDLR minigene contains the poly(A) addition signal sequences of the human growth hormone gene. Because exon 1 codes only for the signal peptide sequence, the mature protein transcribed from the chimeric gene is entirely human protein. The modified locus was transmitted to the F1 generation from chimeras that were made from one of the targeted ES cell lines. All F1 heterozygote matings produced normal litter sizes with a normal Mendelian segregation pattern of the modified locus. We designate the modified allele as hLDLR (or h) to distinguish it from the wild type mouse Ldlr allele (or ϩ).
Increased Steady State LDLR mRNA Levels by Enhanced Message Stability in Heterozygous Mice-The expression of the LDLR was confirmed by Northern blot analysis of mRNA with a probe specific for the human LDLR mRNA. The human message of ϳ3 kb was present only in animals heterozygous for the hLDLR (Ldlr h/ϩ ) with no band detected in wild type mice (Ldlr ϩ/ϩ ) ( Fig. 2A, upper band). The presence of a 4.5-kb mouse Ldlr message was confirmed in both Ldlr ϩ/ϩ and Ldlr h/ϩ mice by hybridizing the same blot with a probe specific for the mouse message ( Fig. 2A, lower panel). The Ldlr h/ϩ mice had ϳ50% of wild type mouse Ldlr message levels, in accordance with the loss of one copy of the mouse Ldlr gene.
To estimate simultaneously the amount of both human and mouse messages in homozygotes, we devised a primer exten-sion strategy utilizing a primer that hybridizes to a sequence in exon 3 shared in both the human and mouse LDLR genes (Fig.  2B). The murine message is extended 5 base pairs, and the human message is extended 3 base pairs before termination with dideoxy-CTP. In heterozygotes, the steady state level of the human LDLR transcript is 3.8 Ϯ 0.3 (n ϭ 18) times the level of the mouse transcript (p Ͻ 0.001). Since the levels of mouse Ldlr message in the Ldlr h/ϩ mice are one-half that in wild type mice, total message (mouse ϩ human) is increased 2.3-fold (Fig. 2C). There is no down-regulation of the mouse message in response to the higher levels of human message, nor adjustment of total mRNA levels in the Ldlr h/ϩ mice, despite the fact that transcriptional regulatory elements in the LDLR gene are intact. The steady state human message is reduced to 68% of the normal chow level in response to dietary cholesterol loading (Fig. 2C) indicating that the sterol-responsive transcriptional regulation of the receptor is intact and functional (26).
To determine if the increased steady state level of the hLDLR mRNA is due to message stabilization, actinomycin D and ␣-amanitin were administered in vivo to Ldlr h/ϩ mice. These drugs inhibit transcription, allowing for the measurement of mRNA decay rates (27). Differentiation between the mouse and human LDLR messages by primer extension was then used to determine relative decay rates of these transcripts in the liver (Fig. 3A). The time course analysis (Fig. 3B) shows that the human message decayed with a half-life of about 160 min while the mouse message decayed with a half-life of 60 min. This confirms that loss of two AREs in the human LDLR transcript results in a large increase in message stability in vivo in the liver of mice.
Reduction of Plasma Lipids in Mice Expressing hLDLR-Both heterozygous (Ldlr h/ϩ ) and homozygous (Ldlr h/h ) female mice expressing the hLDLR exhibited approximately a 4-fold reduction in steady state plasma cholesterol (mainly HDL cholesterol) when fed normal chow (21 Ϯ 3 mg/dl in Ldlr h/ϩ , n ϭ 10; 28 Ϯ 8 mg/dl in Ldlr h/h , n ϭ 8, compared with 85 Ϯ 9 mg/dl in Ldlr ϩ/ϩ , n ϭ 12 p Ͻ 0.00001). The reduction in plasma cholesterol due to expression of the stabilized hLDLR transcript can be viewed as a dominant trait, as there is little difference in plasma cholesterol between heterozygotes and homozygotes for the hLDLR. Consequently, all the remaining characterization was with heterozygous mice and their wild type littermates. There was a trend toward reduction of plasma triglycerides in Ldlr h/ϩ mice (29 Ϯ 6 mg/dl, n ϭ 10) compared with wild type mice (35 Ϯ 7 mg/dl, n ϭ 12), but this reduction was not significant (p ϭ 0.23). Both agarose gel electrophoresis (Fig. 4A) and FPLC analysis (Fig. 4B) of plasma showed pat-  terns consistent with a reduction of plasma lipid and HDL cholesterol. In the Ldlr h/ϩ mice, both ␤and pre-␤-migrating particles, as well as ␣-migrating particles, were dramatically reduced. SDS-PAGE analysis of plasma lipoproteins isolated by sequential density ultracentrifugation showed that the reduction in plasma lipid was accompanied by reductions in plasma apoproteins B, E, and AI (Fig. 4C). Although apoB100 is present in IDL-LDL fractions (p ϭ 1.04 -1.06 g/ml) of the wild type mice, it was virtually absent in the IDL-LDL fractions of the Ldlr h/ϩ mice (the protein band corresponding to apoB100 was slightly degraded in this example). Plasma apoE was also reduced in all fractions of plasma from Ldlr h/ϩ mice, with only a small amount of apoE remaining in the largest, lowest density particles (VLDL in fraction p ϭ 1.006 g/ml).
When fed a diet containing 0.15% (w/w) cholesterol and 21% (w/w) fat, plasma cholesterol levels in Ldlr ϩ/ϩ and Ldlr h/ϩ mice increased to 128 Ϯ 10 and 81 Ϯ 6 mg/dl, respectively. This increase was due mainly to the increase in HDL cholesterol, as shown by FPLC analysis of plasma from both groups of mice (Fig. 4D). This suggests that both wild type and heterozygous mice respond similarly to dietary cholesterol overload.
Increased LDLR Level Ameliorates Type III Hyperlipoproteinemia in Apoe 2/2 Ldlr h/ϩ Mice-When a copy of the hLDLR allele was introduced into the mice expressing human APOE*2 in place of mouse Apoe, these mice (Apoe 2/2 Ldlr h/ϩ ) exhibited a 2.5-fold increase in steady state total LDL receptor mRNA levels when compared with Apoe 2/2 mice (data not shown). The ratios between the mouse Ldlr and hLDLR mRNA levels were similar in Apoe ϩ/ϩ Ldlr h/ϩ and Apoe 2/2 Ldlr h/ϩ mice.
The Apoe 2/2 Ldlr ϩ/ϩ mice maintained on normal chow had type III HLP, with plasma cholesterol levels of 268 Ϯ 12 mg/dl and triglycerides of 157 Ϯ 22 mg/dl. In contrast, their Apoe 2/2 Ldlr h/ϩ littermates had normal plasma cholesterol and triglyc-eride levels of 83 Ϯ 4 and 40 Ϯ 5 mg/dl, respectively. These values in the Apoe 2/2 Ldlr h/ϩ mice are similar to those of wild type mice, suggesting that the 2.5-fold increase in LDLR expression is sufficient to normalize the plasma lipids of the The numerical values represent the total plasma levels in mg/dl Ϯ S.D. from six animals. C, pooled plasma (1 ml) from Apoe ϩ/ϩ Ldlr ϩ/ϩ (ϩ/ϩ). and Apoe ϩ/ϩ Ldlr h/ϩ (h/ϩ) mice fed normal chow was fractionated by sequential density ultracentrifugation, and each fraction was electrophoresed in a denaturing 3-20% SDS-PAGE gradient gel and stained with Coomassie Brilliant Blue. Density fractions are as follows: 1.006, Ͻ1.006 g/ml; 1.02, 1.006 -1.02 g/ml; 1.04, 1.02-1.04 g/ml; 1.06, 1.04 -1.06 g/ml; 1.08, 1.06 -1.08 g/ml; 1.10, 1.08 -1.10 g/ml; and 1.21, 1.10 -1.21 g/ml. D, 100 l of plasma pooled from at least five male mice fed western-style diet was analyzed by FPLC on a Superose 6B column. 0.5-ml fractions were collected, and lipid was measured as g/fraction. Apoe 2/2 mice. The marked reductions in plasma cholesterol and triglycerides observed in the Apoe 2/2 Ldlr h/ϩ mice can be accounted for by the reduction in ␤-VLDL, as shown by agarose gel electrophoresis of whole plasma (Fig. 5A). The lipid content of ␣-migrating HDL particles in Apoe 2/2 Ldlr h/ϩ mice was not reduced in the presence of the hLDLR allele, which is in distinct contrast to the marked reduction of HDL in Apoe ϩ/ϩ Ldlr h/ϩ mice, which have wild type mouse apoE protein.
FPLC analysis of plasma from the Apoe 2/2 Ldlr h/ϩ mice (Fig.  5B) confirmed that reductions in plasma cholesterol were due to the reduction of cholesterol in large lipoproteins (in the VLDL and IDL range). There was no change in HDL cholesterol compared with Apoe 2/2 Ldlr ϩ/ϩ mice. In addition, VLDL triglycerides are dramatically reduced in Apoe 2/2 Ldlr h/ϩ mice, whereas triglycerides in remnant particles (fractions 19 -25) are virtually eliminated. One diagnostic criterion for type III HLP is a ratio of cholesterol/triglyceride in VLDL larger than 0.3. Although the ratio in Apoe 2/2 Ldlr ϩ/ϩ mice is 0.67, the ratio in Apoe 2/2 Ldlr h/ϩ mice is 0.2, less than the value required for the diagnosis of type III HLP in humans.
The amelioration of type III HLP in the mice with modestly increased LDL receptor levels is further shown by SDS-PAGE analysis of apoproteins in lipoproteins isolated by ultracentrifugation (Fig. 5C). As previously reported, Apoe 2/2 Ldlr ϩ/ϩ mice fed normal chow have large amounts of plasma apoB, especially apoB48, as well as apoE in the VLDL fractions (8). In contrast, Apoe 2/2 Ldlr h/ϩ mice have markedly reduced apoB and apoE levels in all the non-HDL lipoprotein classes. One important difference between mice expressing mouse apoE and mice expressing human APOE*2 is that the Apoe 2/2 Ldlr h/ϩ mice do not show any reduction in apoAI (or cholesterol) in the HDL fractions, whereas Apoe ϩ/ϩ Ldlr h/ϩ mice have significant reductions in both.
When challenged by western style diet for 3 weeks, Apoe 2/2 Ldlr h/ϩ mice were still protected from type III HLP. Cholesterol levels increased in Apoe 2/2 Ldlr h/ϩ mice (172 Ϯ 43 mg/dl n ϭ 5) but were significantly lower than in Apoe 2/2 Ldlr ϩ/ϩ mice (546 Ϯ 30 mg/dl, n ϭ 3, p Ͻ 0.0001). The triglycerides in Apoe 2/2 Ldlr h/ϩ mice did not increase compared with levels on normal chow (35 Ϯ 3 mg/dl n ϭ 5), whereas they doubled in the Apoe 2/2 Ldlr ϩ/ϩ mice (310 Ϯ 84 mg/dl n ϭ 3). FPLC analysis of plasma from Apoe 2/2 Ldlr h/ϩ mice (Fig. 5D) fed a high cholesterol diet demonstrated the continued protection against type III HLP, with increased cholesterol predominantly in the HDL fractions. In contrast, the Apoe 2/2 Ldlr ϩ/ϩ mice had markedly increased VLDL triglyceride as well as VLDL-LDL cholesterol with little change in HDL cholesterol (Fig. 5D). Taken together, these results demonstrate that a modest increase in the expression of LDL receptor mRNA can ameliorate the type III HLP phenotype in mice, making them more resistant to diet-induced hyperlipidemia.
Increased Clearance of Non-HDL Lipoproteins from Plasma of Apoe 2/2 Ldlr h/ϩ Mice-To ascertain whether the normalized plasma lipid and lipoprotein levels in Apoe 2/2 Ldlr h/ϩ mice are due to an increase in functional LDL receptor activity, the clearance of radiolabeled VLDL and LDL obtained from mice deficient in apoE was measured. Fig. 6 shows that at 3 h after injection, significantly more VLDL and LDL remain in the plasma of Apoe 2/2 Ldlr ϩ/ϩ mice compared with Apoe 2/2 Ldlr h/ϩ mice. This is consistent with our previous observations that Apoe 2/2 mice are unable to clear VLDL from the plasma completely 4 h post-injection (8). We conclude that the increased steady state LDLR mRNA levels in Apoe 2/2 Ldlr h/ϩ mice results in an increase in functional LDL receptor activity, with an increased fractional catabolic rate of VLDL, as well as LDL, and normalization of plasma lipid levels. DISCUSSION In marked contrast to humans, in which type III HLP affects 5-10% of APOE*2 homozygotes, all mice homozygous for targeted replacement of human APOE*2 exhibit features of type FIG. 5. Plasma lipid and apolipoprotein distribution in mice expressing human apoE2 without (2/2 ؉/؉) and with a human LDLR allele (2/2 h/؉). A, fasted plasma (1 l) from males fed normal chow was electrophoresed in a 1% agarose gel and stained with Fat Red 7B. B, pooled plasma (100 l) from 6 male mice fed normal chow was analyzed by FPLC on a Superose 6B column. 0.5-ml fractions were collected, and lipid was measured as g/fraction. TC, total cholesterol (open circles); TG, triglycerides (closed circles). The numerical values represent the total plasma levels in mg/dl Ϯ S.D. C, pooled plasma (1 ml) from Apoe 2/2 Ldlr ϩ/ϩ (2/2 ϩ/ϩ) and Apoe 2/2 Ldlr h/ϩ (2/2 h/ϩ) mice fed normal chow was fractionated by sequential density ultracentrifugation. Each fraction was electrophoresed in a denaturing 3-20% SDS-PAGE gradient gel and stained with Coomassie Brilliant Blue. Density fractions are as follows: 1.006, Ͻ1.006 g/ml; 1.02, 1.006 -1.02 g/ml; 1.04, 1.02-1.04 g/ml; 1.06, 1.04 -1.06 g/ml; 1.08, 1.06 -1.08 g/ml; 1.10, 1.08 -1.10 g/ml; and 1.21, 1.10 -1.21 g/ml. D, 100 l of plasma pooled from at least five male mice fed western-style diet was analyzed by FPLC on a Superose 6B column. 0.5-ml fractions were collected, and lipid was measured as g/fraction. TC, total cholesterol (open circles); TG, triglycerides (closed circles). The numerical values represent the total plasma levels in mg/dl Ϯ S.D. III hyperlipoproteinemia regardless of age or gender (8). In the current study, we have modestly increased steady state hepatic LDLR mRNA levels by replacing the endogenous mouse Ldlr gene with an hLDLR minigene with increased mRNA stability. The Apoe 2/2 Ldlr h/ϩ mice carrying this allele exhibit a 2.5-fold increase in total hepatic LDLR message, increased clearance of VLDL and LDL particles from the plasma, and a normal plasma lipid phenotype. Not only is this modest increase in hepatic LDLR mRNA sufficient to ameliorate the type III hyperlipoproteinemia, the Apoe 2/2 Ldlr h/ϩ mice are more resistant to diet-induced hyperlipidemia than Apoe 2/2 Ldlr ϩ/ϩ mice. Brown and Goldstein (26) have demonstrated a regulatory pathway that results in the down-regulation of LDLR gene transcription in response to increased intracellular sterol levels. However, regulation of transcription is only one method by which steady state levels of mRNA are altered. While the 5Ј regulatory elements of the human LDLR have been studied in detail, the significance of 3Ј regulatory elements in the regulation of cellular LDL receptor activity is only beginning to be appreciated. A few studies of the LDLR 3Ј-UTR indicate that it may play a significant role in regulation of LDLR mRNA levels. Wilson et al. (28) have identified three AU-rich elements (AREs) at positions 2690, 3257, and 3438 of the human transcript (14) based on sequence homology with the nonameric sequence UUAUUUAUU. This sequence is the minimal element contributing to a rapid turnover of several mRNAs such as those encoding immediate early genes and cytokines (24,25). By fusing the 3Ј-UTR sequence of the human LDLR to the coding region of human ␤-globin gene, Wilson et al. (28) demonstrated that the 3Ј-UTR sequence of the LDLR gene confers a short constitutive half-life to the otherwise stable ␤-globin transcript in cultured cells. They showed that fusion constructs containing all three AREs have a 10-fold higher mRNA turnover rate compared with constructs lacking all three AREs. The three AREs contributed in an additive fashion to the mRNA destabilization rate, with constructs containing only the 5Јmost ARE having just a 3-fold increase in the mRNA degradation rate. Since the chimeric transcript produced by Ldlr h/ϩ mice contains only human sequence up to nucleotide 2804, it lacks two of the AREs. The lack of these two AREs results in a 3-fold increase in the stability of the LDLR message in vitro (28), an increase in stability similar to what is seen in the mice in vivo in our current study. Thus in Ldlr h/ϩ mice there was a 3.8-fold higher steady state level of chimeric human message compared with wild type mouse message. Since both mLDLR and hLDLR alleles have identical transcriptional regulatory sequences, the increase in steady state levels of hLDLR message over the wild type mLDLR message must be the direct result of this increase in message stability.
It is important to note that we found neither down-regulation of the mouse gene accompanying the stabilized human gene nor normalization of the total LDLR mRNA levels in heterozygotes. Furthermore, the messages for both mLDLR and hLDLR alleles are reduced in response to increased dietary cholesterol to similar degrees, maintaining the same steady state ratio observed in mice fed normal chow. This suggests that there is no feedback mechanism to adjust the LDLR mRNA levels through transcriptional regulation when message stability is increased in vivo.
Transgenic mice overproducing the human LDL receptor were reported previously by Yokode et al. (13). These authors used the mouse metallothionein-I promoter sequence to drive expression of the human minigene in the transgenic mice. Similar to our Ldlr h/ϩ mice, the transgenic mice on low fat diet had markedly reduced levels of plasma cholesterol and HDL cholesterol as well as apolipoprotein B and E. The most likely explanation for the low levels of HDL cholesterol in mice with increased LDL receptor activity is increased receptor-mediated clearance of HDL particles with apoE. However, whether other mechanisms such as inhibition of lipoprotein secretion by increased hepatic LDL receptor activity (29) contribute to lower HDL cholesterol in these mice require further studies. The transgenic mice generated by Yokode et al. (13) are completely protected against diet-induced hypercholesterolemia because of their unregulated overexpression of the LDLR. In contrast, HDL cholesterol increased in Ldlr h/ϩ mice in response to increased dietary cholesterol, as in Ldlr ϩ/ϩ mice. This suggests the regulation of transcription of the LDLR gene by increased cellular sterol levels in Ldlr h/ϩ mice is intact.
Whereas many factors besides decreased LDL receptor activity may precipitate type III HLP in human APOE*2 homozygotes (30), it is clear that high hepatic expression of the LDL receptor is sufficient to overcome this phenotype in mice. Previous studies that test the effect of LDL receptor overexpression on type III HLP in mice have used very high levels of uncontrolled overproduction of the LDL receptor (9,31). Our study is the first to show that as little as a 2.5-fold increase in LDL receptor mRNA results in increased clearance of VLDL and a 3-4-fold reduction of plasma cholesterol and triglycerides in mice expressing apoE2 fed normal chow. The fact that most human APOE*2 homozygotes are slightly hypolipidemic, but all Apoe 2/2 Ldlr ϩ/ϩ mice are hyperlipidemic, suggests a possibility that the set point of expression of the LDLR may be higher in humans than in mice. Murine apoE can efficiently mediate clearance of plasma lipoproteins by LDL receptorindependent mechanisms 2 Thus, a relatively low set point of LDLR expression may not influence the overall metabolism of apoE-containing lipoproteins in mice except when mouse apoE is replaced with human apoE2, which is far less effective in mediating LDL receptor-dependent uptake of apoE-containing lipoproteins.
Three-quarters of LDL receptor proteins in Apoe 2/2 Ldlr h/ϩ heterozygotes are of human sequence. Our unpublished studies using mouse fibroblast cells show that VLDL containing mouse apoE or human apoE3 or apoE4 isoforms bind to the mouse LDL receptor equally well, whereas VLDL with apoE2 binds 2 C. Knouff, V. Clavey, and N. Maeda, unpublished observations. FIG. 6. Plasma VLDL and LDL clearance rates in mice fed normal chow. VLDL and LDL were isolated from apoE-deficient mice by ultracentrifugation. VLDL labeled with 125 I and LDL labeled with 131 I were injected via jugular vein into Apoe 2/2 Ldlr ϩ/ϩ (2/2 ϩ/ϩ) or Apoe 2/2 Ldlr h/ϩ (2/2 h/ϩ) mice. At 3 h post-injection, mice were bled retro-orbitally, and the remaining radiolabel in plasma was measured. Values shown are percentage of injected radiolabel amount remaining at 3 h in mean Ϯ S.D.
with similar affinity but with 50% B max compared with VLDL containing the other apoE isoforms. 2 The results were similar when human cells were used. 2 Thus, species differences in the affinity of apoE for the LDL receptor do not appear to be playing a significant role. In contrast, Corsini et al. (32) have shown that human LDL has much lower affinity to mouse receptor than to human receptor, suggesting significant species effects are present in the interaction between apoB100 and the LDL receptor. Although we did not observe any increase of LDL in either heterozygotes (Fig. 4) or homozygotes (data not shown), our present study was carried out in Apoe 2/2 Ldlr h/ϩ heterozygotes to reduce the potential complexities induced by interactions between mouse LDL and the human LDL receptor.
Several studies have shown that cellular LDL receptor activity can be regulated by post-transcriptional mechanisms, including changes in mRNA stability as well as altered protein stability. For example, treatment of HepG2 cells with phorbol-12-myristate-13-acetate has been shown to increase the stability of LDLR mRNA 2-2.5-fold by an indirect effect of destabilization of the actin cytoskeleton by this agent (33). Sequences in the extreme 3Ј-UTR of the LDLR have been shown to confer association of this RNA with the actin cytoskeleton. Lack of this region inhibits the PMA-induced stabilization of this RNA (28). Gemfibrozil, a fibrate drug used in the treatment of type III hyperlipoproteinemia in humans, also has been shown to enhance LDLR mRNA stability in human hepatoma cells (34). In this study pravastatin, an inhibitor of 3-hydroxy-3-methylglutaryl-coenzyme A reductase, significantly increased sterolresponsive element-dependent transcription (with no effect on LDLR mRNA stability), whereas gemfibrozil increased LDLR mRNA stability 4 -6-fold with no change in transcription rate. Whereas both drugs have efficacy in treating type III hyperlipoproteinemia, pravastatin lowers LDL cholesterol preferentially, and gemfibrozil treatment results in larger reductions in VLDL cholesterol and triglycerides (35). Whether these differences are the result of different mechanisms leading to increased LDL receptor activity or are due to other effects of these drugs on peripheral lipolysis or hepatic VLDL secretion has not been resolved.
That the regulation of LDLR mRNA stability may have a physiological role is suggested by the fact that depletion of hepatic sterol by the inhibition of squalene synthase results in increased LDLR mRNA transcription, stability, and translation in rats (36). Whether alterations of LDLR mRNA levels in humans due to regulated alterations in mRNA stability can be a "precipitating factor" for type III hyperlipoproteinemia is unclear. Nevertheless, our present study makes clear that changes in hepatic LDLR levels due solely to differences on LDLR mRNA stability can have a profound effect on plasma lipid levels in a mouse model of type III HLP.